fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/nfs_idmap.h>
  55 #include <linux/sunrpc/bc_xprt.h>
  56 #include <linux/xattr.h>
  57 #include <linux/utsname.h>
  58 #include <linux/freezer.h>
  59
  60 #include "nfs4_fs.h"
  61 #include "delegation.h"
  62 #include "internal.h"
  63 #include "iostat.h"
  64 #include "callback.h"
  65 #include "pnfs.h"
  66 #include "netns.h"
  67
  68 #define NFSDBG_FACILITY         NFSDBG_PROC
  69
  70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  72
  73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  74
  75 struct nfs4_opendata;
  76 static int _nfs4_proc_open(struct nfs4_opendata *data);
  77 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  78 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  79 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  80 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  81 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *);
  82 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  83 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  84                             struct nfs_fattr *fattr, struct iattr *sattr,
  85                             struct nfs4_state *state);
  86 #ifdef CONFIG_NFS_V4_1
  87 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  88 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  89 #endif
  90 /* Prevent leaks of NFSv4 errors into userland */
  91 static int nfs4_map_errors(int err)
  92 {
  93         if (err >= -1000)
  94                 return err;
  95         switch (err) {
  96         case -NFS4ERR_RESOURCE:
  97                 return -EREMOTEIO;
  98         case -NFS4ERR_WRONGSEC:
  99                 return -EPERM;
 100         case -NFS4ERR_BADOWNER:
 101         case -NFS4ERR_BADNAME:
 102                 return -EINVAL;
 103         case -NFS4ERR_SHARE_DENIED:
 104                 return -EACCES;
 105         case -NFS4ERR_MINOR_VERS_MISMATCH:
 106                 return -EPROTONOSUPPORT;
 107         case -NFS4ERR_ACCESS:
 108                 return -EACCES;
 109         default:
 110                 dprintk("%s could not handle NFSv4 error %d\n",
 111                                 __func__, -err);
 112                 break;
 113         }
 114         return -EIO;
 115 }
 116
 117 /*
 118  * This is our standard bitmap for GETATTR requests.
 119  */
 120 const u32 nfs4_fattr_bitmap[3] = {
 121         FATTR4_WORD0_TYPE
 122         | FATTR4_WORD0_CHANGE
 123         | FATTR4_WORD0_SIZE
 124         | FATTR4_WORD0_FSID
 125         | FATTR4_WORD0_FILEID,
 126         FATTR4_WORD1_MODE
 127         | FATTR4_WORD1_NUMLINKS
 128         | FATTR4_WORD1_OWNER
 129         | FATTR4_WORD1_OWNER_GROUP
 130         | FATTR4_WORD1_RAWDEV
 131         | FATTR4_WORD1_SPACE_USED
 132         | FATTR4_WORD1_TIME_ACCESS
 133         | FATTR4_WORD1_TIME_METADATA
 134         | FATTR4_WORD1_TIME_MODIFY
 135 };
 136
 137 static const u32 nfs4_pnfs_open_bitmap[3] = {
 138         FATTR4_WORD0_TYPE
 139         | FATTR4_WORD0_CHANGE
 140         | FATTR4_WORD0_SIZE
 141         | FATTR4_WORD0_FSID
 142         | FATTR4_WORD0_FILEID,
 143         FATTR4_WORD1_MODE
 144         | FATTR4_WORD1_NUMLINKS
 145         | FATTR4_WORD1_OWNER
 146         | FATTR4_WORD1_OWNER_GROUP
 147         | FATTR4_WORD1_RAWDEV
 148         | FATTR4_WORD1_SPACE_USED
 149         | FATTR4_WORD1_TIME_ACCESS
 150         | FATTR4_WORD1_TIME_METADATA
 151         | FATTR4_WORD1_TIME_MODIFY,
 152         FATTR4_WORD2_MDSTHRESHOLD
 153 };
 154
 155 const u32 nfs4_statfs_bitmap[2] = {
 156         FATTR4_WORD0_FILES_AVAIL
 157         | FATTR4_WORD0_FILES_FREE
 158         | FATTR4_WORD0_FILES_TOTAL,
 159         FATTR4_WORD1_SPACE_AVAIL
 160         | FATTR4_WORD1_SPACE_FREE
 161         | FATTR4_WORD1_SPACE_TOTAL
 162 };
 163
 164 const u32 nfs4_pathconf_bitmap[2] = {
 165         FATTR4_WORD0_MAXLINK
 166         | FATTR4_WORD0_MAXNAME,
 167         0
 168 };
 169
 170 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 171                         | FATTR4_WORD0_MAXREAD
 172                         | FATTR4_WORD0_MAXWRITE
 173                         | FATTR4_WORD0_LEASE_TIME,
 174                         FATTR4_WORD1_TIME_DELTA
 175                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 176                         FATTR4_WORD2_LAYOUT_BLKSIZE
 177 };
 178
 179 const u32 nfs4_fs_locations_bitmap[2] = {
 180         FATTR4_WORD0_TYPE
 181         | FATTR4_WORD0_CHANGE
 182         | FATTR4_WORD0_SIZE
 183         | FATTR4_WORD0_FSID
 184         | FATTR4_WORD0_FILEID
 185         | FATTR4_WORD0_FS_LOCATIONS,
 186         FATTR4_WORD1_MODE
 187         | FATTR4_WORD1_NUMLINKS
 188         | FATTR4_WORD1_OWNER
 189         | FATTR4_WORD1_OWNER_GROUP
 190         | FATTR4_WORD1_RAWDEV
 191         | FATTR4_WORD1_SPACE_USED
 192         | FATTR4_WORD1_TIME_ACCESS
 193         | FATTR4_WORD1_TIME_METADATA
 194         | FATTR4_WORD1_TIME_MODIFY
 195         | FATTR4_WORD1_MOUNTED_ON_FILEID
 196 };
 197
 198 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 199                 struct nfs4_readdir_arg *readdir)
 200 {
 201         __be32 *start, *p;
 202
 203         BUG_ON(readdir->count < 80);
 204         if (cookie > 2) {
 205                 readdir->cookie = cookie;
 206                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 207                 return;
 208         }
 209
 210         readdir->cookie = 0;
 211         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 212         if (cookie == 2)
 213                 return;
 214
 215         /*
 216          * NFSv4 servers do not return entries for '.' and '..'
 217          * Therefore, we fake these entries here.  We let '.'
 218          * have cookie 0 and '..' have cookie 1.  Note that
 219          * when talking to the server, we always send cookie 0
 220          * instead of 1 or 2.
 221          */
 222         start = p = kmap_atomic(*readdir->pages);
 223
 224         if (cookie == 0) {
 225                 *p++ = xdr_one;                                  /* next */
 226                 *p++ = xdr_zero;                   /* cookie, first word */
 227                 *p++ = xdr_one;                   /* cookie, second word */
 228                 *p++ = xdr_one;                             /* entry len */
 229                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 230                 p++;
 231                 *p++ = xdr_one;                         /* bitmap length */
 232                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 233                 *p++ = htonl(8);              /* attribute buffer length */
 234                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 235         }
 236
 237         *p++ = xdr_one;                                  /* next */
 238         *p++ = xdr_zero;                   /* cookie, first word */
 239         *p++ = xdr_two;                   /* cookie, second word */
 240         *p++ = xdr_two;                             /* entry len */
 241         memcpy(p, "..\0\0", 4);                         /* entry */
 242         p++;
 243         *p++ = xdr_one;                         /* bitmap length */
 244         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 245         *p++ = htonl(8);              /* attribute buffer length */
 246         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 247
 248         readdir->pgbase = (char *)p - (char *)start;
 249         readdir->count -= readdir->pgbase;
 250         kunmap_atomic(start);
 251 }
 252
 253 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 254 {
 255         int res;
 256
 257         might_sleep();
 258
 259         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 260                         nfs_wait_bit_killable, TASK_KILLABLE);
 261         if (res)
 262                 return res;
 263
 264         if (clp->cl_cons_state < 0)
 265                 return clp->cl_cons_state;
 266         return 0;
 267 }
 268
 269 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 270 {
 271         int res = 0;
 272
 273         might_sleep();
 274
 275         if (*timeout <= 0)
 276                 *timeout = NFS4_POLL_RETRY_MIN;
 277         if (*timeout > NFS4_POLL_RETRY_MAX)
 278                 *timeout = NFS4_POLL_RETRY_MAX;
 279         freezable_schedule_timeout_killable(*timeout);
 280         if (fatal_signal_pending(current))
 281                 res = -ERESTARTSYS;
 282         *timeout <<= 1;
 283         return res;
 284 }
 285
 286 /* This is the error handling routine for processes that are allowed
 287  * to sleep.
 288  */
 289 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 290 {
 291         struct nfs_client *clp = server->nfs_client;
 292         struct nfs4_state *state = exception->state;
 293         struct inode *inode = exception->inode;
 294         int ret = errorcode;
 295
 296         exception->retry = 0;
 297         switch(errorcode) {
 298                 case 0:
 299                         return 0;
 300                 case -NFS4ERR_OPENMODE:
 301                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 302                                 nfs4_inode_return_delegation(inode);
 303                                 exception->retry = 1;
 304                                 return 0;
 305                         }
 306                         if (state == NULL)
 307                                 break;
 308                         nfs4_schedule_stateid_recovery(server, state);
 309                         goto wait_on_recovery;
 310                 case -NFS4ERR_DELEG_REVOKED:
 311                 case -NFS4ERR_ADMIN_REVOKED:
 312                 case -NFS4ERR_BAD_STATEID:
 313                         if (state == NULL)
 314                                 break;
 315                         nfs_remove_bad_delegation(state->inode);
 316                         nfs4_schedule_stateid_recovery(server, state);
 317                         goto wait_on_recovery;
 318                 case -NFS4ERR_EXPIRED:
 319                         if (state != NULL)
 320                                 nfs4_schedule_stateid_recovery(server, state);
 321                 case -NFS4ERR_STALE_STATEID:
 322                 case -NFS4ERR_STALE_CLIENTID:
 323                         nfs4_schedule_lease_recovery(clp);
 324                         goto wait_on_recovery;
 325 #if defined(CONFIG_NFS_V4_1)
 326                 case -NFS4ERR_BADSESSION:
 327                 case -NFS4ERR_BADSLOT:
 328                 case -NFS4ERR_BAD_HIGH_SLOT:
 329                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 330                 case -NFS4ERR_DEADSESSION:
 331                 case -NFS4ERR_SEQ_FALSE_RETRY:
 332                 case -NFS4ERR_SEQ_MISORDERED:
 333                         dprintk("%s ERROR: %d Reset session\n", __func__,
 334                                 errorcode);
 335                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 336                         exception->retry = 1;
 337                         break;
 338 #endif /* defined(CONFIG_NFS_V4_1) */
 339                 case -NFS4ERR_FILE_OPEN:
 340                         if (exception->timeout > HZ) {
 341                                 /* We have retried a decent amount, time to
 342                                  * fail
 343                                  */
 344                                 ret = -EBUSY;
 345                                 break;
 346                         }
 347                 case -NFS4ERR_GRACE:
 348                 case -NFS4ERR_DELAY:
 349                 case -EKEYEXPIRED:
 350                         ret = nfs4_delay(server->client, &exception->timeout);
 351                         if (ret != 0)
 352                                 break;
 353                 case -NFS4ERR_RETRY_UNCACHED_REP:
 354                 case -NFS4ERR_OLD_STATEID:
 355                         exception->retry = 1;
 356                         break;
 357                 case -NFS4ERR_BADOWNER:
 358                         /* The following works around a Linux server bug! */
 359                 case -NFS4ERR_BADNAME:
 360                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 361                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 362                                 exception->retry = 1;
 363                                 printk(KERN_WARNING "NFS: v4 server %s "
 364                                                 "does not accept raw "
 365                                                 "uid/gids. "
 366                                                 "Reenabling the idmapper.\n",
 367                                                 server->nfs_client->cl_hostname);
 368                         }
 369         }
 370         /* We failed to handle the error */
 371         return nfs4_map_errors(ret);
 372 wait_on_recovery:
 373         ret = nfs4_wait_clnt_recover(clp);
 374         if (ret == 0)
 375                 exception->retry = 1;
 376         return ret;
 377 }
 378
 379
 380 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 381 {
 382         spin_lock(&clp->cl_lock);
 383         if (time_before(clp->cl_last_renewal,timestamp))
 384                 clp->cl_last_renewal = timestamp;
 385         spin_unlock(&clp->cl_lock);
 386 }
 387
 388 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 389 {
 390         do_renew_lease(server->nfs_client, timestamp);
 391 }
 392
 393 #if defined(CONFIG_NFS_V4_1)
 394
 395 /*
 396  * nfs4_free_slot - free a slot and efficiently update slot table.
 397  *
 398  * freeing a slot is trivially done by clearing its respective bit
 399  * in the bitmap.
 400  * If the freed slotid equals highest_used_slotid we want to update it
 401  * so that the server would be able to size down the slot table if needed,
 402  * otherwise we know that the highest_used_slotid is still in use.
 403  * When updating highest_used_slotid there may be "holes" in the bitmap
 404  * so we need to scan down from highest_used_slotid to 0 looking for the now
 405  * highest slotid in use.
 406  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 407  *
 408  * Must be called while holding tbl->slot_tbl_lock
 409  */
 410 static void
 411 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 412 {
 413         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 414         /* clear used bit in bitmap */
 415         __clear_bit(slotid, tbl->used_slots);
 416
 417         /* update highest_used_slotid when it is freed */
 418         if (slotid == tbl->highest_used_slotid) {
 419                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 420                 if (slotid < tbl->max_slots)
 421                         tbl->highest_used_slotid = slotid;
 422                 else
 423                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 424         }
 425         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 426                 slotid, tbl->highest_used_slotid);
 427 }
 428
 429 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 430 {
 431         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 432         return true;
 433 }
 434
 435 /*
 436  * Signal state manager thread if session fore channel is drained
 437  */
 438 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 439 {
 440         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 441                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 442                                 nfs4_set_task_privileged, NULL);
 443                 return;
 444         }
 445
 446         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 447                 return;
 448
 449         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 450         complete(&ses->fc_slot_table.complete);
 451 }
 452
 453 /*
 454  * Signal state manager thread if session back channel is drained
 455  */
 456 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 457 {
 458         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 459             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 460                 return;
 461         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 462         complete(&ses->bc_slot_table.complete);
 463 }
 464
 465 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 466 {
 467         struct nfs4_slot_table *tbl;
 468
 469         tbl = &res->sr_session->fc_slot_table;
 470         if (!res->sr_slot) {
 471                 /* just wake up the next guy waiting since
 472                  * we may have not consumed a slot after all */
 473                 dprintk("%s: No slot\n", __func__);
 474                 return;
 475         }
 476
 477         spin_lock(&tbl->slot_tbl_lock);
 478         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 479         nfs4_check_drain_fc_complete(res->sr_session);
 480         spin_unlock(&tbl->slot_tbl_lock);
 481         res->sr_slot = NULL;
 482 }
 483
 484 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 485 {
 486         unsigned long timestamp;
 487         struct nfs_client *clp;
 488
 489         /*
 490          * sr_status remains 1 if an RPC level error occurred. The server
 491          * may or may not have processed the sequence operation..
 492          * Proceed as if the server received and processed the sequence
 493          * operation.
 494          */
 495         if (res->sr_status == 1)
 496                 res->sr_status = NFS_OK;
 497
 498         /* don't increment the sequence number if the task wasn't sent */
 499         if (!RPC_WAS_SENT(task))
 500                 goto out;
 501
 502         /* Check the SEQUENCE operation status */
 503         switch (res->sr_status) {
 504         case 0:
 505                 /* Update the slot's sequence and clientid lease timer */
 506                 ++res->sr_slot->seq_nr;
 507                 timestamp = res->sr_renewal_time;
 508                 clp = res->sr_session->clp;
 509                 do_renew_lease(clp, timestamp);
 510                 /* Check sequence flags */
 511                 if (res->sr_status_flags != 0)
 512                         nfs4_schedule_lease_recovery(clp);
 513                 break;
 514         case -NFS4ERR_DELAY:
 515                 /* The server detected a resend of the RPC call and
 516                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 517                  * of RFC5661.
 518                  */
 519                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 520                         __func__,
 521                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 522                         res->sr_slot->seq_nr);
 523                 goto out_retry;
 524         default:
 525                 /* Just update the slot sequence no. */
 526                 ++res->sr_slot->seq_nr;
 527         }
 528 out:
 529         /* The session may be reset by one of the error handlers. */
 530         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 531         nfs41_sequence_free_slot(res);
 532         return 1;
 533 out_retry:
 534         if (!rpc_restart_call(task))
 535                 goto out;
 536         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 537         return 0;
 538 }
 539
 540 static int nfs4_sequence_done(struct rpc_task *task,
 541                                struct nfs4_sequence_res *res)
 542 {
 543         if (res->sr_session == NULL)
 544                 return 1;
 545         return nfs41_sequence_done(task, res);
 546 }
 547
 548 /*
 549  * nfs4_find_slot - efficiently look for a free slot
 550  *
 551  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 552  * If found, we mark the slot as used, update the highest_used_slotid,
 553  * and respectively set up the sequence operation args.
 554  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 555  *
 556  * Note: must be called with under the slot_tbl_lock.
 557  */
 558 static u32
 559 nfs4_find_slot(struct nfs4_slot_table *tbl)
 560 {
 561         u32 slotid;
 562         u32 ret_id = NFS4_NO_SLOT;
 563
 564         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 565                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 566                 tbl->max_slots);
 567         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 568         if (slotid >= tbl->max_slots)
 569                 goto out;
 570         __set_bit(slotid, tbl->used_slots);
 571         if (slotid > tbl->highest_used_slotid ||
 572                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 573                 tbl->highest_used_slotid = slotid;
 574         ret_id = slotid;
 575 out:
 576         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 577                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 578         return ret_id;
 579 }
 580
 581 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 582                 struct nfs4_sequence_res *res, int cache_reply)
 583 {
 584         args->sa_session = NULL;
 585         args->sa_cache_this = 0;
 586         if (cache_reply)
 587                 args->sa_cache_this = 1;
 588         res->sr_session = NULL;
 589         res->sr_slot = NULL;
 590 }
 591
 592 int nfs41_setup_sequence(struct nfs4_session *session,
 593                                 struct nfs4_sequence_args *args,
 594                                 struct nfs4_sequence_res *res,
 595                                 struct rpc_task *task)
 596 {
 597         struct nfs4_slot *slot;
 598         struct nfs4_slot_table *tbl;
 599         u32 slotid;
 600
 601         dprintk("--> %s\n", __func__);
 602         /* slot already allocated? */
 603         if (res->sr_slot != NULL)
 604                 return 0;
 605
 606         tbl = &session->fc_slot_table;
 607
 608         spin_lock(&tbl->slot_tbl_lock);
 609         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 610             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 611                 /* The state manager will wait until the slot table is empty */
 612                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 613                 spin_unlock(&tbl->slot_tbl_lock);
 614                 dprintk("%s session is draining\n", __func__);
 615                 return -EAGAIN;
 616         }
 617
 618         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 619             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 620                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 621                 spin_unlock(&tbl->slot_tbl_lock);
 622                 dprintk("%s enforce FIFO order\n", __func__);
 623                 return -EAGAIN;
 624         }
 625
 626         slotid = nfs4_find_slot(tbl);
 627         if (slotid == NFS4_NO_SLOT) {
 628                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 629                 spin_unlock(&tbl->slot_tbl_lock);
 630                 dprintk("<-- %s: no free slots\n", __func__);
 631                 return -EAGAIN;
 632         }
 633         spin_unlock(&tbl->slot_tbl_lock);
 634
 635         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 636         slot = tbl->slots + slotid;
 637         args->sa_session = session;
 638         args->sa_slotid = slotid;
 639
 640         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 641
 642         res->sr_session = session;
 643         res->sr_slot = slot;
 644         res->sr_renewal_time = jiffies;
 645         res->sr_status_flags = 0;
 646         /*
 647          * sr_status is only set in decode_sequence, and so will remain
 648          * set to 1 if an rpc level failure occurs.
 649          */
 650         res->sr_status = 1;
 651         return 0;
 652 }
 653 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 654
 655 int nfs4_setup_sequence(const struct nfs_server *server,
 656                         struct nfs4_sequence_args *args,
 657                         struct nfs4_sequence_res *res,
 658                         struct rpc_task *task)
 659 {
 660         struct nfs4_session *session = nfs4_get_session(server);
 661         int ret = 0;
 662
 663         if (session == NULL)
 664                 goto out;
 665
 666         dprintk("--> %s clp %p session %p sr_slot %td\n",
 667                 __func__, session->clp, session, res->sr_slot ?
