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                 /* block layout checks this! */
3407                 server->pnfs_blksize = fsinfo->blksize;
3408                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
3409         }
3410
3411         return error;
3412 }
3413
3414 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3415                 struct nfs_pathconf *pathconf)
3416 {
3417         struct nfs4_pathconf_arg args = {
3418                 .fh = fhandle,
3419                 .bitmask = server->attr_bitmask,
3420         };
3421         struct nfs4_pathconf_res res = {
3422                 .pathconf = pathconf,
3423         };
3424         struct rpc_message msg = {
3425                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3426                 .rpc_argp = &args,
3427                 .rpc_resp = &res,
3428         };
3429
3430         /* None of the pathconf attributes are mandatory to implement */
3431         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3432                 memset(pathconf, 0, sizeof(*pathconf));
3433                 return 0;
3434         }
3435
3436         nfs_fattr_init(pathconf->fattr);
3437         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3438 }
3439
3440 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3441                 struct nfs_pathconf *pathconf)
3442 {
3443         struct nfs4_exception exception = { };
3444         int err;
3445
3446         do {
3447                 err = nfs4_handle_exception(server,
3448                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3449                                 &exception);
3450         } while (exception.retry);
3451         return err;
3452 }
3453
3454 void __nfs4_read_done_cb(struct nfs_read_data *data)
3455 {
3456         nfs_invalidate_atime(data->header->inode);
3457 }
3458
3459 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3460 {
3461         struct nfs_server *server = NFS_SERVER(data->header->inode);
3462
3463         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3464                 rpc_restart_call_prepare(task);
3465                 return -EAGAIN;
3466         }
3467
3468         __nfs4_read_done_cb(data);
3469         if (task->tk_status > 0)
3470                 renew_lease(server, data->timestamp);
3471         return 0;
3472 }
3473
3474 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3475 {
3476
3477         dprintk("--> %s\n", __func__);
3478
3479         if (!nfs4_sequence_done(task, &data->res.seq_res))
3480                 return -EAGAIN;
3481
3482         return data->read_done_cb ? data->read_done_cb(task, data) :
3483                                     nfs4_read_done_cb(task, data);
3484 }
3485
3486 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3487 {
3488         data->timestamp   = jiffies;
3489         data->read_done_cb = nfs4_read_done_cb;
3490         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3491         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3492 }
3493
3494 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3495 {
3496         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3497                                 &data->args.seq_args,
3498                                 &data->res.seq_res,
3499                                 task))
3500                 return;
3501         rpc_call_start(task);
3502 }
3503
3504 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3505 {
3506         struct inode *inode = data->header->inode;
3507
3508         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3509                 rpc_restart_call_prepare(task);
3510                 return -EAGAIN;
3511         }
3512         if (task->tk_status >= 0) {
3513                 renew_lease(NFS_SERVER(inode), data->timestamp);
3514                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
3515         }
3516         return 0;
3517 }
3518
3519 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3520 {
3521         if (!nfs4_sequence_done(task, &data->res.seq_res))
3522                 return -EAGAIN;
3523         return data->write_done_cb ? data->write_done_cb(task, data) :
3524                 nfs4_write_done_cb(task, data);
3525 }
3526
3527 static
3528 bool nfs4_write_need_cache_consistency_data(const struct nfs_write_data *data)
3529 {
3530         const struct nfs_pgio_header *hdr = data->header;
3531
3532         /* Don't request attributes for pNFS or O_DIRECT writes */
3533         if (data->ds_clp != NULL || hdr->dreq != NULL)
3534                 return false;
3535         /* Otherwise, request attributes if and only if we don't hold
3536          * a delegation
3537          */
3538         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
3539 }
3540
3541 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3542 {
3543         struct nfs_server *server = NFS_SERVER(data->header->inode);
3544
3545         if (!nfs4_write_need_cache_consistency_data(data)) {
3546                 data->args.bitmask = NULL;
3547                 data->res.fattr = NULL;
3548         } else
3549                 data->args.bitmask = server->cache_consistency_bitmask;
3550
3551         if (!data->write_done_cb)
3552                 data->write_done_cb = nfs4_write_done_cb;
3553         data->res.server = server;
3554         data->timestamp   = jiffies;
3555
3556         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3557         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3558 }
3559
3560 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3561 {
3562         if (nfs4_setup_sequence(NFS_SERVER(data->header->inode),
3563                                 &data->args.seq_args,
3564                                 &data->res.seq_res,
3565                                 task))
3566                 return;
3567         rpc_call_start(task);
3568 }
3569
3570 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
3571 {
3572         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3573                                 &data->args.seq_args,
3574                                 &data->res.seq_res,
3575                                 task))
3576                 return;
3577         rpc_call_start(task);
3578 }
3579
3580 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
3581 {
3582         struct inode *inode = data->inode;
3583
3584         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3585                 rpc_restart_call_prepare(task);
3586                 return -EAGAIN;
3587         }
3588         return 0;
3589 }
3590
3591 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
3592 {
3593         if (!nfs4_sequence_done(task, &data->res.seq_res))
3594                 return -EAGAIN;
3595         return data->commit_done_cb(task, data);
3596 }
3597
3598 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
3599 {
3600         struct nfs_server *server = NFS_SERVER(data->inode);
3601
3602         if (data->commit_done_cb == NULL)
3603                 data->commit_done_cb = nfs4_commit_done_cb;
3604         data->res.server = server;
3605         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3606         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3607 }
3608
3609 struct nfs4_renewdata {
3610         struct nfs_client       *client;
3611         unsigned long           timestamp;
3612 };
3613
3614 /*
3615  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3616  * standalone procedure for queueing an asynchronous RENEW.
