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.create_time = ktime_to_ns(sp->so_seqid.create_time);
 842         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 843         p->o_arg.name = &dentry->d_name;
 844         p->o_arg.server = server;
 845         p->o_arg.bitmask = server->attr_bitmask;
 846         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 847         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 848         if (attrs != NULL && attrs->ia_valid != 0) {
 849                 __be32 verf[2];
 850
 851                 p->o_arg.u.attrs = &p->attrs;
 852                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 853
 854                 verf[0] = jiffies;
 855                 verf[1] = current->pid;
 856                 memcpy(p->o_arg.u.verifier.data, verf,
 857                                 sizeof(p->o_arg.u.verifier.data));
 858         }
 859         p->c_arg.fh = &p->o_res.fh;
 860         p->c_arg.stateid = &p->o_res.stateid;
 861         p->c_arg.seqid = p->o_arg.seqid;
 862         nfs4_init_opendata_res(p);
 863         kref_init(&p->kref);
 864         return p;
 865 err_free:
 866         kfree(p);
 867 err:
 868         dput(parent);
 869         return NULL;
 870 }
 871
 872 static void nfs4_opendata_free(struct kref *kref)
 873 {
 874         struct nfs4_opendata *p = container_of(kref,
 875                         struct nfs4_opendata, kref);
 876         struct super_block *sb = p->dentry->d_sb;
 877
 878         nfs_free_seqid(p->o_arg.seqid);
 879         if (p->state != NULL)
 880                 nfs4_put_open_state(p->state);
 881         nfs4_put_state_owner(p->owner);
 882         dput(p->dir);
 883         dput(p->dentry);
 884         nfs_sb_deactive(sb);
 885         nfs_fattr_free_names(&p->f_attr);
 886         kfree(p);
 887 }
 888
 889 static void nfs4_opendata_put(struct nfs4_opendata *p)
 890 {
 891         if (p != NULL)
 892                 kref_put(&p->kref, nfs4_opendata_free);
 893 }
 894
 895 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 896 {
 897         int ret;
 898
 899         ret = rpc_wait_for_completion_task(task);
 900         return ret;
 901 }
 902
 903 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 904 {
 905         int ret = 0;
 906
 907         if (open_mode & (O_EXCL|O_TRUNC))
 908                 goto out;
 909         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 910                 case FMODE_READ:
 911                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 912                                 && state->n_rdonly != 0;
 913                         break;
 914                 case FMODE_WRITE:
 915                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 916                                 && state->n_wronly != 0;
 917                         break;
 918                 case FMODE_READ|FMODE_WRITE:
 919                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 920                                 && state->n_rdwr != 0;
 921         }
 922 out:
 923         return ret;
 924 }
 925
 926 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 927 {
 928         if (delegation == NULL)
 929                 return 0;
 930         if ((delegation->type & fmode) != fmode)
 931                 return 0;
 932         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 933                 return 0;
 934         nfs_mark_delegation_referenced(delegation);
 935         return 1;
 936 }
 937
 938 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 939 {
 940         switch (fmode) {
 941                 case FMODE_WRITE:
 942                         state->n_wronly++;
 943                         break;
 944                 case FMODE_READ:
 945                         state->n_rdonly++;
 946                         break;
 947                 case FMODE_READ|FMODE_WRITE:
 948                         state->n_rdwr++;
 949         }
 950         nfs4_state_set_mode_locked(state, state->state | fmode);
 951 }
 952
 953 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 954 {
 955         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 956                 nfs4_stateid_copy(&state->stateid, stateid);
 957         nfs4_stateid_copy(&state->open_stateid, stateid);
 958         switch (fmode) {
 959                 case FMODE_READ:
 960                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 961                         break;
 962                 case FMODE_WRITE:
 963                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 964                         break;
 965                 case FMODE_READ|FMODE_WRITE:
 966                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 967         }
 968 }
 969
 970 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 971 {
 972         write_seqlock(&state->seqlock);
 973         nfs_set_open_stateid_locked(state, stateid, fmode);
 974         write_sequnlock(&state->seqlock);
 975 }
 976
 977 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 978 {
 979         /*
 980          * Protect the call to nfs4_state_set_mode_locked and
 981          * serialise the stateid update
 982          */
 983         write_seqlock(&state->seqlock);
 984         if (deleg_stateid != NULL) {
 985                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
 986                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 987         }
 988         if (open_stateid != NULL)
 989                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 990         write_sequnlock(&state->seqlock);
 991         spin_lock(&state->owner->so_lock);
 992         update_open_stateflags(state, fmode);
 993         spin_unlock(&state->owner->so_lock);
 994 }
 995
 996 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 997 {
 998         struct nfs_inode *nfsi = NFS_I(state->inode);
 999         struct nfs_delegation *deleg_cur;
1000         int ret = 0;
1001
1002         fmode &= (FMODE_READ|FMODE_WRITE);
1003
1004         rcu_read_lock();
1005         deleg_cur = rcu_dereference(nfsi->delegation);
1006         if (deleg_cur == NULL)
1007                 goto no_delegation;
1008
1009         spin_lock(&deleg_cur->lock);
1010         if (nfsi->delegation != deleg_cur ||
1011             (deleg_cur->type & fmode) != fmode)
1012                 goto no_delegation_unlock;
1013
1014         if (delegation == NULL)
1015                 delegation = &deleg_cur->stateid;
1016         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1017                 goto no_delegation_unlock;
1018
1019         nfs_mark_delegation_referenced(deleg_cur);
1020         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1021         ret = 1;
1022 no_delegation_unlock:
1023         spin_unlock(&deleg_cur->lock);
1024 no_delegation:
1025         rcu_read_unlock();
1026
1027         if (!ret && open_stateid != NULL) {
1028                 __update_open_stateid(state, open_stateid, NULL, fmode);
1029                 ret = 1;
1030         }
1031
1032         return ret;
1033 }
1034
1035
1036 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1037 {
1038         struct nfs_delegation *delegation;
1039
1040         rcu_read_lock();
1041         delegation = rcu_dereference(NFS_I(inode)->delegation);
1042         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1043                 rcu_read_unlock();
1044                 return;
1045         }
1046         rcu_read_unlock();
1047         nfs_inode_return_delegation(inode);
1048 }
1049
1050 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1051 {
1052         struct nfs4_state *state = opendata->state;
1053         struct nfs_inode *nfsi = NFS_I(state->inode);
1054         struct nfs_delegation *delegation;
1055         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1056         fmode_t fmode = opendata->o_arg.fmode;
1057         nfs4_stateid stateid;
1058         int ret = -EAGAIN;
1059
1060         for (;;) {
1061                 if (can_open_cached(state, fmode, open_mode)) {
1062                         spin_lock(&state->owner->so_lock);
1063                         if (can_open_cached(state, fmode, open_mode)) {
1064                                 update_open_stateflags(state, fmode);
1065                                 spin_unlock(&state->owner->so_lock);
1066                                 goto out_return_state;
1067                         }
1068                         spin_unlock(&state->owner->so_lock);
1069                 }
1070                 rcu_read_lock();
1071                 delegation = rcu_dereference(nfsi->delegation);
1072                 if (!can_open_delegated(delegation, fmode)) {
1073                         rcu_read_unlock();
1074                         break;
1075                 }
1076                 /* Save the delegation */
1077                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1078                 rcu_read_unlock();
1079                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1080                 if (ret != 0)
1081                         goto out;
1082                 ret = -EAGAIN;
1083
1084                 /* Try to update the stateid using the delegation */
1085                 if (update_open_stateid(state, NULL, &stateid, fmode))
1086                         goto out_return_state;
1087         }
1088 out:
1089         return ERR_PTR(ret);
1090 out_return_state:
1091         atomic_inc(&state->count);
1092         return state;
1093 }
1094
1095 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1096 {
1097         struct inode *inode;
1098         struct nfs4_state *state = NULL;
1099         struct nfs_delegation *delegation;
1100         int ret;
1101
1102         if (!data->rpc_done) {
1103                 state = nfs4_try_open_cached(data);
1104                 goto out;
1105         }
1106
1107         ret = -EAGAIN;
1108         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1109                 goto err;
1110         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1111         ret = PTR_ERR(inode);
1112         if (IS_ERR(inode))
1113                 goto err;
1114         ret = -ENOMEM;
1115         state = nfs4_get_open_state(inode, data->owner);
1116         if (state == NULL)
1117                 goto err_put_inode;
1118         if (data->o_res.delegation_type != 0) {
1119                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1120                 int delegation_flags = 0;
1121
1122                 rcu_read_lock();
1123                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1124                 if (delegation)
1125                         delegation_flags = delegation->flags;
1126                 rcu_read_unlock();
1127                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1128                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1129                                         "returning a delegation for "
1130                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1131                                         clp->cl_hostname);
1132                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1133                         nfs_inode_set_delegation(state->inode,
1134                                         data->owner->so_cred,
1135                                         &data->o_res);
1136                 else
1137                         nfs_inode_reclaim_delegation(state->inode,
1138                                         data->owner->so_cred,
1139                                         &data->o_res);
1140         }
1141
1142         update_open_stateid(state, &data->o_res.stateid, NULL,
1143                         data->o_arg.fmode);
1144         iput(inode);
1145 out:
1146         return state;
1147 err_put_inode:
1148         iput(inode);
1149 err:
1150         return ERR_PTR(ret);
1151 }
1152
1153 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1154 {
1155         struct nfs_inode *nfsi = NFS_I(state->inode);
1156         struct nfs_open_context *ctx;
1157
1158         spin_lock(&state->inode->i_lock);
1159         list_for_each_entry(ctx, &nfsi->open_files, list) {
1160                 if (ctx->state != state)
1161                         continue;
1162                 get_nfs_open_context(ctx);
1163                 spin_unlock(&state->inode->i_lock);
1164                 return ctx;
1165         }
1166         spin_unlock(&state->inode->i_lock);
1167         return ERR_PTR(-ENOENT);
1168 }
1169
1170 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1171 {
1172         struct nfs4_opendata *opendata;
1173
1174         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1175         if (opendata == NULL)
1176                 return ERR_PTR(-ENOMEM);
1177         opendata->state = state;
1178         atomic_inc(&state->count);
1179         return opendata;
1180 }
1181
1182 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1183 {
1184         struct nfs4_state *newstate;
1185         int ret;
1186
1187         opendata->o_arg.open_flags = 0;
1188         opendata->o_arg.fmode = fmode;
1189         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1190         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1191         nfs4_init_opendata_res(opendata);
1192         ret = _nfs4_recover_proc_open(opendata);
1193         if (ret != 0)
1194                 return ret;
1195         newstate = nfs4_opendata_to_nfs4_state(opendata);
1196         if (IS_ERR(newstate))
1197                 return PTR_ERR(newstate);
1198         nfs4_close_state(newstate, fmode);
1199         *res = newstate;
1200         return 0;
1201 }
1202
1203 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1204 {
1205         struct nfs4_state *newstate;
1206         int ret;
1207
1208         /* memory barrier prior to reading state->n_* */
1209         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1210         smp_rmb();
1211         if (state->n_rdwr != 0) {
1212                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1213                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1214                 if (ret != 0)
1215                         return ret;
1216                 if (newstate != state)
1217                         return -ESTALE;
1218         }
1219         if (state->n_wronly != 0) {
1220                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1221                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1222                 if (ret != 0)
1223                         return ret;
1224                 if (newstate != state)
1225                         return -ESTALE;
1226         }
1227         if (state->n_rdonly != 0) {
1228                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1229                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1230                 if (ret != 0)
1231                         return ret;
1232                 if (newstate != state)
1233                         return -ESTALE;
1234         }
1235         /*
1236          * We may have performed cached opens for all three recoveries.
1237          * Check if we need to update the current stateid.
1238          */
1239         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1240             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1241                 write_seqlock(&state->seqlock);
1242                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1243                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1244                 write_sequnlock(&state->seqlock);
1245         }
1246         return 0;
1247 }
1248
1249 /*
1250  * OPEN_RECLAIM:
1251  *      reclaim state on the server after a reboot.
1252  */
1253 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1254 {
1255         struct nfs_delegation *delegation;
1256         struct nfs4_opendata *opendata;
1257         fmode_t delegation_type = 0;
1258         int status;
1259
1260         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1261         if (IS_ERR(opendata))
1262                 return PTR_ERR(opendata);
1263         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1264         opendata->o_arg.fh = NFS_FH(state->inode);
1265         rcu_read_lock();
1266         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1267         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1268                 delegation_type = delegation->type;
1269         rcu_read_unlock();
1270         opendata->o_arg.u.delegation_type = delegation_type;
1271         status = nfs4_open_recover(opendata, state);
1272         nfs4_opendata_put(opendata);
1273         return status;
1274 }
1275
1276 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1277 {
1278         struct nfs_server *server = NFS_SERVER(state->inode);
1279         struct nfs4_exception exception = { };
1280         int err;
1281         do {
1282                 err = _nfs4_do_open_reclaim(ctx, state);
1283                 if (err != -NFS4ERR_DELAY)
1284                         break;
1285                 nfs4_handle_exception(server, err, &exception);
1286         } while (exception.retry);
1287         return err;
1288 }
1289
1290 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1291 {
1292         struct nfs_open_context *ctx;
1293         int ret;
1294
1295         ctx = nfs4_state_find_open_context(state);
1296         if (IS_ERR(ctx))
1297                 return PTR_ERR(ctx);
1298         ret = nfs4_do_open_reclaim(ctx, state);
1299         put_nfs_open_context(ctx);
1300         return ret;
1301 }
1302
1303 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1304 {
1305         struct nfs4_opendata *opendata;
1306         int ret;
1307
1308         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1309         if (IS_ERR(opendata))
1310                 return PTR_ERR(opendata);
1311         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1312         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1313         ret = nfs4_open_recover(opendata, state);
1314         nfs4_opendata_put(opendata);
1315         return ret;
1316 }
1317
1318 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1319 {
1320         struct nfs4_exception exception = { };
1321         struct nfs_server *server = NFS_SERVER(state->inode);
1322         int err;
1323         do {
1324                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1325                 switch (err) {
1326                         case 0:
1327                         case -ENOENT:
1328                         case -ESTALE:
1329                                 goto out;
1330                         case -NFS4ERR_BADSESSION:
1331                         case -NFS4ERR_BADSLOT:
1332                         case -NFS4ERR_BAD_HIGH_SLOT:
1333                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1334                         case -NFS4ERR_DEADSESSION:
1335                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1336                                 goto out;
1337                         case -NFS4ERR_STALE_CLIENTID:
1338                         case -NFS4ERR_STALE_STATEID:
1339                         case -NFS4ERR_EXPIRED:
1340                                 /* Don't recall a delegation if it was lost */
1341                                 nfs4_schedule_lease_recovery(server->nfs_client);
1342                                 goto out;
1343                         case -ERESTARTSYS:
1344                                 /*
1345                                  * The show must go on: exit, but mark the
1346                                  * stateid as needing recovery.
1347                                  */
1348                         case -NFS4ERR_DELEG_REVOKED:
1349                         case -NFS4ERR_ADMIN_REVOKED:
1350                         case -NFS4ERR_BAD_STATEID:
1351                                 nfs_inode_find_state_and_recover(state->inode,
1352                                                 stateid);
1353                                 nfs4_schedule_stateid_recovery(server, state);
1354                         case -EKEYEXPIRED:
1355                                 /*
1356                                  * User RPCSEC_GSS context has expired.
1357                                  * We cannot recover this stateid now, so
1358                                  * skip it and allow recovery thread to
1359                                  * proceed.
1360                                  */
1361                         case -ENOMEM:
1362                                 err = 0;
1363                                 goto out;
1364                 }
1365                 err = nfs4_handle_exception(server, err, &exception);
1366         } while (exception.retry);
1367 out:
1368         return err;
1369 }
1370
1371 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1372 {
1373         struct nfs4_opendata *data = calldata;
1374
1375         data->rpc_status = task->tk_status;
1376         if (data->rpc_status == 0) {
1377                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1378                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1379                 renew_lease(data->o_res.server, data->timestamp);
1380                 data->rpc_done = 1;
1381         }
1382 }
1383
1384 static void nfs4_open_confirm_release(void *calldata)
1385 {
1386         struct nfs4_opendata *data = calldata;
1387         struct nfs4_state *state = NULL;
1388
1389         /* If this request hasn't been cancelled, do nothing */
1390         if (data->cancelled == 0)
1391                 goto out_free;
1392         /* In case of error, no cleanup! */
1393         if (!data->rpc_done)
1394                 goto out_free;
1395         state = nfs4_opendata_to_nfs4_state(data);
1396         if (!IS_ERR(state))
1397                 nfs4_close_state(state, data->o_arg.fmode);
1398 out_free:
1399         nfs4_opendata_put(data);
1400 }
1401
1402 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1403         .rpc_call_done = nfs4_open_confirm_done,
1404         .rpc_release = nfs4_open_confirm_release,
1405 };
1406
1407 /*
1408  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1409  */
1410 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1411 {
1412         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1413         struct rpc_task *task;
1414         struct  rpc_message msg = {
1415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1416                 .rpc_argp = &data->c_arg,
1417                 .rpc_resp = &data->c_res,
1418                 .rpc_cred = data->owner->so_cred,
1419         };
1420         struct rpc_task_setup task_setup_data = {
1421                 .rpc_client = server->client,
1422                 .rpc_message = &msg,
1423                 .callback_ops = &nfs4_open_confirm_ops,
1424                 .callback_data = data,
1425                 .workqueue = nfsiod_workqueue,
1426                 .flags = RPC_TASK_ASYNC,
1427         };
1428         int status;
1429
1430         kref_get(&data->kref);
1431         data->rpc_done = 0;
1432         data->rpc_status = 0;
1433         data->timestamp = jiffies;
1434         task = rpc_run_task(&task_setup_data);
1435         if (IS_ERR(task))
1436                 return PTR_ERR(task);
1437         status = nfs4_wait_for_completion_rpc_task(task);
1438         if (status != 0) {
1439                 data->cancelled = 1;
1440                 smp_wmb();
1441         } else
1442                 status = data->rpc_status;
1443         rpc_put_task(task);
1444         return status;
1445 }
1446
1447 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1448 {
1449         struct nfs4_opendata *data = calldata;
1450         struct nfs4_state_owner *sp = data->owner;
1451
1452         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1453                 return;
1454         /*
1455          * Check if we still need to send an OPEN call, or if we can use
1456          * a delegation instead.