 668                         res->sr_slot - session->fc_slot_table.slots : -1);
 669
 670         ret = nfs41_setup_sequence(session, args, res, task);
 671 out:
 672         dprintk("<-- %s status=%d\n", __func__, ret);
 673         return ret;
 674 }
 675
 676 struct nfs41_call_sync_data {
 677         const struct nfs_server *seq_server;
 678         struct nfs4_sequence_args *seq_args;
 679         struct nfs4_sequence_res *seq_res;
 680 };
 681
 682 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 683 {
 684         struct nfs41_call_sync_data *data = calldata;
 685
 686         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 687
 688         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 689                                 data->seq_res, task))
 690                 return;
 691         rpc_call_start(task);
 692 }
 693
 694 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 695 {
 696         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 697         nfs41_call_sync_prepare(task, calldata);
 698 }
 699
 700 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 701 {
 702         struct nfs41_call_sync_data *data = calldata;
 703
 704         nfs41_sequence_done(task, data->seq_res);
 705 }
 706
 707 static const struct rpc_call_ops nfs41_call_sync_ops = {
 708         .rpc_call_prepare = nfs41_call_sync_prepare,
 709         .rpc_call_done = nfs41_call_sync_done,
 710 };
 711
 712 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 713         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 714         .rpc_call_done = nfs41_call_sync_done,
 715 };
 716
 717 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 718                                    struct nfs_server *server,
 719                                    struct rpc_message *msg,
 720                                    struct nfs4_sequence_args *args,
 721                                    struct nfs4_sequence_res *res,
 722                                    int privileged)
 723 {
 724         int ret;
 725         struct rpc_task *task;
 726         struct nfs41_call_sync_data data = {
 727                 .seq_server = server,
 728                 .seq_args = args,
 729                 .seq_res = res,
 730         };
 731         struct rpc_task_setup task_setup = {
 732                 .rpc_client = clnt,
 733                 .rpc_message = msg,
 734                 .callback_ops = &nfs41_call_sync_ops,
 735                 .callback_data = &data
 736         };
 737
 738         if (privileged)
 739                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 740         task = rpc_run_task(&task_setup);
 741         if (IS_ERR(task))
 742                 ret = PTR_ERR(task);
 743         else {
 744                 ret = task->tk_status;
 745                 rpc_put_task(task);
 746         }
 747         return ret;
 748 }
 749
 750 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 751                             struct nfs_server *server,
 752                             struct rpc_message *msg,
 753                             struct nfs4_sequence_args *args,
 754                             struct nfs4_sequence_res *res,
 755                             int cache_reply)
 756 {
 757         nfs41_init_sequence(args, res, cache_reply);
 758         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 759 }
 760
 761 #else
 762 static inline
 763 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 764                 struct nfs4_sequence_res *res, int cache_reply)
 765 {
 766 }
 767
 768 static int nfs4_sequence_done(struct rpc_task *task,
 769                                struct nfs4_sequence_res *res)
 770 {
 771         return 1;
 772 }
 773 #endif /* CONFIG_NFS_V4_1 */
 774
 775 int _nfs4_call_sync(struct rpc_clnt *clnt,
 776                     struct nfs_server *server,
 777                     struct rpc_message *msg,
 778                     struct nfs4_sequence_args *args,
 779                     struct nfs4_sequence_res *res,
 780                     int cache_reply)
 781 {
 782         nfs41_init_sequence(args, res, cache_reply);
 783         return rpc_call_sync(clnt, msg, 0);
 784 }
 785
 786 static inline
 787 int nfs4_call_sync(struct rpc_clnt *clnt,
 788                    struct nfs_server *server,
 789                    struct rpc_message *msg,
 790                    struct nfs4_sequence_args *args,
 791                    struct nfs4_sequence_res *res,
 792                    int cache_reply)
 793 {
 794         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 795                                                 args, res, cache_reply);
 796 }
 797
 798 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 799 {
 800         struct nfs_inode *nfsi = NFS_I(dir);
 801
 802         spin_lock(&dir->i_lock);
 803         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 804         if (!cinfo->atomic || cinfo->before != dir->i_version)
 805                 nfs_force_lookup_revalidate(dir);
 806         dir->i_version = cinfo->after;
 807         spin_unlock(&dir->i_lock);
 808 }
 809
 810 struct nfs4_opendata {
 811         struct kref kref;
 812         struct nfs_openargs o_arg;
 813         struct nfs_openres o_res;
 814         struct nfs_open_confirmargs c_arg;
 815         struct nfs_open_confirmres c_res;
 816         struct nfs4_string owner_name;
 817         struct nfs4_string group_name;
 818         struct nfs_fattr f_attr;
 819         struct dentry *dir;
 820         struct dentry *dentry;
 821         struct nfs4_state_owner *owner;
 822         struct nfs4_state *state;
 823         struct iattr attrs;
 824         unsigned long timestamp;
 825         unsigned int rpc_done : 1;
 826         int rpc_status;
 827         int cancelled;
 828 };
 829
 830
 831 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 832 {
 833         p->o_res.f_attr = &p->f_attr;
 834         p->o_res.seqid = p->o_arg.seqid;
 835         p->c_res.seqid = p->c_arg.seqid;
 836         p->o_res.server = p->o_arg.server;
 837         nfs_fattr_init(&p->f_attr);
 838         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 839 }
 840
 841 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 842                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 843                 const struct iattr *attrs,
 844                 gfp_t gfp_mask)
 845 {
 846         struct dentry *parent = dget_parent(dentry);
 847         struct inode *dir = parent->d_inode;
 848         struct nfs_server *server = NFS_SERVER(dir);
 849         struct nfs4_opendata *p;
 850
 851         p = kzalloc(sizeof(*p), gfp_mask);
 852         if (p == NULL)
 853                 goto err;
 854         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 855         if (p->o_arg.seqid == NULL)
 856                 goto err_free;
 857         nfs_sb_active(dentry->d_sb);
 858         p->dentry = dget(dentry);
 859         p->dir = parent;
 860         p->owner = sp;
 861         atomic_inc(&sp->so_count);
 862         p->o_arg.fh = NFS_FH(dir);
 863         p->o_arg.open_flags = flags;
 864         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 865         /* ask server to check for all possible rights as results are cached */
 866         p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
 867                           NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
 868         p->o_arg.clientid = server->nfs_client->cl_clientid;
 869         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
 870         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
 871         p->o_arg.name = &dentry->d_name;
 872         p->o_arg.server = server;
 873         p->o_arg.bitmask = server->attr_bitmask;
 874         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
 875         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 876         if (attrs != NULL && attrs->ia_valid != 0) {
 877                 __be32 verf[2];
 878
 879                 p->o_arg.u.attrs = &p->attrs;
 880                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 881
 882                 verf[0] = jiffies;
 883                 verf[1] = current->pid;
 884                 memcpy(p->o_arg.u.verifier.data, verf,
 885                                 sizeof(p->o_arg.u.verifier.data));
 886         }
 887         p->c_arg.fh = &p->o_res.fh;
 888         p->c_arg.stateid = &p->o_res.stateid;
 889         p->c_arg.seqid = p->o_arg.seqid;
 890         nfs4_init_opendata_res(p);
 891         kref_init(&p->kref);
 892         return p;
 893 err_free:
 894         kfree(p);
 895 err:
 896         dput(parent);
 897         return NULL;
 898 }
 899
 900 static void nfs4_opendata_free(struct kref *kref)
 901 {
 902         struct nfs4_opendata *p = container_of(kref,
 903                         struct nfs4_opendata, kref);
 904         struct super_block *sb = p->dentry->d_sb;
 905
 906         nfs_free_seqid(p->o_arg.seqid);
 907         if (p->state != NULL)
 908                 nfs4_put_open_state(p->state);
 909         nfs4_put_state_owner(p->owner);
 910         dput(p->dir);
 911         dput(p->dentry);
 912         nfs_sb_deactive(sb);
 913         nfs_fattr_free_names(&p->f_attr);
 914         kfree(p);
 915 }
 916
 917 static void nfs4_opendata_put(struct nfs4_opendata *p)
 918 {
 919         if (p != NULL)
 920                 kref_put(&p->kref, nfs4_opendata_free);
 921 }
 922
 923 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 924 {
 925         int ret;
 926
 927         ret = rpc_wait_for_completion_task(task);
 928         return ret;
 929 }
 930
 931 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 932 {
 933         int ret = 0;
 934
 935         if (open_mode & (O_EXCL|O_TRUNC))
 936                 goto out;
 937         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 938                 case FMODE_READ:
 939                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 940                                 && state->n_rdonly != 0;
 941                         break;
 942                 case FMODE_WRITE:
 943                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 944                                 && state->n_wronly != 0;
 945                         break;
 946                 case FMODE_READ|FMODE_WRITE:
 947                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 948                                 && state->n_rdwr != 0;
 949         }
 950 out:
 951         return ret;
 952 }
 953
 954 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 955 {
 956         if (delegation == NULL)
 957                 return 0;
 958         if ((delegation->type & fmode) != fmode)
 959                 return 0;
 960         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 961                 return 0;
 962         nfs_mark_delegation_referenced(delegation);
 963         return 1;
 964 }
 965
 966 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 967 {
 968         switch (fmode) {
 969                 case FMODE_WRITE:
 970                         state->n_wronly++;
 971                         break;
 972                 case FMODE_READ:
 973                         state->n_rdonly++;
 974                         break;
 975                 case FMODE_READ|FMODE_WRITE:
 976                         state->n_rdwr++;
 977         }
 978         nfs4_state_set_mode_locked(state, state->state | fmode);
 979 }
 980
 981 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 982 {
 983         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 984                 nfs4_stateid_copy(&state->stateid, stateid);
 985         nfs4_stateid_copy(&state->open_stateid, stateid);
 986         switch (fmode) {
 987                 case FMODE_READ:
 988                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 989                         break;
 990                 case FMODE_WRITE:
 991                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 992                         break;
 993                 case FMODE_READ|FMODE_WRITE:
 994                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 995         }
 996 }
 997
 998 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 999 {
1000         write_seqlock(&state->seqlock);
1001         nfs_set_open_stateid_locked(state, stateid, fmode);
1002         write_sequnlock(&state->seqlock);
1003 }
1004
1005 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1006 {
1007         /*
1008          * Protect the call to nfs4_state_set_mode_locked and
1009          * serialise the stateid update
1010          */
1011         write_seqlock(&state->seqlock);
1012         if (deleg_stateid != NULL) {
1013                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1014                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1015         }
1016         if (open_stateid != NULL)
1017                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1018         write_sequnlock(&state->seqlock);
1019         spin_lock(&state->owner->so_lock);
1020         update_open_stateflags(state, fmode);
1021         spin_unlock(&state->owner->so_lock);
1022 }
1023
1024 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1025 {
1026         struct nfs_inode *nfsi = NFS_I(state->inode);
1027         struct nfs_delegation *deleg_cur;
1028         int ret = 0;
1029
1030         fmode &= (FMODE_READ|FMODE_WRITE);
1031
1032         rcu_read_lock();
1033         deleg_cur = rcu_dereference(nfsi->delegation);
1034         if (deleg_cur == NULL)
1035                 goto no_delegation;
1036
1037         spin_lock(&deleg_cur->lock);
1038         if (nfsi->delegation != deleg_cur ||
1039             (deleg_cur->type & fmode) != fmode)
1040                 goto no_delegation_unlock;
1041
1042         if (delegation == NULL)
1043                 delegation = &deleg_cur->stateid;
1044         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1045                 goto no_delegation_unlock;
1046
1047         nfs_mark_delegation_referenced(deleg_cur);
1048         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1049         ret = 1;
1050 no_delegation_unlock:
1051         spin_unlock(&deleg_cur->lock);
1052 no_delegation:
1053         rcu_read_unlock();
1054
1055         if (!ret && open_stateid != NULL) {
1056                 __update_open_stateid(state, open_stateid, NULL, fmode);
1057                 ret = 1;
1058         }
1059
1060         return ret;
1061 }
1062
1063
1064 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1065 {
1066         struct nfs_delegation *delegation;
1067
1068         rcu_read_lock();
1069         delegation = rcu_dereference(NFS_I(inode)->delegation);
1070         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1071                 rcu_read_unlock();
1072                 return;
1073         }
1074         rcu_read_unlock();
1075         nfs4_inode_return_delegation(inode);
1076 }
1077
1078 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1079 {
1080         struct nfs4_state *state = opendata->state;
1081         struct nfs_inode *nfsi = NFS_I(state->inode);
1082         struct nfs_delegation *delegation;
1083         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1084         fmode_t fmode = opendata->o_arg.fmode;
1085         nfs4_stateid stateid;
1086         int ret = -EAGAIN;
1087
1088         for (;;) {
1089                 if (can_open_cached(state, fmode, open_mode)) {
1090                         spin_lock(&state->owner->so_lock);
1091                         if (can_open_cached(state, fmode, open_mode)) {
1092                                 update_open_stateflags(state, fmode);
1093                                 spin_unlock(&state->owner->so_lock);
1094                                 goto out_return_state;
1095                         }
1096                         spin_unlock(&state->owner->so_lock);
1097                 }
1098                 rcu_read_lock();
1099                 delegation = rcu_dereference(nfsi->delegation);
1100                 if (!can_open_delegated(delegation, fmode)) {
1101                         rcu_read_unlock();
1102                         break;
1103                 }
1104                 /* Save the delegation */
1105                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1106                 rcu_read_unlock();
1107                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1108                 if (ret != 0)
1109                         goto out;
1110                 ret = -EAGAIN;
1111
1112                 /* Try to update the stateid using the delegation */
1113                 if (update_open_stateid(state, NULL, &stateid, fmode))
1114                         goto out_return_state;
1115         }
1116 out:
1117         return ERR_PTR(ret);
1118 out_return_state:
1119         atomic_inc(&state->count);
1120         return state;
1121 }
1122
1123 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1124 {
1125         struct inode *inode;
1126         struct nfs4_state *state = NULL;
1127         struct nfs_delegation *delegation;
1128         int ret;
1129
1130         if (!data->rpc_done) {
1131                 state = nfs4_try_open_cached(data);
1132                 goto out;
1133         }
1134
1135         ret = -EAGAIN;
1136         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1137                 goto err;
1138         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1139         ret = PTR_ERR(inode);
1140         if (IS_ERR(inode))
1141                 goto err;
1142         ret = -ENOMEM;
1143         state = nfs4_get_open_state(inode, data->owner);
1144         if (state == NULL)
1145                 goto err_put_inode;
1146         if (data->o_res.delegation_type != 0) {
1147                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1148                 int delegation_flags = 0;
1149
1150                 rcu_read_lock();
1151                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1152                 if (delegation)
1153                         delegation_flags = delegation->flags;
1154                 rcu_read_unlock();
1155                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1156                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1157                                         "returning a delegation for "
1158                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1159                                         clp->cl_hostname);
1160                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1161                         nfs_inode_set_delegation(state->inode,
1162                                         data->owner->so_cred,
1163                                         &data->o_res);
1164                 else
1165                         nfs_inode_reclaim_delegation(state->inode,
1166                                         data->owner->so_cred,
1167                                         &data->o_res);
1168         }
1169
1170         update_open_stateid(state, &data->o_res.stateid, NULL,
1171                         data->o_arg.fmode);
1172         iput(inode);
1173 out:
1174         return state;
1175 err_put_inode:
1176         iput(inode);
1177 err:
1178         return ERR_PTR(ret);
1179 }
1180
1181 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1182 {
1183         struct nfs_inode *nfsi = NFS_I(state->inode);
1184         struct nfs_open_context *ctx;
1185
1186         spin_lock(&state->inode->i_lock);
1187         list_for_each_entry(ctx, &nfsi->open_files, list) {
1188                 if (ctx->state != state)
1189                         continue;
1190                 get_nfs_open_context(ctx);
1191                 spin_unlock(&state->inode->i_lock);
1192                 return ctx;
1193         }
1194         spin_unlock(&state->inode->i_lock);
1195         return ERR_PTR(-ENOENT);
1196 }
1197
1198 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1199 {
1200         struct nfs4_opendata *opendata;
1201
1202         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1203         if (opendata == NULL)
1204                 return ERR_PTR(-ENOMEM);
1205         opendata->state = state;
1206         atomic_inc(&state->count);
1207         return opendata;
1208 }
1209
1210 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1211 {
1212         struct nfs4_state *newstate;
1213         int ret;
1214
1215         opendata->o_arg.open_flags = 0;
1216         opendata->o_arg.fmode = fmode;
1217         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1218         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1219         nfs4_init_opendata_res(opendata);
1220         ret = _nfs4_recover_proc_open(opendata);
1221         if (ret != 0)
1222                 return ret;
1223         newstate = nfs4_opendata_to_nfs4_state(opendata);
1224         if (IS_ERR(newstate))
1225                 return PTR_ERR(newstate);
1226         nfs4_close_state(newstate, fmode);
1227         *res = newstate;
1228         return 0;
1229 }
1230
1231 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1232 {
1233         struct nfs4_state *newstate;
1234         int ret;
1235
1236         /* memory barrier prior to reading state->n_* */
1237         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1238         smp_rmb();
1239         if (state->n_rdwr != 0) {
1240                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1241                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1242                 if (ret != 0)
1243                         return ret;
1244                 if (newstate != state)
1245                         return -ESTALE;
1246         }
1247         if (state->n_wronly != 0) {
1248                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1249                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1250                 if (ret != 0)
1251                         return ret;
1252                 if (newstate != state)
1253                         return -ESTALE;
1254         }
1255         if (state->n_rdonly != 0) {
1256                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1257                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1258                 if (ret != 0)
1259                         return ret;
1260                 if (newstate != state)
1261                         return -ESTALE;
1262         }
1263         /*
1264          * We may have performed cached opens for all three recoveries.
1265          * Check if we need to update the current stateid.
1266          */
1267         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1268             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1269                 write_seqlock(&state->seqlock);
1270                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1271                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1272                 write_sequnlock(&state->seqlock);
1273         }
1274         return 0;
1275 }
1276
1277 /*
1278  * OPEN_RECLAIM:
1279  *      reclaim state on the server after a reboot.
1280  */
1281 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1282 {
1283         struct nfs_delegation *delegation;
1284         struct nfs4_opendata *opendata;
1285         fmode_t delegation_type = 0;
1286         int status;
1287
1288         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1289         if (IS_ERR(opendata))
1290                 return PTR_ERR(opendata);
1291         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1292         opendata->o_arg.fh = NFS_FH(state->inode);
1293         rcu_read_lock();
1294         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1295         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1296                 delegation_type = delegation->type;
1297         rcu_read_unlock();
1298         opendata->o_arg.u.delegation_type = delegation_type;
1299         status = nfs4_open_recover(opendata, state);
1300         nfs4_opendata_put(opendata);
1301         return status;
1302 }
1303
1304 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1305 {
1306         struct nfs_server *server = NFS_SERVER(state->inode);
1307         struct nfs4_exception exception = { };
1308         int err;
1309         do {
1310                 err = _nfs4_do_open_reclaim(ctx, state);
1311                 if (err != -NFS4ERR_DELAY)
1312                         break;
1313                 nfs4_handle_exception(server, err, &exception);
1314         } while (exception.retry);
1315         return err;
1316 }
1317
1318 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1319 {
1320         struct nfs_open_context *ctx;
1321         int ret;
1322
1323         ctx = nfs4_state_find_open_context(state);
1324         if (IS_ERR(ctx))
1325                 return PTR_ERR(ctx);
1326         ret = nfs4_do_open_reclaim(ctx, state);
1327         put_nfs_open_context(ctx);
1328         return ret;
1329 }
1330
1331 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1332 {
1333         struct nfs4_opendata *opendata;
1334         int ret;
1335
1336         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1337         if (IS_ERR(opendata))
1338                 return PTR_ERR(opendata);
1339         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1340         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1341         ret = nfs4_open_recover(opendata, state);
1342         nfs4_opendata_put(opendata);
1343         return ret;
1344 }
1345
1346 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1347 {
1348         struct nfs4_exception exception = { };
1349         struct nfs_server *server = NFS_SERVER(state->inode);
1350         int err;
1351         do {
1352                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1353                 switch (err) {
1354                         case 0:
1355                         case -ENOENT:
1356                         case -ESTALE:
1357                                 goto out;
1358                         case -NFS4ERR_BADSESSION:
1359                         case -NFS4ERR_BADSLOT:
1360                         case -NFS4ERR_BAD_HIGH_SLOT:
1361                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1362                         case -NFS4ERR_DEADSESSION:
1363                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1364                                 goto out;
1365                         case -NFS4ERR_STALE_CLIENTID:
1366                         case -NFS4ERR_STALE_STATEID:
1367                         case -NFS4ERR_EXPIRED:
1368                                 /* Don't recall a delegation if it was lost */
1369                                 nfs4_schedule_lease_recovery(server->nfs_client);
1370                                 goto out;
1371                         case -ERESTARTSYS:
1372                                 /*
1373                                  * The show must go on: exit, but mark the
1374                                  * stateid as needing recovery.
1375                                  */
1376                         case -NFS4ERR_DELEG_REVOKED:
1377                         case -NFS4ERR_ADMIN_REVOKED:
1378                         case -NFS4ERR_BAD_STATEID:
1379                                 nfs_inode_find_state_and_recover(state->inode,
1380                                                 stateid);
1381                                 nfs4_schedule_stateid_recovery(server, state);
1382                         case -EKEYEXPIRED:
1383                                 /*
1384                                  * User RPCSEC_GSS context has expired.
1385                                  * We cannot recover this stateid now, so
1386                                  * skip it and allow recovery thread to
1387                                  * proceed.
1388                                  */
1389                         case -ENOMEM:
1390                                 err = 0;
1391                                 goto out;
1392                 }
1393                 err = nfs4_handle_exception(server, err, &exception);
1394         } while (exception.retry);
1395 out:
1396         return err;
1397 }
1398
1399 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1400 {
1401         struct nfs4_opendata *data = calldata;
1402
1403         data->rpc_status = task->tk_status;
1404         if (data->rpc_status == 0) {
1405                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1406                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1407                 renew_lease(data->o_res.server, data->timestamp);
1408                 data->rpc_done = 1;
1409         }
1410 }
1411
1412 static void nfs4_open_confirm_release(void *calldata)
1413 {
1414         struct nfs4_opendata *data = calldata;
1415         struct nfs4_state *state = NULL;
1416
1417         /* If this request hasn't been cancelled, do nothing */
1418         if (data->cancelled == 0)
1419                 goto out_free;
1420         /* In case of error, no cleanup! */
1421         if (!data->rpc_done)
1422                 goto out_free;
1423         state = nfs4_opendata_to_nfs4_state(data);
1424         if (!IS_ERR(state))
1425                 nfs4_close_state(state, data->o_arg.fmode);
1426 out_free:
1427         nfs4_opendata_put(data);
1428 }
1429
1430 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1431         .rpc_call_done = nfs4_open_confirm_done,
1432         .rpc_release = nfs4_open_confirm_release,
1433 };
1434
1435 /*
1436  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1437  */
1438 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1439 {
1440         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1441         struct rpc_task *task;
1442         struct  rpc_message msg = {
1443                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1444                 .rpc_argp = &data->c_arg,
1445                 .rpc_resp = &data->c_res,
1446                 .rpc_cred = data->owner->so_cred,
1447         };
1448         struct rpc_task_setup task_setup_data = {
1449                 .rpc_client = server->client,
1450                 .rpc_message = &msg,
1451                 .callback_ops = &nfs4_open_confirm_ops,
1452                 .callback_data = data,
1453                 .workqueue = nfsiod_workqueue,
1454                 .flags = RPC_TASK_ASYNC,
1455         };
1456         int status;
1457
1458         kref_get(&data->kref);
1459         data->rpc_done = 0;
1460         data->rpc_status = 0;
1461         data->timestamp = jiffies;
1462         task = rpc_run_task(&task_setup_data);
1463         if (IS_ERR(task))
1464                 return PTR_ERR(task);
1465         status = nfs4_wait_for_completion_rpc_task(task);
1466         if (status != 0) {
1467                 data->cancelled = 1;
1468                 smp_wmb();
1469         } else
1470                 status = data->rpc_status;
1471         rpc_put_task(task);
1472         return status;
1473 }
1474
1475 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1476 {
1477         struct nfs4_opendata *data = calldata;
1478         struct nfs4_state_owner *sp = data->owner;
1479
1480         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1481                 return;
1482         /*
1483          * Check if we still need to send an OPEN call, or if we can use
1484          * a delegation instead.
1485          */
1486         if (data->state != NULL) {
1487                 struct nfs_delegation *delegation;
1488
1489                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1490                         goto out_no_action;
1491                 rcu_read_lock();
1492                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1493                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1494                     can_open_delegated(delegation, data->o_arg.fmode))
1495                         goto unlock_no_action;
1496                 rcu_read_unlock();
1497         }
1498         /* Update client id. */
1499         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1500         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1501                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1502                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1503         }
1504         data->timestamp = jiffies;
1505         if (nfs4_setup_sequence(data->o_arg.server,
1506                                 &data->o_arg.seq_args,
1507                                 &data->o_res.seq_res, task))
1508                 return;
1509         rpc_call_start(task);
1510         return;
1511 unlock_no_action:
1512         rcu_read_unlock();
1513 out_no_action:
1514         task->tk_action = NULL;
1515
1516 }
1517
1518 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1519 {
1520         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1521         nfs4_open_prepare(task, calldata);
1522 }
1523
1524 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1525 {
1526         struct nfs4_opendata *data = calldata;
1527
1528         data->rpc_status = task->tk_status;
1529
1530         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1531                 return;
1532
1533         if (task->tk_status == 0) {
1534                 switch (data->o_res.f_attr->mode & S_IFMT) {
1535                         case S_IFREG:
1536                                 break;
1537                         case S_IFLNK:
1538                                 data->rpc_status = -ELOOP;
1539                                 break;
1540                         case S_IFDIR:
1541                                 data->rpc_status = -EISDIR;
1542                                 break;
1543                         default:
1544                                 data->rpc_status = -ENOTDIR;
1545                 }
1546                 renew_lease(data->o_res.server, data->timestamp);
1547                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1548                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1549         }
1550         data->rpc_done = 1;
1551 }
1552
1553 static void nfs4_open_release(void *calldata)
1554 {
1555         struct nfs4_opendata *data = calldata;
1556         struct nfs4_state *state = NULL;
1557
1558         /* If this request hasn't been cancelled, do nothing */
1559         if (data->cancelled == 0)
1560                 goto out_free;
1561         /* In case of error, no cleanup! */
1562         if (data->rpc_status != 0 || !data->rpc_done)
1563                 goto out_free;
1564         /* In case we need an open_confirm, no cleanup! */
1565         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1566                 goto out_free;
1567         state = nfs4_opendata_to_nfs4_state(data);
1568         if (!IS_ERR(state))
1569                 nfs4_close_state(state, data->o_arg.fmode);
1570 out_free:
1571         nfs4_opendata_put(data);
1572 }
1573
1574 static const struct rpc_call_ops nfs4_open_ops = {
1575         .rpc_call_prepare = nfs4_open_prepare,
1576         .rpc_call_done = nfs4_open_done,
1577         .rpc_release = nfs4_open_release,
1578 };
1579
1580 static const struct rpc_call_ops nfs4_recover_open_ops = {
1581         .rpc_call_prepare = nfs4_recover_open_prepare,
1582         .rpc_call_done = nfs4_open_done,
1583         .rpc_release = nfs4_open_release,
1584 };
1585
1586 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1587 {
1588         struct inode *dir = data->dir->d_inode;
1589         struct nfs_server *server = NFS_SERVER(dir);
1590         struct nfs_openargs *o_arg = &data->o_arg;
1591         struct nfs_openres *o_res = &data->o_res;
1592         struct rpc_task *task;
1593         struct rpc_message msg = {
1594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1595                 .rpc_argp = o_arg,
1596                 .rpc_resp = o_res,
1597                 .rpc_cred = data->owner->so_cred,
1598         };
1599         struct rpc_task_setup task_setup_data = {
1600                 .rpc_client = server->client,
1601                 .rpc_message = &msg,
1602                 .callback_ops = &nfs4_open_ops,
1603                 .callback_data = data,
1604                 .workqueue = nfsiod_workqueue,
1605                 .flags = RPC_TASK_ASYNC,
1606         };
1607         int status;
1608
1609         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1610         kref_get(&data->kref);
1611         data->rpc_done = 0;
1612         data->rpc_status = 0;
1613         data->cancelled = 0;
1614         if (isrecover)
1615                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1616         task = rpc_run_task(&task_setup_data);
1617         if (IS_ERR(task))
1618                 return PTR_ERR(task);
1619         status = nfs4_wait_for_completion_rpc_task(task);
1620         if (status != 0) {
1621                 data->cancelled = 1;
1622                 smp_wmb();
1623         } else
1624                 status = data->rpc_status;
1625         rpc_put_task(task);
1626
1627         return status;
1628 }
1629
1630 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1631 {
1632         struct inode *dir = data->dir->d_inode;
1633         struct nfs_openres *o_res = &data->o_res;
1634         int status;
1635
1636         status = nfs4_run_open_task(data, 1);
1637         if (status != 0 || !data->rpc_done)
1638                 return status;
1639
1640         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1641
1642         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1643                 status = _nfs4_proc_open_confirm(data);
1644                 if (status != 0)
1645                         return status;
1646         }
1647
1648         return status;
1649 }
1650
1651 static int nfs4_opendata_access(struct rpc_cred *cred,
1652                                 struct nfs4_opendata *opendata,
1653                                 struct nfs4_state *state, fmode_t fmode)
1654 {
1655         struct nfs_access_entry cache;
1656         u32 mask;
1657
1658         /* access call failed or for some reason the server doesn't
1659          * support any access modes -- defer access call until later */
1660         if (opendata->o_res.access_supported == 0)
1661                 return 0;
1662
1663         mask = 0;
1664         if (fmode & FMODE_READ)
1665                 mask |= MAY_READ;
1666         if (fmode & FMODE_WRITE)
1667                 mask |= MAY_WRITE;
1668         if (fmode & FMODE_EXEC)
1669                 mask |= MAY_EXEC;
1670
1671         cache.cred = cred;
1672         cache.jiffies = jiffies;
1673         nfs_access_set_mask(&cache, opendata->o_res.access_result);
1674         nfs_access_add_cache(state->inode, &cache);
1675
1676         if ((mask & ~cache.mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0)
1677                 return 0;
1678
1679         /* even though OPEN succeeded, access is denied. Close the file */
1680         nfs4_close_state(state, fmode);
1681         return -NFS4ERR_ACCESS;
1682 }
1683
1684 /*
1685  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1686  */
1687 static int _nfs4_proc_open(struct nfs4_opendata *data)
1688 {
1689         struct inode *dir = data->dir->d_inode;
1690         struct nfs_server *server = NFS_SERVER(dir);
1691         struct nfs_openargs *o_arg = &data->o_arg;
1692         struct nfs_openres *o_res = &data->o_res;
1693         int status;
1694
1695         status = nfs4_run_open_task(data, 0);
1696         if (!data->rpc_done)
1697                 return status;
1698         if (status != 0) {
1699                 if (status == -NFS4ERR_BADNAME &&
1700                                 !(o_arg->open_flags & O_CREAT))
1701                         return -ENOENT;
1702                 return status;
1703         }
1704
1705         nfs_fattr_map_and_free_names(server, &data->f_attr);
1706
1707         if (o_arg->open_flags & O_CREAT)
1708                 update_changeattr(dir, &o_res->cinfo);
1709         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1710                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1711         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1712                 status = _nfs4_proc_open_confirm(data);
1713                 if (status != 0)
1714                         return status;
1715         }
1716         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1717                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1718         return 0;
1719 }
1720
1721 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1722 {
1723         unsigned int loop;
1724         int ret;
1725
1726         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1727                 ret = nfs4_wait_clnt_recover(clp);
1728                 if (ret != 0)
1729                         break;
1730                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1731                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1732                         break;
1733                 nfs4_schedule_state_manager(clp);
1734                 ret = -EIO;
1735         }
1736         return ret;
1737 }
1738
1739 static int nfs4_recover_expired_lease(struct nfs_server *server)
1740 {
1741         return nfs4_client_recover_expired_lease(server->nfs_client);
1742 }
1743
1744 /*
1745  * OPEN_EXPIRED:
1746  *      reclaim state on the server after a network partition.
