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