1457          */
1458         if (data->state != NULL) {
1459                 struct nfs_delegation *delegation;
1460
1461                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1462                         goto out_no_action;
1463                 rcu_read_lock();
1464                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1465                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1466                     can_open_delegated(delegation, data->o_arg.fmode))
1467                         goto unlock_no_action;
1468                 rcu_read_unlock();
1469         }
1470         /* Update client 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_do_setattr(inode, cred, fattr, sattr, state);
1958                 switch (err) {
1959                 case -NFS4ERR_OPENMODE:
1960                         if (state && !(state->state & FMODE_WRITE)) {
1961                                 err = -EBADF;
1962                                 if (sattr->ia_valid & ATTR_OPEN)
1963                                         err = -EACCES;
1964                                 goto out;
1965                         }
1966                 }
1967                 err = nfs4_handle_exception(server, err, &exception);
1968         } while (exception.retry);
1969 out:
1970         return err;
1971 }
1972
1973 struct nfs4_closedata {
1974         struct inode *inode;
1975         struct nfs4_state *state;
1976         struct nfs_closeargs arg;
1977         struct nfs_closeres res;
1978         struct nfs_fattr fattr;
1979         unsigned long timestamp;
1980         bool roc;
1981         u32 roc_barrier;
1982 };
1983
1984 static void nfs4_free_closedata(void *data)
1985 {
1986         struct nfs4_closedata *calldata = data;
1987         struct nfs4_state_owner *sp = calldata->state->owner;
1988         struct super_block *sb = calldata->state->inode->i_sb;
1989
1990         if (calldata->roc)
1991                 pnfs_roc_release(calldata->state->inode);
1992         nfs4_put_open_state(calldata->state);
1993         nfs_free_seqid(calldata->arg.seqid);
1994         nfs4_put_state_owner(sp);
1995         nfs_sb_deactive(sb);
1996         kfree(calldata);
1997 }
1998
1999 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2000                 fmode_t fmode)
2001 {
2002         spin_lock(&state->owner->so_lock);
2003         if (!(fmode & FMODE_READ))
2004                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2005         if (!(fmode & FMODE_WRITE))
2006                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2007         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2008         spin_unlock(&state->owner->so_lock);
2009 }
2010
2011 static void nfs4_close_done(struct rpc_task *task, void *data)
2012 {
2013         struct nfs4_closedata *calldata = data;
2014         struct nfs4_state *state = calldata->state;
2015         struct nfs_server *server = NFS_SERVER(calldata->inode);
2016
2017         dprintk("%s: begin!\n", __func__);
2018         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2019                 return;
2020         /* hmm. we are done with the inode, and in the process of freeing
2021          * the state_owner. we keep this around to process errors
2022          */
2023         switch (task->tk_status) {
2024                 case 0:
2025                         if (calldata->roc)
2026                                 pnfs_roc_set_barrier(state->inode,
2027                                                      calldata->roc_barrier);
2028                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2029                         renew_lease(server, calldata->timestamp);
2030                         nfs4_close_clear_stateid_flags(state,
2031                                         calldata->arg.fmode);
2032                         break;
2033                 case -NFS4ERR_STALE_STATEID:
2034                 case -NFS4ERR_OLD_STATEID:
2035                 case -NFS4ERR_BAD_STATEID:
2036                 case -NFS4ERR_EXPIRED:
2037                         if (calldata->arg.fmode == 0)
2038                                 break;
2039                 default:
2040                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2041                                 rpc_restart_call_prepare(task);
2042         }
2043         nfs_release_seqid(calldata->arg.seqid);
2044         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2045         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2046 }
2047
2048 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2049 {
2050         struct nfs4_closedata *calldata = data;
2051         struct nfs4_state *state = calldata->state;
2052         int call_close = 0;
2053
2054         dprintk("%s: begin!\n", __func__);
2055         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2056                 return;
2057
2058         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2059         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2060         spin_lock(&state->owner->so_lock);
2061         /* Calculate the change in open mode */
2062         if (state->n_rdwr == 0) {
2063                 if (state->n_rdonly == 0) {
2064                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2065                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2066                         calldata->arg.fmode &= ~FMODE_READ;
2067                 }
2068                 if (state->n_wronly == 0) {
2069                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2070                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2071                         calldata->arg.fmode &= ~FMODE_WRITE;
2072                 }
2073         }
2074         spin_unlock(&state->owner->so_lock);
2075
2076         if (!call_close) {
2077                 /* Note: exit _without_ calling nfs4_close_done */
2078                 task->tk_action = NULL;
2079                 goto out;
2080         }
2081
2082         if (calldata->arg.fmode == 0) {
2083                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2084                 if (calldata->roc &&
2085                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2086                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2087                                      task, NULL);
2088                         goto out;
2089                 }
2090         }
2091
2092         nfs_fattr_init(calldata->res.fattr);
2093         calldata->timestamp = jiffies;
2094         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2095                                 &calldata->arg.seq_args,
2096                                 &calldata->res.seq_res,
2097                                 task))
2098                 goto out;
2099         rpc_call_start(task);
2100 out:
2101         dprintk("%s: done!\n", __func__);
2102 }
2103
2104 static const struct rpc_call_ops nfs4_close_ops = {
2105         .rpc_call_prepare = nfs4_close_prepare,
2106         .rpc_call_done = nfs4_close_done,
2107         .rpc_release = nfs4_free_closedata,
2108 };
2109
2110 /*
2111  * It is possible for data to be read/written from a mem-mapped file
2112  * after the sys_close call (which hits the vfs layer as a flush).
2113  * This means that we can't safely call nfsv4 close on a file until
2114  * the inode is cleared. This in turn means that we are not good
2115  * NFSv4 citizens - we do not indicate to the server to update the file's
2116  * share state even when we are done with one of the three share
2117  * stateid's in the inode.
2118  *
2119  * NOTE: Caller must be holding the sp->so_owner semaphore!
2120  */
2121 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2122 {
2123         struct nfs_server *server = NFS_SERVER(state->inode);
2124         struct nfs4_closedata *calldata;
2125         struct nfs4_state_owner *sp = state->owner;
2126         struct rpc_task *task;
2127         struct rpc_message msg = {
2128                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2129                 .rpc_cred = state->owner->so_cred,
2130         };
2131         struct rpc_task_setup task_setup_data = {
2132                 .rpc_client = server->client,
2133                 .rpc_message = &msg,
2134                 .callback_ops = &nfs4_close_ops,
2135                 .workqueue = nfsiod_workqueue,
2136                 .flags = RPC_TASK_ASYNC,
2137         };
2138         int status = -ENOMEM;
2139
2140         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2141         if (calldata == NULL)
2142                 goto out;
2143         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2144         calldata->inode = state->inode;
2145         calldata->state = state;
2146         calldata->arg.fh = NFS_FH(state->inode);
2147         calldata->arg.stateid = &state->open_stateid;
2148         /* Serialization for the sequence id */
2149         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2150         if (calldata->arg.seqid == NULL)
2151                 goto out_free_calldata;
2152         calldata->arg.fmode = 0;
2153         calldata->arg.bitmask = server->cache_consistency_bitmask;
2154         calldata->res.fattr = &calldata->fattr;
2155         calldata->res.seqid = calldata->arg.seqid;
2156         calldata->res.server = server;
2157         calldata->roc = roc;
2158         nfs_sb_active(calldata->inode->i_sb);
2159
2160         msg.rpc_argp = &calldata->arg;
2161         msg.rpc_resp = &calldata->res;
2162         task_setup_data.callback_data = calldata;
2163         task = rpc_run_task(&task_setup_data);
2164         if (IS_ERR(task))
2165                 return PTR_ERR(task);
2166         status = 0;
2167         if (wait)
2168                 status = rpc_wait_for_completion_task(task);
2169         rpc_put_task(task);
2170         return status;
2171 out_free_calldata:
2172         kfree(calldata);
2173 out:
2174         if (roc)
2175                 pnfs_roc_release(state->inode);
2176         nfs4_put_open_state(state);
2177         nfs4_put_state_owner(sp);
2178         return status;
2179 }
2180
2181 static struct inode *
2182 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2183 {
2184         struct nfs4_state *state;
2185
2186         /* Protect against concurrent sillydeletes */
2187         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2188         if (IS_ERR(state))
2189                 return ERR_CAST(state);
2190         ctx->state = state;
2191         return igrab(state->inode);
2192 }
2193
2194 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2195 {
2196         if (ctx->state == NULL)
2197                 return;
2198         if (is_sync)
2199                 nfs4_close_sync(ctx->state, ctx->mode);
2200         else
2201                 nfs4_close_state(ctx->state, ctx->mode);
2202 }
2203
2204 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2205 {
2206         struct nfs4_server_caps_arg args = {
2207                 .fhandle = fhandle,
2208         };
2209         struct nfs4_server_caps_res res = {};
2210         struct rpc_message msg = {
2211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2212                 .rpc_argp = &args,
2213                 .rpc_resp = &res,
2214         };
2215         int status;
2216
2217         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2218         if (status == 0) {
2219                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2220                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2221                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2222                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2223                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2224                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2225                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2226                         server->caps |= NFS_CAP_ACLS;
2227                 if (res.has_links != 0)
2228                         server->caps |= NFS_CAP_HARDLINKS;
2229                 if (res.has_symlinks != 0)
2230                         server->caps |= NFS_CAP_SYMLINKS;
2231                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2232                         server->caps |= NFS_CAP_FILEID;
2233                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2234                         server->caps |= NFS_CAP_MODE;
2235                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2236                         server->caps |= NFS_CAP_NLINK;
2237                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2238                         server->caps |= NFS_CAP_OWNER;
2239                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2240                         server->caps |= NFS_CAP_OWNER_GROUP;
2241                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2242                         server->caps |= NFS_CAP_ATIME;
2243                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2244                         server->caps |= NFS_CAP_CTIME;
2245                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2246                         server->caps |= NFS_CAP_MTIME;
2247
2248                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2249                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2250                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2251                 server->acl_bitmask = res.acl_bitmask;
2252                 server->fh_expire_type = res.fh_expire_type;
2253         }
2254
2255         return status;
2256 }
2257
2258 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2259 {
2260         struct nfs4_exception exception = { };
2261         int err;
2262         do {
2263                 err = nfs4_handle_exception(server,
2264                                 _nfs4_server_capabilities(server, fhandle),
2265                                 &exception);
2266         } while (exception.retry);
2267         return err;
2268 }
2269
2270 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2271                 struct nfs_fsinfo *info)
2272 {
2273         struct nfs4_lookup_root_arg args = {
2274                 .bitmask = nfs4_fattr_bitmap,
2275         };
2276         struct nfs4_lookup_res res = {
2277                 .server = server,
2278                 .fattr = info->fattr,
2279                 .fh = fhandle,
2280         };
2281         struct rpc_message msg = {
2282                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2283                 .rpc_argp = &args,
2284                 .rpc_resp = &res,
2285         };
2286
2287         nfs_fattr_init(info->fattr);
2288         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2289 }
2290
2291 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2292                 struct nfs_fsinfo *info)
2293 {
2294         struct nfs4_exception exception = { };
2295         int err;
2296         do {
2297                 err = _nfs4_lookup_root(server, fhandle, info);
2298                 switch (err) {
2299                 case 0:
2300                 case -NFS4ERR_WRONGSEC:
2301                         goto out;
2302                 default:
2303                         err = nfs4_handle_exception(server, err, &exception);
2304                 }
2305         } while (exception.retry);
2306 out:
2307         return err;
2308 }
2309
2310 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2311                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2312 {
2313         struct rpc_auth *auth;
2314         int ret;
2315
2316         auth = rpcauth_create(flavor, server->client);
2317         if (!auth) {
2318                 ret = -EIO;
2319                 goto out;
2320         }
2321         ret = nfs4_lookup_root(server, fhandle, info);
2322 out:
2323         return ret;
2324 }
2325
2326 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2327                               struct nfs_fsinfo *info)
2328 {
2329         int i, len, status = 0;
2330         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2331
2332         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2333         flav_array[len] = RPC_AUTH_NULL;
2334         len += 1;
2335
2336         for (i = 0; i < len; i++) {
2337                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2338                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2339                         continue;
2340                 break;
2341         }
2342         /*
2343          * -EACCESS could mean that the user doesn't have correct permissions
2344          * to access the mount.  It could also mean that we tried to mount
2345          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2346          * existing mount programs don't handle -EACCES very well so it should
2347          * be mapped to -EPERM instead.
2348          */
2349         if (status == -EACCES)
2350                 status = -EPERM;
2351         return status;
2352 }
2353
2354 /*
2355  * get the file handle for the "/" directory on the server
2356  */
2357 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2358                               struct nfs_fsinfo *info)
2359 {
2360         int minor_version = server->nfs_client->cl_minorversion;
2361         int status = nfs4_lookup_root(server, fhandle, info);
2362         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2363                 /*
2364                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2365                  * by nfs4_map_errors() as this function exits.
2366                  */
2367                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2368         if (status == 0)
2369                 status = nfs4_server_capabilities(server, fhandle);
2370         if (status == 0)
2371                 status = nfs4_do_fsinfo(server, fhandle, info);
2372         return nfs4_map_errors(status);
2373 }
2374
2375 /*
2376  * Get locations and (maybe) other attributes of a referral.
2377  * Note that we'll actually follow the referral later when
2378  * we detect fsid mismatch in inode revalidation
2379  */
2380 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2381                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2382 {
2383         int status = -ENOMEM;
2384         struct page *page = NULL;
2385         struct nfs4_fs_locations *locations = NULL;
2386
2387         page = alloc_page(GFP_KERNEL);
2388         if (page == NULL)
2389                 goto out;
2390         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2391         if (locations == NULL)
2392                 goto out;
2393
2394         status = nfs4_proc_fs_locations(dir, name, locations, page);
2395         if (status != 0)
2396                 goto out;
2397         /* Make sure server returned a different fsid for the referral */
2398         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2399                 dprintk("%s: server did not return a different fsid for"
2400                         " a referral at %s\n", __func__, name->name);
2401                 status = -EIO;
2402                 goto out;
2403         }
2404         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2405         nfs_fixup_referral_attributes(&locations->fattr);
2406
2407         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2408         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2409         memset(fhandle, 0, sizeof(struct nfs_fh));
2410 out:
2411         if (page)
2412                 __free_page(page);
2413         kfree(locations);
2414         return status;
2415 }
2416
2417 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2418 {
2419         struct nfs4_getattr_arg args = {
2420                 .fh = fhandle,
2421                 .bitmask = server->attr_bitmask,
2422         };
2423         struct nfs4_getattr_res res = {
2424                 .fattr = fattr,
2425                 .server = server,
2426         };
2427         struct rpc_message msg = {
2428                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2429                 .rpc_argp = &args,
2430                 .rpc_resp = &res,
2431         };
2432
2433         nfs_fattr_init(fattr);
2434         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2435 }
2436
2437 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2438 {
2439         struct nfs4_exception exception = { };
2440         int err;
2441         do {
2442                 err = nfs4_handle_exception(server,
2443                                 _nfs4_proc_getattr(server, fhandle, fattr),
2444                                 &exception);
2445         } while (exception.retry);
2446         return err;
2447 }
2448
2449 /*
2450  * The file is not closed if it is opened due to the a request to change
2451  * the size of the file. The open call will not be needed once the
2452  * VFS layer lookup-intents are implemented.
2453  *
2454  * Close is called when the inode is destroyed.
2455  * If we haven't opened the file for O_WRONLY, we
2456  * need to in the size_change case to obtain a stateid.
2457  *
2458  * Got race?
2459  * Because OPEN is always done by name in nfsv4, it is
2460  * possible that we opened a different file by the same
2461  * name.  We can recognize this race condition, but we
2462  * can't do anything about it besides returning an error.
2463  *
2464  * This will be fixed with VFS changes (lookup-intent).
2465  */
2466 static int
2467 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2468                   struct iattr *sattr)
2469 {
2470         struct inode *inode = dentry->d_inode;
2471         struct rpc_cred *cred = NULL;
2472         struct nfs4_state *state = NULL;
2473         int status;
2474
2475         if (pnfs_ld_layoutret_on_setattr(inode))
2476                 pnfs_return_layout(inode);
2477
2478         nfs_fattr_init(fattr);
2479
2480         /* Search for an existing open(O_WRITE) file */
2481         if (sattr->ia_valid & ATTR_FILE) {
2482                 struct nfs_open_context *ctx;
2483
2484                 ctx = nfs_file_open_context(sattr->ia_file);
2485                 if (ctx) {
2486                         cred = ctx->cred;
2487                         state = ctx->state;
2488                 }
2489         }
2490
2491         /* Deal with open(O_TRUNC) */
2492         if (sattr->ia_valid & ATTR_OPEN)
2493                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2494
2495         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2496         if (status == 0)
2497                 nfs_setattr_update_inode(inode, sattr);
2498         return status;
2499 }
2500
2501 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2502                 const struct qstr *name, struct nfs_fh *fhandle,
2503                 struct nfs_fattr *fattr)
2504 {
2505         struct nfs_server *server = NFS_SERVER(dir);
2506         int                    status;
2507         struct nfs4_lookup_arg args = {
2508                 .bitmask = server->attr_bitmask,
2509                 .dir_fh = NFS_FH(dir),
2510                 .name = name,
2511         };
2512         struct nfs4_lookup_res res = {
2513                 .server = server,
2514                 .fattr = fattr,
2515                 .fh = fhandle,
2516         };
2517         struct rpc_message msg = {
2518                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2519                 .rpc_argp = &args,
2520                 .rpc_resp = &res,
2521         };
2522
2523         nfs_fattr_init(fattr);
2524
2525         dprintk("NFS call  lookup %s\n", name->name);
2526         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2527         dprintk("NFS reply lookup: %d\n", status);
2528         return status;
2529 }
2530
2531 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2532 {
2533         memset(fh, 0, sizeof(struct nfs_fh));
2534         fattr->fsid.major = 1;
2535         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2536                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2537         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2538         fattr->nlink = 2;
2539 }
2540
2541 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2542                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2543 {
2544         struct nfs4_exception exception = { };
2545         int err;
2546         do {
2547                 int status;
2548
2549                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2550                 switch (status) {
2551                 case -NFS4ERR_BADNAME:
2552                         return -ENOENT;
2553                 case -NFS4ERR_MOVED:
2554                         return nfs4_get_referral(dir, name, fattr, fhandle);
2555                 case -NFS4ERR_WRONGSEC:
2556                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2557                 }
2558                 err = nfs4_handle_exception(NFS_SERVER(dir),
2559                                 status, &exception);
2560         } while (exception.retry);
2561         return err;
2562 }
2563
2564 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2565 {
2566         struct nfs_server *server = NFS_SERVER(inode);
2567         struct nfs4_accessargs args = {
2568                 .fh = NFS_FH(inode),
2569                 .bitmask = server->cache_consistency_bitmask,
2570         };
2571         struct nfs4_accessres res = {
2572                 .server = server,
2573         };
2574         struct rpc_message msg = {
2575                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2576                 .rpc_argp = &args,
2577                 .rpc_resp = &res,
2578                 .rpc_cred = entry->cred,
2579         };
2580         int mode = entry->mask;
2581         int status;
2582
2583         /*
2584          * Determine which access bits we want to ask for...
2585          */
2586         if (mode & MAY_READ)
2587                 args.access |= NFS4_ACCESS_READ;
2588         if (S_ISDIR(inode->i_mode)) {
2589                 if (mode & MAY_WRITE)
2590                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2591                 if (mode & MAY_EXEC)
2592                         args.access |= NFS4_ACCESS_LOOKUP;
2593         } else {
2594                 if (mode & MAY_WRITE)
2595                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2596                 if (mode & MAY_EXEC)
2597                         args.access |= NFS4_ACCESS_EXECUTE;
2598         }
2599
2600         res.fattr = nfs_alloc_fattr();
2601         if (res.fattr == NULL)
2602                 return -ENOMEM;
2603
2604         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2605         if (!status) {
2606                 entry->mask = 0;
2607                 if (res.access & NFS4_ACCESS_READ)
2608                         entry->mask |= MAY_READ;
2609                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2610                         entry->mask |= MAY_WRITE;
2611                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2612                         entry->mask |= MAY_EXEC;
2613                 nfs_refresh_inode(inode, res.fattr);
2614         }
2615         nfs_free_fattr(res.fattr);
2616         return status;
2617 }
2618
2619 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2620 {
2621         struct nfs4_exception exception = { };
2622         int err;
2623         do {
2624                 err = nfs4_handle_exception(NFS_SERVER(inode),
2625                                 _nfs4_proc_access(inode, entry),
2626                                 &exception);
2627         } while (exception.retry);
2628         return err;
2629 }
2630
2631 /*
2632  * TODO: For the time being, we don't try to get any attributes
2633  * along with any of the zero-copy operations READ, READDIR,
2634  * READLINK, WRITE.