1747  *      Assumes caller holds the appropriate lock
1748  */
1749 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1750 {
1751         struct nfs4_opendata *opendata;
1752         int ret;
1753
1754         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1755         if (IS_ERR(opendata))
1756                 return PTR_ERR(opendata);
1757         ret = nfs4_open_recover(opendata, state);
1758         if (ret == -ESTALE)
1759                 d_drop(ctx->dentry);
1760         nfs4_opendata_put(opendata);
1761         return ret;
1762 }
1763
1764 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1765 {
1766         struct nfs_server *server = NFS_SERVER(state->inode);
1767         struct nfs4_exception exception = { };
1768         int err;
1769
1770         do {
1771                 err = _nfs4_open_expired(ctx, state);
1772                 switch (err) {
1773                 default:
1774                         goto out;
1775                 case -NFS4ERR_GRACE:
1776                 case -NFS4ERR_DELAY:
1777                         nfs4_handle_exception(server, err, &exception);
1778                         err = 0;
1779                 }
1780         } while (exception.retry);
1781 out:
1782         return err;
1783 }
1784
1785 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1786 {
1787         struct nfs_open_context *ctx;
1788         int ret;
1789
1790         ctx = nfs4_state_find_open_context(state);
1791         if (IS_ERR(ctx))
1792                 return PTR_ERR(ctx);
1793         ret = nfs4_do_open_expired(ctx, state);
1794         put_nfs_open_context(ctx);
1795         return ret;
1796 }
1797
1798 #if defined(CONFIG_NFS_V4_1)
1799 static void nfs41_clear_delegation_stateid(struct nfs4_state *state)
1800 {
1801         struct nfs_server *server = NFS_SERVER(state->inode);
1802         nfs4_stateid *stateid = &state->stateid;
1803         int status;
1804
1805         /* If a state reset has been done, test_stateid is unneeded */
1806         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1807                 return;
1808
1809         status = nfs41_test_stateid(server, stateid);
1810         if (status != NFS_OK) {
1811                 /* Free the stateid unless the server explicitly
1812                  * informs us the stateid is unrecognized. */
1813                 if (status != -NFS4ERR_BAD_STATEID)
1814                         nfs41_free_stateid(server, stateid);
1815                 nfs_remove_bad_delegation(state->inode);
1816
1817                 write_seqlock(&state->seqlock);
1818                 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1819                 write_sequnlock(&state->seqlock);
1820                 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1821         }
1822 }
1823
1824 /**
1825  * nfs41_check_open_stateid - possibly free an open stateid
1826  *
1827  * @state: NFSv4 state for an inode
1828  *
1829  * Returns NFS_OK if recovery for this stateid is now finished.
1830  * Otherwise a negative NFS4ERR value is returned.
1831  */
1832 static int nfs41_check_open_stateid(struct nfs4_state *state)
1833 {
1834         struct nfs_server *server = NFS_SERVER(state->inode);
1835         nfs4_stateid *stateid = &state->open_stateid;
1836         int status;
1837
1838         /* If a state reset has been done, test_stateid is unneeded */
1839         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
1840             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
1841             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
1842                 return -NFS4ERR_BAD_STATEID;
1843
1844         status = nfs41_test_stateid(server, stateid);
1845         if (status != NFS_OK) {
1846                 /* Free the stateid unless the server explicitly
1847                  * informs us the stateid is unrecognized. */
1848                 if (status != -NFS4ERR_BAD_STATEID)
1849                         nfs41_free_stateid(server, stateid);
1850
1851                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1852                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1853                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1854         }
1855         return status;
1856 }
1857
1858 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1859 {
1860         int status;
1861
1862         nfs41_clear_delegation_stateid(state);
1863         status = nfs41_check_open_stateid(state);
1864         if (status != NFS_OK)
1865                 status = nfs4_open_expired(sp, state);
1866         return status;
1867 }
1868 #endif
1869
1870 /*
1871  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1872  * fields corresponding to attributes that were used to store the verifier.
1873  * Make sure we clobber those fields in the later setattr call
1874  */
1875 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1876 {
1877         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1878             !(sattr->ia_valid & ATTR_ATIME_SET))
1879                 sattr->ia_valid |= ATTR_ATIME;
1880
1881         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1882             !(sattr->ia_valid & ATTR_MTIME_SET))
1883                 sattr->ia_valid |= ATTR_MTIME;
1884 }
1885
1886 /*
1887  * Returns a referenced nfs4_state
1888  */
1889 static int _nfs4_do_open(struct inode *dir,
1890                         struct dentry *dentry,
1891                         fmode_t fmode,
1892                         int flags,
1893                         struct iattr *sattr,
1894                         struct rpc_cred *cred,
1895                         struct nfs4_state **res,
1896                         struct nfs4_threshold **ctx_th)
1897 {
1898         struct nfs4_state_owner  *sp;
1899         struct nfs4_state     *state = NULL;
1900         struct nfs_server       *server = NFS_SERVER(dir);
1901         struct nfs4_opendata *opendata;
1902         int status;
1903
1904         /* Protect against reboot recovery conflicts */
1905         status = -ENOMEM;
1906         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1907         if (sp == NULL) {
1908                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1909                 goto out_err;
1910         }
1911         status = nfs4_recover_expired_lease(server);
1912         if (status != 0)
1913                 goto err_put_state_owner;
1914         if (dentry->d_inode != NULL)
1915                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1916         status = -ENOMEM;
1917         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1918         if (opendata == NULL)
1919                 goto err_put_state_owner;
1920
1921         if (ctx_th && server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
1922                 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
1923                 if (!opendata->f_attr.mdsthreshold)
1924                         goto err_opendata_put;
1925                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
1926         }
1927         if (dentry->d_inode != NULL)
1928                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1929
1930         status = _nfs4_proc_open(opendata);
1931         if (status != 0)
1932                 goto err_opendata_put;
1933
1934         state = nfs4_opendata_to_nfs4_state(opendata);
1935         status = PTR_ERR(state);
1936         if (IS_ERR(state))
1937                 goto err_opendata_put;
1938         if (server->caps & NFS_CAP_POSIX_LOCK)
1939                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1940
1941         status = nfs4_opendata_access(cred, opendata, state, fmode);
1942         if (status != 0)
1943                 goto err_opendata_put;
1944
1945         if (opendata->o_arg.open_flags & O_EXCL) {
1946                 nfs4_exclusive_attrset(opendata, sattr);
1947
1948                 nfs_fattr_init(opendata->o_res.f_attr);
1949                 status = nfs4_do_setattr(state->inode, cred,
1950                                 opendata->o_res.f_attr, sattr,
1951                                 state);
1952                 if (status == 0)
1953                         nfs_setattr_update_inode(state->inode, sattr);
1954                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1955         }
1956
1957         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server))
1958                 *ctx_th = opendata->f_attr.mdsthreshold;
1959         else
1960                 kfree(opendata->f_attr.mdsthreshold);
1961         opendata->f_attr.mdsthreshold = NULL;
1962
1963         nfs4_opendata_put(opendata);
1964         nfs4_put_state_owner(sp);
1965         *res = state;
1966         return 0;
1967 err_opendata_put:
1968         kfree(opendata->f_attr.mdsthreshold);
1969         nfs4_opendata_put(opendata);
1970 err_put_state_owner:
1971         nfs4_put_state_owner(sp);
1972 out_err:
1973         *res = NULL;
1974         return status;
1975 }
1976
1977
1978 static struct nfs4_state *nfs4_do_open(struct inode *dir,
1979                                         struct dentry *dentry,
1980                                         fmode_t fmode,
1981                                         int flags,
1982                                         struct iattr *sattr,
1983                                         struct rpc_cred *cred,
1984                                         struct nfs4_threshold **ctx_th)
1985 {
1986         struct nfs4_exception exception = { };
1987         struct nfs4_state *res;
1988         int status;
1989
1990         fmode &= FMODE_READ|FMODE_WRITE|FMODE_EXEC;
1991         do {
1992                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred,
1993                                        &res, ctx_th);
1994                 if (status == 0)
1995                         break;
1996                 /* NOTE: BAD_SEQID means the server and client disagree about the
1997                  * book-keeping w.r.t. state-changing operations
1998                  * (OPEN/CLOSE/LOCK/LOCKU...)
1999                  * It is actually a sign of a bug on the client or on the server.
2000                  *
2001                  * If we receive a BAD_SEQID error in the particular case of
2002                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2003                  * have unhashed the old state_owner for us, and that we can
2004                  * therefore safely retry using a new one. We should still warn
2005                  * the user though...
2006                  */
2007                 if (status == -NFS4ERR_BAD_SEQID) {
2008                         pr_warn_ratelimited("NFS: v4 server %s "
2009                                         " returned a bad sequence-id error!\n",
2010                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2011                         exception.retry = 1;
2012                         continue;
2013                 }
2014                 /*
2015                  * BAD_STATEID on OPEN means that the server cancelled our
2016                  * state before it received the OPEN_CONFIRM.
2017                  * Recover by retrying the request as per the discussion
2018                  * on Page 181 of RFC3530.
2019                  */
2020                 if (status == -NFS4ERR_BAD_STATEID) {
2021                         exception.retry = 1;
2022                         continue;
2023                 }
2024                 if (status == -EAGAIN) {
2025                         /* We must have found a delegation */
2026                         exception.retry = 1;
2027                         continue;
2028                 }
2029                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
2030                                         status, &exception));
2031         } while (exception.retry);
2032         return res;
2033 }
2034
2035 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2036                             struct nfs_fattr *fattr, struct iattr *sattr,
2037                             struct nfs4_state *state)
2038 {
2039         struct nfs_server *server = NFS_SERVER(inode);
2040         struct nfs_setattrargs  arg = {
2041                 .fh             = NFS_FH(inode),
2042                 .iap            = sattr,
2043                 .server         = server,
2044                 .bitmask = server->attr_bitmask,
2045         };
2046         struct nfs_setattrres  res = {
2047                 .fattr          = fattr,
2048                 .server         = server,
2049         };
2050         struct rpc_message msg = {
2051                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2052                 .rpc_argp       = &arg,
2053                 .rpc_resp       = &res,
2054                 .rpc_cred       = cred,
2055         };
2056         unsigned long timestamp = jiffies;
2057         int status;
2058
2059         nfs_fattr_init(fattr);
2060
2061         if (state != NULL) {
2062                 struct nfs_lockowner lockowner = {
2063                         .l_owner = current->files,
2064                         .l_pid = current->tgid,
2065                 };
2066                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2067                                 &lockowner);
2068         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
2069                                 FMODE_WRITE)) {
2070                 /* Use that stateid */
2071         } else
2072                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2073
2074         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2075         if (status == 0 && state != NULL)
2076                 renew_lease(server, timestamp);
2077         return status;
2078 }
2079
2080 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2081                            struct nfs_fattr *fattr, struct iattr *sattr,
2082                            struct nfs4_state *state)
2083 {
2084         struct nfs_server *server = NFS_SERVER(inode);
2085         struct nfs4_exception exception = {
2086                 .state = state,
2087                 .inode = inode,
2088         };
2089         int err;
2090         do {
2091                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state);
2092                 switch (err) {
2093                 case -NFS4ERR_OPENMODE:
2094                         if (state && !(state->state & FMODE_WRITE)) {
2095                                 err = -EBADF;
2096                                 if (sattr->ia_valid & ATTR_OPEN)
2097                                         err = -EACCES;
2098                                 goto out;
2099                         }
2100                 }
2101                 err = nfs4_handle_exception(server, err, &exception);
2102         } while (exception.retry);
2103 out:
2104         return err;
2105 }
2106
2107 struct nfs4_closedata {
2108         struct inode *inode;
2109         struct nfs4_state *state;
2110         struct nfs_closeargs arg;
2111         struct nfs_closeres res;
2112         struct nfs_fattr fattr;
2113         unsigned long timestamp;
2114         bool roc;
2115         u32 roc_barrier;
2116 };
2117
2118 static void nfs4_free_closedata(void *data)
2119 {
2120         struct nfs4_closedata *calldata = data;
2121         struct nfs4_state_owner *sp = calldata->state->owner;
2122         struct super_block *sb = calldata->state->inode->i_sb;
2123
2124         if (calldata->roc)
2125                 pnfs_roc_release(calldata->state->inode);
2126         nfs4_put_open_state(calldata->state);
2127         nfs_free_seqid(calldata->arg.seqid);
2128         nfs4_put_state_owner(sp);
2129         nfs_sb_deactive(sb);
2130         kfree(calldata);
2131 }
2132
2133 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
2134                 fmode_t fmode)
2135 {
2136         spin_lock(&state->owner->so_lock);
2137         if (!(fmode & FMODE_READ))
2138                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2139         if (!(fmode & FMODE_WRITE))
2140                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2141         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2142         spin_unlock(&state->owner->so_lock);
2143 }
2144
2145 static void nfs4_close_done(struct rpc_task *task, void *data)
2146 {
2147         struct nfs4_closedata *calldata = data;
2148         struct nfs4_state *state = calldata->state;
2149         struct nfs_server *server = NFS_SERVER(calldata->inode);
2150
2151         dprintk("%s: begin!\n", __func__);
2152         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2153                 return;
2154         /* hmm. we are done with the inode, and in the process of freeing
2155          * the state_owner. we keep this around to process errors
2156          */
2157         switch (task->tk_status) {
2158                 case 0:
2159                         if (calldata->roc)
2160                                 pnfs_roc_set_barrier(state->inode,
2161                                                      calldata->roc_barrier);
2162                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2163                         renew_lease(server, calldata->timestamp);
2164                         nfs4_close_clear_stateid_flags(state,
2165                                         calldata->arg.fmode);
2166                         break;
2167                 case -NFS4ERR_STALE_STATEID:
2168                 case -NFS4ERR_OLD_STATEID:
2169                 case -NFS4ERR_BAD_STATEID:
2170                 case -NFS4ERR_EXPIRED:
2171                         if (calldata->arg.fmode == 0)
2172                                 break;
2173                 default:
2174                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2175                                 rpc_restart_call_prepare(task);
2176         }
2177         nfs_release_seqid(calldata->arg.seqid);
2178         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2179         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2180 }
2181
2182 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2183 {
2184         struct nfs4_closedata *calldata = data;
2185         struct nfs4_state *state = calldata->state;
2186         struct inode *inode = calldata->inode;
2187         int call_close = 0;
2188
2189         dprintk("%s: begin!\n", __func__);
2190         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2191                 return;
2192
2193         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2194         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2195         spin_lock(&state->owner->so_lock);
2196         /* Calculate the change in open mode */
2197         if (state->n_rdwr == 0) {
2198                 if (state->n_rdonly == 0) {
2199                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2200                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2201                         calldata->arg.fmode &= ~FMODE_READ;
2202                 }
2203                 if (state->n_wronly == 0) {
2204                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2205                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2206                         calldata->arg.fmode &= ~FMODE_WRITE;
2207                 }
2208         }
2209         spin_unlock(&state->owner->so_lock);
2210
2211         if (!call_close) {
2212                 /* Note: exit _without_ calling nfs4_close_done */
2213                 task->tk_action = NULL;
2214                 goto out;
2215         }
2216
2217         if (calldata->arg.fmode == 0) {
2218                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2219                 if (calldata->roc &&
2220                     pnfs_roc_drain(inode, &calldata->roc_barrier, task))
2221                         goto out;
2222         }
2223
2224         nfs_fattr_init(calldata->res.fattr);
2225         calldata->timestamp = jiffies;
2226         if (nfs4_setup_sequence(NFS_SERVER(inode),
2227                                 &calldata->arg.seq_args,
2228                                 &calldata->res.seq_res,
2229                                 task))
2230                 goto out;
2231         rpc_call_start(task);
2232 out:
2233         dprintk("%s: done!\n", __func__);
2234 }
2235
2236 static const struct rpc_call_ops nfs4_close_ops = {
2237         .rpc_call_prepare = nfs4_close_prepare,
2238         .rpc_call_done = nfs4_close_done,
2239         .rpc_release = nfs4_free_closedata,
2240 };
2241
2242 /*
2243  * It is possible for data to be read/written from a mem-mapped file
2244  * after the sys_close call (which hits the vfs layer as a flush).
2245  * This means that we can't safely call nfsv4 close on a file until
2246  * the inode is cleared. This in turn means that we are not good
2247  * NFSv4 citizens - we do not indicate to the server to update the file's
2248  * share state even when we are done with one of the three share
2249  * stateid's in the inode.
2250  *
2251  * NOTE: Caller must be holding the sp->so_owner semaphore!
2252  */
2253 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2254 {
2255         struct nfs_server *server = NFS_SERVER(state->inode);
2256         struct nfs4_closedata *calldata;
2257         struct nfs4_state_owner *sp = state->owner;
2258         struct rpc_task *task;
2259         struct rpc_message msg = {
2260                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2261                 .rpc_cred = state->owner->so_cred,
2262         };
2263         struct rpc_task_setup task_setup_data = {
2264                 .rpc_client = server->client,
2265                 .rpc_message = &msg,
2266                 .callback_ops = &nfs4_close_ops,
2267                 .workqueue = nfsiod_workqueue,
2268                 .flags = RPC_TASK_ASYNC,
2269         };
2270         int status = -ENOMEM;
2271
2272         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2273         if (calldata == NULL)
2274                 goto out;
2275         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2276         calldata->inode = state->inode;
2277         calldata->state = state;
2278         calldata->arg.fh = NFS_FH(state->inode);
2279         calldata->arg.stateid = &state->open_stateid;
2280         /* Serialization for the sequence id */
2281         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2282         if (calldata->arg.seqid == NULL)
2283                 goto out_free_calldata;
2284         calldata->arg.fmode = 0;
2285         calldata->arg.bitmask = server->cache_consistency_bitmask;
2286         calldata->res.fattr = &calldata->fattr;
2287         calldata->res.seqid = calldata->arg.seqid;
2288         calldata->res.server = server;
2289         calldata->roc = pnfs_roc(state->inode);
2290         nfs_sb_active(calldata->inode->i_sb);
2291
2292         msg.rpc_argp = &calldata->arg;
2293         msg.rpc_resp = &calldata->res;
2294         task_setup_data.callback_data = calldata;
2295         task = rpc_run_task(&task_setup_data);
2296         if (IS_ERR(task))
2297                 return PTR_ERR(task);
2298         status = 0;
2299         if (wait)
2300                 status = rpc_wait_for_completion_task(task);
2301         rpc_put_task(task);
2302         return status;
2303 out_free_calldata:
2304         kfree(calldata);
2305 out:
2306         nfs4_put_open_state(state);
2307         nfs4_put_state_owner(sp);
2308         return status;
2309 }
2310
2311 static struct inode *
2312 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2313 {
2314         struct nfs4_state *state;
2315
2316         /* Protect against concurrent sillydeletes */
2317         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr,
2318                              ctx->cred, &ctx->mdsthreshold);
2319         if (IS_ERR(state))
2320                 return ERR_CAST(state);
2321         ctx->state = state;
2322         return igrab(state->inode);
2323 }
2324
2325 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2326 {
2327         if (ctx->state == NULL)
2328                 return;
2329         if (is_sync)
2330                 nfs4_close_sync(ctx->state, ctx->mode);
2331         else
2332                 nfs4_close_state(ctx->state, ctx->mode);
2333 }
2334
2335 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2336 {
2337         struct nfs4_server_caps_arg args = {
2338                 .fhandle = fhandle,
2339         };
2340         struct nfs4_server_caps_res res = {};
2341         struct rpc_message msg = {
2342                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2343                 .rpc_argp = &args,
2344                 .rpc_resp = &res,
2345         };
2346         int status;
2347
2348         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2349         if (status == 0) {
2350                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2351                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2352                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2353                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2354                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2355                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2356                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2357                         server->caps |= NFS_CAP_ACLS;
2358                 if (res.has_links != 0)
2359                         server->caps |= NFS_CAP_HARDLINKS;
2360                 if (res.has_symlinks != 0)
2361                         server->caps |= NFS_CAP_SYMLINKS;
2362                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2363                         server->caps |= NFS_CAP_FILEID;
2364                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2365                         server->caps |= NFS_CAP_MODE;
2366                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2367                         server->caps |= NFS_CAP_NLINK;
2368                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2369                         server->caps |= NFS_CAP_OWNER;
2370                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2371                         server->caps |= NFS_CAP_OWNER_GROUP;
2372                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2373                         server->caps |= NFS_CAP_ATIME;
2374                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2375                         server->caps |= NFS_CAP_CTIME;
2376                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2377                         server->caps |= NFS_CAP_MTIME;
2378
2379                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2380                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2381                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2382                 server->acl_bitmask = res.acl_bitmask;
2383                 server->fh_expire_type = res.fh_expire_type;
2384         }
2385
2386         return status;
2387 }
2388
2389 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2390 {
2391         struct nfs4_exception exception = { };
2392         int err;
2393         do {
2394                 err = nfs4_handle_exception(server,
2395                                 _nfs4_server_capabilities(server, fhandle),
2396                                 &exception);
2397         } while (exception.retry);
2398         return err;
2399 }
2400
2401 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2402                 struct nfs_fsinfo *info)
2403 {
2404         struct nfs4_lookup_root_arg args = {
2405                 .bitmask = nfs4_fattr_bitmap,
2406         };
2407         struct nfs4_lookup_res res = {
2408                 .server = server,
2409                 .fattr = info->fattr,
2410                 .fh = fhandle,
2411         };
2412         struct rpc_message msg = {
2413                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2414                 .rpc_argp = &args,
2415                 .rpc_resp = &res,
2416         };
2417
2418         nfs_fattr_init(info->fattr);
2419         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2420 }
2421
2422 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2423                 struct nfs_fsinfo *info)
2424 {
2425         struct nfs4_exception exception = { };
2426         int err;
2427         do {
2428                 err = _nfs4_lookup_root(server, fhandle, info);
2429                 switch (err) {
2430                 case 0:
2431                 case -NFS4ERR_WRONGSEC:
2432                         goto out;
2433                 default:
2434                         err = nfs4_handle_exception(server, err, &exception);
2435                 }
2436         } while (exception.retry);
2437 out:
2438         return err;
2439 }
2440
2441 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2442                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2443 {
2444         struct rpc_auth *auth;
2445         int ret;
2446
2447         auth = rpcauth_create(flavor, server->client);
2448         if (IS_ERR(auth)) {
2449                 ret = -EIO;
2450                 goto out;
2451         }
2452         ret = nfs4_lookup_root(server, fhandle, info);
2453 out:
2454         return ret;
2455 }
2456
2457 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2458                               struct nfs_fsinfo *info)
2459 {
2460         int i, len, status = 0;
2461         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2462
2463         len = rpcauth_list_flavors(flav_array, ARRAY_SIZE(flav_array));
2464         BUG_ON(len < 0);
2465
2466         for (i = 0; i < len; i++) {
2467                 /* AUTH_UNIX is the default flavor if none was specified,
2468                  * thus has already been tried. */
2469                 if (flav_array[i] == RPC_AUTH_UNIX)
2470                         continue;
2471
2472                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2473                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2474                         continue;
2475                 break;
2476         }
2477         /*
2478          * -EACCESS could mean that the user doesn't have correct permissions
2479          * to access the mount.  It could also mean that we tried to mount
2480          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2481          * existing mount programs don't handle -EACCES very well so it should
2482          * be mapped to -EPERM instead.
2483          */
2484         if (status == -EACCES)
2485                 status = -EPERM;
2486         return status;
2487 }
2488
2489 /*
2490  * get the file handle for the "/" directory on the server
2491  */
2492 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
2493                          struct nfs_fsinfo *info)
2494 {
2495         int minor_version = server->nfs_client->cl_minorversion;
2496         int status = nfs4_lookup_root(server, fhandle, info);
2497         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2498                 /*
2499                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2500                  * by nfs4_map_errors() as this function exits.
2501                  */
2502                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2503         if (status == 0)
2504                 status = nfs4_server_capabilities(server, fhandle);
2505         if (status == 0)
2506                 status = nfs4_do_fsinfo(server, fhandle, info);
2507         return nfs4_map_errors(status);
2508 }
2509
2510 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
2511                               struct nfs_fsinfo *info)
2512 {
2513         int error;
2514         struct nfs_fattr *fattr = info->fattr;
2515
2516         error = nfs4_server_capabilities(server, mntfh);
2517         if (error < 0) {
2518                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
2519                 return error;
2520         }
2521
2522         error = nfs4_proc_getattr(server, mntfh, fattr);
2523         if (error < 0) {
2524                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
2525                 return error;
2526         }
2527
2528         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
2529             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
2530                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
2531
2532         return error;
2533 }
2534
2535 /*
2536  * Get locations and (maybe) other attributes of a referral.