3617  */
3618 static void nfs4_renew_release(void *calldata)
3619 {
3620         struct nfs4_renewdata *data = calldata;
3621         struct nfs_client *clp = data->client;
3622
3623         if (atomic_read(&clp->cl_count) > 1)
3624                 nfs4_schedule_state_renewal(clp);
3625         nfs_put_client(clp);
3626         kfree(data);
3627 }
3628
3629 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3630 {
3631         struct nfs4_renewdata *data = calldata;
3632         struct nfs_client *clp = data->client;
3633         unsigned long timestamp = data->timestamp;
3634
3635         if (task->tk_status < 0) {
3636                 /* Unless we're shutting down, schedule state recovery! */
3637                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3638                         return;
3639                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3640                         nfs4_schedule_lease_recovery(clp);
3641                         return;
3642                 }
3643                 nfs4_schedule_path_down_recovery(clp);
3644         }
3645         do_renew_lease(clp, timestamp);
3646 }
3647
3648 static const struct rpc_call_ops nfs4_renew_ops = {
3649         .rpc_call_done = nfs4_renew_done,
3650         .rpc_release = nfs4_renew_release,
3651 };
3652
3653 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3654 {
3655         struct rpc_message msg = {
3656                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3657                 .rpc_argp       = clp,
3658                 .rpc_cred       = cred,
3659         };
3660         struct nfs4_renewdata *data;
3661
3662         if (renew_flags == 0)
3663                 return 0;
3664         if (!atomic_inc_not_zero(&clp->cl_count))
3665                 return -EIO;
3666         data = kmalloc(sizeof(*data), GFP_NOFS);
3667         if (data == NULL)
3668                 return -ENOMEM;
3669         data->client = clp;
3670         data->timestamp = jiffies;
3671         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3672                         &nfs4_renew_ops, data);
3673 }
3674
3675 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3676 {
3677         struct rpc_message msg = {
3678                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3679                 .rpc_argp       = clp,
3680                 .rpc_cred       = cred,
3681         };
3682         unsigned long now = jiffies;
3683         int status;
3684
3685         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3686         if (status < 0)
3687                 return status;
3688         do_renew_lease(clp, now);
3689         return 0;
3690 }
3691
3692 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3693 {
3694         return (server->caps & NFS_CAP_ACLS)
3695                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3696                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3697 }
3698
3699 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
3700  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
3701  * the stack.