2635  *
2636  * In the case of the first three, we want to put the GETATTR
2637  * after the read-type operation -- this is because it is hard
2638  * to predict the length of a GETATTR response in v4, and thus
2639  * align the READ data correctly.  This means that the GETATTR
2640  * may end up partially falling into the page cache, and we should
2641  * shift it into the 'tail' of the xdr_buf before processing.
2642  * To do this efficiently, we need to know the total length
2643  * of data received, which doesn't seem to be available outside
2644  * of the RPC layer.
2645  *
2646  * In the case of WRITE, we also want to put the GETATTR after
2647  * the operation -- in this case because we want to make sure
2648  * we get the post-operation mtime and size.  This means that
2649  * we can't use xdr_encode_pages() as written: we need a variant
2650  * of it which would leave room in the 'tail' iovec.
2651  *
2652  * Both of these changes to the XDR layer would in fact be quite
2653  * minor, but I decided to leave them for a subsequent patch.
2654  */
2655 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2656                 unsigned int pgbase, unsigned int pglen)
2657 {
2658         struct nfs4_readlink args = {
2659                 .fh       = NFS_FH(inode),
2660                 .pgbase   = pgbase,
2661                 .pglen    = pglen,
2662                 .pages    = &page,
2663         };
2664         struct nfs4_readlink_res res;
2665         struct rpc_message msg = {
2666                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2667                 .rpc_argp = &args,
2668                 .rpc_resp = &res,
2669         };
2670
2671         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2672 }
2673
2674 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2675                 unsigned int pgbase, unsigned int pglen)
2676 {
2677         struct nfs4_exception exception = { };
2678         int err;
2679         do {
2680                 err = nfs4_handle_exception(NFS_SERVER(inode),
2681                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2682                                 &exception);
2683         } while (exception.retry);
2684         return err;
2685 }
2686
2687 /*
2688  * Got race?
2689  * We will need to arrange for the VFS layer to provide an atomic open.
2690  * Until then, this create/open method is prone to inefficiency and race
2691  * conditions due to the lookup, create, and open VFS calls from sys_open()
2692  * placed on the wire.
2693  *
2694  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2695  * The file will be opened again in the subsequent VFS open call
2696  * (nfs4_proc_file_open).
2697  *
2698  * The open for read will just hang around to be used by any process that
2699  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2700  */
2701
2702 static int
2703 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2704                  int flags, struct nfs_open_context *ctx)
2705 {
2706         struct dentry *de = dentry;
2707         struct nfs4_state *state;
2708         struct rpc_cred *cred = NULL;
2709         fmode_t fmode = 0;
2710         int status = 0;
2711
2712         if (ctx != NULL) {
2713                 cred = ctx->cred;
2714                 de = ctx->dentry;
2715                 fmode = ctx->mode;
2716         }
2717         sattr->ia_mode &= ~current_umask();
2718         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2719         d_drop(dentry);
2720         if (IS_ERR(state)) {
2721                 status = PTR_ERR(state);
2722                 goto out;
2723         }
2724         d_add(dentry, igrab(state->inode));
2725         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2726         if (ctx != NULL)
2727                 ctx->state = state;
2728         else
2729                 nfs4_close_sync(state, fmode);
2730 out:
2731         return status;
2732 }
2733
2734 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2735 {
2736         struct nfs_server *server = NFS_SERVER(dir);
2737         struct nfs_removeargs args = {
2738                 .fh = NFS_FH(dir),
2739                 .name.len = name->len,
2740                 .name.name = name->name,
2741                 .bitmask = server->attr_bitmask,
2742         };
2743         struct nfs_removeres res = {
2744                 .server = server,
2745         };
2746         struct rpc_message msg = {
2747                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2748                 .rpc_argp = &args,
2749                 .rpc_resp = &res,
2750         };
2751         int status = -ENOMEM;
2752
2753         res.dir_attr = nfs_alloc_fattr();
2754         if (res.dir_attr == NULL)
2755                 goto out;
2756
2757         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2758         if (status == 0) {
2759                 update_changeattr(dir, &res.cinfo);
2760                 nfs_post_op_update_inode(dir, res.dir_attr);
2761         }
2762         nfs_free_fattr(res.dir_attr);
2763 out:
2764         return status;
2765 }
2766
2767 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2768 {
2769         struct nfs4_exception exception = { };
2770         int err;
2771         do {
2772                 err = nfs4_handle_exception(NFS_SERVER(dir),
2773                                 _nfs4_proc_remove(dir, name),
2774                                 &exception);
2775         } while (exception.retry);
2776         return err;
2777 }
2778
2779 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2780 {
2781         struct nfs_server *server = NFS_SERVER(dir);
2782         struct nfs_removeargs *args = msg->rpc_argp;
2783         struct nfs_removeres *res = msg->rpc_resp;
2784
2785         args->bitmask = server->cache_consistency_bitmask;
2786         res->server = server;
2787         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2788         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2789 }
2790
2791 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2792 {
2793         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2794                                 &data->args.seq_args,
2795                                 &data->res.seq_res,
2796                                 task))
2797                 return;
2798         rpc_call_start(task);
2799 }
2800
2801 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2802 {
2803         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2804
2805         if (!nfs4_sequence_done(task, &res->seq_res))
2806                 return 0;
2807         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2808                 return 0;
2809         update_changeattr(dir, &res->cinfo);
2810         nfs_post_op_update_inode(dir, res->dir_attr);
2811         return 1;
2812 }
2813
2814 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2815 {
2816         struct nfs_server *server = NFS_SERVER(dir);
2817         struct nfs_renameargs *arg = msg->rpc_argp;
2818         struct nfs_renameres *res = msg->rpc_resp;
2819
2820         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2821         arg->bitmask = server->attr_bitmask;
2822         res->server = server;
2823         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2824 }
2825
2826 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
2827 {
2828         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
2829                                 &data->args.seq_args,
2830                                 &data->res.seq_res,
2831                                 task))
2832                 return;
2833         rpc_call_start(task);
2834 }
2835
2836 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2837                                  struct inode *new_dir)
2838 {
2839         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2840
2841         if (!nfs4_sequence_done(task, &res->seq_res))
2842                 return 0;
2843         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2844                 return 0;
2845
2846         update_changeattr(old_dir, &res->old_cinfo);
2847         nfs_post_op_update_inode(old_dir, res->old_fattr);
2848         update_changeattr(new_dir, &res->new_cinfo);
2849         nfs_post_op_update_inode(new_dir, res->new_fattr);
2850         return 1;
2851 }
2852
2853 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2854                 struct inode *new_dir, struct qstr *new_name)
2855 {
2856         struct nfs_server *server = NFS_SERVER(old_dir);
2857         struct nfs_renameargs arg = {
2858                 .old_dir = NFS_FH(old_dir),
2859                 .new_dir = NFS_FH(new_dir),
2860                 .old_name = old_name,
2861                 .new_name = new_name,
2862                 .bitmask = server->attr_bitmask,
2863         };
2864         struct nfs_renameres res = {
2865                 .server = server,
2866         };
2867         struct rpc_message msg = {
2868                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2869                 .rpc_argp = &arg,
2870                 .rpc_resp = &res,
2871         };
2872         int status = -ENOMEM;
2873
2874         res.old_fattr = nfs_alloc_fattr();
2875         res.new_fattr = nfs_alloc_fattr();
2876         if (res.old_fattr == NULL || res.new_fattr == NULL)
2877                 goto out;
2878
2879         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2880         if (!status) {
2881                 update_changeattr(old_dir, &res.old_cinfo);
2882                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2883                 update_changeattr(new_dir, &res.new_cinfo);
2884                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2885         }
2886 out:
2887         nfs_free_fattr(res.new_fattr);
2888         nfs_free_fattr(res.old_fattr);
2889         return status;
2890 }
2891
2892 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2893                 struct inode *new_dir, struct qstr *new_name)
2894 {
2895         struct nfs4_exception exception = { };
2896         int err;
2897         do {
2898                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2899                                 _nfs4_proc_rename(old_dir, old_name,
2900                                         new_dir, new_name),
2901                                 &exception);
2902         } while (exception.retry);
2903         return err;
2904 }
2905
2906 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2907 {
2908         struct nfs_server *server = NFS_SERVER(inode);
2909         struct nfs4_link_arg arg = {
2910                 .fh     = NFS_FH(inode),
2911                 .dir_fh = NFS_FH(dir),
2912                 .name   = name,
2913                 .bitmask = server->attr_bitmask,
2914         };
2915         struct nfs4_link_res res = {
2916                 .server = server,
2917         };
2918         struct rpc_message msg = {
2919                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2920                 .rpc_argp = &arg,
2921                 .rpc_resp = &res,
2922         };
2923         int status = -ENOMEM;
2924
2925         res.fattr = nfs_alloc_fattr();
2926         res.dir_attr = nfs_alloc_fattr();
2927         if (res.fattr == NULL || res.dir_attr == NULL)
2928                 goto out;
2929
2930         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2931         if (!status) {
2932                 update_changeattr(dir, &res.cinfo);
2933                 nfs_post_op_update_inode(dir, res.dir_attr);
2934                 nfs_post_op_update_inode(inode, res.fattr);
2935         }
2936 out:
2937         nfs_free_fattr(res.dir_attr);
2938         nfs_free_fattr(res.fattr);
2939         return status;
2940 }
2941
2942 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2943 {
2944         struct nfs4_exception exception = { };
2945         int err;
2946         do {
2947                 err = nfs4_handle_exception(NFS_SERVER(inode),
2948                                 _nfs4_proc_link(inode, dir, name),
2949                                 &exception);
2950         } while (exception.retry);
2951         return err;
2952 }
2953
2954 struct nfs4_createdata {
2955         struct rpc_message msg;
2956         struct nfs4_create_arg arg;
2957         struct nfs4_create_res res;
2958         struct nfs_fh fh;
2959         struct nfs_fattr fattr;
2960         struct nfs_fattr dir_fattr;
2961 };
2962
2963 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2964                 struct qstr *name, struct iattr *sattr, u32 ftype)
2965 {
2966         struct nfs4_createdata *data;
2967
2968         data = kzalloc(sizeof(*data), GFP_KERNEL);
2969         if (data != NULL) {
2970                 struct nfs_server *server = NFS_SERVER(dir);
2971
2972                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2973                 data->msg.rpc_argp = &data->arg;
2974                 data->msg.rpc_resp = &data->res;
2975                 data->arg.dir_fh = NFS_FH(dir);
2976                 data->arg.server = server;
2977                 data->arg.name = name;
2978                 data->arg.attrs = sattr;
2979                 data->arg.ftype = ftype;
2980                 data->arg.bitmask = server->attr_bitmask;
2981                 data->res.server = server;
2982                 data->res.fh = &data->fh;
2983                 data->res.fattr = &data->fattr;
2984                 data->res.dir_fattr = &data->dir_fattr;
2985                 nfs_fattr_init(data->res.fattr);
2986                 nfs_fattr_init(data->res.dir_fattr);
2987         }
2988         return data;
2989 }
2990
2991 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2992 {
2993         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2994                                     &data->arg.seq_args, &data->res.seq_res, 1);
2995         if (status == 0) {
2996                 update_changeattr(dir, &data->res.dir_cinfo);
2997                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2998                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2999         }
3000         return status;
3001 }
3002
3003 static void nfs4_free_createdata(struct nfs4_createdata *data)
3004 {
3005         kfree(data);
3006 }
3007
3008 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3009                 struct page *page, unsigned int len, struct iattr *sattr)
3010 {
3011         struct nfs4_createdata *data;
3012         int status = -ENAMETOOLONG;
3013
3014         if (len > NFS4_MAXPATHLEN)
3015                 goto out;
3016
3017         status = -ENOMEM;
3018         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3019         if (data == NULL)
3020                 goto out;
3021
3022         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3023         data->arg.u.symlink.pages = &page;
3024         data->arg.u.symlink.len = len;
3025
3026         status = nfs4_do_create(dir, dentry, data);
3027
3028         nfs4_free_createdata(data);
3029 out:
3030         return status;
3031 }
3032
3033 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3034                 struct page *page, unsigned int len, struct iattr *sattr)
3035 {
3036         struct nfs4_exception exception = { };
3037         int err;
3038         do {
3039                 err = nfs4_handle_exception(NFS_SERVER(dir),
3040                                 _nfs4_proc_symlink(dir, dentry, page,
3041                                                         len, sattr),
3042                                 &exception);
3043         } while (exception.retry);
3044         return err;
3045 }
3046
3047 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3048                 struct iattr *sattr)
3049 {
3050         struct nfs4_createdata *data;
3051         int status = -ENOMEM;
3052
3053         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3054         if (data == NULL)
3055                 goto out;
3056
3057         status = nfs4_do_create(dir, dentry, data);
3058
3059         nfs4_free_createdata(data);
3060 out:
3061         return status;
3062 }
3063
3064 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3065                 struct iattr *sattr)
3066 {
3067         struct nfs4_exception exception = { };
3068         int err;
3069
3070         sattr->ia_mode &= ~current_umask();
3071         do {
3072                 err = nfs4_handle_exception(NFS_SERVER(dir),
3073                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3074                                 &exception);
3075         } while (exception.retry);
3076         return err;
3077 }
3078
3079 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3080                 u64 cookie, struct page **pages, unsigned int count, int plus)
3081 {
3082         struct inode            *dir = dentry->d_inode;
3083         struct nfs4_readdir_arg args = {
3084                 .fh = NFS_FH(dir),
3085                 .pages = pages,
3086                 .pgbase = 0,
3087                 .count = count,
3088                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3089                 .plus = plus,
3090         };
3091         struct nfs4_readdir_res res;
3092         struct rpc_message msg = {
3093                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3094                 .rpc_argp = &args,
3095                 .rpc_resp = &res,
3096                 .rpc_cred = cred,
3097         };
3098         int                     status;
3099
3100         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3101                         dentry->d_parent->d_name.name,
3102                         dentry->d_name.name,
3103                         (unsigned long long)cookie);
3104         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3105         res.pgbase = args.pgbase;
3106         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3107         if (status >= 0) {
3108                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3109                 status += args.pgbase;
3110         }
3111
3112         nfs_invalidate_atime(dir);
3113
3114         dprintk("%s: returns %d\n", __func__, status);
3115         return status;
3116 }
3117
3118 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3119                 u64 cookie, struct page **pages, unsigned int count, int plus)
3120 {
3121         struct nfs4_exception exception = { };
3122         int err;
3123         do {
3124                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3125                                 _nfs4_proc_readdir(dentry, cred, cookie,
3126                                         pages, count, plus),
3127                                 &exception);
3128         } while (exception.retry);
3129         return err;
3130 }
3131
3132 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3133                 struct iattr *sattr, dev_t rdev)
3134 {
3135         struct nfs4_createdata *data;
3136         int mode = sattr->ia_mode;
3137         int status = -ENOMEM;
3138
3139         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3140         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3141
3142         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3143         if (data == NULL)
3144                 goto out;
3145
3146         if (S_ISFIFO(mode))
3147                 data->arg.ftype = NF4FIFO;
3148         else if (S_ISBLK(mode)) {
3149                 data->arg.ftype = NF4BLK;
3150                 data->arg.u.device.specdata1 = MAJOR(rdev);
3151                 data->arg.u.device.specdata2 = MINOR(rdev);
3152         }
3153         else if (S_ISCHR(mode)) {
3154                 data->arg.ftype = NF4CHR;
3155                 data->arg.u.device.specdata1 = MAJOR(rdev);
3156                 data->arg.u.device.specdata2 = MINOR(rdev);
3157         }
3158
3159         status = nfs4_do_create(dir, dentry, data);
3160
3161         nfs4_free_createdata(data);
3162 out:
3163         return status;
3164 }
3165
3166 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3167                 struct iattr *sattr, dev_t rdev)
3168 {
3169         struct nfs4_exception exception = { };
3170         int err;
3171
3172         sattr->ia_mode &= ~current_umask();
3173         do {
3174                 err = nfs4_handle_exception(NFS_SERVER(dir),
3175                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3176                                 &exception);
3177         } while (exception.retry);
3178         return err;
3179 }
3180
3181 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3182                  struct nfs_fsstat *fsstat)
3183 {
3184         struct nfs4_statfs_arg args = {
3185                 .fh = fhandle,
3186                 .bitmask = server->attr_bitmask,
3187         };
3188         struct nfs4_statfs_res res = {
3189                 .fsstat = fsstat,
3190         };
3191         struct rpc_message msg = {
3192                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3193                 .rpc_argp = &args,
3194                 .rpc_resp = &res,
3195         };
3196
3197         nfs_fattr_init(fsstat->fattr);
3198         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3199 }
3200
3201 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3202 {
3203         struct nfs4_exception exception = { };
3204         int err;
3205         do {
3206                 err = nfs4_handle_exception(server,
3207                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3208                                 &exception);
3209         } while (exception.retry);
3210         return err;
3211 }
3212
3213 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3214                 struct nfs_fsinfo *fsinfo)
3215 {
3216         struct nfs4_fsinfo_arg args = {
3217                 .fh = fhandle,
3218                 .bitmask = server->attr_bitmask,
3219         };
3220         struct nfs4_fsinfo_res res = {
3221                 .fsinfo = fsinfo,
3222         };
3223         struct rpc_message msg = {
3224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3225                 .rpc_argp = &args,
3226                 .rpc_resp = &res,
3227         };
3228
3229         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3230 }
3231
3232 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3233 {
3234         struct nfs4_exception exception = { };
3235         int err;
3236
3237         do {
3238                 err = nfs4_handle_exception(server,
3239                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3240                                 &exception);
3241         } while (exception.retry);
3242         return err;
3243 }
3244
3245 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3246 {
3247         nfs_fattr_init(fsinfo->fattr);
3248         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3249 }
3250
3251 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3252                 struct nfs_pathconf *pathconf)
3253 {
3254         struct nfs4_pathconf_arg args = {
3255                 .fh = fhandle,
3256                 .bitmask = server->attr_bitmask,
3257         };
3258         struct nfs4_pathconf_res res = {
3259                 .pathconf = pathconf,
3260         };
3261         struct rpc_message msg = {
3262                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3263                 .rpc_argp = &args,
3264                 .rpc_resp = &res,
3265         };
3266
3267         /* None of the pathconf attributes are mandatory to implement */
3268         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3269                 memset(pathconf, 0, sizeof(*pathconf));
3270                 return 0;
3271         }
3272
3273         nfs_fattr_init(pathconf->fattr);
3274         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3275 }
3276
3277 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3278                 struct nfs_pathconf *pathconf)
3279 {
3280         struct nfs4_exception exception = { };
3281         int err;
3282
3283         do {
3284                 err = nfs4_handle_exception(server,
3285                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3286                                 &exception);
3287         } while (exception.retry);
3288         return err;
3289 }
3290
3291 void __nfs4_read_done_cb(struct nfs_read_data *data)
3292 {
3293         nfs_invalidate_atime(data->inode);
3294 }
3295
3296 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3297 {
3298         struct nfs_server *server = NFS_SERVER(data->inode);
3299
3300         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3301                 rpc_restart_call_prepare(task);
3302                 return -EAGAIN;
3303         }
3304
3305         __nfs4_read_done_cb(data);
3306         if (task->tk_status > 0)
3307                 renew_lease(server, data->timestamp);
3308         return 0;
3309 }
3310
3311 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3312 {
3313
3314         dprintk("--> %s\n", __func__);
3315
3316         if (!nfs4_sequence_done(task, &data->res.seq_res))
3317                 return -EAGAIN;
3318
3319         return data->read_done_cb ? data->read_done_cb(task, data) :
3320                                     nfs4_read_done_cb(task, data);
3321 }
3322
3323 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3324 {
3325         data->timestamp   = jiffies;
3326         data->read_done_cb = nfs4_read_done_cb;
3327         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3328         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3329 }
3330
3331 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3332 {
3333         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3334                                 &data->args.seq_args,
3335                                 &data->res.seq_res,
3336                                 task))
3337                 return;
3338         rpc_call_start(task);
3339 }
3340
3341 /* Reset the the nfs_read_data to send the read to the MDS. */
3342 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3343 {
3344         dprintk("%s Reset task for i/o through\n", __func__);
3345         put_lseg(data->lseg);
3346         data->lseg = NULL;
3347         /* offsets will differ in the dense stripe case */
3348         data->args.offset = data->mds_offset;
3349         data->ds_clp = NULL;
3350         data->args.fh     = NFS_FH(data->inode);
3351         data->read_done_cb = nfs4_read_done_cb;
3352         task->tk_ops = data->mds_ops;
3353         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3354 }
3355 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3356
3357 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3358 {
3359         struct inode *inode = data->inode;
3360
3361         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3362                 rpc_restart_call_prepare(task);
3363                 return -EAGAIN;
3364         }
3365         if (task->tk_status >= 0) {
3366                 renew_lease(NFS_SERVER(inode), data->timestamp);
3367                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3368         }
3369         return 0;
3370 }
3371
3372 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3373 {
3374         if (!nfs4_sequence_done(task, &data->res.seq_res))
3375                 return -EAGAIN;
3376         return data->write_done_cb ? data->write_done_cb(task, data) :
3377                 nfs4_write_done_cb(task, data);
3378 }
3379
3380 /* Reset the the nfs_write_data to send the write to the MDS. */
3381 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3382 {
3383         dprintk("%s Reset task for i/o through\n", __func__);
3384         put_lseg(data->lseg);
3385         data->lseg          = NULL;
3386         data->ds_clp        = NULL;
3387         data->write_done_cb = nfs4_write_done_cb;
3388         data->args.fh       = NFS_FH(data->inode);
3389         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3390         data->args.offset   = data->mds_offset;
3391         data->res.fattr     = &data->fattr;
3392         task->tk_ops        = data->mds_ops;
3393         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3394 }
3395 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3396
3397 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3398 {
3399         struct nfs_server *server = NFS_SERVER(data->inode);
3400
3401         if (data->lseg) {
3402                 data->args.bitmask = NULL;
3403                 data->res.fattr = NULL;
3404         } else
3405                 data->args.bitmask = server->cache_consistency_bitmask;
3406         if (!data->write_done_cb)
3407                 data->write_done_cb = nfs4_write_done_cb;
3408         data->res.server = server;
3409         data->timestamp   = jiffies;
3410
3411         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3412         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3413 }
3414
3415 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3416 {
3417         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3418                                 &data->args.seq_args,
3419                                 &data->res.seq_res,
3420                                 task))
3421                 return;
3422         rpc_call_start(task);
3423 }
3424
3425 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3426 {
3427         struct inode *inode = data->inode;
3428
3429         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3430                 rpc_restart_call_prepare(task);
3431                 return -EAGAIN;
3432         }
3433         nfs_refresh_inode(inode, data->res.fattr);
3434         return 0;
3435 }
3436
3437 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3438 {
3439         if (!nfs4_sequence_done(task, &data->res.seq_res))
3440                 return -EAGAIN;
3441         return data->write_done_cb(task, data);
3442 }
3443
3444 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3445 {
3446         struct nfs_server *server = NFS_SERVER(data->inode);
3447
3448         if (data->lseg) {
3449                 data->args.bitmask = NULL;
3450                 data->res.fattr = NULL;
3451         } else
3452                 data->args.bitmask = server->cache_consistency_bitmask;
3453         if (!data->write_done_cb)
3454                 data->write_done_cb = nfs4_commit_done_cb;
3455         data->res.server = server;
3456         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3457         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3458 }
3459
3460 struct nfs4_renewdata {
3461         struct nfs_client       *client;
3462         unsigned long           timestamp;
3463 };
3464
3465 /*
3466  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3467  * standalone procedure for queueing an asynchronous RENEW.