2537  * Note that we'll actually follow the referral later when
2538  * we detect fsid mismatch in inode revalidation
2539  */
2540 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
2541                              const struct qstr *name, struct nfs_fattr *fattr,
2542                              struct nfs_fh *fhandle)
2543 {
2544         int status = -ENOMEM;
2545         struct page *page = NULL;
2546         struct nfs4_fs_locations *locations = NULL;
2547
2548         page = alloc_page(GFP_KERNEL);
2549         if (page == NULL)
2550                 goto out;
2551         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2552         if (locations == NULL)
2553                 goto out;
2554
2555         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
2556         if (status != 0)
2557                 goto out;
2558         /* Make sure server returned a different fsid for the referral */
2559         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2560                 dprintk("%s: server did not return a different fsid for"
2561                         " a referral at %s\n", __func__, name->name);
2562                 status = -EIO;
2563                 goto out;
2564         }
2565         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2566         nfs_fixup_referral_attributes(&locations->fattr);
2567
2568         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2569         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2570         memset(fhandle, 0, sizeof(struct nfs_fh));
2571 out:
2572         if (page)
2573                 __free_page(page);
2574         kfree(locations);
2575         return status;
2576 }
2577
2578 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2579 {
2580         struct nfs4_getattr_arg args = {
2581                 .fh = fhandle,
2582                 .bitmask = server->attr_bitmask,
2583         };
2584         struct nfs4_getattr_res res = {
2585                 .fattr = fattr,
2586                 .server = server,
2587         };
2588         struct rpc_message msg = {
2589                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2590                 .rpc_argp = &args,
2591                 .rpc_resp = &res,
2592         };
2593
2594         nfs_fattr_init(fattr);
2595         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2596 }
2597
2598 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2599 {
2600         struct nfs4_exception exception = { };
2601         int err;
2602         do {
2603                 err = nfs4_handle_exception(server,
2604                                 _nfs4_proc_getattr(server, fhandle, fattr),
2605                                 &exception);
2606         } while (exception.retry);
2607         return err;
2608 }
2609
2610 /*
2611  * The file is not closed if it is opened due to the a request to change
2612  * the size of the file. The open call will not be needed once the
2613  * VFS layer lookup-intents are implemented.
2614  *
2615  * Close is called when the inode is destroyed.
2616  * If we haven't opened the file for O_WRONLY, we
2617  * need to in the size_change case to obtain a stateid.
2618  *
2619  * Got race?
2620  * Because OPEN is always done by name in nfsv4, it is
2621  * possible that we opened a different file by the same
2622  * name.  We can recognize this race condition, but we
2623  * can't do anything about it besides returning an error.
2624  *
2625  * This will be fixed with VFS changes (lookup-intent).
2626  */
2627 static int
2628 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2629                   struct iattr *sattr)
2630 {
2631         struct inode *inode = dentry->d_inode;
2632         struct rpc_cred *cred = NULL;
2633         struct nfs4_state *state = NULL;
2634         int status;
2635
2636         if (pnfs_ld_layoutret_on_setattr(inode))
2637                 pnfs_return_layout(inode);
2638
2639         nfs_fattr_init(fattr);
2640
2641         /* Deal with open(O_TRUNC) */
2642         if (sattr->ia_valid & ATTR_OPEN)
2643                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2644
2645         /* Optimization: if the end result is no change, don't RPC */
2646         if ((sattr->ia_valid & ~(ATTR_FILE)) == 0)
2647                 return 0;
2648
2649         /* Search for an existing open(O_WRITE) file */
2650         if (sattr->ia_valid & ATTR_FILE) {
2651                 struct nfs_open_context *ctx;
2652
2653                 ctx = nfs_file_open_context(sattr->ia_file);
2654                 if (ctx) {
2655                         cred = ctx->cred;
2656                         state = ctx->state;
2657                 }
2658         }
2659
2660         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2661         if (status == 0)
2662                 nfs_setattr_update_inode(inode, sattr);
2663         return status;
2664 }
2665
2666 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2667                 const struct qstr *name, struct nfs_fh *fhandle,
2668                 struct nfs_fattr *fattr)
2669 {
2670         struct nfs_server *server = NFS_SERVER(dir);
2671         int                    status;
2672         struct nfs4_lookup_arg args = {
2673                 .bitmask = server->attr_bitmask,
2674                 .dir_fh = NFS_FH(dir),
2675                 .name = name,
2676         };
2677         struct nfs4_lookup_res res = {
2678                 .server = server,
2679                 .fattr = fattr,
2680                 .fh = fhandle,
2681         };
2682         struct rpc_message msg = {
2683                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2684                 .rpc_argp = &args,
2685                 .rpc_resp = &res,
2686         };
2687
2688         nfs_fattr_init(fattr);
2689
2690         dprintk("NFS call  lookup %s\n", name->name);
2691         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2692         dprintk("NFS reply lookup: %d\n", status);
2693         return status;
2694 }
2695
2696 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
2697 {
2698         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2699                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
2700         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2701         fattr->nlink = 2;
2702 }
2703
2704 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
2705                                    struct qstr *name, struct nfs_fh *fhandle,
2706                                    struct nfs_fattr *fattr)
2707 {
2708         struct nfs4_exception exception = { };
2709         struct rpc_clnt *client = *clnt;
2710         int err;
2711         do {
2712                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr);
2713                 switch (err) {
2714                 case -NFS4ERR_BADNAME:
2715                         err = -ENOENT;
2716                         goto out;
2717                 case -NFS4ERR_MOVED:
2718                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
2719                         goto out;
2720                 case -NFS4ERR_WRONGSEC:
2721                         err = -EPERM;
2722                         if (client != *clnt)
2723                                 goto out;
2724
2725                         client = nfs4_create_sec_client(client, dir, name);
2726                         if (IS_ERR(client))
2727                                 return PTR_ERR(client);
2728
2729                         exception.retry = 1;
2730                         break;
2731                 default:
2732                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
2733                 }
2734         } while (exception.retry);
2735
2736 out:
2737         if (err == 0)
2738                 *clnt = client;
2739         else if (client != *clnt)
2740                 rpc_shutdown_client(client);
2741
2742         return err;
2743 }
2744
2745 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
2746                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2747 {
2748         int status;
2749         struct rpc_clnt *client = NFS_CLIENT(dir);
2750
2751         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2752         if (client != NFS_CLIENT(dir)) {
2753                 rpc_shutdown_client(client);
2754                 nfs_fixup_secinfo_attributes(fattr);
2755         }
2756         return status;
2757 }
2758
2759 struct rpc_clnt *
2760 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
2761                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2762 {
2763         int status;
2764         struct rpc_clnt *client = rpc_clone_client(NFS_CLIENT(dir));
2765
2766         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr);
2767         if (status < 0) {
2768                 rpc_shutdown_client(client);
2769                 return ERR_PTR(status);
2770         }
2771         return client;
2772 }
2773
2774 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2775 {
2776         struct nfs_server *server = NFS_SERVER(inode);
2777         struct nfs4_accessargs args = {
2778                 .fh = NFS_FH(inode),
2779                 .bitmask = server->cache_consistency_bitmask,
2780         };
2781         struct nfs4_accessres res = {
2782                 .server = server,
2783         };
2784         struct rpc_message msg = {
2785                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2786                 .rpc_argp = &args,
2787                 .rpc_resp = &res,
2788                 .rpc_cred = entry->cred,
2789         };
2790         int mode = entry->mask;
2791         int status;
2792
2793         /*
2794          * Determine which access bits we want to ask for...
2795          */
2796         if (mode & MAY_READ)
2797                 args.access |= NFS4_ACCESS_READ;
2798         if (S_ISDIR(inode->i_mode)) {
2799                 if (mode & MAY_WRITE)
2800                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2801                 if (mode & MAY_EXEC)
2802                         args.access |= NFS4_ACCESS_LOOKUP;
2803         } else {
2804                 if (mode & MAY_WRITE)
2805                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2806                 if (mode & MAY_EXEC)
2807                         args.access |= NFS4_ACCESS_EXECUTE;
2808         }
2809
2810         res.fattr = nfs_alloc_fattr();
2811         if (res.fattr == NULL)
2812                 return -ENOMEM;
2813
2814         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2815         if (!status) {
2816                 nfs_access_set_mask(entry, res.access);
2817                 nfs_refresh_inode(inode, res.fattr);
2818         }
2819         nfs_free_fattr(res.fattr);
2820         return status;
2821 }
2822
2823 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2824 {
2825         struct nfs4_exception exception = { };
2826         int err;
2827         do {
2828                 err = nfs4_handle_exception(NFS_SERVER(inode),
2829                                 _nfs4_proc_access(inode, entry),
2830                                 &exception);
2831         } while (exception.retry);
2832         return err;
2833 }
2834
2835 /*
2836  * TODO: For the time being, we don't try to get any attributes
2837  * along with any of the zero-copy operations READ, READDIR,
2838  * READLINK, WRITE.
2839  *
2840  * In the case of the first three, we want to put the GETATTR
2841  * after the read-type operation -- this is because it is hard
2842  * to predict the length of a GETATTR response in v4, and thus
2843  * align the READ data correctly.  This means that the GETATTR
2844  * may end up partially falling into the page cache, and we should
2845  * shift it into the 'tail' of the xdr_buf before processing.
2846  * To do this efficiently, we need to know the total length
2847  * of data received, which doesn't seem to be available outside
2848  * of the RPC layer.
2849  *
2850  * In the case of WRITE, we also want to put the GETATTR after
2851  * the operation -- in this case because we want to make sure
2852  * we get the post-operation mtime and size.
2853  *
2854  * Both of these changes to the XDR layer would in fact be quite
2855  * minor, but I decided to leave them for a subsequent patch.
2856  */
2857 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2858                 unsigned int pgbase, unsigned int pglen)
2859 {
2860         struct nfs4_readlink args = {
2861                 .fh       = NFS_FH(inode),
2862                 .pgbase   = pgbase,
2863                 .pglen    = pglen,
2864                 .pages    = &page,
2865         };
2866         struct nfs4_readlink_res res;
2867         struct rpc_message msg = {
2868                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2869                 .rpc_argp = &args,
2870                 .rpc_resp = &res,
2871         };
2872
2873         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2874 }
2875
2876 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2877                 unsigned int pgbase, unsigned int pglen)
2878 {
2879         struct nfs4_exception exception = { };
2880         int err;
2881         do {
2882                 err = nfs4_handle_exception(NFS_SERVER(inode),
2883                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2884                                 &exception);
2885         } while (exception.retry);
2886         return err;
2887 }
2888
2889 /*
2890  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
2891  */
2892 static int
2893 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2894                  int flags)
2895 {
2896         struct nfs_open_context *ctx;
2897         struct nfs4_state *state;
2898         int status = 0;
2899
2900         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
2901         if (IS_ERR(ctx))
2902                 return PTR_ERR(ctx);
2903
2904         sattr->ia_mode &= ~current_umask();
2905         state = nfs4_do_open(dir, dentry, ctx->mode,
2906                         flags, sattr, ctx->cred,
2907                         &ctx->mdsthreshold);
2908         d_drop(dentry);
2909         if (IS_ERR(state)) {
2910                 status = PTR_ERR(state);
2911                 goto out;
2912         }
2913         d_add(dentry, igrab(state->inode));
2914         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2915         ctx->state = state;
2916 out:
2917         put_nfs_open_context(ctx);
2918         return status;
2919 }
2920
2921 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2922 {
2923         struct nfs_server *server = NFS_SERVER(dir);
2924         struct nfs_removeargs args = {
2925                 .fh = NFS_FH(dir),
2926                 .name = *name,
2927         };
2928         struct nfs_removeres res = {
2929                 .server = server,
2930         };
2931         struct rpc_message msg = {
2932                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2933                 .rpc_argp = &args,
2934                 .rpc_resp = &res,
2935         };
2936         int status;
2937
2938         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2939         if (status == 0)
2940                 update_changeattr(dir, &res.cinfo);
2941         return status;
2942 }
2943
2944 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2945 {
2946         struct nfs4_exception exception = { };
2947         int err;
2948         do {
2949                 err = nfs4_handle_exception(NFS_SERVER(dir),
2950                                 _nfs4_proc_remove(dir, name),
2951                                 &exception);
2952         } while (exception.retry);
2953         return err;
2954 }
2955
2956 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2957 {
2958         struct nfs_server *server = NFS_SERVER(dir);
2959         struct nfs_removeargs *args = msg->rpc_argp;
2960         struct nfs_removeres *res = msg->rpc_resp;
2961
2962         res->server = server;
2963         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2964         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2965 }
2966
2967 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
2968 {
2969         if (nfs4_setup_sequence(NFS_SERVER(data->dir),
2970                                 &data->args.seq_args,
2971                                 &data->res.seq_res,
2972                                 task))
2973                 return;
2974         rpc_call_start(task);
2975 }
2976
2977 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2978 {
2979         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2980
2981         if (!nfs4_sequence_done(task, &res->seq_res))
2982                 return 0;
2983         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2984                 return 0;
2985         update_changeattr(dir, &res->cinfo);
2986         return 1;
2987 }
2988
2989 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2990 {
2991         struct nfs_server *server = NFS_SERVER(dir);
2992         struct nfs_renameargs *arg = msg->rpc_argp;
2993         struct nfs_renameres *res = msg->rpc_resp;
2994
2995         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2996         res->server = server;
2997         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2998 }
2999
3000 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3001 {
3002         if (nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3003                                 &data->args.seq_args,
3004                                 &data->res.seq_res,
3005                                 task))
3006                 return;
3007         rpc_call_start(task);
3008 }
3009
3010 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3011                                  struct inode *new_dir)
3012 {
3013         struct nfs_renameres *res = task->tk_msg.rpc_resp;
3014
3015         if (!nfs4_sequence_done(task, &res->seq_res))
3016                 return 0;
3017         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
3018                 return 0;
3019
3020         update_changeattr(old_dir, &res->old_cinfo);
3021         update_changeattr(new_dir, &res->new_cinfo);
3022         return 1;
3023 }
3024
3025 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3026                 struct inode *new_dir, struct qstr *new_name)
3027 {
3028         struct nfs_server *server = NFS_SERVER(old_dir);
3029         struct nfs_renameargs arg = {
3030                 .old_dir = NFS_FH(old_dir),
3031                 .new_dir = NFS_FH(new_dir),
3032                 .old_name = old_name,
3033                 .new_name = new_name,
3034         };
3035         struct nfs_renameres res = {
3036                 .server = server,
3037         };
3038         struct rpc_message msg = {
3039                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
3040                 .rpc_argp = &arg,
3041                 .rpc_resp = &res,
3042         };
3043         int status = -ENOMEM;
3044
3045         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3046         if (!status) {
3047                 update_changeattr(old_dir, &res.old_cinfo);
3048                 update_changeattr(new_dir, &res.new_cinfo);
3049         }
3050         return status;
3051 }
3052
3053 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
3054                 struct inode *new_dir, struct qstr *new_name)
3055 {
3056         struct nfs4_exception exception = { };
3057         int err;
3058         do {
3059                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
3060                                 _nfs4_proc_rename(old_dir, old_name,
3061                                         new_dir, new_name),
3062                                 &exception);
3063         } while (exception.retry);
3064         return err;
3065 }
3066
3067 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3068 {
3069         struct nfs_server *server = NFS_SERVER(inode);
3070         struct nfs4_link_arg arg = {
3071                 .fh     = NFS_FH(inode),
3072                 .dir_fh = NFS_FH(dir),
3073                 .name   = name,
3074                 .bitmask = server->attr_bitmask,
3075         };
3076         struct nfs4_link_res res = {
3077                 .server = server,
3078         };
3079         struct rpc_message msg = {
3080                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3081                 .rpc_argp = &arg,
3082                 .rpc_resp = &res,
3083         };
3084         int status = -ENOMEM;
3085
3086         res.fattr = nfs_alloc_fattr();
3087         if (res.fattr == NULL)
3088                 goto out;
3089
3090         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3091         if (!status) {
3092                 update_changeattr(dir, &res.cinfo);
3093                 nfs_post_op_update_inode(inode, res.fattr);
3094         }
3095 out:
3096         nfs_free_fattr(res.fattr);
3097         return status;
3098 }
3099
3100 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3101 {
3102         struct nfs4_exception exception = { };
3103         int err;
3104         do {
3105                 err = nfs4_handle_exception(NFS_SERVER(inode),
3106                                 _nfs4_proc_link(inode, dir, name),
3107                                 &exception);
3108         } while (exception.retry);
3109         return err;
3110 }
3111
3112 struct nfs4_createdata {
3113         struct rpc_message msg;
3114         struct nfs4_create_arg arg;
3115         struct nfs4_create_res res;
3116         struct nfs_fh fh;
3117         struct nfs_fattr fattr;
3118 };
3119
3120 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3121                 struct qstr *name, struct iattr *sattr, u32 ftype)
3122 {
3123         struct nfs4_createdata *data;
3124
3125         data = kzalloc(sizeof(*data), GFP_KERNEL);
3126         if (data != NULL) {
3127                 struct nfs_server *server = NFS_SERVER(dir);
3128
3129                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3130                 data->msg.rpc_argp = &data->arg;
3131                 data->msg.rpc_resp = &data->res;
3132                 data->arg.dir_fh = NFS_FH(dir);
3133                 data->arg.server = server;
3134                 data->arg.name = name;
3135                 data->arg.attrs = sattr;
3136                 data->arg.ftype = ftype;
3137                 data->arg.bitmask = server->attr_bitmask;
3138                 data->res.server = server;
3139                 data->res.fh = &data->fh;
3140                 data->res.fattr = &data->fattr;
3141                 nfs_fattr_init(data->res.fattr);
3142         }
3143         return data;
3144 }
3145
3146 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3147 {
3148         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3149                                     &data->arg.seq_args, &data->res.seq_res, 1);
3150         if (status == 0) {
3151                 update_changeattr(dir, &data->res.dir_cinfo);
3152                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
3153         }
3154         return status;
3155 }
3156
3157 static void nfs4_free_createdata(struct nfs4_createdata *data)
3158 {
3159         kfree(data);
3160 }
3161
3162 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3163                 struct page *page, unsigned int len, struct iattr *sattr)
3164 {
3165         struct nfs4_createdata *data;
3166         int status = -ENAMETOOLONG;
3167
3168         if (len > NFS4_MAXPATHLEN)
3169                 goto out;
3170
3171         status = -ENOMEM;
3172         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3173         if (data == NULL)
3174                 goto out;
3175
3176         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3177         data->arg.u.symlink.pages = &page;
3178         data->arg.u.symlink.len = len;
3179
3180         status = nfs4_do_create(dir, dentry, data);
3181
3182         nfs4_free_createdata(data);
3183 out:
3184         return status;
3185 }
3186
3187 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3188                 struct page *page, unsigned int len, struct iattr *sattr)
3189 {
3190         struct nfs4_exception exception = { };
3191         int err;
3192         do {
3193                 err = nfs4_handle_exception(NFS_SERVER(dir),
3194                                 _nfs4_proc_symlink(dir, dentry, page,
3195                                                         len, sattr),
3196                                 &exception);
3197         } while (exception.retry);
3198         return err;
3199 }
3200
3201 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3202                 struct iattr *sattr)
3203 {
3204         struct nfs4_createdata *data;
3205         int status = -ENOMEM;
3206
3207         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3208         if (data == NULL)
3209                 goto out;
3210
3211         status = nfs4_do_create(dir, dentry, data);
3212
3213         nfs4_free_createdata(data);
3214 out:
3215         return status;
3216 }
3217
3218 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3219                 struct iattr *sattr)
3220 {
3221         struct nfs4_exception exception = { };
3222         int err;
3223
3224         sattr->ia_mode &= ~current_umask();
3225         do {
3226                 err = nfs4_handle_exception(NFS_SERVER(dir),
3227                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3228                                 &exception);
3229         } while (exception.retry);
3230         return err;
3231 }
3232
3233 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3234                 u64 cookie, struct page **pages, unsigned int count, int plus)
3235 {
3236         struct inode            *dir = dentry->d_inode;
3237         struct nfs4_readdir_arg args = {
3238                 .fh = NFS_FH(dir),
3239                 .pages = pages,
3240                 .pgbase = 0,
3241                 .count = count,
3242                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3243                 .plus = plus,
3244         };
3245         struct nfs4_readdir_res res;
3246         struct rpc_message msg = {
3247                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3248                 .rpc_argp = &args,
3249                 .rpc_resp = &res,
3250                 .rpc_cred = cred,
3251         };
3252         int                     status;
3253
3254         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3255                         dentry->d_parent->d_name.name,
3256                         dentry->d_name.name,
3257                         (unsigned long long)cookie);
3258         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3259         res.pgbase = args.pgbase;
3260         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3261         if (status >= 0) {
3262                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3263                 status += args.pgbase;
3264         }
3265
3266         nfs_invalidate_atime(dir);
3267
3268         dprintk("%s: returns %d\n", __func__, status);
3269         return status;
3270 }
3271
3272 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3273                 u64 cookie, struct page **pages, unsigned int count, int plus)
3274 {
3275         struct nfs4_exception exception = { };
3276         int err;
3277         do {
3278                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3279                                 _nfs4_proc_readdir(dentry, cred, cookie,
3280                                         pages, count, plus),
3281                                 &exception);
3282         } while (exception.retry);
3283         return err;
3284 }
3285
3286 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3287                 struct iattr *sattr, dev_t rdev)
3288 {
3289         struct nfs4_createdata *data;
3290         int mode = sattr->ia_mode;
3291         int status = -ENOMEM;
3292
3293         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3294         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3295
3296         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3297         if (data == NULL)
3298                 goto out;
3299
3300         if (S_ISFIFO(mode))
3301                 data->arg.ftype = NF4FIFO;
3302         else if (S_ISBLK(mode)) {
3303                 data->arg.ftype = NF4BLK;
3304                 data->arg.u.device.specdata1 = MAJOR(rdev);
3305                 data->arg.u.device.specdata2 = MINOR(rdev);
3306         }
3307         else if (S_ISCHR(mode)) {
3308                 data->arg.ftype = NF4CHR;
3309                 data->arg.u.device.specdata1 = MAJOR(rdev);
3310                 data->arg.u.device.specdata2 = MINOR(rdev);
3311         }
3312
3313         status = nfs4_do_create(dir, dentry, data);
3314
3315         nfs4_free_createdata(data);
3316 out:
3317         return status;
3318 }
3319
3320 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3321                 struct iattr *sattr, dev_t rdev)
3322 {
3323         struct nfs4_exception exception = { };
3324         int err;
3325
3326         sattr->ia_mode &= ~current_umask();
3327         do {
3328                 err = nfs4_handle_exception(NFS_SERVER(dir),
3329                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3330                                 &exception);
3331         } while (exception.retry);
3332         return err;
3333 }
3334
3335 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3336                  struct nfs_fsstat *fsstat)
3337 {
3338         struct nfs4_statfs_arg args = {
3339                 .fh = fhandle,
3340                 .bitmask = server->attr_bitmask,
3341         };
3342         struct nfs4_statfs_res res = {
3343                 .fsstat = fsstat,
3344         };
3345         struct rpc_message msg = {
3346                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3347                 .rpc_argp = &args,
3348                 .rpc_resp = &res,
3349         };
3350
3351         nfs_fattr_init(fsstat->fattr);
3352         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3353 }
3354
3355 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3356 {
3357         struct nfs4_exception exception = { };
3358         int err;
3359         do {
3360                 err = nfs4_handle_exception(server,
3361                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3362                                 &exception);
3363         } while (exception.retry);
3364         return err;
3365 }
3366
3367 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3368                 struct nfs_fsinfo *fsinfo)
3369 {
3370         struct nfs4_fsinfo_arg args = {
3371                 .fh = fhandle,
3372                 .bitmask = server->attr_bitmask,
3373         };
3374         struct nfs4_fsinfo_res res = {
3375                 .fsinfo = fsinfo,
3376         };
3377         struct rpc_message msg = {
3378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3379                 .rpc_argp = &args,
3380                 .rpc_resp = &res,
3381         };
3382
3383         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3384 }
3385
3386 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3387 {
3388         struct nfs4_exception exception = { };
3389         int err;
3390
3391         do {
3392                 err = nfs4_handle_exception(server,
3393                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3394                                 &exception);
3395         } while (exception.retry);
3396         return err;
3397 }
3398
3399 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3400 {
3401         int error;
3402
3403         nfs_fattr_init(fsinfo->fattr);
3404         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
3405         if (error == 0)
3406                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3407
3408         return error;
3409 }
3410
3411 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3412                 struct nfs_pathconf *pathconf)
3413 {
3414         struct nfs4_pathconf_arg args = {
3415                 .fh = fhandle,
3416                 .bitmask = server->attr_bitmask,
3417         };
3418         struct nfs4_pathconf_res res = {
3419                 .pathconf = pathconf,
3420         };
3421         struct rpc_message msg = {
3422                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3423                 .rpc_argp = &args,
3424                 .rpc_resp = &res,
3425         };
3426
3427         /* None of the pathconf attributes are mandatory to implement */
3428         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3429                 memset(pathconf, 0, sizeof(*pathconf));
3430                 return 0;
3431         }
3432
3433         nfs_fattr_init(pathconf->fattr);
3434         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3435 }
3436
3437 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3438                 struct nfs_pathconf *pathconf)
3439 {
3440         struct nfs4_exception exception = { };
3441         int err;
3442
3443         do {
3444                 err = nfs4_handle_exception(server,
3445                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3446                                 &exception);
3447         } while (exception.retry);
3448         return err;
3449 }
3450
3451 void __nfs4_read_done_cb(struct nfs_read_data *data)
3452 {
3453         nfs_invalidate_atime(data->header->inode);
3454 }
3455
3456 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3457 {
3458         struct nfs_server *server = NFS_SERVER(data->header->inode);
3459
3460         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3461                 rpc_restart_call_prepare(task);
3462                 return -EAGAIN;
3463         }
3464
3465         __nfs4_read_done_cb(data);
3466         if (task->tk_status > 0)
3467                 renew_lease(server, data->timestamp);
3468         return 0;
3469 }
3470
3471 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3472 {
3473
3474         dprintk("--> %s\n", __func__);
3475
3476         if (!nfs4_sequence_done(task, &data->res.seq_res))
3477                 return -EAGAIN;
3478
3479         return data->read_done_cb ? data->read_done_cb(task, data) :
3480                                     nfs4_read_done_cb(task, data);
3481 }
3482
3483 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3484 {
3485         data->timestamp   = jiffies;
3486         data->read_done_cb = nfs4_read_done_cb;
3487         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3488         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3489 }
3490
3491 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3492 {
3493         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3494                                 &data->args.seq_args,
3495                                 &data->res.seq_res,
3496                                 task))
3497                 return;
3498         rpc_call_start(task);
3499 }
3500
3501 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3502 {
3503         struct inode *inode = data->header->inode;
3504
3505         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3506                 rpc_restart_call_prepare(task);
3507                 return -EAGAIN;
3508         }
3509         if (task->tk_status >= 0) {
3510                 renew_lease(NFS_SERVER(inode), data->timestamp);
3511                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3512         }
3513         return 0;
3514 }
3515
3516 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3517 {
3518         if (!nfs4_sequence_done(task, &data->res.seq_res))
3519                 return -EAGAIN;
3520         return data->write_done_cb ? data->write_done_cb(task, data) :
3521                 nfs4_write_done_cb(task, data);
3522 }
3523
3524 static
3525 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3526 {
3527         const struct nfs_pgio_header *hdr = data->header;
3528
3529         /* Don't request attributes for pNFS or O_DIRECT writes */
3530         if (data->ds_clp != NULL || hdr->dreq != NULL)
3531                 return false;
3532         /* Otherwise, request attributes if and only if we don't hold
3533          * a delegation
3534          */
3535         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3536 }
3537
3538 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3539 {
3540         struct nfs_server *server = NFS_SERVER(data->header->inode);
3541
3542         if (!nfs4_write_need_cache_consistency_data(data)) {
3543                 data->args.bitmask = NULL;
3544                 data->res.fattr = NULL;
3545         } else
3546                 data->args.bitmask = server->cache_consistency_bitmask;
3547
3548         if (!data->write_done_cb)
3549                 data->write_done_cb = nfs4_write_done_cb;
3550         data->res.server = server;
3551         data->timestamp   = jiffies;
3552
3553         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3554         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3555 }
3556
3557 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3558 {
3559         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3560                                 &data->args.seq_args,
3561                                 &data->res.seq_res,
3562                                 task))
3563                 return;
3564         rpc_call_start(task);
3565 }
3566
3567 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3568 {
3569         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3570                                 &data->args.seq_args,
3571                                 &data->res.seq_res,
3572                                 task))
3573                 return;
3574         rpc_call_start(task);
3575 }
3576
3577 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3578 {
3579         struct inode *inode = data->inode;
3580
3581         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3582                 rpc_restart_call_prepare(task);
3583                 return -EAGAIN;
3584         }
3585         return 0;
3586 }
3587
3588 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3589 {
3590         if (!nfs4_sequence_done(task, &data->res.seq_res))
3591                 return -EAGAIN;
3592         return data->commit_done_cb(task, data);
3593 }
3594
3595 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3596 {
3597         struct nfs_server *server = NFS_SERVER(data->inode);
3598
3599         if (data->commit_done_cb == NULL)
3600                 data->commit_done_cb = nfs4_commit_done_cb;
3601         data->res.server = server;
3602         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3603         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3604 }
3605
3606 struct nfs4_renewdata {
3607         struct nfs_client       *client;
3608         unsigned long           timestamp;
3609 };
3610
3611 /*
3612  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3613  * standalone procedure for queueing an asynchronous RENEW.
3614  */
3615 static void nfs4_renew_release(void *calldata)
3616 {
3617         struct nfs4_renewdata *data = calldata;
3618         struct nfs_client *clp = data->client;
3619
3620         if (atomic_read(&clp->cl_count) > 1)
3621                 nfs4_schedule_state_renewal(clp);
3622         nfs_put_client(clp);
3623         kfree(data);
3624 }
3625
3626 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3627 {
3628         struct nfs4_renewdata *data = calldata;
3629         struct nfs_client *clp = data->client;
3630         unsigned long timestamp = data->timestamp;
3631
3632         if (task->tk_status < 0) {
3633                 /* Unless we're shutting down, schedule state recovery! */
3634                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3635                         return;
3636                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3637                         nfs4_schedule_lease_recovery(clp);
3638                         return;
3639                 }
3640                 nfs4_schedule_path_down_recovery(clp);
3641         }
3642         do_renew_lease(clp, timestamp);
3643 }
3644
3645 static const struct rpc_call_ops nfs4_renew_ops = {
3646         .rpc_call_done = nfs4_renew_done,
3647         .rpc_release = nfs4_renew_release,
3648 };
3649
3650 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3651 {
3652         struct rpc_message msg = {
3653                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3654                 .rpc_argp       = clp,
3655                 .rpc_cred       = cred,
3656         };
3657         struct nfs4_renewdata *data;
3658
3659         if (renew_flags == 0)
3660                 return 0;
3661         if (!atomic_inc_not_zero(&clp->cl_count))
3662                 return -EIO;
3663         data = kmalloc(sizeof(*data), GFP_NOFS);
3664         if (data == NULL)
3665                 return -ENOMEM;
3666         data->client = clp;
3667         data->timestamp = jiffies;
3668         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3669                         &nfs4_renew_ops, data);
3670 }
3671
3672 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3673 {
3674         struct rpc_message msg = {
3675                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3676                 .rpc_argp       = clp,
3677                 .rpc_cred       = cred,
3678         };
3679         unsigned long now = jiffies;
3680         int status;
3681
3682         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3683         if (status < 0)
3684                 return status;
3685         do_renew_lease(clp, now);
3686         return 0;
3687 }
3688
3689 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3690 {
3691         return (server->caps & NFS_CAP_ACLS)
3692                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3693                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3694 }
3695
3696 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3697  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3698  * the stack.
3699  */
3700 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3701
3702 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3703                 struct page **pages, unsigned int *pgbase)
3704 {
3705         struct page *newpage, **spages;
3706         int rc = 0;
3707         size_t len;
3708         spages = pages;
3709
3710         do {
3711                 len = min_t(size_t, PAGE_SIZE, buflen);
3712                 newpage = alloc_page(GFP_KERNEL);
3713
3714                 if (newpage == NULL)
3715                         goto unwind;
3716                 memcpy(page_address(newpage), buf, len);
3717                 buf += len;
3718                 buflen -= len;
3719                 *pages++ = newpage;
3720                 rc++;
3721         } while (buflen != 0);
3722
3723         return rc;
3724
3725 unwind:
3726         for(; rc > 0; rc--)
3727                 __free_page(spages[rc-1]);
3728         return -ENOMEM;
3729 }
3730
3731 struct nfs4_cached_acl {
3732         int cached;
3733         size_t len;
3734         char data[0];
3735 };
3736
3737 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3738 {
3739         struct nfs_inode *nfsi = NFS_I(inode);
3740
3741         spin_lock(&inode->i_lock);
3742         kfree(nfsi->nfs4_acl);
3743         nfsi->nfs4_acl = acl;
3744         spin_unlock(&inode->i_lock);
3745 }
3746
3747 static void nfs4_zap_acl_attr(struct inode *inode)
3748 {
3749         nfs4_set_cached_acl(inode, NULL);
3750 }
3751
3752 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3753 {
3754         struct nfs_inode *nfsi = NFS_I(inode);
3755         struct nfs4_cached_acl *acl;
3756         int ret = -ENOENT;
3757
3758         spin_lock(&inode->i_lock);
3759         acl = nfsi->nfs4_acl;
3760         if (acl == NULL)
3761                 goto out;
3762         if (buf == NULL) /* user is just asking for length */
3763                 goto out_len;
3764         if (acl->cached == 0)
3765                 goto out;
3766         ret = -ERANGE; /* see getxattr(2) man page */
3767         if (acl->len > buflen)
3768                 goto out;
3769         memcpy(buf, acl->data, acl->len);
3770 out_len:
3771         ret = acl->len;
3772 out:
3773         spin_unlock(&inode->i_lock);
3774         return ret;
3775 }
3776
3777 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3778 {
3779         struct nfs4_cached_acl *acl;
3780         size_t buflen = sizeof(*acl) + acl_len;
3781
3782         if (buflen <= PAGE_SIZE) {
3783                 acl = kmalloc(buflen, GFP_KERNEL);
3784                 if (acl == NULL)
3785                         goto out;
3786                 acl->cached = 1;
3787                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3788         } else {
3789                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3790                 if (acl == NULL)
3791                         goto out;
3792                 acl->cached = 0;
3793         }
3794         acl->len = acl_len;
3795 out:
3796         nfs4_set_cached_acl(inode, acl);
3797 }
3798
3799 /*
3800  * The getxattr API returns the required buffer length when called with a
3801  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3802  * the required buf.  On a NULL buf, we send a page of data to the server
3803  * guessing that the ACL request can be serviced by a page. If so, we cache
3804  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3805  * the cache. If not so, we throw away the page, and cache the required
3806  * length. The next getxattr call will then produce another round trip to
3807  * the server, this time with the input buf of the required size.
3808  */
3809 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3810 {
3811         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3812         struct nfs_getaclargs args = {
3813                 .fh = NFS_FH(inode),
3814                 .acl_pages = pages,
3815                 .acl_len = buflen,
3816         };
3817         struct nfs_getaclres res = {
3818                 .acl_len = buflen,
3819         };
3820         struct rpc_message msg = {
3821                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3822                 .rpc_argp = &args,
3823                 .rpc_resp = &res,
3824         };
3825         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3826         int ret = -ENOMEM, i;
3827
3828         /* As long as we're doing a round trip to the server anyway,
3829          * let's be prepared for a page of acl data. */
3830         if (npages == 0)
3831                 npages = 1;
3832         if (npages > ARRAY_SIZE(pages))
3833                 return -ERANGE;
3834
3835         for (i = 0; i < npages; i++) {
3836                 pages[i] = alloc_page(GFP_KERNEL);
3837                 if (!pages[i])
3838                         goto out_free;
3839         }
3840
3841         /* for decoding across pages */
3842         res.acl_scratch = alloc_page(GFP_KERNEL);
3843         if (!res.acl_scratch)
3844                 goto out_free;
3845
3846         args.acl_len = npages * PAGE_SIZE;
3847         args.acl_pgbase = 0;
3848
3849         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3850                 __func__, buf, buflen, npages, args.acl_len);
3851         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3852                              &msg, &args.seq_args, &res.seq_res, 0);
3853         if (ret)
3854                 goto out_free;
3855
3856         /* Handle the case where the passed-in buffer is too short */
3857         if (res.acl_flags & NFS4_ACL_TRUNC) {
3858                 /* Did the user only issue a request for the acl length? */
3859                 if (buf == NULL)
3860                         goto out_ok;
3861                 ret = -ERANGE;
3862                 goto out_free;
3863         }
3864         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3865         if (buf)
3866                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3867 out_ok:
3868         ret = res.acl_len;
3869 out_free:
3870         for (i = 0; i < npages; i++)
3871                 if (pages[i])
3872                         __free_page(pages[i]);
3873         if (res.acl_scratch)
3874                 __free_page(res.acl_scratch);
3875         return ret;
3876 }
3877
3878 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3879 {
3880         struct nfs4_exception exception = { };
3881         ssize_t ret;
3882         do {
3883                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3884                 if (ret >= 0)
3885                         break;
3886                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3887         } while (exception.retry);
3888         return ret;
3889 }
3890
3891 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3892 {
3893         struct nfs_server *server = NFS_SERVER(inode);
3894         int ret;
3895
3896         if (!nfs4_server_supports_acls(server))
3897                 return -EOPNOTSUPP;
3898         ret = nfs_revalidate_inode(server, inode);
3899         if (ret < 0)
3900                 return ret;
3901         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3902                 nfs_zap_acl_cache(inode);
3903         ret = nfs4_read_cached_acl(inode, buf, buflen);
3904         if (ret != -ENOENT)
3905                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3906                  * but no cached acl data, just the acl length */
3907                 return ret;
3908         return nfs4_get_acl_uncached(inode, buf, buflen);
3909 }
3910
3911 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3912 {
3913         struct nfs_server *server = NFS_SERVER(inode);
3914         struct page *pages[NFS4ACL_MAXPAGES];
3915         struct nfs_setaclargs arg = {
3916                 .fh             = NFS_FH(inode),
3917                 .acl_pages      = pages,
3918                 .acl_len        = buflen,
3919         };
3920         struct nfs_setaclres res;
3921         struct rpc_message msg = {
3922                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3923                 .rpc_argp       = &arg,
3924                 .rpc_resp       = &res,
3925         };
3926         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3927         int ret, i;
3928
3929         if (!nfs4_server_supports_acls(server))
3930                 return -EOPNOTSUPP;
3931         if (npages > ARRAY_SIZE(pages))
3932                 return -ERANGE;
3933         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3934         if (i < 0)
3935                 return i;
3936         nfs4_inode_return_delegation(inode);
3937         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3938
3939         /*
3940          * Free each page after tx, so the only ref left is
3941          * held by the network stack
3942          */
3943         for (; i > 0; i--)
3944                 put_page(pages[i-1]);
3945
3946         /*
3947          * Acl update can result in inode attribute update.
3948          * so mark the attribute cache invalid.
3949          */
3950         spin_lock(&inode->i_lock);
3951         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3952         spin_unlock(&inode->i_lock);
3953         nfs_access_zap_cache(inode);
3954         nfs_zap_acl_cache(inode);
3955         return ret;
3956 }
3957
3958 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3959 {
3960         struct nfs4_exception exception = { };
3961         int err;
3962         do {
3963                 err = nfs4_handle_exception(NFS_SERVER(inode),
3964                                 __nfs4_proc_set_acl(inode, buf, buflen),
3965                                 &exception);
3966         } while (exception.retry);
3967         return err;
3968 }
3969
3970 static int
3971 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3972 {
3973         struct nfs_client *clp = server->nfs_client;
3974
3975         if (task->tk_status >= 0)
3976                 return 0;
3977         switch(task->tk_status) {
3978                 case -NFS4ERR_DELEG_REVOKED:
3979                 case -NFS4ERR_ADMIN_REVOKED:
3980                 case -NFS4ERR_BAD_STATEID:
3981                         if (state == NULL)
3982                                 break;
3983                         nfs_remove_bad_delegation(state->inode);
3984                 case -NFS4ERR_OPENMODE:
3985                         if (state == NULL)
3986                                 break;
3987                         nfs4_schedule_stateid_recovery(server, state);
3988                         goto wait_on_recovery;
3989                 case -NFS4ERR_EXPIRED:
3990                         if (state != NULL)
3991                                 nfs4_schedule_stateid_recovery(server, state);
3992                 case -NFS4ERR_STALE_STATEID:
3993                 case -NFS4ERR_STALE_CLIENTID:
3994                         nfs4_schedule_lease_recovery(clp);
3995                         goto wait_on_recovery;
3996 #if defined(CONFIG_NFS_V4_1)
3997                 case -NFS4ERR_BADSESSION:
3998                 case -NFS4ERR_BADSLOT:
3999                 case -NFS4ERR_BAD_HIGH_SLOT:
4000                 case -NFS4ERR_DEADSESSION:
4001                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4002                 case -NFS4ERR_SEQ_FALSE_RETRY:
4003                 case -NFS4ERR_SEQ_MISORDERED:
4004                         dprintk("%s ERROR %d, Reset session\n", __func__,
4005                                 task->tk_status);
4006                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4007                         task->tk_status = 0;
4008                         return -EAGAIN;
4009 #endif /* CONFIG_NFS_V4_1 */
4010                 case -NFS4ERR_DELAY:
4011                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4012                 case -NFS4ERR_GRACE:
4013                 case -EKEYEXPIRED:
4014                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4015                         task->tk_status = 0;
4016                         return -EAGAIN;
4017                 case -NFS4ERR_RETRY_UNCACHED_REP:
4018                 case -NFS4ERR_OLD_STATEID:
4019                         task->tk_status = 0;
4020                         return -EAGAIN;
4021         }
4022         task->tk_status = nfs4_map_errors(task->tk_status);
4023         return 0;
4024 wait_on_recovery:
4025         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4026         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4027                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4028         task->tk_status = 0;
4029         return -EAGAIN;
4030 }
4031
4032 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4033                                     nfs4_verifier *bootverf)
4034 {
4035         __be32 verf[2];
4036
4037         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4038                 /* An impossible timestamp guarantees this value
4039                  * will never match a generated boot time. */
4040                 verf[0] = 0;
4041                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4042         } else {
4043                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4044                 verf[0] = (__be32)nn->boot_time.tv_sec;
4045                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4046         }
4047         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4048 }
4049
4050 /**
4051  * nfs4_proc_setclientid - Negotiate client ID
4052  * @clp: state data structure
4053  * @program: RPC program for NFSv4 callback service
4054  * @port: IP port number for NFS4 callback service
4055  * @cred: RPC credential to use for this call
4056  * @res: where to place the result
4057  *
4058  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4059  */
4060 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4061                 unsigned short port, struct rpc_cred *cred,
4062                 struct nfs4_setclientid_res *res)
4063 {
4064         nfs4_verifier sc_verifier;
4065         struct nfs4_setclientid setclientid = {
4066                 .sc_verifier = &sc_verifier,
4067                 .sc_prog = program,
4068                 .sc_cb_ident = clp->cl_cb_ident,
4069         };
4070         struct rpc_message msg = {
4071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4072                 .rpc_argp = &setclientid,
4073                 .rpc_resp = res,
4074                 .rpc_cred = cred,
4075         };
4076         int status;
4077
4078         /* nfs_client_id4 */
4079         nfs4_init_boot_verifier(clp, &sc_verifier);
4080         rcu_read_lock();
4081         setclientid.sc_name_len = scnprintf(setclientid.sc_name,
4082                         sizeof(setclientid.sc_name), "%s/%s %s",
4083                         clp->cl_ipaddr,
4084                         rpc_peeraddr2str(clp->cl_rpcclient,
4085                                                 RPC_DISPLAY_ADDR),
4086                         rpc_peeraddr2str(clp->cl_rpcclient,
4087                                                 RPC_DISPLAY_PROTO));
4088         /* cb_client4 */
4089         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4090                                 sizeof(setclientid.sc_netid),
4091                                 rpc_peeraddr2str(clp->cl_rpcclient,
4092                                                         RPC_DISPLAY_NETID));
4093         rcu_read_unlock();
4094         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4095                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4096                                 clp->cl_ipaddr, port >> 8, port & 255);
4097
4098         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4099                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4100                 setclientid.sc_name_len, setclientid.sc_name);
4101         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4102         dprintk("NFS reply setclientid: %d\n", status);
4103         return status;
4104 }
4105
4106 /**
4107  * nfs4_proc_setclientid_confirm - Confirm client ID
4108  * @clp: state data structure
4109  * @res: result of a previous SETCLIENTID
4110  * @cred: RPC credential to use for this call
4111  *
4112  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4113  */
4114 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4115                 struct nfs4_setclientid_res *arg,
4116                 struct rpc_cred *cred)
4117 {
4118         struct nfs_fsinfo fsinfo;
4119         struct rpc_message msg = {
4120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4121                 .rpc_argp = arg,
4122                 .rpc_resp = &fsinfo,
4123                 .rpc_cred = cred,
4124         };
4125         unsigned long now;
4126         int status;
4127
4128         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4129                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4130                 clp->cl_clientid);
4131         now = jiffies;
4132         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4133         if (status == 0) {
4134                 spin_lock(&clp->cl_lock);
4135                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4136                 clp->cl_last_renewal = now;
4137                 spin_unlock(&clp->cl_lock);
4138         }
4139         dprintk("NFS reply setclientid_confirm: %d\n", status);
4140         return status;
4141 }
4142
4143 struct nfs4_delegreturndata {
4144         struct nfs4_delegreturnargs args;
4145         struct nfs4_delegreturnres res;
4146         struct nfs_fh fh;
4147         nfs4_stateid stateid;
4148         unsigned long timestamp;
4149         struct nfs_fattr fattr;
4150         int rpc_status;
4151 };
4152
4153 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4154 {
4155         struct nfs4_delegreturndata *data = calldata;
4156
4157         if (!nfs4_sequence_done(task, &data->res.seq_res))
4158                 return;
4159
4160         switch (task->tk_status) {
4161         case -NFS4ERR_STALE_STATEID:
4162         case -NFS4ERR_EXPIRED:
4163         case 0:
4164                 renew_lease(data->res.server, data->timestamp);
4165                 break;
4166         default:
4167                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4168                                 -EAGAIN) {
4169                         rpc_restart_call_prepare(task);
4170                         return;
4171                 }
4172         }
4173         data->rpc_status = task->tk_status;
4174 }
4175
4176 static void nfs4_delegreturn_release(void *calldata)
4177 {
4178         kfree(calldata);
4179 }
4180
4181 #if defined(CONFIG_NFS_V4_1)
4182 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4183 {
4184         struct nfs4_delegreturndata *d_data;
4185
4186         d_data = (struct nfs4_delegreturndata *)data;
4187
4188         if (nfs4_setup_sequence(d_data->res.server,
4189                                 &d_data->args.seq_args,
4190                                 &d_data->res.seq_res, task))
4191                 return;
4192         rpc_call_start(task);
4193 }
4194 #endif /* CONFIG_NFS_V4_1 */
4195
4196 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4197 #if defined(CONFIG_NFS_V4_1)
4198         .rpc_call_prepare = nfs4_delegreturn_prepare,
4199 #endif /* CONFIG_NFS_V4_1 */
4200         .rpc_call_done = nfs4_delegreturn_done,
4201         .rpc_release = nfs4_delegreturn_release,
4202 };
4203
4204 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4205 {
4206         struct nfs4_delegreturndata *data;
4207         struct nfs_server *server = NFS_SERVER(inode);
4208         struct rpc_task *task;
4209         struct rpc_message msg = {
4210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4211                 .rpc_cred = cred,
4212         };
4213         struct rpc_task_setup task_setup_data = {
4214                 .rpc_client = server->client,
4215                 .rpc_message = &msg,
4216                 .callback_ops = &nfs4_delegreturn_ops,
4217                 .flags = RPC_TASK_ASYNC,
4218         };
4219         int status = 0;
4220
4221         data = kzalloc(sizeof(*data), GFP_NOFS);
4222         if (data == NULL)
4223                 return -ENOMEM;
4224         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4225         data->args.fhandle = &data->fh;
4226         data->args.stateid = &data->stateid;
4227         data->args.bitmask = server->cache_consistency_bitmask;
4228         nfs_copy_fh(&data->fh, NFS_FH(inode));
4229         nfs4_stateid_copy(&data->stateid, stateid);
4230         data->res.fattr = &data->fattr;
4231         data->res.server = server;
4232         nfs_fattr_init(data->res.fattr);
4233         data->timestamp = jiffies;
4234         data->rpc_status = 0;
4235
4236         task_setup_data.callback_data = data;
4237         msg.rpc_argp = &data->args;
4238         msg.rpc_resp = &data->res;
4239         task = rpc_run_task(&task_setup_data);
4240         if (IS_ERR(task))
4241                 return PTR_ERR(task);
4242         if (!issync)
4243                 goto out;
4244         status = nfs4_wait_for_completion_rpc_task(task);
4245         if (status != 0)
4246                 goto out;
4247         status = data->rpc_status;
4248         if (status == 0)
4249                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4250         else
4251                 nfs_refresh_inode(inode, &data->fattr);
4252 out:
4253         rpc_put_task(task);
4254         return status;
4255 }
4256
4257 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4258 {
4259         struct nfs_server *server = NFS_SERVER(inode);
4260         struct nfs4_exception exception = { };
4261         int err;
4262         do {
4263                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4264                 switch (err) {
4265                         case -NFS4ERR_STALE_STATEID:
4266                         case -NFS4ERR_EXPIRED:
4267                         case 0:
4268                                 return 0;
4269                 }
4270                 err = nfs4_handle_exception(server, err, &exception);
4271         } while (exception.retry);
4272         return err;
4273 }
4274
4275 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4276 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4277
4278 /*
4279  * sleep, with exponential backoff, and retry the LOCK operation.