3468  */
3469 static void nfs4_renew_release(void *calldata)
3470 {
3471         struct nfs4_renewdata *data = calldata;
3472         struct nfs_client *clp = data->client;
3473
3474         if (atomic_read(&clp->cl_count) > 1)
3475                 nfs4_schedule_state_renewal(clp);
3476         nfs_put_client(clp);
3477         kfree(data);
3478 }
3479
3480 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3481 {
3482         struct nfs4_renewdata *data = calldata;
3483         struct nfs_client *clp = data->client;
3484         unsigned long timestamp = data->timestamp;
3485
3486         if (task->tk_status < 0) {
3487                 /* Unless we're shutting down, schedule state recovery! */
3488                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3489                         return;
3490                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3491                         nfs4_schedule_lease_recovery(clp);
3492                         return;
3493                 }
3494                 nfs4_schedule_path_down_recovery(clp);
3495         }
3496         do_renew_lease(clp, timestamp);
3497 }
3498
3499 static const struct rpc_call_ops nfs4_renew_ops = {
3500         .rpc_call_done = nfs4_renew_done,
3501         .rpc_release = nfs4_renew_release,
3502 };
3503
3504 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3505 {
3506         struct rpc_message msg = {
3507                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3508                 .rpc_argp       = clp,
3509                 .rpc_cred       = cred,
3510         };
3511         struct nfs4_renewdata *data;
3512
3513         if (renew_flags == 0)
3514                 return 0;
3515         if (!atomic_inc_not_zero(&clp->cl_count))
3516                 return -EIO;
3517         data = kmalloc(sizeof(*data), GFP_NOFS);
3518         if (data == NULL)
3519                 return -ENOMEM;
3520         data->client = clp;
3521         data->timestamp = jiffies;
3522         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3523                         &nfs4_renew_ops, data);
3524 }
3525
3526 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3527 {
3528         struct rpc_message msg = {
3529                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3530                 .rpc_argp       = clp,
3531                 .rpc_cred       = cred,
3532         };
3533         unsigned long now = jiffies;
3534         int status;
3535
3536         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3537         if (status < 0)
3538                 return status;
3539         do_renew_lease(clp, now);
3540         return 0;
3541 }
3542
3543 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3544 {
3545         return (server->caps & NFS_CAP_ACLS)
3546                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3547                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3548 }
3549
3550 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3551  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3552  * the stack.
3553  */
3554 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3555
3556 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3557                 struct page **pages, unsigned int *pgbase)
3558 {
3559         struct page *newpage, **spages;
3560         int rc = 0;
3561         size_t len;
3562         spages = pages;
3563
3564         do {
3565                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3566                 newpage = alloc_page(GFP_KERNEL);
3567
3568                 if (newpage == NULL)
3569                         goto unwind;
3570                 memcpy(page_address(newpage), buf, len);
3571                 buf += len;
3572                 buflen -= len;
3573                 *pages++ = newpage;
3574                 rc++;
3575         } while (buflen != 0);
3576
3577         return rc;
3578
3579 unwind:
3580         for(; rc > 0; rc--)
3581                 __free_page(spages[rc-1]);
3582         return -ENOMEM;
3583 }
3584
3585 struct nfs4_cached_acl {
3586         int cached;
3587         size_t len;
3588         char data[0];
3589 };
3590
3591 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3592 {
3593         struct nfs_inode *nfsi = NFS_I(inode);
3594
3595         spin_lock(&inode->i_lock);
3596         kfree(nfsi->nfs4_acl);
3597         nfsi->nfs4_acl = acl;
3598         spin_unlock(&inode->i_lock);
3599 }
3600
3601 static void nfs4_zap_acl_attr(struct inode *inode)
3602 {
3603         nfs4_set_cached_acl(inode, NULL);
3604 }
3605
3606 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3607 {
3608         struct nfs_inode *nfsi = NFS_I(inode);
3609         struct nfs4_cached_acl *acl;
3610         int ret = -ENOENT;
3611
3612         spin_lock(&inode->i_lock);
3613         acl = nfsi->nfs4_acl;
3614         if (acl == NULL)
3615                 goto out;
3616         if (buf == NULL) /* user is just asking for length */
3617                 goto out_len;
3618         if (acl->cached == 0)
3619                 goto out;
3620         ret = -ERANGE; /* see getxattr(2) man page */
3621         if (acl->len > buflen)
3622                 goto out;
3623         memcpy(buf, acl->data, acl->len);
3624 out_len:
3625         ret = acl->len;
3626 out:
3627         spin_unlock(&inode->i_lock);
3628         return ret;
3629 }
3630
3631 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3632 {
3633         struct nfs4_cached_acl *acl;
3634
3635         if (pages && acl_len <= PAGE_SIZE) {
3636                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3637                 if (acl == NULL)
3638                         goto out;
3639                 acl->cached = 1;
3640                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3641         } else {
3642                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3643                 if (acl == NULL)
3644                         goto out;
3645                 acl->cached = 0;
3646         }
3647         acl->len = acl_len;
3648 out:
3649         nfs4_set_cached_acl(inode, acl);
3650 }
3651
3652 /*
3653  * The getxattr API returns the required buffer length when called with a
3654  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3655  * the required buf.  On a NULL buf, we send a page of data to the server
3656  * guessing that the ACL request can be serviced by a page. If so, we cache
3657  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3658  * the cache. If not so, we throw away the page, and cache the required
3659  * length. The next getxattr call will then produce another round trip to
3660  * the server, this time with the input buf of the required size.
3661  */
3662 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3663 {
3664         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3665         struct nfs_getaclargs args = {
3666                 .fh = NFS_FH(inode),
3667                 .acl_pages = pages,
3668                 .acl_len = buflen,
3669         };
3670         struct nfs_getaclres res = {
3671                 .acl_len = buflen,
3672         };
3673         struct rpc_message msg = {
3674                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3675                 .rpc_argp = &args,
3676                 .rpc_resp = &res,
3677         };
3678         int ret = -ENOMEM, npages, i, acl_len = 0;
3679
3680         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3681         /* As long as we're doing a round trip to the server anyway,
3682          * let's be prepared for a page of acl data. */
3683         if (npages == 0)
3684                 npages = 1;
3685
3686         /* Add an extra page to handle the bitmap returned */
3687         npages++;
3688
3689         for (i = 0; i < npages; i++) {
3690                 pages[i] = alloc_page(GFP_KERNEL);
3691                 if (!pages[i])
3692                         goto out_free;
3693         }
3694
3695         /* for decoding across pages */
3696         res.acl_scratch = alloc_page(GFP_KERNEL);
3697         if (!res.acl_scratch)
3698                 goto out_free;
3699
3700         args.acl_len = npages * PAGE_SIZE;
3701         args.acl_pgbase = 0;
3702
3703         /* Let decode_getfacl know not to fail if the ACL data is larger than
3704          * the page we send as a guess */
3705         if (buf == NULL)
3706                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3707
3708         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3709                 __func__, buf, buflen, npages, args.acl_len);
3710         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3711                              &msg, &args.seq_args, &res.seq_res, 0);
3712         if (ret)
3713                 goto out_free;
3714
3715         acl_len = res.acl_len - res.acl_data_offset;
3716         if (acl_len > args.acl_len)
3717                 nfs4_write_cached_acl(inode, NULL, 0, acl_len);
3718         else
3719                 nfs4_write_cached_acl(inode, pages, res.acl_data_offset,
3720                                       acl_len);
3721         if (buf) {
3722                 ret = -ERANGE;
3723                 if (acl_len > buflen)
3724                         goto out_free;
3725                 _copy_from_pages(buf, pages, res.acl_data_offset,
3726                                 acl_len);
3727         }
3728         ret = acl_len;
3729 out_free:
3730         for (i = 0; i < npages; i++)
3731                 if (pages[i])
3732                         __free_page(pages[i]);
3733         if (res.acl_scratch)
3734                 __free_page(res.acl_scratch);
3735         return ret;
3736 }
3737
3738 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3739 {
3740         struct nfs4_exception exception = { };
3741         ssize_t ret;
3742         do {
3743                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3744                 if (ret >= 0)
3745                         break;
3746                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3747         } while (exception.retry);
3748         return ret;
3749 }
3750
3751 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3752 {
3753         struct nfs_server *server = NFS_SERVER(inode);
3754         int ret;
3755
3756         if (!nfs4_server_supports_acls(server))
3757                 return -EOPNOTSUPP;
3758         ret = nfs_revalidate_inode(server, inode);
3759         if (ret < 0)
3760                 return ret;
3761         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3762                 nfs_zap_acl_cache(inode);
3763         ret = nfs4_read_cached_acl(inode, buf, buflen);
3764         if (ret != -ENOENT)
3765                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3766                  * but no cached acl data, just the acl length */
3767                 return ret;
3768         return nfs4_get_acl_uncached(inode, buf, buflen);
3769 }
3770
3771 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3772 {
3773         struct nfs_server *server = NFS_SERVER(inode);
3774         struct page *pages[NFS4ACL_MAXPAGES];
3775         struct nfs_setaclargs arg = {
3776                 .fh             = NFS_FH(inode),
3777                 .acl_pages      = pages,
3778                 .acl_len        = buflen,
3779         };
3780         struct nfs_setaclres res;
3781         struct rpc_message msg = {
3782                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3783                 .rpc_argp       = &arg,
3784                 .rpc_resp       = &res,
3785         };
3786         int ret, i;
3787
3788         if (!nfs4_server_supports_acls(server))
3789                 return -EOPNOTSUPP;
3790         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3791         if (i < 0)
3792                 return i;
3793         nfs_inode_return_delegation(inode);
3794         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3795
3796         /*
3797          * Free each page after tx, so the only ref left is
3798          * held by the network stack
3799          */
3800         for (; i > 0; i--)
3801                 put_page(pages[i-1]);
3802
3803         /*
3804          * Acl update can result in inode attribute update.
3805          * so mark the attribute cache invalid.
3806          */
3807         spin_lock(&inode->i_lock);
3808         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3809         spin_unlock(&inode->i_lock);
3810         nfs_access_zap_cache(inode);
3811         nfs_zap_acl_cache(inode);
3812         return ret;
3813 }
3814
3815 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3816 {
3817         struct nfs4_exception exception = { };
3818         int err;
3819         do {
3820                 err = nfs4_handle_exception(NFS_SERVER(inode),
3821                                 __nfs4_proc_set_acl(inode, buf, buflen),
3822                                 &exception);
3823         } while (exception.retry);
3824         return err;
3825 }
3826
3827 static int
3828 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3829 {
3830         struct nfs_client *clp = server->nfs_client;
3831
3832         if (task->tk_status >= 0)
3833                 return 0;
3834         switch(task->tk_status) {
3835                 case -NFS4ERR_DELEG_REVOKED:
3836                 case -NFS4ERR_ADMIN_REVOKED:
3837                 case -NFS4ERR_BAD_STATEID:
3838                         if (state == NULL)
3839                                 break;
3840                         nfs_remove_bad_delegation(state->inode);
3841                 case -NFS4ERR_OPENMODE:
3842                         if (state == NULL)
3843                                 break;
3844                         nfs4_schedule_stateid_recovery(server, state);
3845                         goto wait_on_recovery;
3846                 case -NFS4ERR_EXPIRED:
3847                         if (state != NULL)
3848                                 nfs4_schedule_stateid_recovery(server, state);
3849                 case -NFS4ERR_STALE_STATEID:
3850                 case -NFS4ERR_STALE_CLIENTID:
3851                         nfs4_schedule_lease_recovery(clp);
3852                         goto wait_on_recovery;
3853 #if defined(CONFIG_NFS_V4_1)
3854                 case -NFS4ERR_BADSESSION:
3855                 case -NFS4ERR_BADSLOT:
3856                 case -NFS4ERR_BAD_HIGH_SLOT:
3857                 case -NFS4ERR_DEADSESSION:
3858                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3859                 case -NFS4ERR_SEQ_FALSE_RETRY:
3860                 case -NFS4ERR_SEQ_MISORDERED:
3861                         dprintk("%s ERROR %d, Reset session\n", __func__,
3862                                 task->tk_status);
3863                         nfs4_schedule_session_recovery(clp->cl_session);
3864                         task->tk_status = 0;
3865                         return -EAGAIN;
3866 #endif /* CONFIG_NFS_V4_1 */
3867                 case -NFS4ERR_DELAY:
3868                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3869                 case -NFS4ERR_GRACE:
3870                 case -EKEYEXPIRED:
3871                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3872                         task->tk_status = 0;
3873                         return -EAGAIN;
3874                 case -NFS4ERR_RETRY_UNCACHED_REP:
3875                 case -NFS4ERR_OLD_STATEID:
3876                         task->tk_status = 0;
3877                         return -EAGAIN;
3878         }
3879         task->tk_status = nfs4_map_errors(task->tk_status);
3880         return 0;
3881 wait_on_recovery:
3882         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3883         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3884                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3885         task->tk_status = 0;
3886         return -EAGAIN;
3887 }
3888
3889 static void nfs4_construct_boot_verifier(struct nfs_client *clp,
3890                                          nfs4_verifier *bootverf)
3891 {
3892         __be32 verf[2];
3893
3894         verf[0] = htonl((u32)clp->cl_boot_time.tv_sec);
3895         verf[1] = htonl((u32)clp->cl_boot_time.tv_nsec);
3896         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3897 }
3898
3899 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3900                 unsigned short port, struct rpc_cred *cred,
3901                 struct nfs4_setclientid_res *res)
3902 {
3903         nfs4_verifier sc_verifier;
3904         struct nfs4_setclientid setclientid = {
3905                 .sc_verifier = &sc_verifier,
3906                 .sc_prog = program,
3907                 .sc_cb_ident = clp->cl_cb_ident,
3908         };
3909         struct rpc_message msg = {
3910                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3911                 .rpc_argp = &setclientid,
3912                 .rpc_resp = res,
3913                 .rpc_cred = cred,
3914         };
3915         int loop = 0;
3916         int status;
3917
3918         nfs4_construct_boot_verifier(clp, &sc_verifier);
3919
3920         for(;;) {
3921                 rcu_read_lock();
3922                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3923                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3924                                 clp->cl_ipaddr,
3925                                 rpc_peeraddr2str(clp->cl_rpcclient,
3926                                                         RPC_DISPLAY_ADDR),
3927                                 rpc_peeraddr2str(clp->cl_rpcclient,
3928                                                         RPC_DISPLAY_PROTO),
3929                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3930                                 clp->cl_id_uniquifier);
3931                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3932                                 sizeof(setclientid.sc_netid),
3933                                 rpc_peeraddr2str(clp->cl_rpcclient,
3934                                                         RPC_DISPLAY_NETID));
3935                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3936                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3937                                 clp->cl_ipaddr, port >> 8, port & 255);
3938                 rcu_read_unlock();
3939
3940                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3941                 if (status != -NFS4ERR_CLID_INUSE)
3942                         break;
3943                 if (loop != 0) {
3944                         ++clp->cl_id_uniquifier;
3945                         break;
3946                 }
3947                 ++loop;
3948                 ssleep(clp->cl_lease_time / HZ + 1);
3949         }
3950         return status;
3951 }
3952
3953 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3954                 struct nfs4_setclientid_res *arg,
3955                 struct rpc_cred *cred)
3956 {
3957         struct nfs_fsinfo fsinfo;
3958         struct rpc_message msg = {
3959                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3960                 .rpc_argp = arg,
3961                 .rpc_resp = &fsinfo,
3962                 .rpc_cred = cred,
3963         };
3964         unsigned long now;
3965         int status;
3966
3967         now = jiffies;
3968         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3969         if (status == 0) {
3970                 spin_lock(&clp->cl_lock);
3971                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3972                 clp->cl_last_renewal = now;
3973                 spin_unlock(&clp->cl_lock);
3974         }
3975         return status;
3976 }
3977
3978 struct nfs4_delegreturndata {
3979         struct nfs4_delegreturnargs args;
3980         struct nfs4_delegreturnres res;
3981         struct nfs_fh fh;
3982         nfs4_stateid stateid;
3983         unsigned long timestamp;
3984         struct nfs_fattr fattr;
3985         int rpc_status;
3986 };
3987
3988 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3989 {
3990         struct nfs4_delegreturndata *data = calldata;
3991
3992         if (!nfs4_sequence_done(task, &data->res.seq_res))
3993                 return;
3994
3995         switch (task->tk_status) {
3996         case -NFS4ERR_STALE_STATEID:
3997         case -NFS4ERR_EXPIRED:
3998         case 0:
3999                 renew_lease(data->res.server, data->timestamp);
4000                 break;
4001         default:
4002                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4003                                 -EAGAIN) {
4004                         rpc_restart_call_prepare(task);
4005                         return;
4006                 }
4007         }
4008         data->rpc_status = task->tk_status;
4009 }
4010
4011 static void nfs4_delegreturn_release(void *calldata)
4012 {
4013         kfree(calldata);
4014 }
4015
4016 #if defined(CONFIG_NFS_V4_1)
4017 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4018 {
4019         struct nfs4_delegreturndata *d_data;
4020
4021         d_data = (struct nfs4_delegreturndata *)data;
4022
4023         if (nfs4_setup_sequence(d_data->res.server,
4024                                 &d_data->args.seq_args,
4025                                 &d_data->res.seq_res, task))
4026                 return;
4027         rpc_call_start(task);
4028 }
4029 #endif /* CONFIG_NFS_V4_1 */
4030
4031 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4032 #if defined(CONFIG_NFS_V4_1)
4033         .rpc_call_prepare = nfs4_delegreturn_prepare,
4034 #endif /* CONFIG_NFS_V4_1 */
4035         .rpc_call_done = nfs4_delegreturn_done,
4036         .rpc_release = nfs4_delegreturn_release,
4037 };
4038
4039 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4040 {
4041         struct nfs4_delegreturndata *data;
4042         struct nfs_server *server = NFS_SERVER(inode);
4043         struct rpc_task *task;
4044         struct rpc_message msg = {
4045                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4046                 .rpc_cred = cred,
4047         };
4048         struct rpc_task_setup task_setup_data = {
4049                 .rpc_client = server->client,
4050                 .rpc_message = &msg,
4051                 .callback_ops = &nfs4_delegreturn_ops,
4052                 .flags = RPC_TASK_ASYNC,
4053         };
4054         int status = 0;
4055
4056         data = kzalloc(sizeof(*data), GFP_NOFS);
4057         if (data == NULL)
4058                 return -ENOMEM;
4059         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4060         data->args.fhandle = &data->fh;
4061         data->args.stateid = &data->stateid;
4062         data->args.bitmask = server->attr_bitmask;
4063         nfs_copy_fh(&data->fh, NFS_FH(inode));
4064         nfs4_stateid_copy(&data->stateid, stateid);
4065         data->res.fattr = &data->fattr;
4066         data->res.server = server;
4067         nfs_fattr_init(data->res.fattr);
4068         data->timestamp = jiffies;
4069         data->rpc_status = 0;
4070
4071         task_setup_data.callback_data = data;
4072         msg.rpc_argp = &data->args;
4073         msg.rpc_resp = &data->res;
4074         task = rpc_run_task(&task_setup_data);
4075         if (IS_ERR(task))
4076                 return PTR_ERR(task);
4077         if (!issync)
4078                 goto out;
4079         status = nfs4_wait_for_completion_rpc_task(task);
4080         if (status != 0)
4081                 goto out;
4082         status = data->rpc_status;
4083         if (status != 0)
4084                 goto out;
4085         nfs_refresh_inode(inode, &data->fattr);
4086 out:
4087         rpc_put_task(task);
4088         return status;
4089 }
4090
4091 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4092 {
4093         struct nfs_server *server = NFS_SERVER(inode);
4094         struct nfs4_exception exception = { };
4095         int err;
4096         do {
4097                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4098                 switch (err) {
4099                         case -NFS4ERR_STALE_STATEID:
4100                         case -NFS4ERR_EXPIRED:
4101                         case 0:
4102                                 return 0;
4103                 }
4104                 err = nfs4_handle_exception(server, err, &exception);
4105         } while (exception.retry);
4106         return err;
4107 }
4108
4109 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4110 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4111
4112 /*
4113  * sleep, with exponential backoff, and retry the LOCK operation.
4114  */
4115 static unsigned long
4116 nfs4_set_lock_task_retry(unsigned long timeout)
4117 {
4118         freezable_schedule_timeout_killable(timeout);
4119         timeout <<= 1;
4120         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4121                 return NFS4_LOCK_MAXTIMEOUT;
4122         return timeout;
4123 }
4124
4125 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4126 {
4127         struct inode *inode = state->inode;
4128         struct nfs_server *server = NFS_SERVER(inode);
4129         struct nfs_client *clp = server->nfs_client;
4130         struct nfs_lockt_args arg = {
4131                 .fh = NFS_FH(inode),
4132                 .fl = request,
4133         };
4134         struct nfs_lockt_res res = {
4135                 .denied = request,
4136         };
4137         struct rpc_message msg = {
4138                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4139                 .rpc_argp       = &arg,
4140                 .rpc_resp       = &res,
4141                 .rpc_cred       = state->owner->so_cred,
4142         };
4143         struct nfs4_lock_state *lsp;
4144         int status;
4145
4146         arg.lock_owner.clientid = clp->cl_clientid;
4147         status = nfs4_set_lock_state(state, request);
4148         if (status != 0)
4149                 goto out;
4150         lsp = request->fl_u.nfs4_fl.owner;
4151         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4152         arg.lock_owner.s_dev = server->s_dev;
4153         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4154         switch (status) {
4155                 case 0:
4156                         request->fl_type = F_UNLCK;
4157                         break;
4158                 case -NFS4ERR_DENIED:
4159                         status = 0;
4160         }
4161         request->fl_ops->fl_release_private(request);
4162 out:
4163         return status;
4164 }
4165
4166 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4167 {
4168         struct nfs4_exception exception = { };
4169         int err;
4170
4171         do {
4172                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4173                                 _nfs4_proc_getlk(state, cmd, request),
4174                                 &exception);
4175         } while (exception.retry);
4176         return err;
4177 }
4178
4179 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4180 {
4181         int res = 0;
4182         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4183                 case FL_POSIX:
4184                         res = posix_lock_file_wait(file, fl);
4185                         break;
4186                 case FL_FLOCK:
4187                         res = flock_lock_file_wait(file, fl);
4188                         break;
4189                 default:
4190                         BUG();
4191         }
4192         return res;
4193 }
4194
4195 struct nfs4_unlockdata {
4196         struct nfs_locku_args arg;
4197         struct nfs_locku_res res;
4198         struct nfs4_lock_state *lsp;
4199         struct nfs_open_context *ctx;
4200         struct file_lock fl;
4201         const struct nfs_server *server;
4202         unsigned long timestamp;
4203 };
4204
4205 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4206                 struct nfs_open_context *ctx,
4207                 struct nfs4_lock_state *lsp,
4208                 struct nfs_seqid *seqid)
4209 {
4210         struct nfs4_unlockdata *p;
4211         struct inode *inode = lsp->ls_state->inode;
4212
4213         p = kzalloc(sizeof(*p), GFP_NOFS);
4214         if (p == NULL)
4215                 return NULL;
4216         p->arg.fh = NFS_FH(inode);
4217         p->arg.fl = &p->fl;
4218         p->arg.seqid = seqid;
4219         p->res.seqid = seqid;
4220         p->arg.stateid = &lsp->ls_stateid;
4221         p->lsp = lsp;
4222         atomic_inc(&lsp->ls_count);
4223         /* Ensure we don't close file until we're done freeing locks! */
4224         p->ctx = get_nfs_open_context(ctx);
4225         memcpy(&p->fl, fl, sizeof(p->fl));
4226         p->server = NFS_SERVER(inode);
4227         return p;
4228 }
4229
4230 static void nfs4_locku_release_calldata(void *data)
4231 {
4232         struct nfs4_unlockdata *calldata = data;
4233         nfs_free_seqid(calldata->arg.seqid);
4234         nfs4_put_lock_state(calldata->lsp);
4235         put_nfs_open_context(calldata->ctx);
4236         kfree(calldata);
4237 }
4238
4239 static void nfs4_locku_done(struct rpc_task *task, void *data)
4240 {
4241         struct nfs4_unlockdata *calldata = data;
4242
4243         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4244                 return;
4245         switch (task->tk_status) {
4246                 case 0:
4247                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4248                                         &calldata->res.stateid);
4249                         renew_lease(calldata->server, calldata->timestamp);
4250                         break;
4251                 case -NFS4ERR_BAD_STATEID:
4252                 case -NFS4ERR_OLD_STATEID:
4253                 case -NFS4ERR_STALE_STATEID:
4254                 case -NFS4ERR_EXPIRED:
4255                         break;
4256                 default:
4257                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4258                                 rpc_restart_call_prepare(task);
4259         }
4260 }
4261
4262 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4263 {
4264         struct nfs4_unlockdata *calldata = data;
4265
4266         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4267                 return;
4268         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4269                 /* Note: exit _without_ running nfs4_locku_done */
4270                 task->tk_action = NULL;
4271                 return;
4272         }
4273         calldata->timestamp = jiffies;
4274         if (nfs4_setup_sequence(calldata->server,
4275                                 &calldata->arg.seq_args,
4276                                 &calldata->res.seq_res, task))
4277                 return;
4278         rpc_call_start(task);
4279 }
4280
4281 static const struct rpc_call_ops nfs4_locku_ops = {
4282         .rpc_call_prepare = nfs4_locku_prepare,
4283         .rpc_call_done = nfs4_locku_done,
4284         .rpc_release = nfs4_locku_release_calldata,
4285 };
4286
4287 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4288                 struct nfs_open_context *ctx,
4289                 struct nfs4_lock_state *lsp,
4290                 struct nfs_seqid *seqid)
4291 {
4292         struct nfs4_unlockdata *data;
4293         struct rpc_message msg = {
4294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4295                 .rpc_cred = ctx->cred,
4296         };
4297         struct rpc_task_setup task_setup_data = {
4298                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4299                 .rpc_message = &msg,
4300                 .callback_ops = &nfs4_locku_ops,
4301                 .workqueue = nfsiod_workqueue,
4302                 .flags = RPC_TASK_ASYNC,
4303         };
4304
4305         /* Ensure this is an unlock - when canceling a lock, the
4306          * canceled lock is passed in, and it won't be an unlock.
4307          */
4308         fl->fl_type = F_UNLCK;
4309
4310         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4311         if (data == NULL) {
4312                 nfs_free_seqid(seqid);
4313                 return ERR_PTR(-ENOMEM);
4314         }
4315
4316         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4317         msg.rpc_argp = &data->arg;
4318         msg.rpc_resp = &data->res;
4319         task_setup_data.callback_data = data;
4320         return rpc_run_task(&task_setup_data);
4321 }
4322
4323 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4324 {
4325         struct nfs_inode *nfsi = NFS_I(state->inode);
4326         struct nfs_seqid *seqid;
4327         struct nfs4_lock_state *lsp;
4328         struct rpc_task *task;
4329         int status = 0;
4330         unsigned char fl_flags = request->fl_flags;
4331
4332         status = nfs4_set_lock_state(state, request);
4333         /* Unlock _before_ we do the RPC call */
4334         request->fl_flags |= FL_EXISTS;
4335         down_read(&nfsi->rwsem);
4336         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4337                 up_read(&nfsi->rwsem);
4338                 goto out;
4339         }
4340         up_read(&nfsi->rwsem);
4341         if (status != 0)
4342                 goto out;
4343         /* Is this a delegated lock? */
4344         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4345                 goto out;
4346         lsp = request->fl_u.nfs4_fl.owner;
4347         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4348         status = -ENOMEM;
4349         if (seqid == NULL)
4350                 goto out;
4351         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4352         status = PTR_ERR(task);
4353         if (IS_ERR(task))
4354                 goto out;
4355         status = nfs4_wait_for_completion_rpc_task(task);
4356         rpc_put_task(task);
4357 out:
4358         request->fl_flags = fl_flags;
4359         return status;
4360 }
4361
4362 struct nfs4_lockdata {
4363         struct nfs_lock_args arg;
4364         struct nfs_lock_res res;
4365         struct nfs4_lock_state *lsp;
4366         struct nfs_open_context *ctx;
4367         struct file_lock fl;
4368         unsigned long timestamp;
4369         int rpc_status;
4370         int cancelled;
4371         struct nfs_server *server;
4372 };
4373
4374 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4375                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4376                 gfp_t gfp_mask)
4377 {
4378         struct nfs4_lockdata *p;
4379         struct inode *inode = lsp->ls_state->inode;
4380         struct nfs_server *server = NFS_SERVER(inode);
4381
4382         p = kzalloc(sizeof(*p), gfp_mask);
4383         if (p == NULL)
4384                 return NULL;
4385
4386         p->arg.fh = NFS_FH(inode);
4387         p->arg.fl = &p->fl;
4388         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4389         if (p->arg.open_seqid == NULL)
4390                 goto out_free;
4391         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4392         if (p->arg.lock_seqid == NULL)
4393                 goto out_free_seqid;
4394         p->arg.lock_stateid = &lsp->ls_stateid;
4395         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4396         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4397         p->arg.lock_owner.s_dev = server->s_dev;
4398         p->res.lock_seqid = p->arg.lock_seqid;
4399         p->lsp = lsp;
4400         p->server = server;
4401         atomic_inc(&lsp->ls_count);
4402         p->ctx = get_nfs_open_context(ctx);
4403         memcpy(&p->fl, fl, sizeof(p->fl));
4404         return p;
4405 out_free_seqid:
4406         nfs_free_seqid(p->arg.open_seqid);
4407 out_free:
4408         kfree(p);
4409         return NULL;
4410 }
4411
4412 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4413 {
4414         struct nfs4_lockdata *data = calldata;
4415         struct nfs4_state *state = data->lsp->ls_state;
4416
4417         dprintk("%s: begin!\n", __func__);
4418         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4419                 return;
4420         /* Do we need to do an open_to_lock_owner? */
4421         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4422                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4423                         return;
4424                 data->arg.open_stateid = &state->stateid;
4425                 data->arg.new_lock_owner = 1;
4426                 data->res.open_seqid = data->arg.open_seqid;
4427         } else
4428                 data->arg.new_lock_owner = 0;
4429         data->timestamp = jiffies;
4430         if (nfs4_setup_sequence(data->server,
4431                                 &data->arg.seq_args,
4432                                 &data->res.seq_res, task))
4433                 return;
4434         rpc_call_start(task);
4435         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4436 }
4437
4438 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4439 {
4440         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4441         nfs4_lock_prepare(task, calldata);
4442 }
4443
4444 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4445 {
4446         struct nfs4_lockdata *data = calldata;
4447
4448         dprintk("%s: begin!\n", __func__);
4449
4450         if (!nfs4_sequence_done(task, &data->res.seq_res))
4451                 return;
4452
4453         data->rpc_status = task->tk_status;
4454         if (data->arg.new_lock_owner != 0) {
4455                 if (data->rpc_status == 0)
4456                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4457                 else
4458                         goto out;
4459         }
4460         if (data->rpc_status == 0) {
4461                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4462                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4463                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4464         }
4465 out:
4466         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4467 }
4468
4469 static void nfs4_lock_release(void *calldata)
4470 {
4471         struct nfs4_lockdata *data = calldata;
4472
4473         dprintk("%s: begin!\n", __func__);
4474         nfs_free_seqid(data->arg.open_seqid);
4475         if (data->cancelled != 0) {
4476                 struct rpc_task *task;
4477                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4478                                 data->arg.lock_seqid);
4479                 if (!IS_ERR(task))
4480                         rpc_put_task_async(task);
4481                 dprintk("%s: cancelling lock!\n", __func__);
4482         } else
4483                 nfs_free_seqid(data->arg.lock_seqid);
4484         nfs4_put_lock_state(data->lsp);
4485         put_nfs_open_context(data->ctx);
4486         kfree(data);
4487         dprintk("%s: done!\n", __func__);
4488 }
4489
4490 static const struct rpc_call_ops nfs4_lock_ops = {
4491         .rpc_call_prepare = nfs4_lock_prepare,
4492         .rpc_call_done = nfs4_lock_done,
4493         .rpc_release = nfs4_lock_release,
4494 };
4495
4496 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4497         .rpc_call_prepare = nfs4_recover_lock_prepare,
4498         .rpc_call_done = nfs4_lock_done,
4499         .rpc_release = nfs4_lock_release,
4500 };
4501
4502 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4503 {
4504         switch (error) {
4505         case -NFS4ERR_ADMIN_REVOKED:
4506         case -NFS4ERR_BAD_STATEID:
4507                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4508                 if (new_lock_owner != 0 ||
4509                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4510                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4511                 break;
4512         case -NFS4ERR_STALE_STATEID:
4513                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4514         case -NFS4ERR_EXPIRED:
4515                 nfs4_schedule_lease_recovery(server->nfs_client);
4516         };
4517 }
4518
4519 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4520 {
4521         struct nfs4_lockdata *data;
4522         struct rpc_task *task;
4523         struct rpc_message msg = {
4524                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4525                 .rpc_cred = state->owner->so_cred,
4526         };
4527         struct rpc_task_setup task_setup_data = {
4528                 .rpc_client = NFS_CLIENT(state->inode),
4529                 .rpc_message = &msg,
4530                 .callback_ops = &nfs4_lock_ops,
4531                 .workqueue = nfsiod_workqueue,
4532                 .flags = RPC_TASK_ASYNC,
4533         };
4534         int ret;
4535
4536         dprintk("%s: begin!\n", __func__);
4537         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4538                         fl->fl_u.nfs4_fl.owner,
4539                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4540         if (data == NULL)
4541                 return -ENOMEM;
4542         if (IS_SETLKW(cmd))
4543                 data->arg.block = 1;
4544         if (recovery_type > NFS_LOCK_NEW) {
4545                 if (recovery_type == NFS_LOCK_RECLAIM)
4546                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4547                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4548         }
4549         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4550         msg.rpc_argp = &data->arg;
4551         msg.rpc_resp = &data->res;
4552         task_setup_data.callback_data = data;
4553         task = rpc_run_task(&task_setup_data);
4554         if (IS_ERR(task))
4555                 return PTR_ERR(task);
4556         ret = nfs4_wait_for_completion_rpc_task(task);
4557         if (ret == 0) {
4558                 ret = data->rpc_status;
4559                 if (ret)
4560                         nfs4_handle_setlk_error(data->server, data->lsp,
4561                                         data->arg.new_lock_owner, ret);
4562         } else
4563                 data->cancelled = 1;
4564         rpc_put_task(task);
4565         dprintk("%s: done, ret = %d!\n", __func__, ret);
4566         return ret;
4567 }
4568
4569 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4570 {
4571         struct nfs_server *server = NFS_SERVER(state->inode);
4572         struct nfs4_exception exception = {
4573                 .inode = state->inode,
4574         };
4575         int err;
4576
4577         do {
4578                 /* Cache the lock if possible... */
4579                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4580                         return 0;
4581                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4582                 if (err != -NFS4ERR_DELAY)
4583                         break;
4584                 nfs4_handle_exception(server, err, &exception);
4585         } while (exception.retry);
4586         return err;
4587 }
4588
4589 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4590 {
4591         struct nfs_server *server = NFS_SERVER(state->inode);
4592         struct nfs4_exception exception = {
4593                 .inode = state->inode,
4594         };
4595         int err;
4596
4597         err = nfs4_set_lock_state(state, request);
4598         if (err != 0)
4599                 return err;
4600         do {
4601                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4602                         return 0;
4603                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4604                 switch (err) {
4605                 default:
4606                         goto out;
4607                 case -NFS4ERR_GRACE:
4608                 case -NFS4ERR_DELAY:
4609                         nfs4_handle_exception(server, err, &exception);
4610                         err = 0;
4611                 }
4612         } while (exception.retry);
4613 out:
4614         return err;
4615 }
4616
4617 #if defined(CONFIG_NFS_V4_1)
4618 static int nfs41_check_expired_locks(struct nfs4_state *state)
4619 {
4620         int status, ret = NFS_OK;
4621         struct nfs4_lock_state *lsp;
4622         struct nfs_server *server = NFS_SERVER(state->inode);
4623
4624         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4625                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4626                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4627                         if (status != NFS_OK) {
4628                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4629                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4630                                 ret = status;
4631                         }
4632                 }
4633         };
4634
4635         return ret;
4636 }
4637
4638 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4639 {
4640         int status = NFS_OK;
4641
4642         if (test_bit(LK_STATE_IN_USE, &state->flags))
4643                 status = nfs41_check_expired_locks(state);
4644         if (status == NFS_OK)
4645                 return status;
4646         return nfs4_lock_expired(state, request);
4647 }
4648 #endif
4649
4650 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4651 {
4652         struct nfs_inode *nfsi = NFS_I(state->inode);
4653         unsigned char fl_flags = request->fl_flags;
4654         int status = -ENOLCK;
4655
4656         if ((fl_flags & FL_POSIX) &&
4657                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4658                 goto out;
4659         /* Is this a delegated open? */
4660         status = nfs4_set_lock_state(state, request);
4661         if (status != 0)
4662                 goto out;
4663         request->fl_flags |= FL_ACCESS;
4664         status = do_vfs_lock(request->fl_file, request);
4665         if (status < 0)
4666                 goto out;
4667         down_read(&nfsi->rwsem);
4668         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4669                 /* Yes: cache locks! */
4670                 /* ...but avoid races with delegation recall... */
4671                 request->fl_flags = fl_flags & ~FL_SLEEP;
4672                 status = do_vfs_lock(request->fl_file, request);
4673                 goto out_unlock;
4674         }
4675         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4676         if (status != 0)
4677                 goto out_unlock;
4678         /* Note: we always want to sleep here! */
4679         request->fl_flags = fl_flags | FL_SLEEP;
4680         if (do_vfs_lock(request->fl_file, request) < 0)
4681                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4682                         "manager!\n", __func__);
4683 out_unlock:
4684         up_read(&nfsi->rwsem);
4685 out:
4686         request->fl_flags = fl_flags;
4687         return status;
4688 }
4689
4690 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4691 {
4692         struct nfs4_exception exception = {
4693                 .state = state,
4694                 .inode = state->inode,
4695         };
4696         int err;
4697
4698         do {
4699                 err = _nfs4_proc_setlk(state, cmd, request);
4700                 if (err == -NFS4ERR_DENIED)
4701                         err = -EAGAIN;
4702                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4703                                 err, &exception);
4704         } while (exception.retry);
4705         return err;
4706 }
4707
4708 static int
4709 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4710 {
4711         struct nfs_open_context *ctx;
4712         struct nfs4_state *state;
4713         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4714         int status;
4715
4716         /* verify open state */
4717         ctx = nfs_file_open_context(filp);
4718         state = ctx->state;
4719
4720         if (request->fl_start < 0 || request->fl_end < 0)
4721                 return -EINVAL;
4722
4723         if (IS_GETLK(cmd)) {
4724                 if (state != NULL)
4725                         return nfs4_proc_getlk(state, F_GETLK, request);
4726                 return 0;
4727         }
4728
4729         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4730                 return -EINVAL;
4731
4732         if (request->fl_type == F_UNLCK) {
4733                 if (state != NULL)
4734                         return nfs4_proc_unlck(state, cmd, request);
4735                 return 0;
4736         }
4737
4738         if (state == NULL)
4739                 return -ENOLCK;
4740         /*
4741          * Don't rely on the VFS having checked the file open mode,
4742          * since it won't do this for flock() locks.