4280  */
4281 static unsigned long
4282 nfs4_set_lock_task_retry(unsigned long timeout)
4283 {
4284         freezable_schedule_timeout_killable(timeout);
4285         timeout <<= 1;
4286         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4287                 return NFS4_LOCK_MAXTIMEOUT;
4288         return timeout;
4289 }
4290
4291 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4292 {
4293         struct inode *inode = state->inode;
4294         struct nfs_server *server = NFS_SERVER(inode);
4295         struct nfs_client *clp = server->nfs_client;
4296         struct nfs_lockt_args arg = {
4297                 .fh = NFS_FH(inode),
4298                 .fl = request,
4299         };
4300         struct nfs_lockt_res res = {
4301                 .denied = request,
4302         };
4303         struct rpc_message msg = {
4304                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4305                 .rpc_argp       = &arg,
4306                 .rpc_resp       = &res,
4307                 .rpc_cred       = state->owner->so_cred,
4308         };
4309         struct nfs4_lock_state *lsp;
4310         int status;
4311
4312         arg.lock_owner.clientid = clp->cl_clientid;
4313         status = nfs4_set_lock_state(state, request);
4314         if (status != 0)
4315                 goto out;
4316         lsp = request->fl_u.nfs4_fl.owner;
4317         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4318         arg.lock_owner.s_dev = server->s_dev;
4319         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4320         switch (status) {
4321                 case 0:
4322                         request->fl_type = F_UNLCK;
4323                         break;
4324                 case -NFS4ERR_DENIED:
4325                         status = 0;
4326         }
4327         request->fl_ops->fl_release_private(request);
4328 out:
4329         return status;
4330 }
4331
4332 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4333 {
4334         struct nfs4_exception exception = { };
4335         int err;
4336
4337         do {
4338                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4339                                 _nfs4_proc_getlk(state, cmd, request),
4340                                 &exception);
4341         } while (exception.retry);
4342         return err;
4343 }
4344
4345 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4346 {
4347         int res = 0;
4348         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4349                 case FL_POSIX:
4350                         res = posix_lock_file_wait(file, fl);
4351                         break;
4352                 case FL_FLOCK:
4353                         res = flock_lock_file_wait(file, fl);
4354                         break;
4355                 default:
4356                         BUG();
4357         }
4358         return res;
4359 }
4360
4361 struct nfs4_unlockdata {
4362         struct nfs_locku_args arg;
4363         struct nfs_locku_res res;
4364         struct nfs4_lock_state *lsp;
4365         struct nfs_open_context *ctx;
4366         struct file_lock fl;
4367         const struct nfs_server *server;
4368         unsigned long timestamp;
4369 };
4370
4371 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4372                 struct nfs_open_context *ctx,
4373                 struct nfs4_lock_state *lsp,
4374                 struct nfs_seqid *seqid)
4375 {
4376         struct nfs4_unlockdata *p;
4377         struct inode *inode = lsp->ls_state->inode;
4378
4379         p = kzalloc(sizeof(*p), GFP_NOFS);
4380         if (p == NULL)
4381                 return NULL;
4382         p->arg.fh = NFS_FH(inode);
4383         p->arg.fl = &p->fl;
4384         p->arg.seqid = seqid;
4385         p->res.seqid = seqid;
4386         p->arg.stateid = &lsp->ls_stateid;
4387         p->lsp = lsp;
4388         atomic_inc(&lsp->ls_count);
4389         /* Ensure we don't close file until we're done freeing locks! */
4390         p->ctx = get_nfs_open_context(ctx);
4391         memcpy(&p->fl, fl, sizeof(p->fl));
4392         p->server = NFS_SERVER(inode);
4393         return p;
4394 }
4395
4396 static void nfs4_locku_release_calldata(void *data)
4397 {
4398         struct nfs4_unlockdata *calldata = data;
4399         nfs_free_seqid(calldata->arg.seqid);
4400         nfs4_put_lock_state(calldata->lsp);
4401         put_nfs_open_context(calldata->ctx);
4402         kfree(calldata);
4403 }
4404
4405 static void nfs4_locku_done(struct rpc_task *task, void *data)
4406 {
4407         struct nfs4_unlockdata *calldata = data;
4408
4409         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4410                 return;
4411         switch (task->tk_status) {
4412                 case 0:
4413                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4414                                         &calldata->res.stateid);
4415                         renew_lease(calldata->server, calldata->timestamp);
4416                         break;
4417                 case -NFS4ERR_BAD_STATEID:
4418                 case -NFS4ERR_OLD_STATEID:
4419                 case -NFS4ERR_STALE_STATEID:
4420                 case -NFS4ERR_EXPIRED:
4421                         break;
4422                 default:
4423                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4424                                 rpc_restart_call_prepare(task);
4425         }
4426 }
4427
4428 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4429 {
4430         struct nfs4_unlockdata *calldata = data;
4431
4432         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4433                 return;
4434         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4435                 /* Note: exit _without_ running nfs4_locku_done */
4436                 task->tk_action = NULL;
4437                 return;
4438         }
4439         calldata->timestamp = jiffies;
4440         if (nfs4_setup_sequence(calldata->server,
4441                                 &calldata->arg.seq_args,
4442                                 &calldata->res.seq_res, task))
4443                 return;
4444         rpc_call_start(task);
4445 }
4446
4447 static const struct rpc_call_ops nfs4_locku_ops = {
4448         .rpc_call_prepare = nfs4_locku_prepare,
4449         .rpc_call_done = nfs4_locku_done,
4450         .rpc_release = nfs4_locku_release_calldata,
4451 };
4452
4453 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4454                 struct nfs_open_context *ctx,
4455                 struct nfs4_lock_state *lsp,
4456                 struct nfs_seqid *seqid)
4457 {
4458         struct nfs4_unlockdata *data;
4459         struct rpc_message msg = {
4460                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4461                 .rpc_cred = ctx->cred,
4462         };
4463         struct rpc_task_setup task_setup_data = {
4464                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4465                 .rpc_message = &msg,
4466                 .callback_ops = &nfs4_locku_ops,
4467                 .workqueue = nfsiod_workqueue,
4468                 .flags = RPC_TASK_ASYNC,
4469         };
4470
4471         /* Ensure this is an unlock - when canceling a lock, the
4472          * canceled lock is passed in, and it won't be an unlock.
4473          */
4474         fl->fl_type = F_UNLCK;
4475
4476         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4477         if (data == NULL) {
4478                 nfs_free_seqid(seqid);
4479                 return ERR_PTR(-ENOMEM);
4480         }
4481
4482         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4483         msg.rpc_argp = &data->arg;
4484         msg.rpc_resp = &data->res;
4485         task_setup_data.callback_data = data;
4486         return rpc_run_task(&task_setup_data);
4487 }
4488
4489 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4490 {
4491         struct nfs_inode *nfsi = NFS_I(state->inode);
4492         struct nfs_seqid *seqid;
4493         struct nfs4_lock_state *lsp;
4494         struct rpc_task *task;
4495         int status = 0;
4496         unsigned char fl_flags = request->fl_flags;
4497
4498         status = nfs4_set_lock_state(state, request);
4499         /* Unlock _before_ we do the RPC call */
4500         request->fl_flags |= FL_EXISTS;
4501         down_read(&nfsi->rwsem);
4502         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4503                 up_read(&nfsi->rwsem);
4504                 goto out;
4505         }
4506         up_read(&nfsi->rwsem);
4507         if (status != 0)
4508                 goto out;
4509         /* Is this a delegated lock? */
4510         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4511                 goto out;
4512         lsp = request->fl_u.nfs4_fl.owner;
4513         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4514         status = -ENOMEM;
4515         if (seqid == NULL)
4516                 goto out;
4517         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4518         status = PTR_ERR(task);
4519         if (IS_ERR(task))
4520                 goto out;
4521         status = nfs4_wait_for_completion_rpc_task(task);
4522         rpc_put_task(task);
4523 out:
4524         request->fl_flags = fl_flags;
4525         return status;
4526 }
4527
4528 struct nfs4_lockdata {
4529         struct nfs_lock_args arg;
4530         struct nfs_lock_res res;
4531         struct nfs4_lock_state *lsp;
4532         struct nfs_open_context *ctx;
4533         struct file_lock fl;
4534         unsigned long timestamp;
4535         int rpc_status;
4536         int cancelled;
4537         struct nfs_server *server;
4538 };
4539
4540 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4541                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4542                 gfp_t gfp_mask)
4543 {
4544         struct nfs4_lockdata *p;
4545         struct inode *inode = lsp->ls_state->inode;
4546         struct nfs_server *server = NFS_SERVER(inode);
4547
4548         p = kzalloc(sizeof(*p), gfp_mask);
4549         if (p == NULL)
4550                 return NULL;
4551
4552         p->arg.fh = NFS_FH(inode);
4553         p->arg.fl = &p->fl;
4554         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4555         if (p->arg.open_seqid == NULL)
4556                 goto out_free;
4557         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4558         if (p->arg.lock_seqid == NULL)
4559                 goto out_free_seqid;
4560         p->arg.lock_stateid = &lsp->ls_stateid;
4561         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4562         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4563         p->arg.lock_owner.s_dev = server->s_dev;
4564         p->res.lock_seqid = p->arg.lock_seqid;
4565         p->lsp = lsp;
4566         p->server = server;
4567         atomic_inc(&lsp->ls_count);
4568         p->ctx = get_nfs_open_context(ctx);
4569         memcpy(&p->fl, fl, sizeof(p->fl));
4570         return p;
4571 out_free_seqid:
4572         nfs_free_seqid(p->arg.open_seqid);
4573 out_free:
4574         kfree(p);
4575         return NULL;
4576 }
4577
4578 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4579 {
4580         struct nfs4_lockdata *data = calldata;
4581         struct nfs4_state *state = data->lsp->ls_state;
4582
4583         dprintk("%s: begin!\n", __func__);
4584         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4585                 return;
4586         /* Do we need to do an open_to_lock_owner? */
4587         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4588                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4589                         return;
4590                 data->arg.open_stateid = &state->stateid;
4591                 data->arg.new_lock_owner = 1;
4592                 data->res.open_seqid = data->arg.open_seqid;
4593         } else
4594                 data->arg.new_lock_owner = 0;
4595         data->timestamp = jiffies;
4596         if (nfs4_setup_sequence(data->server,
4597                                 &data->arg.seq_args,
4598                                 &data->res.seq_res, task))
4599                 return;
4600         rpc_call_start(task);
4601         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4602 }
4603
4604 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4605 {
4606         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4607         nfs4_lock_prepare(task, calldata);
4608 }
4609
4610 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4611 {
4612         struct nfs4_lockdata *data = calldata;
4613
4614         dprintk("%s: begin!\n", __func__);
4615
4616         if (!nfs4_sequence_done(task, &data->res.seq_res))
4617                 return;
4618
4619         data->rpc_status = task->tk_status;
4620         if (data->arg.new_lock_owner != 0) {
4621                 if (data->rpc_status == 0)
4622                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4623                 else
4624                         goto out;
4625         }
4626         if (data->rpc_status == 0) {
4627                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4628                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4629                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4630         }
4631 out:
4632         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4633 }
4634
4635 static void nfs4_lock_release(void *calldata)
4636 {
4637         struct nfs4_lockdata *data = calldata;
4638
4639         dprintk("%s: begin!\n", __func__);
4640         nfs_free_seqid(data->arg.open_seqid);
4641         if (data->cancelled != 0) {
4642                 struct rpc_task *task;
4643                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4644                                 data->arg.lock_seqid);
4645                 if (!IS_ERR(task))
4646                         rpc_put_task_async(task);
4647                 dprintk("%s: cancelling lock!\n", __func__);
4648         } else
4649                 nfs_free_seqid(data->arg.lock_seqid);
4650         nfs4_put_lock_state(data->lsp);
4651         put_nfs_open_context(data->ctx);
4652         kfree(data);
4653         dprintk("%s: done!\n", __func__);
4654 }
4655
4656 static const struct rpc_call_ops nfs4_lock_ops = {
4657         .rpc_call_prepare = nfs4_lock_prepare,
4658         .rpc_call_done = nfs4_lock_done,
4659         .rpc_release = nfs4_lock_release,
4660 };
4661
4662 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4663         .rpc_call_prepare = nfs4_recover_lock_prepare,
4664         .rpc_call_done = nfs4_lock_done,
4665         .rpc_release = nfs4_lock_release,
4666 };
4667
4668 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4669 {
4670         switch (error) {
4671         case -NFS4ERR_ADMIN_REVOKED:
4672         case -NFS4ERR_BAD_STATEID:
4673                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4674                 if (new_lock_owner != 0 ||
4675                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4676                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4677                 break;
4678         case -NFS4ERR_STALE_STATEID:
4679                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4680         case -NFS4ERR_EXPIRED:
4681                 nfs4_schedule_lease_recovery(server->nfs_client);
4682         };
4683 }
4684
4685 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4686 {
4687         struct nfs4_lockdata *data;
4688         struct rpc_task *task;
4689         struct rpc_message msg = {
4690                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4691                 .rpc_cred = state->owner->so_cred,
4692         };
4693         struct rpc_task_setup task_setup_data = {
4694                 .rpc_client = NFS_CLIENT(state->inode),
4695                 .rpc_message = &msg,
4696                 .callback_ops = &nfs4_lock_ops,
4697                 .workqueue = nfsiod_workqueue,
4698                 .flags = RPC_TASK_ASYNC,
4699         };
4700         int ret;
4701
4702         dprintk("%s: begin!\n", __func__);
4703         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4704                         fl->fl_u.nfs4_fl.owner,
4705                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4706         if (data == NULL)
4707                 return -ENOMEM;
4708         if (IS_SETLKW(cmd))
4709                 data->arg.block = 1;
4710         if (recovery_type > NFS_LOCK_NEW) {
4711                 if (recovery_type == NFS_LOCK_RECLAIM)
4712                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4713                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4714         }
4715         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4716         msg.rpc_argp = &data->arg;
4717         msg.rpc_resp = &data->res;
4718         task_setup_data.callback_data = data;
4719         task = rpc_run_task(&task_setup_data);
4720         if (IS_ERR(task))
4721                 return PTR_ERR(task);
4722         ret = nfs4_wait_for_completion_rpc_task(task);
4723         if (ret == 0) {
4724                 ret = data->rpc_status;
4725                 if (ret)
4726                         nfs4_handle_setlk_error(data->server, data->lsp,
4727                                         data->arg.new_lock_owner, ret);
4728         } else
4729                 data->cancelled = 1;
4730         rpc_put_task(task);
4731         dprintk("%s: done, ret = %d!\n", __func__, ret);
4732         return ret;
4733 }
4734
4735 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4736 {
4737         struct nfs_server *server = NFS_SERVER(state->inode);
4738         struct nfs4_exception exception = {
4739                 .inode = state->inode,
4740         };
4741         int err;
4742
4743         do {
4744                 /* Cache the lock if possible... */
4745                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4746                         return 0;
4747                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4748                 if (err != -NFS4ERR_DELAY)
4749                         break;
4750                 nfs4_handle_exception(server, err, &exception);
4751         } while (exception.retry);
4752         return err;
4753 }
4754
4755 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4756 {
4757         struct nfs_server *server = NFS_SERVER(state->inode);
4758         struct nfs4_exception exception = {
4759                 .inode = state->inode,
4760         };
4761         int err;
4762
4763         err = nfs4_set_lock_state(state, request);
4764         if (err != 0)
4765                 return err;
4766         do {
4767                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4768                         return 0;
4769                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4770                 switch (err) {
4771                 default:
4772                         goto out;
4773                 case -NFS4ERR_GRACE:
4774                 case -NFS4ERR_DELAY:
4775                         nfs4_handle_exception(server, err, &exception);
4776                         err = 0;
4777                 }
4778         } while (exception.retry);
4779 out:
4780         return err;
4781 }
4782
4783 #if defined(CONFIG_NFS_V4_1)
4784 /**
4785  * nfs41_check_expired_locks - possibly free a lock stateid
4786  *
4787  * @state: NFSv4 state for an inode
4788  *
4789  * Returns NFS_OK if recovery for this stateid is now finished.
4790  * Otherwise a negative NFS4ERR value is returned.
4791  */
4792 static int nfs41_check_expired_locks(struct nfs4_state *state)
4793 {
4794         int status, ret = -NFS4ERR_BAD_STATEID;
4795         struct nfs4_lock_state *lsp;
4796         struct nfs_server *server = NFS_SERVER(state->inode);
4797
4798         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4799                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4800                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4801                         if (status != NFS_OK) {
4802                                 /* Free the stateid unless the server
4803                                  * informs us the stateid is unrecognized. */
4804                                 if (status != -NFS4ERR_BAD_STATEID)
4805                                         nfs41_free_stateid(server,
4806                                                         &lsp->ls_stateid);
4807                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4808                                 ret = status;
4809                         }
4810                 }
4811         };
4812
4813         return ret;
4814 }
4815
4816 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4817 {
4818         int status = NFS_OK;
4819
4820         if (test_bit(LK_STATE_IN_USE, &state->flags))
4821                 status = nfs41_check_expired_locks(state);
4822         if (status != NFS_OK)
4823                 status = nfs4_lock_expired(state, request);
4824         return status;
4825 }
4826 #endif
4827
4828 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4829 {
4830         struct nfs_inode *nfsi = NFS_I(state->inode);
4831         unsigned char fl_flags = request->fl_flags;
4832         int status = -ENOLCK;
4833
4834         if ((fl_flags & FL_POSIX) &&
4835                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4836                 goto out;
4837         /* Is this a delegated open? */
4838         status = nfs4_set_lock_state(state, request);
4839         if (status != 0)
4840                 goto out;
4841         request->fl_flags |= FL_ACCESS;
4842         status = do_vfs_lock(request->fl_file, request);
4843         if (status < 0)
4844                 goto out;
4845         down_read(&nfsi->rwsem);
4846         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4847                 /* Yes: cache locks! */
4848                 /* ...but avoid races with delegation recall... */
4849                 request->fl_flags = fl_flags & ~FL_SLEEP;
4850                 status = do_vfs_lock(request->fl_file, request);
4851                 goto out_unlock;
4852         }
4853         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4854         if (status != 0)
4855                 goto out_unlock;
4856         /* Note: we always want to sleep here! */
4857         request->fl_flags = fl_flags | FL_SLEEP;
4858         if (do_vfs_lock(request->fl_file, request) < 0)
4859                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4860                         "manager!\n", __func__);
4861 out_unlock:
4862         up_read(&nfsi->rwsem);
4863 out:
4864         request->fl_flags = fl_flags;
4865         return status;
4866 }
4867
4868 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4869 {
4870         struct nfs4_exception exception = {
4871                 .state = state,
4872                 .inode = state->inode,
4873         };
4874         int err;
4875
4876         do {
4877                 err = _nfs4_proc_setlk(state, cmd, request);
4878                 if (err == -NFS4ERR_DENIED)
4879                         err = -EAGAIN;
4880                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4881                                 err, &exception);
4882         } while (exception.retry);
4883         return err;
4884 }
4885
4886 static int
4887 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4888 {
4889         struct nfs_open_context *ctx;
4890         struct nfs4_state *state;
4891         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4892         int status;
4893
4894         /* verify open state */
4895         ctx = nfs_file_open_context(filp);
4896         state = ctx->state;
4897
4898         if (request->fl_start < 0 || request->fl_end < 0)
4899                 return -EINVAL;
4900
4901         if (IS_GETLK(cmd)) {
4902                 if (state != NULL)
4903                         return nfs4_proc_getlk(state, F_GETLK, request);
4904                 return 0;
4905         }
4906
4907         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4908                 return -EINVAL;
4909
4910         if (request->fl_type == F_UNLCK) {
4911                 if (state != NULL)
4912                         return nfs4_proc_unlck(state, cmd, request);
4913                 return 0;
4914         }
4915
4916         if (state == NULL)
4917                 return -ENOLCK;
4918         /*
4919          * Don't rely on the VFS having checked the file open mode,
4920          * since it won't do this for flock() locks.
4921          */
4922         switch (request->fl_type) {
4923         case F_RDLCK:
4924                 if (!(filp->f_mode & FMODE_READ))
4925                         return -EBADF;
4926                 break;
4927         case F_WRLCK:
4928                 if (!(filp->f_mode & FMODE_WRITE))
4929                         return -EBADF;
4930         }
4931
4932         do {
4933                 status = nfs4_proc_setlk(state, cmd, request);
4934                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4935                         break;
4936                 timeout = nfs4_set_lock_task_retry(timeout);
4937                 status = -ERESTARTSYS;
4938                 if (signalled())
4939                         break;
4940         } while(status < 0);
4941         return status;
4942 }
4943
4944 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4945 {
4946         struct nfs_server *server = NFS_SERVER(state->inode);
4947         struct nfs4_exception exception = { };
4948         int err;
4949
4950         err = nfs4_set_lock_state(state, fl);
4951         if (err != 0)
4952                 goto out;
4953         do {
4954                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4955                 switch (err) {
4956                         default:
4957                                 printk(KERN_ERR "NFS: %s: unhandled error "
4958                                         "%d.\n", __func__, err);
4959                         case 0:
4960                         case -ESTALE:
4961                                 goto out;
4962                         case -NFS4ERR_EXPIRED:
4963                                 nfs4_schedule_stateid_recovery(server, state);
4964                         case -NFS4ERR_STALE_CLIENTID:
4965                         case -NFS4ERR_STALE_STATEID:
4966                                 nfs4_schedule_lease_recovery(server->nfs_client);
4967                                 goto out;
4968                         case -NFS4ERR_BADSESSION:
4969                         case -NFS4ERR_BADSLOT:
4970                         case -NFS4ERR_BAD_HIGH_SLOT:
4971                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4972                         case -NFS4ERR_DEADSESSION:
4973                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
4974                                 goto out;
4975                         case -ERESTARTSYS:
4976                                 /*
4977                                  * The show must go on: exit, but mark the
4978                                  * stateid as needing recovery.
4979                                  */
4980                         case -NFS4ERR_DELEG_REVOKED:
4981                         case -NFS4ERR_ADMIN_REVOKED:
4982                         case -NFS4ERR_BAD_STATEID:
4983                         case -NFS4ERR_OPENMODE:
4984                                 nfs4_schedule_stateid_recovery(server, state);
4985                                 err = 0;
4986                                 goto out;
4987                         case -EKEYEXPIRED:
4988                                 /*
4989                                  * User RPCSEC_GSS context has expired.
4990                                  * We cannot recover this stateid now, so
4991                                  * skip it and allow recovery thread to
4992                                  * proceed.
4993                                  */
4994                                 err = 0;
4995                                 goto out;
4996                         case -ENOMEM:
4997                         case -NFS4ERR_DENIED:
4998                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4999                                 err = 0;
5000                                 goto out;
5001                         case -NFS4ERR_DELAY:
5002                                 break;
5003                 }
5004                 err = nfs4_handle_exception(server, err, &exception);
5005         } while (exception.retry);
5006 out:
5007         return err;
5008 }
5009
5010 struct nfs_release_lockowner_data {
5011         struct nfs4_lock_state *lsp;
5012         struct nfs_server *server;
5013         struct nfs_release_lockowner_args args;
5014 };
5015
5016 static void nfs4_release_lockowner_release(void *calldata)
5017 {
5018         struct nfs_release_lockowner_data *data = calldata;
5019         nfs4_free_lock_state(data->server, data->lsp);
5020         kfree(calldata);
5021 }
5022
5023 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5024         .rpc_release = nfs4_release_lockowner_release,
5025 };
5026
5027 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5028 {
5029         struct nfs_server *server = lsp->ls_state->owner->so_server;
5030         struct nfs_release_lockowner_data *data;
5031         struct rpc_message msg = {
5032                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5033         };
5034
5035         if (server->nfs_client->cl_mvops->minor_version != 0)
5036                 return -EINVAL;
5037         data = kmalloc(sizeof(*data), GFP_NOFS);
5038         if (!data)
5039                 return -ENOMEM;
5040         data->lsp = lsp;
5041         data->server = server;
5042         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5043         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5044         data->args.lock_owner.s_dev = server->s_dev;
5045         msg.rpc_argp = &data->args;
5046         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5047         return 0;
5048 }
5049
5050 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5051
5052 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5053                                    const void *buf, size_t buflen,
5054                                    int flags, int type)
5055 {
5056         if (strcmp(key, "") != 0)
5057                 return -EINVAL;
5058
5059         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5060 }
5061
5062 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5063                                    void *buf, size_t buflen, int type)
5064 {
5065         if (strcmp(key, "") != 0)
5066                 return -EINVAL;
5067
5068         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5069 }
5070
5071 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5072                                        size_t list_len, const char *name,
5073                                        size_t name_len, int type)
5074 {
5075         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5076
5077         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5078                 return 0;
5079
5080         if (list && len <= list_len)
5081                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5082         return len;
5083 }
5084
5085 /*
5086  * nfs_fhget will use either the mounted_on_fileid or the fileid
5087  */
5088 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5089 {
5090         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5091                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5092               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5093               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5094                 return;
5095
5096         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5097                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5098         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5099         fattr->nlink = 2;
5100 }
5101
5102 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5103                                    const struct qstr *name,
5104                                    struct nfs4_fs_locations *fs_locations,
5105                                    struct page *page)
5106 {
5107         struct nfs_server *server = NFS_SERVER(dir);
5108         u32 bitmask[2] = {
5109                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5110         };
5111         struct nfs4_fs_locations_arg args = {
5112                 .dir_fh = NFS_FH(dir),
5113                 .name = name,
5114                 .page = page,
5115                 .bitmask = bitmask,
5116         };
5117         struct nfs4_fs_locations_res res = {
5118                 .fs_locations = fs_locations,
5119         };
5120         struct rpc_message msg = {
5121                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5122                 .rpc_argp = &args,
5123                 .rpc_resp = &res,
5124         };
5125         int status;
5126
5127         dprintk("%s: start\n", __func__);
5128
5129         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5130          * is not supported */
5131         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5132                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5133         else
5134                 bitmask[0] |= FATTR4_WORD0_FILEID;
5135
5136         nfs_fattr_init(&fs_locations->fattr);
5137         fs_locations->server = server;
5138         fs_locations->nlocations = 0;
5139         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5140         dprintk("%s: returned status = %d\n", __func__, status);
5141         return status;
5142 }
5143
5144 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5145                            const struct qstr *name,
5146                            struct nfs4_fs_locations *fs_locations,
5147                            struct page *page)
5148 {
5149         struct nfs4_exception exception = { };
5150         int err;
5151         do {
5152                 err = nfs4_handle_exception(NFS_SERVER(dir),
5153                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5154                                 &exception);
5155         } while (exception.retry);
5156         return err;
5157 }
5158
5159 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5160 {
5161         int status;
5162         struct nfs4_secinfo_arg args = {
5163                 .dir_fh = NFS_FH(dir),
5164                 .name   = name,
5165         };
5166         struct nfs4_secinfo_res res = {
5167                 .flavors     = flavors,
5168         };
5169         struct rpc_message msg = {
5170                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5171                 .rpc_argp = &args,
5172                 .rpc_resp = &res,
5173         };
5174
5175         dprintk("NFS call  secinfo %s\n", name->name);
5176         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5177         dprintk("NFS reply  secinfo: %d\n", status);
5178         return status;
5179 }
5180
5181 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5182                       struct nfs4_secinfo_flavors *flavors)
5183 {
5184         struct nfs4_exception exception = { };
5185         int err;
5186         do {
5187                 err = nfs4_handle_exception(NFS_SERVER(dir),
5188                                 _nfs4_proc_secinfo(dir, name, flavors),
5189                                 &exception);
5190         } while (exception.retry);
5191         return err;
5192 }
5193
5194 #ifdef CONFIG_NFS_V4_1
5195 /*
5196  * Check the exchange flags returned by the server for invalid flags, having
5197  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5198  * DS flags set.
5199  */
5200 static int nfs4_check_cl_exchange_flags(u32 flags)
5201 {
5202         if (flags & ~EXCHGID4_FLAG_MASK_R)
5203                 goto out_inval;
5204         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5205             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5206                 goto out_inval;
5207         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5208                 goto out_inval;
5209         return NFS_OK;
5210 out_inval:
5211         return -NFS4ERR_INVAL;
5212 }
5213
5214 static bool
5215 nfs41_same_server_scope(struct nfs41_server_scope *a,
5216                         struct nfs41_server_scope *b)
5217 {
5218         if (a->server_scope_sz == b->server_scope_sz &&
5219             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5220                 return true;
5221
5222         return false;
5223 }
5224
5225 /*
5226  * nfs4_proc_bind_conn_to_session()
5227  *
5228  * The 4.1 client currently uses the same TCP connection for the
5229  * fore and backchannel.
5230  */
5231 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5232 {
5233         int status;
5234         struct nfs41_bind_conn_to_session_res res;
5235         struct rpc_message msg = {
5236                 .rpc_proc =
5237                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5238                 .rpc_argp = clp,
5239                 .rpc_resp = &res,
5240                 .rpc_cred = cred,
5241         };
5242
5243         dprintk("--> %s\n", __func__);
5244         BUG_ON(clp == NULL);
5245
5246         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5247         if (unlikely(res.session == NULL)) {
5248                 status = -ENOMEM;
5249                 goto out;
5250         }
5251
5252         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5253         if (status == 0) {
5254                 if (memcmp(res.session->sess_id.data,
5255                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5256                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5257                         status = -EIO;
5258                         goto out_session;
5259                 }
5260                 if (res.dir != NFS4_CDFS4_BOTH) {
5261                         dprintk("NFS: %s: Unexpected direction from server\n",
5262                                 __func__);
5263                         status = -EIO;
5264                         goto out_session;
5265                 }
5266                 if (res.use_conn_in_rdma_mode) {
5267                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5268                                 __func__);
5269                         status = -EIO;
5270                         goto out_session;
5271                 }
5272         }
5273 out_session:
5274         kfree(res.session);
5275 out:
5276         dprintk("<-- %s status= %d\n", __func__, status);
5277         return status;
5278 }
5279
5280 /*
5281  * nfs4_proc_exchange_id()
5282  *
5283  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5284  *
5285  * Since the clientid has expired, all compounds using sessions
5286  * associated with the stale clientid will be returning
5287  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5288  * be in some phase of session reset.
5289  */
5290 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5291 {
5292         nfs4_verifier verifier;
5293         struct nfs41_exchange_id_args args = {
5294                 .verifier = &verifier,
5295                 .client = clp,
5296                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5297         };
5298         struct nfs41_exchange_id_res res = {
5299                 0
5300         };
5301         int status;
5302         struct rpc_message msg = {
5303                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5304                 .rpc_argp = &args,
5305                 .rpc_resp = &res,
5306                 .rpc_cred = cred,
5307         };
5308
5309         nfs4_init_boot_verifier(clp, &verifier);
5310         args.id_len = scnprintf(args.id, sizeof(args.id),
5311                                 "%s/%s",
5312                                 clp->cl_ipaddr,
5313                                 clp->cl_rpcclient->cl_nodename);
5314         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5315                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5316                 args.id_len, args.id);
5317
5318         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5319                                         GFP_NOFS);
5320         if (unlikely(res.server_owner == NULL)) {
5321                 status = -ENOMEM;
5322                 goto out;
5323         }
5324
5325         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5326                                         GFP_NOFS);
5327         if (unlikely(res.server_scope == NULL)) {
5328                 status = -ENOMEM;
5329                 goto out_server_owner;
5330         }
5331
5332         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5333         if (unlikely(res.impl_id == NULL)) {
5334                 status = -ENOMEM;
5335                 goto out_server_scope;
5336         }
5337
5338         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5339         if (status == 0)
5340                 status = nfs4_check_cl_exchange_flags(res.flags);
5341
5342         if (status == 0) {
5343                 clp->cl_clientid = res.clientid;
5344                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5345                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5346                         clp->cl_seqid = res.seqid;
5347
5348                 kfree(clp->cl_serverowner);
5349                 clp->cl_serverowner = res.server_owner;
5350                 res.server_owner = NULL;
5351
5352                 /* use the most recent implementation id */
5353                 kfree(clp->cl_implid);
5354                 clp->cl_implid = res.impl_id;
5355
5356                 if (clp->cl_serverscope != NULL &&
5357                     !nfs41_same_server_scope(clp->cl_serverscope,
5358                                              res.server_scope)) {
5359                         dprintk("%s: server_scope mismatch detected\n",
5360                                 __func__);
5361                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5362                         kfree(clp->cl_serverscope);
5363                         clp->cl_serverscope = NULL;
5364                 }
5365
5366                 if (clp->cl_serverscope == NULL) {
5367                         clp->cl_serverscope = res.server_scope;
5368                         goto out;
5369                 }
5370         } else
5371                 kfree(res.impl_id);
5372
5373 out_server_owner:
5374         kfree(res.server_owner);
5375 out_server_scope:
5376         kfree(res.server_scope);
5377 out:
5378         if (clp->cl_implid != NULL)
5379                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5380                         "domain: %s, name: %s, date: %llu,%u\n",
5381                         clp->cl_implid->domain, clp->cl_implid->name,
5382                         clp->cl_implid->date.seconds,
5383                         clp->cl_implid->date.nseconds);
5384         dprintk("NFS reply exchange_id: %d\n", status);
5385         return status;
5386 }
5387
5388 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5389                 struct rpc_cred *cred)
5390 {
5391         struct rpc_message msg = {
5392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5393                 .rpc_argp = clp,
5394                 .rpc_cred = cred,
5395         };
5396         int status;
5397
5398         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5399         if (status)
5400                 dprintk("NFS: Got error %d from the server %s on "
5401                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5402         return status;
5403 }
5404
5405 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5406                 struct rpc_cred *cred)
5407 {
5408         unsigned int loop;
5409         int ret;
5410
5411         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5412                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5413                 switch (ret) {
5414                 case -NFS4ERR_DELAY:
5415                 case -NFS4ERR_CLIENTID_BUSY:
5416                         ssleep(1);
5417                         break;
5418                 default:
5419                         return ret;
5420                 }
5421         }
5422         return 0;
5423 }
5424
5425 int nfs4_destroy_clientid(struct nfs_client *clp)
5426 {
5427         struct rpc_cred *cred;
5428         int ret = 0;
5429
5430         if (clp->cl_mvops->minor_version < 1)
5431                 goto out;
5432         if (clp->cl_exchange_flags == 0)
5433                 goto out;
5434         cred = nfs4_get_exchange_id_cred(clp);
5435         ret = nfs4_proc_destroy_clientid(clp, cred);
5436         if (cred)
5437                 put_rpccred(cred);
5438         switch (ret) {
5439         case 0:
5440         case -NFS4ERR_STALE_CLIENTID:
5441                 clp->cl_exchange_flags = 0;
5442         }
5443 out:
5444         return ret;
5445 }
5446
5447 struct nfs4_get_lease_time_data {
5448         struct nfs4_get_lease_time_args *args;
5449         struct nfs4_get_lease_time_res *res;
5450         struct nfs_client *clp;
5451 };
5452
5453 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5454                                         void *calldata)
5455 {
5456         int ret;
5457         struct nfs4_get_lease_time_data *data =
5458                         (struct nfs4_get_lease_time_data *)calldata;
5459
5460         dprintk("--> %s\n", __func__);
5461         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5462         /* just setup sequence, do not trigger session recovery
5463            since we're invoked within one */
5464         ret = nfs41_setup_sequence(data->clp->cl_session,
5465                                    &data->args->la_seq_args,
5466                                    &data->res->lr_seq_res, task);
5467
5468         BUG_ON(ret == -EAGAIN);
5469         rpc_call_start(task);
5470         dprintk("<-- %s\n", __func__);
5471 }
5472
5473 /*
5474  * Called from nfs4_state_manager thread for session setup, so don't recover
5475  * from sequence operation or clientid errors.
5476  */
5477 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5478 {
5479         struct nfs4_get_lease_time_data *data =
5480                         (struct nfs4_get_lease_time_data *)calldata;
5481
5482         dprintk("--> %s\n", __func__);
5483         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5484                 return;
5485         switch (task->tk_status) {
5486         case -NFS4ERR_DELAY:
5487         case -NFS4ERR_GRACE:
5488                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5489                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5490                 task->tk_status = 0;
5491                 /* fall through */
5492         case -NFS4ERR_RETRY_UNCACHED_REP:
5493                 rpc_restart_call_prepare(task);
5494                 return;
5495         }
5496         dprintk("<-- %s\n", __func__);
5497 }
5498
5499 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5500         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5501         .rpc_call_done = nfs4_get_lease_time_done,
5502 };
5503
5504 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5505 {
5506         struct rpc_task *task;
5507         struct nfs4_get_lease_time_args args;
5508         struct nfs4_get_lease_time_res res = {
5509                 .lr_fsinfo = fsinfo,
5510         };
5511         struct nfs4_get_lease_time_data data = {
5512                 .args = &args,
5513                 .res = &res,
5514                 .clp = clp,
5515         };
5516         struct rpc_message msg = {
5517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5518                 .rpc_argp = &args,
5519                 .rpc_resp = &res,
5520         };
5521         struct rpc_task_setup task_setup = {
5522                 .rpc_client = clp->cl_rpcclient,
5523                 .rpc_message = &msg,
5524                 .callback_ops = &nfs4_get_lease_time_ops,
5525                 .callback_data = &data,
5526                 .flags = RPC_TASK_TIMEOUT,
5527         };
5528         int status;
5529
5530         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5531         dprintk("--> %s\n", __func__);
5532         task = rpc_run_task(&task_setup);
5533
5534         if (IS_ERR(task))
5535                 status = PTR_ERR(task);
5536         else {
5537                 status = task->tk_status;
5538                 rpc_put_task(task);
5539         }
5540         dprintk("<-- %s return %d\n", __func__, status);
5541
5542         return status;
5543 }
5544
5545 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5546 {
5547         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5548 }
5549
5550 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5551                 struct nfs4_slot *new,
5552                 u32 max_slots,
5553                 u32 ivalue)
5554 {
5555         struct nfs4_slot *old = NULL;
5556         u32 i;
5557
5558         spin_lock(&tbl->slot_tbl_lock);
5559         if (new) {
5560                 old = tbl->slots;
5561                 tbl->slots = new;
5562                 tbl->max_slots = max_slots;
5563         }
5564         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5565         for (i = 0; i < tbl->max_slots; i++)
5566                 tbl->slots[i].seq_nr = ivalue;
5567         spin_unlock(&tbl->slot_tbl_lock);
5568         kfree(old);
5569 }
5570
5571 /*
5572  * (re)Initialise a slot table
5573  */
5574 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5575                                  u32 ivalue)
5576 {
5577         struct nfs4_slot *new = NULL;
5578         int ret = -ENOMEM;
5579
5580         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5581                 max_reqs, tbl->max_slots);
5582
5583         /* Does the newly negotiated max_reqs match the existing slot table? */
5584         if (max_reqs != tbl->max_slots) {
5585                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5586                 if (!new)
5587                         goto out;
5588         }
5589         ret = 0;
5590
5591         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5592         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5593                 tbl, tbl->slots, tbl->max_slots);
5594 out:
5595         dprintk("<-- %s: return %d\n", __func__, ret);
5596         return ret;
5597 }
5598
5599 /* Destroy the slot table */
5600 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5601 {
5602         if (session->fc_slot_table.slots != NULL) {
5603                 kfree(session->fc_slot_table.slots);
5604                 session->fc_slot_table.slots = NULL;
5605         }
5606         if (session->bc_slot_table.slots != NULL) {
5607                 kfree(session->bc_slot_table.slots);
5608                 session->bc_slot_table.slots = NULL;
5609         }
5610         return;
5611 }
5612
5613 /*
5614  * Initialize or reset the forechannel and backchannel tables
5615  */
5616 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5617 {
5618         struct nfs4_slot_table *tbl;
5619         int status;
5620
5621         dprintk("--> %s\n", __func__);
5622         /* Fore channel */
5623         tbl = &ses->fc_slot_table;
5624         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5625         if (status) /* -ENOMEM */
5626                 return status;
5627         /* Back channel */
5628         tbl = &ses->bc_slot_table;
5629         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5630         if (status && tbl->slots == NULL)
5631                 /* Fore and back channel share a connection so get
5632                  * both slot tables or neither */
5633                 nfs4_destroy_slot_tables(ses);
5634         return status;
5635 }
5636
5637 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5638 {
5639         struct nfs4_session *session;
5640         struct nfs4_slot_table *tbl;
5641
5642         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5643         if (!session)
5644                 return NULL;
5645
5646         tbl = &session->fc_slot_table;
5647         tbl->highest_used_slotid = NFS4_NO_SLOT;
5648         spin_lock_init(&tbl->slot_tbl_lock);
5649         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5650         init_completion(&tbl->complete);
5651
5652         tbl = &session->bc_slot_table;
5653         tbl->highest_used_slotid = NFS4_NO_SLOT;
5654         spin_lock_init(&tbl->slot_tbl_lock);
5655         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5656         init_completion(&tbl->complete);
5657
5658         session->session_state = 1<<NFS4_SESSION_INITING;
5659
5660         session->clp = clp;
5661         return session;
5662 }
5663
5664 void nfs4_destroy_session(struct nfs4_session *session)
5665 {
5666         struct rpc_xprt *xprt;
5667         struct rpc_cred *cred;
5668
5669         cred = nfs4_get_exchange_id_cred(session->clp);
5670         nfs4_proc_destroy_session(session, cred);
5671         if (cred)
5672                 put_rpccred(cred);
5673
5674         rcu_read_lock();
5675         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5676         rcu_read_unlock();
5677         dprintk("%s Destroy backchannel for xprt %p\n",
5678                 __func__, xprt);
5679         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5680         nfs4_destroy_slot_tables(session);
5681         kfree(session);
5682 }
5683
5684 /*
5685  * Initialize the values to be used by the client in CREATE_SESSION
5686  * If nfs4_init_session set the fore channel request and response sizes,
5687  * use them.
5688  *
5689  * Set the back channel max_resp_sz_cached to zero to force the client to
5690  * always set csa_cachethis to FALSE because the current implementation
5691  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5692  */
5693 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5694 {
5695         struct nfs4_session *session = args->client->cl_session;
5696         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5697                      mxresp_sz = session->fc_attrs.max_resp_sz;
5698
5699         if (mxrqst_sz == 0)
5700                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5701         if (mxresp_sz == 0)
5702                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5703         /* Fore channel attributes */
5704         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5705         args->fc_attrs.max_resp_sz = mxresp_sz;
5706         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5707         args->fc_attrs.max_reqs = max_session_slots;
5708
5709         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5710                 "max_ops=%u max_reqs=%u\n",
5711                 __func__,
5712                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5713                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5714
5715         /* Back channel attributes */
5716         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5717         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5718         args->bc_attrs.max_resp_sz_cached = 0;
5719         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5720         args->bc_attrs.max_reqs = 1;
5721
5722         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5723                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5724                 __func__,
5725                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5726                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5727                 args->bc_attrs.max_reqs);
5728 }
5729
5730 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5731 {
5732         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5733         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5734
5735         if (rcvd->max_resp_sz > sent->max_resp_sz)
5736                 return -EINVAL;
5737         /*
5738          * Our requested max_ops is the minimum we need; we're not
5739          * prepared to break up compounds into smaller pieces than that.
5740          * So, no point even trying to continue if the server won't
5741          * cooperate:
5742          */
5743         if (rcvd->max_ops < sent->max_ops)
5744                 return -EINVAL;
5745         if (rcvd->max_reqs == 0)
5746                 return -EINVAL;
5747         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5748                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5749         return 0;
5750 }
5751
5752 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5753 {
5754         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5755         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5756
5757         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5758                 return -EINVAL;
5759         if (rcvd->max_resp_sz < sent->max_resp_sz)
5760                 return -EINVAL;
5761         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5762                 return -EINVAL;
5763         /* These would render the backchannel useless: */
5764         if (rcvd->max_ops != sent->max_ops)
5765                 return -EINVAL;
5766         if (rcvd->max_reqs != sent->max_reqs)
5767                 return -EINVAL;
5768         return 0;
5769 }
5770
5771 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5772                                      struct nfs4_session *session)
5773 {
5774         int ret;
5775
5776         ret = nfs4_verify_fore_channel_attrs(args, session);
5777         if (ret)
5778                 return ret;
5779         return nfs4_verify_back_channel_attrs(args, session);
5780 }
5781
5782 static int _nfs4_proc_create_session(struct nfs_client *clp,
5783                 struct rpc_cred *cred)
5784 {
5785         struct nfs4_session *session = clp->cl_session;
5786         struct nfs41_create_session_args args = {
5787                 .client = clp,
5788                 .cb_program = NFS4_CALLBACK,
5789         };
5790         struct nfs41_create_session_res res = {
5791                 .client = clp,
5792         };
5793         struct rpc_message msg = {
5794                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5795                 .rpc_argp = &args,
5796                 .rpc_resp = &res,
5797                 .rpc_cred = cred,
5798         };
5799         int status;
5800
5801         nfs4_init_channel_attrs(&args);
5802         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5803
5804         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5805
5806         if (!status)
5807                 /* Verify the session's negotiated channel_attrs values */
5808                 status = nfs4_verify_channel_attrs(&args, session);
5809         if (!status) {
5810                 /* Increment the clientid slot sequence id */
5811                 clp->cl_seqid++;
5812         }
5813
5814         return status;
5815 }
5816
5817 /*
5818  * Issues a CREATE_SESSION operation to the server.
5819  * It is the responsibility of the caller to verify the session is
5820  * expired before calling this routine.
5821  */
5822 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5823 {
5824         int status;
5825         unsigned *ptr;
5826         struct nfs4_session *session = clp->cl_session;
5827
5828         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5829
5830         status = _nfs4_proc_create_session(clp, cred);
5831         if (status)
5832                 goto out;
5833
5834         /* Init or reset the session slot tables */
5835         status = nfs4_setup_session_slot_tables(session);
5836         dprintk("slot table setup returned %d\n", status);
5837         if (status)
5838                 goto out;
5839
5840         ptr = (unsigned *)&session->sess_id.data[0];
5841         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5842                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5843 out:
5844         dprintk("<-- %s\n", __func__);
5845         return status;
5846 }
5847
5848 /*
5849  * Issue the over-the-wire RPC DESTROY_SESSION.
5850  * The caller must serialize access to this routine.
5851  */
5852 int nfs4_proc_destroy_session(struct nfs4_session *session,
5853                 struct rpc_cred *cred)
5854 {
5855         struct rpc_message msg = {
5856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5857                 .rpc_argp = session,
5858                 .rpc_cred = cred,
5859         };
5860         int status = 0;
5861
5862         dprintk("--> nfs4_proc_destroy_session\n");
5863
5864         /* session is still being setup */
5865         if (session->clp->cl_cons_state != NFS_CS_READY)
5866                 return status;
5867
5868         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5869
5870         if (status)
5871                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5872                         "Session has been destroyed regardless...\n", status);
5873
5874         dprintk("<-- nfs4_proc_destroy_session\n");
5875         return status;
5876 }
5877
5878 /*
5879  * With sessions, the client is not marked ready until after a
5880  * successful EXCHANGE_ID and CREATE_SESSION.
5881  *
5882  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
5883  * other versions of NFS can be tried.
5884  */
5885 static int nfs41_check_session_ready(struct nfs_client *clp)
5886 {
5887         int ret;
5888
5889         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
5890                 ret = nfs4_client_recover_expired_lease(clp);
5891                 if (ret)
5892                         return ret;
5893         }
5894         if (clp->cl_cons_state < NFS_CS_READY)
5895                 return -EPROTONOSUPPORT;
5896         smp_rmb();
5897         return 0;
5898 }
5899
5900 int nfs4_init_session(struct nfs_server *server)
5901 {
5902         struct nfs_client *clp = server->nfs_client;
5903         struct nfs4_session *session;
5904         unsigned int rsize, wsize;
5905
5906         if (!nfs4_has_session(clp))
5907                 return 0;
5908
5909         session = clp->cl_session;
5910         spin_lock(&clp->cl_lock);
5911         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5912
5913                 rsize = server->rsize;
5914                 if (rsize == 0)
5915                         rsize = NFS_MAX_FILE_IO_SIZE;
5916                 wsize = server->wsize;
5917                 if (wsize == 0)
5918                         wsize = NFS_MAX_FILE_IO_SIZE;
5919
5920                 session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5921                 session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5922         }
5923         spin_unlock(&clp->cl_lock);
5924
5925         return nfs41_check_session_ready(clp);
5926 }
5927
5928 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
5929 {
5930         struct nfs4_session *session = clp->cl_session;
5931         int ret;
5932
5933         spin_lock(&clp->cl_lock);
5934         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5935                 /*
5936                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
5937                  * DS lease to be equal to the MDS lease.
5938                  */
5939                 clp->cl_lease_time = lease_time;
5940                 clp->cl_last_renewal = jiffies;
5941         }
5942         spin_unlock(&clp->cl_lock);
5943
5944         ret = nfs41_check_session_ready(clp);
5945         if (ret)
5946                 return ret;
5947         /* Test for the DS role */
5948         if (!is_ds_client(clp))
5949                 return -ENODEV;
5950         return 0;
5951 }
5952 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5953
5954
5955 /*
5956  * Renew the cl_session lease.
5957  */
5958 struct nfs4_sequence_data {
5959         struct nfs_client *clp;
5960         struct nfs4_sequence_args args;
5961         struct nfs4_sequence_res res;
5962 };
5963
5964 static void nfs41_sequence_release(void *data)
5965 {
5966         struct nfs4_sequence_data *calldata = data;
5967         struct nfs_client *clp = calldata->clp;
5968
5969         if (atomic_read(&clp->cl_count) > 1)
5970                 nfs4_schedule_state_renewal(clp);
5971         nfs_put_client(clp);
5972         kfree(calldata);
5973 }
5974
5975 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5976 {
5977         switch(task->tk_status) {
5978         case -NFS4ERR_DELAY:
5979                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5980                 return -EAGAIN;
5981         default:
5982                 nfs4_schedule_lease_recovery(clp);
5983         }
5984         return 0;
5985 }
5986
5987 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5988 {
5989         struct nfs4_sequence_data *calldata = data;
5990         struct nfs_client *clp = calldata->clp;
5991
5992         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5993                 return;
5994
5995         if (task->tk_status < 0) {
5996                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5997                 if (atomic_read(&clp->cl_count) == 1)
5998                         goto out;
5999
6000                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6001                         rpc_restart_call_prepare(task);
6002                         return;
6003                 }
6004         }
6005         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6006 out:
6007         dprintk("<-- %s\n", __func__);
6008 }
6009
6010 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6011 {
6012         struct nfs4_sequence_data *calldata = data;
6013         struct nfs_client *clp = calldata->clp;
6014         struct nfs4_sequence_args *args;
6015         struct nfs4_sequence_res *res;
6016
6017         args = task->tk_msg.rpc_argp;
6018         res = task->tk_msg.rpc_resp;
6019
6020         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
6021                 return;
6022         rpc_call_start(task);
6023 }
6024
6025 static const struct rpc_call_ops nfs41_sequence_ops = {
6026         .rpc_call_done = nfs41_sequence_call_done,
6027         .rpc_call_prepare = nfs41_sequence_prepare,
6028         .rpc_release = nfs41_sequence_release,
6029 };
6030
6031 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6032 {
6033         struct nfs4_sequence_data *calldata;
6034         struct rpc_message msg = {
6035                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6036                 .rpc_cred = cred,
6037         };
6038         struct rpc_task_setup task_setup_data = {
6039                 .rpc_client = clp->cl_rpcclient,
6040                 .rpc_message = &msg,
6041                 .callback_ops = &nfs41_sequence_ops,
6042                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6043         };
6044
6045         if (!atomic_inc_not_zero(&clp->cl_count))
6046                 return ERR_PTR(-EIO);
6047         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6048         if (calldata == NULL) {
6049                 nfs_put_client(clp);
6050                 return ERR_PTR(-ENOMEM);
6051         }
6052         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6053         msg.rpc_argp = &calldata->args;
6054         msg.rpc_resp = &calldata->res;
6055         calldata->clp = clp;
6056         task_setup_data.callback_data = calldata;
6057
6058         return rpc_run_task(&task_setup_data);
6059 }
6060
6061 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6062 {
6063         struct rpc_task *task;
6064         int ret = 0;
6065
6066         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6067                 return 0;
6068         task = _nfs41_proc_sequence(clp, cred);
6069         if (IS_ERR(task))
6070                 ret = PTR_ERR(task);
6071         else
6072                 rpc_put_task_async(task);
6073         dprintk("<-- %s status=%d\n", __func__, ret);
6074         return ret;
6075 }
6076
6077 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6078 {
6079         struct rpc_task *task;
6080         int ret;
6081
6082         task = _nfs41_proc_sequence(clp, cred);
6083         if (IS_ERR(task)) {
6084                 ret = PTR_ERR(task);
6085                 goto out;
6086         }
6087         ret = rpc_wait_for_completion_task(task);
6088         if (!ret) {
6089                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6090
6091                 if (task->tk_status == 0)
6092                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6093                 ret = task->tk_status;
6094         }
6095         rpc_put_task(task);
6096 out:
6097         dprintk("<-- %s status=%d\n", __func__, ret);
6098         return ret;
6099 }
6100
6101 struct nfs4_reclaim_complete_data {
6102         struct nfs_client *clp;
6103         struct nfs41_reclaim_complete_args arg;
6104         struct nfs41_reclaim_complete_res res;
6105 };
6106
6107 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6108 {
6109         struct nfs4_reclaim_complete_data *calldata = data;
6110
6111         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6112         if (nfs41_setup_sequence(calldata->clp->cl_session,
6113                                 &calldata->arg.seq_args,
6114                                 &calldata->res.seq_res, task))
6115                 return;
6116
6117         rpc_call_start(task);
6118 }
6119
6120 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6121 {
6122         switch(task->tk_status) {
6123         case 0:
6124         case -NFS4ERR_COMPLETE_ALREADY:
6125         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6126                 break;
6127         case -NFS4ERR_DELAY:
6128                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6129                 /* fall through */
6130         case -NFS4ERR_RETRY_UNCACHED_REP:
6131                 return -EAGAIN;
6132         default:
6133                 nfs4_schedule_lease_recovery(clp);
6134         }
6135         return 0;
6136 }
6137
6138 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6139 {
6140         struct nfs4_reclaim_complete_data *calldata = data;
6141         struct nfs_client *clp = calldata->clp;
6142         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6143
6144         dprintk("--> %s\n", __func__);
6145         if (!nfs41_sequence_done(task, res))
6146                 return;
6147
6148         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6149                 rpc_restart_call_prepare(task);
6150                 return;
6151         }
6152         dprintk("<-- %s\n", __func__);
6153 }
6154
6155 static void nfs4_free_reclaim_complete_data(void *data)
6156 {
6157         struct nfs4_reclaim_complete_data *calldata = data;
6158
6159         kfree(calldata);
6160 }
6161
6162 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6163         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6164         .rpc_call_done = nfs4_reclaim_complete_done,
6165         .rpc_release = nfs4_free_reclaim_complete_data,
6166 };
6167
6168 /*
6169  * Issue a global reclaim complete.