4743          */
4744         switch (request->fl_type & (F_RDLCK|F_WRLCK|F_UNLCK)) {
4745         case F_RDLCK:
4746                 if (!(filp->f_mode & FMODE_READ))
4747                         return -EBADF;
4748                 break;
4749         case F_WRLCK:
4750                 if (!(filp->f_mode & FMODE_WRITE))
4751                         return -EBADF;
4752         }
4753
4754         do {
4755                 status = nfs4_proc_setlk(state, cmd, request);
4756                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4757                         break;
4758                 timeout = nfs4_set_lock_task_retry(timeout);
4759                 status = -ERESTARTSYS;
4760                 if (signalled())
4761                         break;
4762         } while(status < 0);
4763         return status;
4764 }
4765
4766 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4767 {
4768         struct nfs_server *server = NFS_SERVER(state->inode);
4769         struct nfs4_exception exception = { };
4770         int err;
4771
4772         err = nfs4_set_lock_state(state, fl);
4773         if (err != 0)
4774                 goto out;
4775         do {
4776                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4777                 switch (err) {
4778                         default:
4779                                 printk(KERN_ERR "NFS: %s: unhandled error "
4780                                         "%d.\n", __func__, err);
4781                         case 0:
4782                         case -ESTALE:
4783                                 goto out;
4784                         case -NFS4ERR_EXPIRED:
4785                                 nfs4_schedule_stateid_recovery(server, state);
4786                         case -NFS4ERR_STALE_CLIENTID:
4787                         case -NFS4ERR_STALE_STATEID:
4788                                 nfs4_schedule_lease_recovery(server->nfs_client);
4789                                 goto out;
4790                         case -NFS4ERR_BADSESSION:
4791                         case -NFS4ERR_BADSLOT:
4792                         case -NFS4ERR_BAD_HIGH_SLOT:
4793                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4794                         case -NFS4ERR_DEADSESSION:
4795                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4796                                 goto out;
4797                         case -ERESTARTSYS:
4798                                 /*
4799                                  * The show must go on: exit, but mark the
4800                                  * stateid as needing recovery.
4801                                  */
4802                         case -NFS4ERR_DELEG_REVOKED:
4803                         case -NFS4ERR_ADMIN_REVOKED:
4804                         case -NFS4ERR_BAD_STATEID:
4805                         case -NFS4ERR_OPENMODE:
4806                                 nfs4_schedule_stateid_recovery(server, state);
4807                                 err = 0;
4808                                 goto out;
4809                         case -EKEYEXPIRED:
4810                                 /*
4811                                  * User RPCSEC_GSS context has expired.
4812                                  * We cannot recover this stateid now, so
4813                                  * skip it and allow recovery thread to
4814                                  * proceed.
4815                                  */
4816                                 err = 0;
4817                                 goto out;
4818                         case -ENOMEM:
4819                         case -NFS4ERR_DENIED:
4820                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4821                                 err = 0;
4822                                 goto out;
4823                         case -NFS4ERR_DELAY:
4824                                 break;
4825                 }
4826                 err = nfs4_handle_exception(server, err, &exception);
4827         } while (exception.retry);
4828 out:
4829         return err;
4830 }
4831
4832 struct nfs_release_lockowner_data {
4833         struct nfs4_lock_state *lsp;
4834         struct nfs_server *server;
4835         struct nfs_release_lockowner_args args;
4836 };
4837
4838 static void nfs4_release_lockowner_release(void *calldata)
4839 {
4840         struct nfs_release_lockowner_data *data = calldata;
4841         nfs4_free_lock_state(data->server, data->lsp);
4842         kfree(calldata);
4843 }
4844
4845 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4846         .rpc_release = nfs4_release_lockowner_release,
4847 };
4848
4849 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4850 {
4851         struct nfs_server *server = lsp->ls_state->owner->so_server;
4852         struct nfs_release_lockowner_data *data;
4853         struct rpc_message msg = {
4854                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4855         };
4856
4857         if (server->nfs_client->cl_mvops->minor_version != 0)
4858                 return -EINVAL;
4859         data = kmalloc(sizeof(*data), GFP_NOFS);
4860         if (!data)
4861                 return -ENOMEM;
4862         data->lsp = lsp;
4863         data->server = server;
4864         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4865         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4866         data->args.lock_owner.s_dev = server->s_dev;
4867         msg.rpc_argp = &data->args;
4868         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4869         return 0;
4870 }
4871
4872 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4873
4874 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4875                                    const void *buf, size_t buflen,
4876                                    int flags, int type)
4877 {
4878         if (strcmp(key, "") != 0)
4879                 return -EINVAL;
4880
4881         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4882 }
4883
4884 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4885                                    void *buf, size_t buflen, int type)
4886 {
4887         if (strcmp(key, "") != 0)
4888                 return -EINVAL;
4889
4890         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4891 }
4892
4893 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4894                                        size_t list_len, const char *name,
4895                                        size_t name_len, int type)
4896 {
4897         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4898
4899         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4900                 return 0;
4901
4902         if (list && len <= list_len)
4903                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4904         return len;
4905 }
4906
4907 /*
4908  * nfs_fhget will use either the mounted_on_fileid or the fileid
4909  */
4910 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4911 {
4912         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4913                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4914               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4915               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4916                 return;
4917
4918         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4919                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4920         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4921         fattr->nlink = 2;
4922 }
4923
4924 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4925                 struct nfs4_fs_locations *fs_locations, struct page *page)
4926 {
4927         struct nfs_server *server = NFS_SERVER(dir);
4928         u32 bitmask[2] = {
4929                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4930         };
4931         struct nfs4_fs_locations_arg args = {
4932                 .dir_fh = NFS_FH(dir),
4933                 .name = name,
4934                 .page = page,
4935                 .bitmask = bitmask,
4936         };
4937         struct nfs4_fs_locations_res res = {
4938                 .fs_locations = fs_locations,
4939         };
4940         struct rpc_message msg = {
4941                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4942                 .rpc_argp = &args,
4943                 .rpc_resp = &res,
4944         };
4945         int status;
4946
4947         dprintk("%s: start\n", __func__);
4948
4949         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4950          * is not supported */
4951         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4952                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4953         else
4954                 bitmask[0] |= FATTR4_WORD0_FILEID;
4955
4956         nfs_fattr_init(&fs_locations->fattr);
4957         fs_locations->server = server;
4958         fs_locations->nlocations = 0;
4959         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4960         dprintk("%s: returned status = %d\n", __func__, status);
4961         return status;
4962 }
4963
4964 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4965 {
4966         int status;
4967         struct nfs4_secinfo_arg args = {
4968                 .dir_fh = NFS_FH(dir),
4969                 .name   = name,
4970         };
4971         struct nfs4_secinfo_res res = {
4972                 .flavors     = flavors,
4973         };
4974         struct rpc_message msg = {
4975                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4976                 .rpc_argp = &args,
4977                 .rpc_resp = &res,
4978         };
4979
4980         dprintk("NFS call  secinfo %s\n", name->name);
4981         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4982         dprintk("NFS reply  secinfo: %d\n", status);
4983         return status;
4984 }
4985
4986 static int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
4987                 struct nfs4_secinfo_flavors *flavors)
4988 {
4989         struct nfs4_exception exception = { };
4990         int err;
4991         do {
4992                 err = nfs4_handle_exception(NFS_SERVER(dir),
4993                                 _nfs4_proc_secinfo(dir, name, flavors),
4994                                 &exception);
4995         } while (exception.retry);
4996         return err;
4997 }
4998
4999 #ifdef CONFIG_NFS_V4_1
5000 /*
5001  * Check the exchange flags returned by the server for invalid flags, having
5002  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5003  * DS flags set.
5004  */
5005 static int nfs4_check_cl_exchange_flags(u32 flags)
5006 {
5007         if (flags & ~EXCHGID4_FLAG_MASK_R)
5008                 goto out_inval;
5009         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5010             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5011                 goto out_inval;
5012         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5013                 goto out_inval;
5014         return NFS_OK;
5015 out_inval:
5016         return -NFS4ERR_INVAL;
5017 }
5018
5019 static bool
5020 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
5021 {
5022         if (a->server_scope_sz == b->server_scope_sz &&
5023             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5024                 return true;
5025
5026         return false;
5027 }
5028
5029 /*
5030  * nfs4_proc_exchange_id()
5031  *
5032  * Since the clientid has expired, all compounds using sessions
5033  * associated with the stale clientid will be returning
5034  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5035  * be in some phase of session reset.
5036  */
5037 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5038 {
5039         nfs4_verifier verifier;
5040         struct nfs41_exchange_id_args args = {
5041                 .verifier = &verifier,
5042                 .client = clp,
5043                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5044         };
5045         struct nfs41_exchange_id_res res = {
5046                 .client = clp,
5047         };
5048         int status;
5049         struct rpc_message msg = {
5050                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5051                 .rpc_argp = &args,
5052                 .rpc_resp = &res,
5053                 .rpc_cred = cred,
5054         };
5055
5056         dprintk("--> %s\n", __func__);
5057         BUG_ON(clp == NULL);
5058
5059         nfs4_construct_boot_verifier(clp, &verifier);
5060
5061         args.id_len = scnprintf(args.id, sizeof(args.id),
5062                                 "%s/%s.%s/%u",
5063                                 clp->cl_ipaddr,
5064                                 init_utsname()->nodename,
5065                                 init_utsname()->domainname,
5066                                 clp->cl_rpcclient->cl_auth->au_flavor);
5067
5068         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
5069         if (unlikely(!res.server_scope)) {
5070                 status = -ENOMEM;
5071                 goto out;
5072         }
5073
5074         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
5075         if (unlikely(!res.impl_id)) {
5076                 status = -ENOMEM;
5077                 goto out_server_scope;
5078         }
5079
5080         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5081         if (!status)
5082                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
5083
5084         if (!status) {
5085                 /* use the most recent implementation id */
5086                 kfree(clp->impl_id);
5087                 clp->impl_id = res.impl_id;
5088         } else
5089                 kfree(res.impl_id);
5090
5091         if (!status) {
5092                 if (clp->server_scope &&
5093                     !nfs41_same_server_scope(clp->server_scope,
5094                                              res.server_scope)) {
5095                         dprintk("%s: server_scope mismatch detected\n",
5096                                 __func__);
5097                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5098                         kfree(clp->server_scope);
5099                         clp->server_scope = NULL;
5100                 }
5101
5102                 if (!clp->server_scope) {
5103                         clp->server_scope = res.server_scope;
5104                         goto out;
5105                 }
5106         }
5107
5108 out_server_scope:
5109         kfree(res.server_scope);
5110 out:
5111         if (clp->impl_id)
5112                 dprintk("%s: Server Implementation ID: "
5113                         "domain: %s, name: %s, date: %llu,%u\n",
5114                         __func__, clp->impl_id->domain, clp->impl_id->name,
5115                         clp->impl_id->date.seconds,
5116                         clp->impl_id->date.nseconds);
5117         dprintk("<-- %s status= %d\n", __func__, status);
5118         return status;
5119 }
5120
5121 struct nfs4_get_lease_time_data {
5122         struct nfs4_get_lease_time_args *args;
5123         struct nfs4_get_lease_time_res *res;
5124         struct nfs_client *clp;
5125 };
5126
5127 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5128                                         void *calldata)
5129 {
5130         int ret;
5131         struct nfs4_get_lease_time_data *data =
5132                         (struct nfs4_get_lease_time_data *)calldata;
5133
5134         dprintk("--> %s\n", __func__);
5135         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5136         /* just setup sequence, do not trigger session recovery
5137            since we're invoked within one */
5138         ret = nfs41_setup_sequence(data->clp->cl_session,
5139                                    &data->args->la_seq_args,
5140                                    &data->res->lr_seq_res, task);
5141
5142         BUG_ON(ret == -EAGAIN);
5143         rpc_call_start(task);
5144         dprintk("<-- %s\n", __func__);
5145 }
5146
5147 /*
5148  * Called from nfs4_state_manager thread for session setup, so don't recover
5149  * from sequence operation or clientid errors.
5150  */
5151 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5152 {
5153         struct nfs4_get_lease_time_data *data =
5154                         (struct nfs4_get_lease_time_data *)calldata;
5155
5156         dprintk("--> %s\n", __func__);
5157         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5158                 return;
5159         switch (task->tk_status) {
5160         case -NFS4ERR_DELAY:
5161         case -NFS4ERR_GRACE:
5162                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5163                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5164                 task->tk_status = 0;
5165                 /* fall through */
5166         case -NFS4ERR_RETRY_UNCACHED_REP:
5167                 rpc_restart_call_prepare(task);
5168                 return;
5169         }
5170         dprintk("<-- %s\n", __func__);
5171 }
5172
5173 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5174         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5175         .rpc_call_done = nfs4_get_lease_time_done,
5176 };
5177
5178 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5179 {
5180         struct rpc_task *task;
5181         struct nfs4_get_lease_time_args args;
5182         struct nfs4_get_lease_time_res res = {
5183                 .lr_fsinfo = fsinfo,
5184         };
5185         struct nfs4_get_lease_time_data data = {
5186                 .args = &args,
5187                 .res = &res,
5188                 .clp = clp,
5189         };
5190         struct rpc_message msg = {
5191                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5192                 .rpc_argp = &args,
5193                 .rpc_resp = &res,
5194         };
5195         struct rpc_task_setup task_setup = {
5196                 .rpc_client = clp->cl_rpcclient,
5197                 .rpc_message = &msg,
5198                 .callback_ops = &nfs4_get_lease_time_ops,
5199                 .callback_data = &data,
5200                 .flags = RPC_TASK_TIMEOUT,
5201         };
5202         int status;
5203
5204         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5205         dprintk("--> %s\n", __func__);
5206         task = rpc_run_task(&task_setup);
5207
5208         if (IS_ERR(task))
5209                 status = PTR_ERR(task);
5210         else {
5211                 status = task->tk_status;
5212                 rpc_put_task(task);
5213         }
5214         dprintk("<-- %s return %d\n", __func__, status);
5215
5216         return status;
5217 }
5218
5219 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5220 {
5221         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5222 }
5223
5224 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5225                 struct nfs4_slot *new,
5226                 u32 max_slots,
5227                 u32 ivalue)
5228 {
5229         struct nfs4_slot *old = NULL;
5230         u32 i;
5231
5232         spin_lock(&tbl->slot_tbl_lock);
5233         if (new) {
5234                 old = tbl->slots;
5235                 tbl->slots = new;
5236                 tbl->max_slots = max_slots;
5237         }
5238         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5239         for (i = 0; i < tbl->max_slots; i++)
5240                 tbl->slots[i].seq_nr = ivalue;
5241         spin_unlock(&tbl->slot_tbl_lock);
5242         kfree(old);
5243 }
5244
5245 /*
5246  * (re)Initialise a slot table
5247  */
5248 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5249                                  u32 ivalue)
5250 {
5251         struct nfs4_slot *new = NULL;
5252         int ret = -ENOMEM;
5253
5254         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5255                 max_reqs, tbl->max_slots);
5256
5257         /* Does the newly negotiated max_reqs match the existing slot table? */
5258         if (max_reqs != tbl->max_slots) {
5259                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5260                 if (!new)
5261                         goto out;
5262         }
5263         ret = 0;
5264
5265         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5266         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5267                 tbl, tbl->slots, tbl->max_slots);
5268 out:
5269         dprintk("<-- %s: return %d\n", __func__, ret);
5270         return ret;
5271 }
5272
5273 /* Destroy the slot table */
5274 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5275 {
5276         if (session->fc_slot_table.slots != NULL) {
5277                 kfree(session->fc_slot_table.slots);
5278                 session->fc_slot_table.slots = NULL;
5279         }
5280         if (session->bc_slot_table.slots != NULL) {
5281                 kfree(session->bc_slot_table.slots);
5282                 session->bc_slot_table.slots = NULL;
5283         }
5284         return;
5285 }
5286
5287 /*
5288  * Initialize or reset the forechannel and backchannel tables
5289  */
5290 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5291 {
5292         struct nfs4_slot_table *tbl;
5293         int status;
5294
5295         dprintk("--> %s\n", __func__);
5296         /* Fore channel */
5297         tbl = &ses->fc_slot_table;
5298         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5299         if (status) /* -ENOMEM */
5300                 return status;
5301         /* Back channel */
5302         tbl = &ses->bc_slot_table;
5303         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5304         if (status && tbl->slots == NULL)
5305                 /* Fore and back channel share a connection so get
5306                  * both slot tables or neither */
5307                 nfs4_destroy_slot_tables(ses);
5308         return status;
5309 }
5310
5311 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5312 {
5313         struct nfs4_session *session;
5314         struct nfs4_slot_table *tbl;
5315
5316         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5317         if (!session)
5318                 return NULL;
5319
5320         tbl = &session->fc_slot_table;
5321         tbl->highest_used_slotid = NFS4_NO_SLOT;
5322         spin_lock_init(&tbl->slot_tbl_lock);
5323         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5324         init_completion(&tbl->complete);
5325
5326         tbl = &session->bc_slot_table;
5327         tbl->highest_used_slotid = NFS4_NO_SLOT;
5328         spin_lock_init(&tbl->slot_tbl_lock);
5329         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5330         init_completion(&tbl->complete);
5331
5332         session->session_state = 1<<NFS4_SESSION_INITING;
5333
5334         session->clp = clp;
5335         return session;
5336 }
5337
5338 void nfs4_destroy_session(struct nfs4_session *session)
5339 {
5340         struct rpc_xprt *xprt;
5341
5342         nfs4_proc_destroy_session(session);
5343
5344         rcu_read_lock();
5345         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5346         rcu_read_unlock();
5347         dprintk("%s Destroy backchannel for xprt %p\n",
5348                 __func__, xprt);
5349         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5350         nfs4_destroy_slot_tables(session);
5351         kfree(session);
5352 }
5353
5354 /*
5355  * Initialize the values to be used by the client in CREATE_SESSION
5356  * If nfs4_init_session set the fore channel request and response sizes,
5357  * use them.
5358  *
5359  * Set the back channel max_resp_sz_cached to zero to force the client to
5360  * always set csa_cachethis to FALSE because the current implementation
5361  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5362  */
5363 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5364 {
5365         struct nfs4_session *session = args->client->cl_session;
5366         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5367                      mxresp_sz = session->fc_attrs.max_resp_sz;
5368
5369         if (mxrqst_sz == 0)
5370                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5371         if (mxresp_sz == 0)
5372                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5373         /* Fore channel attributes */
5374         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5375         args->fc_attrs.max_resp_sz = mxresp_sz;
5376         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5377         args->fc_attrs.max_reqs = max_session_slots;
5378
5379         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5380                 "max_ops=%u max_reqs=%u\n",
5381                 __func__,
5382                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5383                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5384
5385         /* Back channel attributes */
5386         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5387         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5388         args->bc_attrs.max_resp_sz_cached = 0;
5389         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5390         args->bc_attrs.max_reqs = 1;
5391
5392         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5393                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5394                 __func__,
5395                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5396                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5397                 args->bc_attrs.max_reqs);
5398 }
5399
5400 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5401 {
5402         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5403         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5404
5405         if (rcvd->max_resp_sz > sent->max_resp_sz)
5406                 return -EINVAL;
5407         /*
5408          * Our requested max_ops is the minimum we need; we're not
5409          * prepared to break up compounds into smaller pieces than that.
5410          * So, no point even trying to continue if the server won't
5411          * cooperate:
5412          */
5413         if (rcvd->max_ops < sent->max_ops)
5414                 return -EINVAL;
5415         if (rcvd->max_reqs == 0)
5416                 return -EINVAL;
5417         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5418                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5419         return 0;
5420 }
5421
5422 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5423 {
5424         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5425         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5426
5427         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5428                 return -EINVAL;
5429         if (rcvd->max_resp_sz < sent->max_resp_sz)
5430                 return -EINVAL;
5431         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5432                 return -EINVAL;
5433         /* These would render the backchannel useless: */
5434         if (rcvd->max_ops != sent->max_ops)
5435                 return -EINVAL;
5436         if (rcvd->max_reqs != sent->max_reqs)
5437                 return -EINVAL;
5438         return 0;
5439 }
5440
5441 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5442                                      struct nfs4_session *session)
5443 {
5444         int ret;
5445
5446         ret = nfs4_verify_fore_channel_attrs(args, session);
5447         if (ret)
5448                 return ret;
5449         return nfs4_verify_back_channel_attrs(args, session);
5450 }
5451
5452 static int _nfs4_proc_create_session(struct nfs_client *clp)
5453 {
5454         struct nfs4_session *session = clp->cl_session;
5455         struct nfs41_create_session_args args = {
5456                 .client = clp,
5457                 .cb_program = NFS4_CALLBACK,
5458         };
5459         struct nfs41_create_session_res res = {
5460                 .client = clp,
5461         };
5462         struct rpc_message msg = {
5463                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5464                 .rpc_argp = &args,
5465                 .rpc_resp = &res,
5466         };
5467         int status;
5468
5469         nfs4_init_channel_attrs(&args);
5470         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5471
5472         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5473
5474         if (!status)
5475                 /* Verify the session's negotiated channel_attrs values */
5476                 status = nfs4_verify_channel_attrs(&args, session);
5477         if (!status) {
5478                 /* Increment the clientid slot sequence id */
5479                 clp->cl_seqid++;
5480         }
5481
5482         return status;
5483 }
5484
5485 /*
5486  * Issues a CREATE_SESSION operation to the server.
5487  * It is the responsibility of the caller to verify the session is
5488  * expired before calling this routine.
5489  */
5490 int nfs4_proc_create_session(struct nfs_client *clp)
5491 {
5492         int status;
5493         unsigned *ptr;
5494         struct nfs4_session *session = clp->cl_session;
5495
5496         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5497
5498         status = _nfs4_proc_create_session(clp);
5499         if (status)
5500                 goto out;
5501
5502         /* Init or reset the session slot tables */
5503         status = nfs4_setup_session_slot_tables(session);
5504         dprintk("slot table setup returned %d\n", status);
5505         if (status)
5506                 goto out;
5507
5508         ptr = (unsigned *)&session->sess_id.data[0];
5509         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5510                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5511 out:
5512         dprintk("<-- %s\n", __func__);
5513         return status;
5514 }
5515
5516 /*
5517  * Issue the over-the-wire RPC DESTROY_SESSION.
5518  * The caller must serialize access to this routine.
5519  */
5520 int nfs4_proc_destroy_session(struct nfs4_session *session)
5521 {
5522         int status = 0;
5523         struct rpc_message msg;
5524
5525         dprintk("--> nfs4_proc_destroy_session\n");
5526
5527         /* session is still being setup */
5528         if (session->clp->cl_cons_state != NFS_CS_READY)
5529                 return status;
5530
5531         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5532         msg.rpc_argp = session;
5533         msg.rpc_resp = NULL;
5534         msg.rpc_cred = NULL;
5535         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5536
5537         if (status)
5538                 printk(KERN_WARNING
5539                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5540                         "Session has been destroyed regardless...\n", status);
5541
5542         dprintk("<-- nfs4_proc_destroy_session\n");
5543         return status;
5544 }
5545
5546 int nfs4_init_session(struct nfs_server *server)
5547 {
5548         struct nfs_client *clp = server->nfs_client;
5549         struct nfs4_session *session;
5550         unsigned int rsize, wsize;
5551         int ret;
5552
5553         if (!nfs4_has_session(clp))
5554                 return 0;
5555
5556         session = clp->cl_session;
5557         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5558                 return 0;
5559
5560         rsize = server->rsize;
5561         if (rsize == 0)
5562                 rsize = NFS_MAX_FILE_IO_SIZE;
5563         wsize = server->wsize;
5564         if (wsize == 0)
5565                 wsize = NFS_MAX_FILE_IO_SIZE;
5566
5567         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5568         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5569
5570         ret = nfs4_recover_expired_lease(server);
5571         if (!ret)
5572                 ret = nfs4_check_client_ready(clp);
5573         return ret;
5574 }
5575
5576 int nfs4_init_ds_session(struct nfs_client *clp)
5577 {
5578         struct nfs4_session *session = clp->cl_session;
5579         int ret;
5580
5581         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5582                 return 0;
5583
5584         ret = nfs4_client_recover_expired_lease(clp);
5585         if (!ret)
5586                 /* Test for the DS role */
5587                 if (!is_ds_client(clp))
5588                         ret = -ENODEV;
5589         if (!ret)
5590                 ret = nfs4_check_client_ready(clp);
5591         return ret;
5592
5593 }
5594 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5595
5596
5597 /*
5598  * Renew the cl_session lease.
5599  */
5600 struct nfs4_sequence_data {
5601         struct nfs_client *clp;
5602         struct nfs4_sequence_args args;
5603         struct nfs4_sequence_res res;
5604 };
5605
5606 static void nfs41_sequence_release(void *data)
5607 {
5608         struct nfs4_sequence_data *calldata = data;
5609         struct nfs_client *clp = calldata->clp;
5610
5611         if (atomic_read(&clp->cl_count) > 1)
5612                 nfs4_schedule_state_renewal(clp);
5613         nfs_put_client(clp);
5614         kfree(calldata);
5615 }
5616
5617 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5618 {
5619         switch(task->tk_status) {
5620         case -NFS4ERR_DELAY:
5621                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5622                 return -EAGAIN;
5623         default:
5624                 nfs4_schedule_lease_recovery(clp);
5625         }
5626         return 0;
5627 }
5628
5629 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5630 {
5631         struct nfs4_sequence_data *calldata = data;
5632         struct nfs_client *clp = calldata->clp;
5633
5634         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5635                 return;
5636
5637         if (task->tk_status < 0) {
5638                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5639                 if (atomic_read(&clp->cl_count) == 1)
5640                         goto out;
5641
5642                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5643                         rpc_restart_call_prepare(task);
5644                         return;
5645                 }
5646         }
5647         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5648 out:
5649         dprintk("<-- %s\n", __func__);
5650 }
5651
5652 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5653 {
5654         struct nfs4_sequence_data *calldata = data;
5655         struct nfs_client *clp = calldata->clp;
5656         struct nfs4_sequence_args *args;
5657         struct nfs4_sequence_res *res;
5658
5659         args = task->tk_msg.rpc_argp;
5660         res = task->tk_msg.rpc_resp;
5661
5662         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5663                 return;
5664         rpc_call_start(task);
5665 }
5666
5667 static const struct rpc_call_ops nfs41_sequence_ops = {
5668         .rpc_call_done = nfs41_sequence_call_done,
5669         .rpc_call_prepare = nfs41_sequence_prepare,
5670         .rpc_release = nfs41_sequence_release,
5671 };
5672
5673 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5674 {
5675         struct nfs4_sequence_data *calldata;
5676         struct rpc_message msg = {
5677                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5678                 .rpc_cred = cred,
5679         };
5680         struct rpc_task_setup task_setup_data = {
5681                 .rpc_client = clp->cl_rpcclient,
5682                 .rpc_message = &msg,
5683                 .callback_ops = &nfs41_sequence_ops,
5684                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5685         };
5686
5687         if (!atomic_inc_not_zero(&clp->cl_count))
5688                 return ERR_PTR(-EIO);
5689         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5690         if (calldata == NULL) {
5691                 nfs_put_client(clp);
5692                 return ERR_PTR(-ENOMEM);
5693         }
5694         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5695         msg.rpc_argp = &calldata->args;
5696         msg.rpc_resp = &calldata->res;
5697         calldata->clp = clp;
5698         task_setup_data.callback_data = calldata;
5699
5700         return rpc_run_task(&task_setup_data);
5701 }
5702
5703 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5704 {
5705         struct rpc_task *task;
5706         int ret = 0;
5707
5708         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5709                 return 0;
5710         task = _nfs41_proc_sequence(clp, cred);
5711         if (IS_ERR(task))
5712                 ret = PTR_ERR(task);
5713         else
5714                 rpc_put_task_async(task);
5715         dprintk("<-- %s status=%d\n", __func__, ret);
5716         return ret;
5717 }
5718
5719 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5720 {
5721         struct rpc_task *task;
5722         int ret;
5723
5724         task = _nfs41_proc_sequence(clp, cred);
5725         if (IS_ERR(task)) {
5726                 ret = PTR_ERR(task);
5727                 goto out;
5728         }
5729         ret = rpc_wait_for_completion_task(task);
5730         if (!ret) {
5731                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5732
5733                 if (task->tk_status == 0)
5734                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5735                 ret = task->tk_status;
5736         }
5737         rpc_put_task(task);
5738 out:
5739         dprintk("<-- %s status=%d\n", __func__, ret);
5740         return ret;
5741 }
5742
5743 struct nfs4_reclaim_complete_data {
5744         struct nfs_client *clp;
5745         struct nfs41_reclaim_complete_args arg;
5746         struct nfs41_reclaim_complete_res res;
5747 };
5748
5749 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5750 {
5751         struct nfs4_reclaim_complete_data *calldata = data;
5752
5753         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5754         if (nfs41_setup_sequence(calldata->clp->cl_session,
5755                                 &calldata->arg.seq_args,
5756                                 &calldata->res.seq_res, task))
5757                 return;
5758
5759         rpc_call_start(task);
5760 }
5761
5762 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5763 {
5764         switch(task->tk_status) {
5765         case 0:
5766         case -NFS4ERR_COMPLETE_ALREADY:
5767         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5768                 break;
5769         case -NFS4ERR_DELAY:
5770                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5771                 /* fall through */
5772         case -NFS4ERR_RETRY_UNCACHED_REP:
5773                 return -EAGAIN;
5774         default:
5775                 nfs4_schedule_lease_recovery(clp);
5776         }
5777         return 0;
5778 }
5779
5780 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5781 {
5782         struct nfs4_reclaim_complete_data *calldata = data;
5783         struct nfs_client *clp = calldata->clp;
5784         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5785
5786         dprintk("--> %s\n", __func__);
5787         if (!nfs41_sequence_done(task, res))
5788                 return;
5789
5790         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5791                 rpc_restart_call_prepare(task);
5792                 return;
5793         }
5794         dprintk("<-- %s\n", __func__);
5795 }
5796
5797 static void nfs4_free_reclaim_complete_data(void *data)
5798 {
5799         struct nfs4_reclaim_complete_data *calldata = data;
5800
5801         kfree(calldata);
5802 }
5803
5804 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5805         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5806         .rpc_call_done = nfs4_reclaim_complete_done,
5807         .rpc_release = nfs4_free_reclaim_complete_data,
5808 };
5809
5810 /*
5811  * Issue a global reclaim complete.
5812  */
5813 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5814 {
5815         struct nfs4_reclaim_complete_data *calldata;
5816         struct rpc_task *task;
5817         struct rpc_message msg = {
5818                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5819         };
5820         struct rpc_task_setup task_setup_data = {
5821                 .rpc_client = clp->cl_rpcclient,
5822                 .rpc_message = &msg,
5823                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5824                 .flags = RPC_TASK_ASYNC,
5825         };
5826         int status = -ENOMEM;
5827
5828         dprintk("--> %s\n", __func__);
5829         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5830         if (calldata == NULL)
5831                 goto out;
5832         calldata->clp = clp;
5833         calldata->arg.one_fs = 0;
5834
5835         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5836         msg.rpc_argp = &calldata->arg;
5837         msg.rpc_resp = &calldata->res;
5838         task_setup_data.callback_data = calldata;
5839         task = rpc_run_task(&task_setup_data);
5840         if (IS_ERR(task)) {
5841                 status = PTR_ERR(task);
5842                 goto out;
5843         }
5844         status = nfs4_wait_for_completion_rpc_task(task);
5845         if (status == 0)
5846                 status = task->tk_status;
5847         rpc_put_task(task);
5848         return 0;
5849 out:
5850         dprintk("<-- %s status=%d\n", __func__, status);
5851         return status;
5852 }
5853
5854 static void
5855 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5856 {
5857         struct nfs4_layoutget *lgp = calldata;
5858         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5859
5860         dprintk("--> %s\n", __func__);
5861         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5862          * right now covering the LAYOUTGET we are about to send.
5863          * However, that is not so catastrophic, and there seems
5864          * to be no way to prevent it completely.