6170  */
6171 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6172 {
6173         struct nfs4_reclaim_complete_data *calldata;
6174         struct rpc_task *task;
6175         struct rpc_message msg = {
6176                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6177         };
6178         struct rpc_task_setup task_setup_data = {
6179                 .rpc_client = clp->cl_rpcclient,
6180                 .rpc_message = &msg,
6181                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6182                 .flags = RPC_TASK_ASYNC,
6183         };
6184         int status = -ENOMEM;
6185
6186         dprintk("--> %s\n", __func__);
6187         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6188         if (calldata == NULL)
6189                 goto out;
6190         calldata->clp = clp;
6191         calldata->arg.one_fs = 0;
6192
6193         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6194         msg.rpc_argp = &calldata->arg;
6195         msg.rpc_resp = &calldata->res;
6196         task_setup_data.callback_data = calldata;
6197         task = rpc_run_task(&task_setup_data);
6198         if (IS_ERR(task)) {
6199                 status = PTR_ERR(task);
6200                 goto out;
6201         }
6202         status = nfs4_wait_for_completion_rpc_task(task);
6203         if (status == 0)
6204                 status = task->tk_status;
6205         rpc_put_task(task);
6206         return 0;
6207 out:
6208         dprintk("<-- %s status=%d\n", __func__, status);
6209         return status;
6210 }
6211
6212 static void
6213 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6214 {
6215         struct nfs4_layoutget *lgp = calldata;
6216         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6217
6218         dprintk("--> %s\n", __func__);
6219         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6220          * right now covering the LAYOUTGET we are about to send.
6221          * However, that is not so catastrophic, and there seems
6222          * to be no way to prevent it completely.
6223          */
6224         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6225                                 &lgp->res.seq_res, task))
6226                 return;
6227         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6228                                           NFS_I(lgp->args.inode)->layout,
6229                                           lgp->args.ctx->state)) {
6230                 rpc_exit(task, NFS4_OK);
6231                 return;
6232         }
6233         rpc_call_start(task);
6234 }
6235
6236 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6237 {
6238         struct nfs4_layoutget *lgp = calldata;
6239         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6240
6241         dprintk("--> %s\n", __func__);
6242
6243         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6244                 return;
6245
6246         switch (task->tk_status) {
6247         case 0:
6248                 break;
6249         case -NFS4ERR_LAYOUTTRYLATER:
6250         case -NFS4ERR_RECALLCONFLICT:
6251                 task->tk_status = -NFS4ERR_DELAY;
6252                 /* Fall through */
6253         default:
6254                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6255                         rpc_restart_call_prepare(task);
6256                         return;
6257                 }
6258         }
6259         dprintk("<-- %s\n", __func__);
6260 }
6261
6262 static size_t max_response_pages(struct nfs_server *server)
6263 {
6264         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6265         return nfs_page_array_len(0, max_resp_sz);
6266 }
6267
6268 static void nfs4_free_pages(struct page **pages, size_t size)
6269 {
6270         int i;
6271
6272         if (!pages)
6273                 return;
6274
6275         for (i = 0; i < size; i++) {
6276                 if (!pages[i])
6277                         break;
6278                 __free_page(pages[i]);
6279         }
6280         kfree(pages);
6281 }
6282
6283 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6284 {
6285         struct page **pages;
6286         int i;
6287
6288         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6289         if (!pages) {
6290                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6291                 return NULL;
6292         }
6293
6294         for (i = 0; i < size; i++) {
6295                 pages[i] = alloc_page(gfp_flags);
6296                 if (!pages[i]) {
6297                         dprintk("%s: failed to allocate page\n", __func__);
6298                         nfs4_free_pages(pages, size);
6299                         return NULL;
6300                 }
6301         }
6302
6303         return pages;
6304 }
6305
6306 static void nfs4_layoutget_release(void *calldata)
6307 {
6308         struct nfs4_layoutget *lgp = calldata;
6309         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6310         size_t max_pages = max_response_pages(server);
6311
6312         dprintk("--> %s\n", __func__);
6313         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6314         put_nfs_open_context(lgp->args.ctx);
6315         kfree(calldata);
6316         dprintk("<-- %s\n", __func__);
6317 }
6318
6319 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6320         .rpc_call_prepare = nfs4_layoutget_prepare,
6321         .rpc_call_done = nfs4_layoutget_done,
6322         .rpc_release = nfs4_layoutget_release,
6323 };
6324
6325 struct pnfs_layout_segment *
6326 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6327 {
6328         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6329         size_t max_pages = max_response_pages(server);
6330         struct rpc_task *task;
6331         struct rpc_message msg = {
6332                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6333                 .rpc_argp = &lgp->args,
6334                 .rpc_resp = &lgp->res,
6335         };
6336         struct rpc_task_setup task_setup_data = {
6337                 .rpc_client = server->client,
6338                 .rpc_message = &msg,
6339                 .callback_ops = &nfs4_layoutget_call_ops,
6340                 .callback_data = lgp,
6341                 .flags = RPC_TASK_ASYNC,
6342         };
6343         struct pnfs_layout_segment *lseg = NULL;
6344         int status = 0;
6345
6346         dprintk("--> %s\n", __func__);
6347
6348         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6349         if (!lgp->args.layout.pages) {
6350                 nfs4_layoutget_release(lgp);
6351                 return ERR_PTR(-ENOMEM);
6352         }
6353         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6354
6355         lgp->res.layoutp = &lgp->args.layout;
6356         lgp->res.seq_res.sr_slot = NULL;
6357         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6358         task = rpc_run_task(&task_setup_data);
6359         if (IS_ERR(task))
6360                 return ERR_CAST(task);
6361         status = nfs4_wait_for_completion_rpc_task(task);
6362         if (status == 0)
6363                 status = task->tk_status;
6364         if (status == 0)
6365                 lseg = pnfs_layout_process(lgp);
6366         rpc_put_task(task);
6367         dprintk("<-- %s status=%d\n", __func__, status);
6368         if (status)
6369                 return ERR_PTR(status);
6370         return lseg;
6371 }
6372
6373 static void
6374 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6375 {
6376         struct nfs4_layoutreturn *lrp = calldata;
6377
6378         dprintk("--> %s\n", __func__);
6379         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6380                                 &lrp->res.seq_res, task))
6381                 return;
6382         rpc_call_start(task);
6383 }
6384
6385 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6386 {
6387         struct nfs4_layoutreturn *lrp = calldata;
6388         struct nfs_server *server;
6389
6390         dprintk("--> %s\n", __func__);
6391
6392         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6393                 return;
6394
6395         server = NFS_SERVER(lrp->args.inode);
6396         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6397                 rpc_restart_call_prepare(task);
6398                 return;
6399         }
6400         dprintk("<-- %s\n", __func__);
6401 }
6402
6403 static void nfs4_layoutreturn_release(void *calldata)
6404 {
6405         struct nfs4_layoutreturn *lrp = calldata;
6406         struct pnfs_layout_hdr *lo = lrp->args.layout;
6407
6408         dprintk("--> %s\n", __func__);
6409         spin_lock(&lo->plh_inode->i_lock);
6410         if (lrp->res.lrs_present)
6411                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6412         lo->plh_block_lgets--;
6413         spin_unlock(&lo->plh_inode->i_lock);
6414         pnfs_put_layout_hdr(lrp->args.layout);
6415         kfree(calldata);
6416         dprintk("<-- %s\n", __func__);
6417 }
6418
6419 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6420         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6421         .rpc_call_done = nfs4_layoutreturn_done,
6422         .rpc_release = nfs4_layoutreturn_release,
6423 };
6424
6425 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6426 {
6427         struct rpc_task *task;
6428         struct rpc_message msg = {
6429                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6430                 .rpc_argp = &lrp->args,
6431                 .rpc_resp = &lrp->res,
6432         };
6433         struct rpc_task_setup task_setup_data = {
6434                 .rpc_client = lrp->clp->cl_rpcclient,
6435                 .rpc_message = &msg,
6436                 .callback_ops = &nfs4_layoutreturn_call_ops,
6437                 .callback_data = lrp,
6438         };
6439         int status;
6440
6441         dprintk("--> %s\n", __func__);
6442         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6443         task = rpc_run_task(&task_setup_data);
6444         if (IS_ERR(task))
6445                 return PTR_ERR(task);
6446         status = task->tk_status;
6447         dprintk("<-- %s status=%d\n", __func__, status);
6448         rpc_put_task(task);
6449         return status;
6450 }
6451
6452 /*
6453  * Retrieve the list of Data Server devices from the MDS.
6454  */
6455 static int _nfs4_getdevicelist(struct nfs_server *server,
6456                                     const struct nfs_fh *fh,
6457                                     struct pnfs_devicelist *devlist)
6458 {
6459         struct nfs4_getdevicelist_args args = {
6460                 .fh = fh,
6461                 .layoutclass = server->pnfs_curr_ld->id,
6462         };
6463         struct nfs4_getdevicelist_res res = {
6464                 .devlist = devlist,
6465         };
6466         struct rpc_message msg = {
6467                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6468                 .rpc_argp = &args,
6469                 .rpc_resp = &res,
6470         };
6471         int status;
6472
6473         dprintk("--> %s\n", __func__);
6474         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6475                                 &res.seq_res, 0);
6476         dprintk("<-- %s status=%d\n", __func__, status);
6477         return status;
6478 }
6479
6480 int nfs4_proc_getdevicelist(struct nfs_server *server,
6481                             const struct nfs_fh *fh,
6482                             struct pnfs_devicelist *devlist)
6483 {
6484         struct nfs4_exception exception = { };
6485         int err;
6486
6487         do {
6488                 err = nfs4_handle_exception(server,
6489                                 _nfs4_getdevicelist(server, fh, devlist),
6490                                 &exception);
6491         } while (exception.retry);
6492
6493         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6494                 err, devlist->num_devs);
6495
6496         return err;
6497 }
6498 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6499
6500 static int
6501 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6502 {
6503         struct nfs4_getdeviceinfo_args args = {
6504                 .pdev = pdev,
6505         };
6506         struct nfs4_getdeviceinfo_res res = {
6507                 .pdev = pdev,
6508         };
6509         struct rpc_message msg = {
6510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6511                 .rpc_argp = &args,
6512                 .rpc_resp = &res,
6513         };
6514         int status;
6515
6516         dprintk("--> %s\n", __func__);
6517         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6518         dprintk("<-- %s status=%d\n", __func__, status);
6519
6520         return status;
6521 }
6522
6523 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6524 {
6525         struct nfs4_exception exception = { };
6526         int err;
6527
6528         do {
6529                 err = nfs4_handle_exception(server,
6530                                         _nfs4_proc_getdeviceinfo(server, pdev),
6531                                         &exception);
6532         } while (exception.retry);
6533         return err;
6534 }
6535 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6536
6537 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6538 {
6539         struct nfs4_layoutcommit_data *data = calldata;
6540         struct nfs_server *server = NFS_SERVER(data->args.inode);
6541
6542         if (nfs4_setup_sequence(server, &data->args.seq_args,
6543                                 &data->res.seq_res, task))
6544                 return;
6545         rpc_call_start(task);
6546 }
6547
6548 static void
6549 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6550 {
6551         struct nfs4_layoutcommit_data *data = calldata;
6552         struct nfs_server *server = NFS_SERVER(data->args.inode);
6553
6554         if (!nfs4_sequence_done(task, &data->res.seq_res))
6555                 return;
6556
6557         switch (task->tk_status) { /* Just ignore these failures */
6558         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6559         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6560         case -NFS4ERR_BADLAYOUT:     /* no layout */
6561         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6562                 task->tk_status = 0;
6563                 break;
6564         case 0:
6565                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6566                                                    data->res.fattr);
6567                 break;
6568         default:
6569                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6570                         rpc_restart_call_prepare(task);
6571                         return;
6572                 }
6573         }
6574 }
6575
6576 static void nfs4_layoutcommit_release(void *calldata)
6577 {
6578         struct nfs4_layoutcommit_data *data = calldata;
6579         struct pnfs_layout_segment *lseg, *tmp;
6580         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6581
6582         pnfs_cleanup_layoutcommit(data);
6583         /* Matched by references in pnfs_set_layoutcommit */
6584         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6585                 list_del_init(&lseg->pls_lc_list);
6586                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6587                                        &lseg->pls_flags))
6588                         pnfs_put_lseg(lseg);
6589         }
6590
6591         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6592         smp_mb__after_clear_bit();
6593         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6594
6595         put_rpccred(data->cred);
6596         kfree(data);
6597 }
6598
6599 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6600         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6601         .rpc_call_done = nfs4_layoutcommit_done,
6602         .rpc_release = nfs4_layoutcommit_release,
6603 };
6604
6605 int
6606 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6607 {
6608         struct rpc_message msg = {
6609                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6610                 .rpc_argp = &data->args,
6611                 .rpc_resp = &data->res,
6612                 .rpc_cred = data->cred,
6613         };
6614         struct rpc_task_setup task_setup_data = {
6615                 .task = &data->task,
6616                 .rpc_client = NFS_CLIENT(data->args.inode),
6617                 .rpc_message = &msg,
6618                 .callback_ops = &nfs4_layoutcommit_ops,
6619                 .callback_data = data,
6620                 .flags = RPC_TASK_ASYNC,
6621         };
6622         struct rpc_task *task;
6623         int status = 0;
6624
6625         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6626                 "lbw: %llu inode %lu\n",
6627                 data->task.tk_pid, sync,
6628                 data->args.lastbytewritten,
6629                 data->args.inode->i_ino);
6630
6631         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6632         task = rpc_run_task(&task_setup_data);
6633         if (IS_ERR(task))
6634                 return PTR_ERR(task);
6635         if (sync == false)
6636                 goto out;
6637         status = nfs4_wait_for_completion_rpc_task(task);
6638         if (status != 0)
6639                 goto out;
6640         status = task->tk_status;
6641 out:
6642         dprintk("%s: status %d\n", __func__, status);
6643         rpc_put_task(task);
6644         return status;
6645 }
6646
6647 static int
6648 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6649                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6650 {
6651         struct nfs41_secinfo_no_name_args args = {
6652                 .style = SECINFO_STYLE_CURRENT_FH,
6653         };
6654         struct nfs4_secinfo_res res = {
6655                 .flavors = flavors,
6656         };
6657         struct rpc_message msg = {
6658                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6659                 .rpc_argp = &args,
6660                 .rpc_resp = &res,
6661         };
6662         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6663 }
6664
6665 static int
6666 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6667                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6668 {
6669         struct nfs4_exception exception = { };
6670         int err;
6671         do {
6672                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6673                 switch (err) {
6674                 case 0:
6675                 case -NFS4ERR_WRONGSEC:
6676                 case -NFS4ERR_NOTSUPP:
6677                         goto out;
6678                 default:
6679                         err = nfs4_handle_exception(server, err, &exception);
6680                 }
6681         } while (exception.retry);
6682 out:
6683         return err;
6684 }
6685
6686 static int
6687 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6688                     struct nfs_fsinfo *info)
6689 {
6690         int err;
6691         struct page *page;
6692         rpc_authflavor_t flavor;
6693         struct nfs4_secinfo_flavors *flavors;
6694
6695         page = alloc_page(GFP_KERNEL);
6696         if (!page) {
6697                 err = -ENOMEM;
6698                 goto out;
6699         }
6700
6701         flavors = page_address(page);
6702         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6703
6704         /*
6705          * Fall back on "guess and check" method if
6706          * the server doesn't support SECINFO_NO_NAME
6707          */
6708         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6709                 err = nfs4_find_root_sec(server, fhandle, info);
6710                 goto out_freepage;
6711         }
6712         if (err)
6713                 goto out_freepage;
6714
6715         flavor = nfs_find_best_sec(flavors);
6716         if (err == 0)
6717                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6718
6719 out_freepage:
6720         put_page(page);
6721         if (err == -EACCES)
6722                 return -EPERM;
6723 out:
6724         return err;
6725 }
6726
6727 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6728 {
6729         int status;
6730         struct nfs41_test_stateid_args args = {
6731                 .stateid = stateid,
6732         };
6733         struct nfs41_test_stateid_res res;
6734         struct rpc_message msg = {
6735                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6736                 .rpc_argp = &args,
6737                 .rpc_resp = &res,
6738         };
6739
6740         dprintk("NFS call  test_stateid %p\n", stateid);
6741         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6742         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6743         if (status != NFS_OK) {
6744                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6745                 return status;
6746         }
6747         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6748         return -res.status;
6749 }
6750
6751 /**
6752  * nfs41_test_stateid - perform a TEST_STATEID operation
6753  *
6754  * @server: server / transport on which to perform the operation
6755  * @stateid: state ID to test
6756  *
6757  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6758  * Otherwise a negative NFS4ERR value is returned if the operation
6759  * failed or the state ID is not currently valid.
6760  */
6761 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6762 {
6763         struct nfs4_exception exception = { };
6764         int err;
6765         do {
6766                 err = _nfs41_test_stateid(server, stateid);
6767                 if (err != -NFS4ERR_DELAY)
6768                         break;
6769                 nfs4_handle_exception(server, err, &exception);
6770         } while (exception.retry);
6771         return err;
6772 }
6773
6774 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6775 {
6776         struct nfs41_free_stateid_args args = {
6777                 .stateid = stateid,
6778         };
6779         struct nfs41_free_stateid_res res;
6780         struct rpc_message msg = {
6781                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6782                 .rpc_argp = &args,
6783                 .rpc_resp = &res,
6784         };
6785         int status;
6786
6787         dprintk("NFS call  free_stateid %p\n", stateid);
6788         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6789         status = nfs4_call_sync_sequence(server->client, server, &msg,
6790                                          &args.seq_args, &res.seq_res, 1);
6791         dprintk("NFS reply free_stateid: %d\n", status);
6792         return status;
6793 }
6794
6795 /**
6796  * nfs41_free_stateid - perform a FREE_STATEID operation
6797  *
6798  * @server: server / transport on which to perform the operation
6799  * @stateid: state ID to release
6800  *
6801  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6802  * negative NFS4ERR value is returned.
6803  */
6804 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6805 {
6806         struct nfs4_exception exception = { };
6807         int err;
6808         do {
6809                 err = _nfs4_free_stateid(server, stateid);
6810                 if (err != -NFS4ERR_DELAY)
6811                         break;
6812                 nfs4_handle_exception(server, err, &exception);
6813         } while (exception.retry);
6814         return err;
6815 }
6816
6817 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6818                 const nfs4_stateid *s2)
6819 {
6820         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6821                 return false;
6822
6823         if (s1->seqid == s2->seqid)
6824                 return true;
6825         if (s1->seqid == 0 || s2->seqid == 0)
6826                 return true;
6827
6828         return false;
6829 }
6830
6831 #endif /* CONFIG_NFS_V4_1 */
6832
6833 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6834                 const nfs4_stateid *s2)
6835 {
6836         return nfs4_stateid_match(s1, s2);
6837 }
6838
6839
6840 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6841         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6842         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6843         .recover_open   = nfs4_open_reclaim,
6844         .recover_lock   = nfs4_lock_reclaim,
6845         .establish_clid = nfs4_init_clientid,
6846         .get_clid_cred  = nfs4_get_setclientid_cred,
6847 };
6848
6849 #if defined(CONFIG_NFS_V4_1)
6850 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6851         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6852         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6853         .recover_open   = nfs4_open_reclaim,
6854         .recover_lock   = nfs4_lock_reclaim,
6855         .establish_clid = nfs41_init_clientid,
6856         .get_clid_cred  = nfs4_get_exchange_id_cred,
6857         .reclaim_complete = nfs41_proc_reclaim_complete,
6858 };
6859 #endif /* CONFIG_NFS_V4_1 */
6860
6861 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6862         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6863         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6864         .recover_open   = nfs4_open_expired,
6865         .recover_lock   = nfs4_lock_expired,
6866         .establish_clid = nfs4_init_clientid,
6867         .get_clid_cred  = nfs4_get_setclientid_cred,
6868 };
6869
6870 #if defined(CONFIG_NFS_V4_1)
6871 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6872         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6873         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6874         .recover_open   = nfs41_open_expired,
6875         .recover_lock   = nfs41_lock_expired,
6876         .establish_clid = nfs41_init_clientid,
6877         .get_clid_cred  = nfs4_get_exchange_id_cred,
6878 };
6879 #endif /* CONFIG_NFS_V4_1 */
6880
6881 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6882         .sched_state_renewal = nfs4_proc_async_renew,
6883         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6884         .renew_lease = nfs4_proc_renew,
6885 };
6886
6887 #if defined(CONFIG_NFS_V4_1)
6888 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6889         .sched_state_renewal = nfs41_proc_async_sequence,
6890         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6891         .renew_lease = nfs4_proc_sequence,
6892 };
6893 #endif
6894
6895 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6896         .minor_version = 0,
6897         .call_sync = _nfs4_call_sync,
6898         .match_stateid = nfs4_match_stateid,
6899         .find_root_sec = nfs4_find_root_sec,
6900         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6901         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6902         .state_renewal_ops = &nfs40_state_renewal_ops,
6903 };
6904
6905 #if defined(CONFIG_NFS_V4_1)
6906 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6907         .minor_version = 1,
6908         .call_sync = _nfs4_call_sync_session,
6909         .match_stateid = nfs41_match_stateid,
6910         .find_root_sec = nfs41_find_root_sec,
6911         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6912         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6913         .state_renewal_ops = &nfs41_state_renewal_ops,
6914 };
6915 #endif
6916
6917 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6918         [0] = &nfs_v4_0_minor_ops,
6919 #if defined(CONFIG_NFS_V4_1)
6920         [1] = &nfs_v4_1_minor_ops,
6921 #endif
6922 };
6923
6924 const struct inode_operations nfs4_dir_inode_operations = {
6925         .create         = nfs_create,
6926         .lookup         = nfs_lookup,
6927         .atomic_open    = nfs_atomic_open,
6928         .link           = nfs_link,
6929         .unlink         = nfs_unlink,
6930         .symlink        = nfs_symlink,
6931         .mkdir          = nfs_mkdir,
6932         .rmdir          = nfs_rmdir,
6933         .mknod          = nfs_mknod,
6934         .rename         = nfs_rename,
6935         .permission     = nfs_permission,
6936         .getattr        = nfs_getattr,
6937         .setattr        = nfs_setattr,
6938         .getxattr       = generic_getxattr,
6939         .setxattr       = generic_setxattr,
6940         .listxattr      = generic_listxattr,
6941         .removexattr    = generic_removexattr,
6942 };
6943
6944 static const struct inode_operations nfs4_file_inode_operations = {
6945         .permission     = nfs_permission,
6946         .getattr        = nfs_getattr,
6947         .setattr        = nfs_setattr,
6948         .getxattr       = generic_getxattr,
6949         .setxattr       = generic_setxattr,
6950         .listxattr      = generic_listxattr,
6951         .removexattr    = generic_removexattr,
6952 };
6953
6954 const struct nfs_rpc_ops nfs_v4_clientops = {
6955         .version        = 4,                    /* protocol version */
6956         .dentry_ops     = &nfs4_dentry_operations,
6957         .dir_inode_ops  = &nfs4_dir_inode_operations,
6958         .file_inode_ops = &nfs4_file_inode_operations,
6959         .file_ops       = &nfs4_file_operations,
6960         .getroot        = nfs4_proc_get_root,
6961         .submount       = nfs4_submount,
6962         .try_mount      = nfs4_try_mount,
6963         .getattr        = nfs4_proc_getattr,
6964         .setattr        = nfs4_proc_setattr,
6965         .lookup         = nfs4_proc_lookup,
6966         .access         = nfs4_proc_access,
6967         .readlink       = nfs4_proc_readlink,
6968         .create         = nfs4_proc_create,
6969         .remove         = nfs4_proc_remove,
6970         .unlink_setup   = nfs4_proc_unlink_setup,
6971         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
6972         .unlink_done    = nfs4_proc_unlink_done,
6973         .rename         = nfs4_proc_rename,
6974         .rename_setup   = nfs4_proc_rename_setup,
6975         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
6976         .rename_done    = nfs4_proc_rename_done,
6977         .link           = nfs4_proc_link,
6978         .symlink        = nfs4_proc_symlink,
6979         .mkdir          = nfs4_proc_mkdir,
6980         .rmdir          = nfs4_proc_remove,
6981         .readdir        = nfs4_proc_readdir,
6982         .mknod          = nfs4_proc_mknod,
6983         .statfs         = nfs4_proc_statfs,
6984         .fsinfo         = nfs4_proc_fsinfo,
6985         .pathconf       = nfs4_proc_pathconf,
6986         .set_capabilities = nfs4_server_capabilities,
6987         .decode_dirent  = nfs4_decode_dirent,
6988         .read_setup     = nfs4_proc_read_setup,
6989         .read_pageio_init = pnfs_pageio_init_read,
6990         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6991         .read_done      = nfs4_read_done,
6992         .write_setup    = nfs4_proc_write_setup,
6993         .write_pageio_init = pnfs_pageio_init_write,
6994         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6995         .write_done     = nfs4_write_done,
6996         .commit_setup   = nfs4_proc_commit_setup,
6997         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
6998         .commit_done    = nfs4_commit_done,
6999         .lock           = nfs4_proc_lock,
7000         .clear_acl_cache = nfs4_zap_acl_attr,
7001         .close_context  = nfs4_close_context,
7002         .open_context   = nfs4_atomic_open,
7003         .have_delegation = nfs4_have_delegation,
7004         .return_delegation = nfs4_inode_return_delegation,
7005         .alloc_client   = nfs4_alloc_client,
7006         .init_client    = nfs4_init_client,
7007         .free_client    = nfs4_free_client,
7008         .create_server  = nfs4_create_server,
7009         .clone_server   = nfs_clone_server,
7010 };
7011
7012 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7013         .prefix = XATTR_NAME_NFSV4_ACL,
7014         .list   = nfs4_xattr_list_nfs4_acl,
7015         .get    = nfs4_xattr_get_nfs4_acl,
7016         .set    = nfs4_xattr_set_nfs4_acl,
7017 };
7018
7019 const struct xattr_handler *nfs4_xattr_handlers[] = {
7020         &nfs4_xattr_nfs4_acl_handler,
7021         NULL
7022 };
7023
7024 /*
7025  * Local variables:
7026  *  c-basic-offset: 8
7027  * End:
7028  */