5865          */
5866         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5867                                 &lgp->res.seq_res, task))
5868                 return;
5869         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5870                                           NFS_I(lgp->args.inode)->layout,
5871                                           lgp->args.ctx->state)) {
5872                 rpc_exit(task, NFS4_OK);
5873                 return;
5874         }
5875         rpc_call_start(task);
5876 }
5877
5878 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5879 {
5880         struct nfs4_layoutget *lgp = calldata;
5881         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5882
5883         dprintk("--> %s\n", __func__);
5884
5885         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5886                 return;
5887
5888         switch (task->tk_status) {
5889         case 0:
5890                 break;
5891         case -NFS4ERR_LAYOUTTRYLATER:
5892         case -NFS4ERR_RECALLCONFLICT:
5893                 task->tk_status = -NFS4ERR_DELAY;
5894                 /* Fall through */
5895         default:
5896                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5897                         rpc_restart_call_prepare(task);
5898                         return;
5899                 }
5900         }
5901         dprintk("<-- %s\n", __func__);
5902 }
5903
5904 static void nfs4_layoutget_release(void *calldata)
5905 {
5906         struct nfs4_layoutget *lgp = calldata;
5907
5908         dprintk("--> %s\n", __func__);
5909         put_nfs_open_context(lgp->args.ctx);
5910         kfree(calldata);
5911         dprintk("<-- %s\n", __func__);
5912 }
5913
5914 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5915         .rpc_call_prepare = nfs4_layoutget_prepare,
5916         .rpc_call_done = nfs4_layoutget_done,
5917         .rpc_release = nfs4_layoutget_release,
5918 };
5919
5920 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5921 {
5922         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5923         struct rpc_task *task;
5924         struct rpc_message msg = {
5925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5926                 .rpc_argp = &lgp->args,
5927                 .rpc_resp = &lgp->res,
5928         };
5929         struct rpc_task_setup task_setup_data = {
5930                 .rpc_client = server->client,
5931                 .rpc_message = &msg,
5932                 .callback_ops = &nfs4_layoutget_call_ops,
5933                 .callback_data = lgp,
5934                 .flags = RPC_TASK_ASYNC,
5935         };
5936         int status = 0;
5937
5938         dprintk("--> %s\n", __func__);
5939
5940         lgp->res.layoutp = &lgp->args.layout;
5941         lgp->res.seq_res.sr_slot = NULL;
5942         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5943         task = rpc_run_task(&task_setup_data);
5944         if (IS_ERR(task))
5945                 return PTR_ERR(task);
5946         status = nfs4_wait_for_completion_rpc_task(task);
5947         if (status == 0)
5948                 status = task->tk_status;
5949         if (status == 0)
5950                 status = pnfs_layout_process(lgp);
5951         rpc_put_task(task);
5952         dprintk("<-- %s status=%d\n", __func__, status);
5953         return status;
5954 }
5955
5956 static void
5957 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5958 {
5959         struct nfs4_layoutreturn *lrp = calldata;
5960
5961         dprintk("--> %s\n", __func__);
5962         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5963                                 &lrp->res.seq_res, task))
5964                 return;
5965         rpc_call_start(task);
5966 }
5967
5968 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5969 {
5970         struct nfs4_layoutreturn *lrp = calldata;
5971         struct nfs_server *server;
5972         struct pnfs_layout_hdr *lo = lrp->args.layout;
5973
5974         dprintk("--> %s\n", __func__);
5975
5976         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5977                 return;
5978
5979         server = NFS_SERVER(lrp->args.inode);
5980         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5981                 rpc_restart_call_prepare(task);
5982                 return;
5983         }
5984         spin_lock(&lo->plh_inode->i_lock);
5985         if (task->tk_status == 0) {
5986                 if (lrp->res.lrs_present) {
5987                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5988                 } else
5989                         BUG_ON(!list_empty(&lo->plh_segs));
5990         }
5991         lo->plh_block_lgets--;
5992         spin_unlock(&lo->plh_inode->i_lock);
5993         dprintk("<-- %s\n", __func__);
5994 }
5995
5996 static void nfs4_layoutreturn_release(void *calldata)
5997 {
5998         struct nfs4_layoutreturn *lrp = calldata;
5999
6000         dprintk("--> %s\n", __func__);
6001         put_layout_hdr(lrp->args.layout);
6002         kfree(calldata);
6003         dprintk("<-- %s\n", __func__);
6004 }
6005
6006 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6007         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6008         .rpc_call_done = nfs4_layoutreturn_done,
6009         .rpc_release = nfs4_layoutreturn_release,
6010 };
6011
6012 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6013 {
6014         struct rpc_task *task;
6015         struct rpc_message msg = {
6016                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6017                 .rpc_argp = &lrp->args,
6018                 .rpc_resp = &lrp->res,
6019         };
6020         struct rpc_task_setup task_setup_data = {
6021                 .rpc_client = lrp->clp->cl_rpcclient,
6022                 .rpc_message = &msg,
6023                 .callback_ops = &nfs4_layoutreturn_call_ops,
6024                 .callback_data = lrp,
6025         };
6026         int status;
6027
6028         dprintk("--> %s\n", __func__);
6029         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6030         task = rpc_run_task(&task_setup_data);
6031         if (IS_ERR(task))
6032                 return PTR_ERR(task);
6033         status = task->tk_status;
6034         dprintk("<-- %s status=%d\n", __func__, status);
6035         rpc_put_task(task);
6036         return status;
6037 }
6038
6039 /*
6040  * Retrieve the list of Data Server devices from the MDS.
6041  */
6042 static int _nfs4_getdevicelist(struct nfs_server *server,
6043                                     const struct nfs_fh *fh,
6044                                     struct pnfs_devicelist *devlist)
6045 {
6046         struct nfs4_getdevicelist_args args = {
6047                 .fh = fh,
6048                 .layoutclass = server->pnfs_curr_ld->id,
6049         };
6050         struct nfs4_getdevicelist_res res = {
6051                 .devlist = devlist,
6052         };
6053         struct rpc_message msg = {
6054                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6055                 .rpc_argp = &args,
6056                 .rpc_resp = &res,
6057         };
6058         int status;
6059
6060         dprintk("--> %s\n", __func__);
6061         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6062                                 &res.seq_res, 0);
6063         dprintk("<-- %s status=%d\n", __func__, status);
6064         return status;
6065 }
6066
6067 int nfs4_proc_getdevicelist(struct nfs_server *server,
6068                             const struct nfs_fh *fh,
6069                             struct pnfs_devicelist *devlist)
6070 {
6071         struct nfs4_exception exception = { };
6072         int err;
6073
6074         do {
6075                 err = nfs4_handle_exception(server,
6076                                 _nfs4_getdevicelist(server, fh, devlist),
6077                                 &exception);
6078         } while (exception.retry);
6079
6080         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6081                 err, devlist->num_devs);
6082
6083         return err;
6084 }
6085 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6086
6087 static int
6088 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6089 {
6090         struct nfs4_getdeviceinfo_args args = {
6091                 .pdev = pdev,
6092         };
6093         struct nfs4_getdeviceinfo_res res = {
6094                 .pdev = pdev,
6095         };
6096         struct rpc_message msg = {
6097                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6098                 .rpc_argp = &args,
6099                 .rpc_resp = &res,
6100         };
6101         int status;
6102
6103         dprintk("--> %s\n", __func__);
6104         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6105         dprintk("<-- %s status=%d\n", __func__, status);
6106
6107         return status;
6108 }
6109
6110 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6111 {
6112         struct nfs4_exception exception = { };
6113         int err;
6114
6115         do {
6116                 err = nfs4_handle_exception(server,
6117                                         _nfs4_proc_getdeviceinfo(server, pdev),
6118                                         &exception);
6119         } while (exception.retry);
6120         return err;
6121 }
6122 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6123
6124 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6125 {
6126         struct nfs4_layoutcommit_data *data = calldata;
6127         struct nfs_server *server = NFS_SERVER(data->args.inode);
6128
6129         if (nfs4_setup_sequence(server, &data->args.seq_args,
6130                                 &data->res.seq_res, task))
6131                 return;
6132         rpc_call_start(task);
6133 }
6134
6135 static void
6136 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6137 {
6138         struct nfs4_layoutcommit_data *data = calldata;
6139         struct nfs_server *server = NFS_SERVER(data->args.inode);
6140
6141         if (!nfs4_sequence_done(task, &data->res.seq_res))
6142                 return;
6143
6144         switch (task->tk_status) { /* Just ignore these failures */
6145         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6146         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6147         case -NFS4ERR_BADLAYOUT:     /* no layout */
6148         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6149                 task->tk_status = 0;
6150                 break;
6151         case 0:
6152                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6153                                                    data->res.fattr);
6154                 break;
6155         default:
6156                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6157                         rpc_restart_call_prepare(task);
6158                         return;
6159                 }
6160         }
6161 }
6162
6163 static void nfs4_layoutcommit_release(void *calldata)
6164 {
6165         struct nfs4_layoutcommit_data *data = calldata;
6166         struct pnfs_layout_segment *lseg, *tmp;
6167         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6168
6169         pnfs_cleanup_layoutcommit(data);
6170         /* Matched by references in pnfs_set_layoutcommit */
6171         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6172                 list_del_init(&lseg->pls_lc_list);
6173                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6174                                        &lseg->pls_flags))
6175                         put_lseg(lseg);
6176         }
6177
6178         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6179         smp_mb__after_clear_bit();
6180         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6181
6182         put_rpccred(data->cred);
6183         kfree(data);
6184 }
6185
6186 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6187         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6188         .rpc_call_done = nfs4_layoutcommit_done,
6189         .rpc_release = nfs4_layoutcommit_release,
6190 };
6191
6192 int
6193 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6194 {
6195         struct rpc_message msg = {
6196                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6197                 .rpc_argp = &data->args,
6198                 .rpc_resp = &data->res,
6199                 .rpc_cred = data->cred,
6200         };
6201         struct rpc_task_setup task_setup_data = {
6202                 .task = &data->task,
6203                 .rpc_client = NFS_CLIENT(data->args.inode),
6204                 .rpc_message = &msg,
6205                 .callback_ops = &nfs4_layoutcommit_ops,
6206                 .callback_data = data,
6207                 .flags = RPC_TASK_ASYNC,
6208         };
6209         struct rpc_task *task;
6210         int status = 0;
6211
6212         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6213                 "lbw: %llu inode %lu\n",
6214                 data->task.tk_pid, sync,
6215                 data->args.lastbytewritten,
6216                 data->args.inode->i_ino);
6217
6218         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6219         task = rpc_run_task(&task_setup_data);
6220         if (IS_ERR(task))
6221                 return PTR_ERR(task);
6222         if (sync == false)
6223                 goto out;
6224         status = nfs4_wait_for_completion_rpc_task(task);
6225         if (status != 0)
6226                 goto out;
6227         status = task->tk_status;
6228 out:
6229         dprintk("%s: status %d\n", __func__, status);
6230         rpc_put_task(task);
6231         return status;
6232 }
6233
6234 static int
6235 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6236                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6237 {
6238         struct nfs41_secinfo_no_name_args args = {
6239                 .style = SECINFO_STYLE_CURRENT_FH,
6240         };
6241         struct nfs4_secinfo_res res = {
6242                 .flavors = flavors,
6243         };
6244         struct rpc_message msg = {
6245                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6246                 .rpc_argp = &args,
6247                 .rpc_resp = &res,
6248         };
6249         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6250 }
6251
6252 static int
6253 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6254                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6255 {
6256         struct nfs4_exception exception = { };
6257         int err;
6258         do {
6259                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6260                 switch (err) {
6261                 case 0:
6262                 case -NFS4ERR_WRONGSEC:
6263                 case -NFS4ERR_NOTSUPP:
6264                         goto out;
6265                 default:
6266                         err = nfs4_handle_exception(server, err, &exception);
6267                 }
6268         } while (exception.retry);
6269 out:
6270         return err;
6271 }
6272
6273 static int
6274 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6275                     struct nfs_fsinfo *info)
6276 {
6277         int err;
6278         struct page *page;
6279         rpc_authflavor_t flavor;
6280         struct nfs4_secinfo_flavors *flavors;
6281
6282         page = alloc_page(GFP_KERNEL);
6283         if (!page) {
6284                 err = -ENOMEM;
6285                 goto out;
6286         }
6287
6288         flavors = page_address(page);
6289         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6290
6291         /*
6292          * Fall back on "guess and check" method if
6293          * the server doesn't support SECINFO_NO_NAME
6294          */
6295         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6296                 err = nfs4_find_root_sec(server, fhandle, info);
6297                 goto out_freepage;
6298         }
6299         if (err)
6300                 goto out_freepage;
6301
6302         flavor = nfs_find_best_sec(flavors);
6303         if (err == 0)
6304                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6305
6306 out_freepage:
6307         put_page(page);
6308         if (err == -EACCES)
6309                 return -EPERM;
6310 out:
6311         return err;
6312 }
6313
6314 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6315 {
6316         int status;
6317         struct nfs41_test_stateid_args args = {
6318                 .stateid = stateid,
6319         };
6320         struct nfs41_test_stateid_res res;
6321         struct rpc_message msg = {
6322                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6323                 .rpc_argp = &args,
6324                 .rpc_resp = &res,
6325         };
6326
6327         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6328         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6329
6330         if (status == NFS_OK)
6331                 return res.status;
6332         return status;
6333 }
6334
6335 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6336 {
6337         struct nfs4_exception exception = { };
6338         int err;
6339         do {
6340                 err = nfs4_handle_exception(server,
6341                                 _nfs41_test_stateid(server, stateid),
6342                                 &exception);
6343         } while (exception.retry);
6344         return err;
6345 }
6346
6347 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6348 {
6349         struct nfs41_free_stateid_args args = {
6350                 .stateid = stateid,
6351         };
6352         struct nfs41_free_stateid_res res;
6353         struct rpc_message msg = {
6354                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6355                 .rpc_argp = &args,
6356                 .rpc_resp = &res,
6357         };
6358
6359         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6360         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6361 }
6362
6363 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6364 {
6365         struct nfs4_exception exception = { };
6366         int err;
6367         do {
6368                 err = nfs4_handle_exception(server,
6369                                 _nfs4_free_stateid(server, stateid),
6370                                 &exception);
6371         } while (exception.retry);
6372         return err;
6373 }
6374
6375 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6376                 const nfs4_stateid *s2)
6377 {
6378         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6379                 return false;
6380
6381         if (s1->seqid == s2->seqid)
6382                 return true;
6383         if (s1->seqid == 0 || s2->seqid == 0)
6384                 return true;
6385
6386         return false;
6387 }
6388
6389 #endif /* CONFIG_NFS_V4_1 */
6390
6391 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6392                 const nfs4_stateid *s2)
6393 {
6394         return nfs4_stateid_match(s1, s2);
6395 }
6396
6397
6398 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6399         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6400         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6401         .recover_open   = nfs4_open_reclaim,
6402         .recover_lock   = nfs4_lock_reclaim,
6403         .establish_clid = nfs4_init_clientid,
6404         .get_clid_cred  = nfs4_get_setclientid_cred,
6405 };
6406
6407 #if defined(CONFIG_NFS_V4_1)
6408 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6409         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6410         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6411         .recover_open   = nfs4_open_reclaim,
6412         .recover_lock   = nfs4_lock_reclaim,
6413         .establish_clid = nfs41_init_clientid,
6414         .get_clid_cred  = nfs4_get_exchange_id_cred,
6415         .reclaim_complete = nfs41_proc_reclaim_complete,
6416 };
6417 #endif /* CONFIG_NFS_V4_1 */
6418
6419 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6420         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6421         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6422         .recover_open   = nfs4_open_expired,
6423         .recover_lock   = nfs4_lock_expired,
6424         .establish_clid = nfs4_init_clientid,
6425         .get_clid_cred  = nfs4_get_setclientid_cred,
6426 };
6427
6428 #if defined(CONFIG_NFS_V4_1)
6429 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6430         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6431         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6432         .recover_open   = nfs41_open_expired,
6433         .recover_lock   = nfs41_lock_expired,
6434         .establish_clid = nfs41_init_clientid,
6435         .get_clid_cred  = nfs4_get_exchange_id_cred,
6436 };
6437 #endif /* CONFIG_NFS_V4_1 */
6438
6439 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6440         .sched_state_renewal = nfs4_proc_async_renew,
6441         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6442         .renew_lease = nfs4_proc_renew,
6443 };
6444
6445 #if defined(CONFIG_NFS_V4_1)
6446 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6447         .sched_state_renewal = nfs41_proc_async_sequence,
6448         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6449         .renew_lease = nfs4_proc_sequence,
6450 };
6451 #endif
6452
6453 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6454         .minor_version = 0,
6455         .call_sync = _nfs4_call_sync,
6456         .match_stateid = nfs4_match_stateid,
6457         .find_root_sec = nfs4_find_root_sec,
6458         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6459         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6460         .state_renewal_ops = &nfs40_state_renewal_ops,
6461 };
6462
6463 #if defined(CONFIG_NFS_V4_1)
6464 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6465         .minor_version = 1,
6466         .call_sync = _nfs4_call_sync_session,
6467         .match_stateid = nfs41_match_stateid,
6468         .find_root_sec = nfs41_find_root_sec,
6469         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6470         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6471         .state_renewal_ops = &nfs41_state_renewal_ops,
6472 };
6473 #endif
6474
6475 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6476         [0] = &nfs_v4_0_minor_ops,
6477 #if defined(CONFIG_NFS_V4_1)
6478         [1] = &nfs_v4_1_minor_ops,
6479 #endif
6480 };
6481
6482 static const struct inode_operations nfs4_file_inode_operations = {
6483         .permission     = nfs_permission,
6484         .getattr        = nfs_getattr,
6485         .setattr        = nfs_setattr,
6486         .getxattr       = generic_getxattr,
6487         .setxattr       = generic_setxattr,
6488         .listxattr      = generic_listxattr,
6489         .removexattr    = generic_removexattr,
6490 };
6491
6492 const struct nfs_rpc_ops nfs_v4_clientops = {
6493         .version        = 4,                    /* protocol version */
6494         .dentry_ops     = &nfs4_dentry_operations,
6495         .dir_inode_ops  = &nfs4_dir_inode_operations,
6496         .file_inode_ops = &nfs4_file_inode_operations,
6497         .file_ops       = &nfs4_file_operations,
6498         .getroot        = nfs4_proc_get_root,
6499         .getattr        = nfs4_proc_getattr,
6500         .setattr        = nfs4_proc_setattr,
6501         .lookup         = nfs4_proc_lookup,
6502         .access         = nfs4_proc_access,
6503         .readlink       = nfs4_proc_readlink,
6504         .create         = nfs4_proc_create,
6505         .remove         = nfs4_proc_remove,
6506         .unlink_setup   = nfs4_proc_unlink_setup,
6507         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6508         .unlink_done    = nfs4_proc_unlink_done,
6509         .rename         = nfs4_proc_rename,
6510         .rename_setup   = nfs4_proc_rename_setup,
6511         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6512         .rename_done    = nfs4_proc_rename_done,
6513         .link           = nfs4_proc_link,
6514         .symlink        = nfs4_proc_symlink,
6515         .mkdir          = nfs4_proc_mkdir,
6516         .rmdir          = nfs4_proc_remove,
6517         .readdir        = nfs4_proc_readdir,
6518         .mknod          = nfs4_proc_mknod,
6519         .statfs         = nfs4_proc_statfs,
6520         .fsinfo         = nfs4_proc_fsinfo,
6521         .pathconf       = nfs4_proc_pathconf,
6522         .set_capabilities = nfs4_server_capabilities,
6523         .decode_dirent  = nfs4_decode_dirent,
6524         .read_setup     = nfs4_proc_read_setup,
6525         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6526         .read_done      = nfs4_read_done,
6527         .write_setup    = nfs4_proc_write_setup,
6528         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6529         .write_done     = nfs4_write_done,
6530         .commit_setup   = nfs4_proc_commit_setup,
6531         .commit_done    = nfs4_commit_done,
6532         .lock           = nfs4_proc_lock,
6533         .clear_acl_cache = nfs4_zap_acl_attr,
6534         .close_context  = nfs4_close_context,
6535         .open_context   = nfs4_atomic_open,
6536         .init_client    = nfs4_init_client,
6537         .secinfo        = nfs4_proc_secinfo,
6538 };
6539
6540 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6541         .prefix = XATTR_NAME_NFSV4_ACL,
6542         .list   = nfs4_xattr_list_nfs4_acl,
6543         .get    = nfs4_xattr_get_nfs4_acl,
6544         .set    = nfs4_xattr_set_nfs4_acl,
6545 };
6546
6547 const struct xattr_handler *nfs4_xattr_handlers[] = {
6548         &nfs4_xattr_nfs4_acl_handler,
6549         NULL
6550 };
6551
6552 module_param(max_session_slots, ushort, 0644);
6553 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6554                 "requests the client will negotiate");
6555
6556 /*
6557  * Local variables:
6558  *  c-basic-offset: 8
6559  * End:
6560  */