3702  */
3703 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
3704
3705 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3706                 struct page **pages, unsigned int *pgbase)
3707 {
3708         struct page *newpage, **spages;
3709         int rc = 0;
3710         size_t len;
3711         spages = pages;
3712
3713         do {
3714                 len = min_t(size_t, PAGE_SIZE, buflen);
3715                 newpage = alloc_page(GFP_KERNEL);
3716
3717                 if (newpage == NULL)
3718                         goto unwind;
3719                 memcpy(page_address(newpage), buf, len);
3720                 buf += len;
3721                 buflen -= len;
3722                 *pages++ = newpage;
3723                 rc++;
3724         } while (buflen != 0);
3725
3726         return rc;
3727
3728 unwind:
3729         for(; rc > 0; rc--)
3730                 __free_page(spages[rc-1]);
3731         return -ENOMEM;
3732 }
3733
3734 struct nfs4_cached_acl {
3735         int cached;
3736         size_t len;
3737         char data[0];
3738 };
3739
3740 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3741 {
3742         struct nfs_inode *nfsi = NFS_I(inode);
3743
3744         spin_lock(&inode->i_lock);
3745         kfree(nfsi->nfs4_acl);
3746         nfsi->nfs4_acl = acl;
3747         spin_unlock(&inode->i_lock);
3748 }
3749
3750 static void nfs4_zap_acl_attr(struct inode *inode)
3751 {
3752         nfs4_set_cached_acl(inode, NULL);
3753 }
3754
3755 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3756 {
3757         struct nfs_inode *nfsi = NFS_I(inode);
3758         struct nfs4_cached_acl *acl;
3759         int ret = -ENOENT;
3760
3761         spin_lock(&inode->i_lock);
3762         acl = nfsi->nfs4_acl;
3763         if (acl == NULL)
3764                 goto out;
3765         if (buf == NULL) /* user is just asking for length */
3766                 goto out_len;
3767         if (acl->cached == 0)
3768                 goto out;
3769         ret = -ERANGE; /* see getxattr(2) man page */
3770         if (acl->len > buflen)
3771                 goto out;
3772         memcpy(buf, acl->data, acl->len);
3773 out_len:
3774         ret = acl->len;
3775 out:
3776         spin_unlock(&inode->i_lock);
3777         return ret;
3778 }
3779
3780 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
3781 {
3782         struct nfs4_cached_acl *acl;
3783         size_t buflen = sizeof(*acl) + acl_len;
3784
3785         if (buflen <= PAGE_SIZE) {
3786                 acl = kmalloc(buflen, GFP_KERNEL);
3787                 if (acl == NULL)
3788                         goto out;
3789                 acl->cached = 1;
3790                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
3791         } else {
3792                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3793                 if (acl == NULL)
3794                         goto out;
3795                 acl->cached = 0;
3796         }
3797         acl->len = acl_len;
3798 out:
3799         nfs4_set_cached_acl(inode, acl);
3800 }
3801
3802 /*
3803  * The getxattr API returns the required buffer length when called with a
3804  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3805  * the required buf.  On a NULL buf, we send a page of data to the server
3806  * guessing that the ACL request can be serviced by a page. If so, we cache
3807  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3808  * the cache. If not so, we throw away the page, and cache the required
3809  * length. The next getxattr call will then produce another round trip to
3810  * the server, this time with the input buf of the required size.
3811  */
3812 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3813 {
3814         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3815         struct nfs_getaclargs args = {
3816                 .fh = NFS_FH(inode),
3817                 .acl_pages = pages,
3818                 .acl_len = buflen,
3819         };
3820         struct nfs_getaclres res = {
3821                 .acl_len = buflen,
3822         };
3823         struct rpc_message msg = {
3824                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3825                 .rpc_argp = &args,
3826                 .rpc_resp = &res,
3827         };
3828         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3829         int ret = -ENOMEM, i;
3830
3831         /* As long as we're doing a round trip to the server anyway,
3832          * let's be prepared for a page of acl data. */
3833         if (npages == 0)
3834                 npages = 1;
3835         if (npages > ARRAY_SIZE(pages))
3836                 return -ERANGE;
3837
3838         for (i = 0; i < npages; i++) {
3839                 pages[i] = alloc_page(GFP_KERNEL);
3840                 if (!pages[i])
3841                         goto out_free;
3842         }
3843
3844         /* for decoding across pages */
3845         res.acl_scratch = alloc_page(GFP_KERNEL);
3846         if (!res.acl_scratch)
3847                 goto out_free;
3848
3849         args.acl_len = npages * PAGE_SIZE;
3850         args.acl_pgbase = 0;
3851
3852         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3853                 __func__, buf, buflen, npages, args.acl_len);
3854         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3855                              &msg, &args.seq_args, &res.seq_res, 0);
3856         if (ret)
3857                 goto out_free;
3858
3859         /* Handle the case where the passed-in buffer is too short */
3860         if (res.acl_flags & NFS4_ACL_TRUNC) {
3861                 /* Did the user only issue a request for the acl length? */
3862                 if (buf == NULL)
3863                         goto out_ok;
3864                 ret = -ERANGE;
3865                 goto out_free;
3866         }
3867         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
3868         if (buf)
3869                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
3870 out_ok:
3871         ret = res.acl_len;
3872 out_free:
3873         for (i = 0; i < npages; i++)
3874                 if (pages[i])
3875                         __free_page(pages[i]);
3876         if (res.acl_scratch)
3877                 __free_page(res.acl_scratch);
3878         return ret;
3879 }
3880
3881 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3882 {
3883         struct nfs4_exception exception = { };
3884         ssize_t ret;
3885         do {
3886                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3887                 if (ret >= 0)
3888                         break;
3889                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3890         } while (exception.retry);
3891         return ret;
3892 }
3893
3894 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3895 {
3896         struct nfs_server *server = NFS_SERVER(inode);
3897         int ret;
3898
3899         if (!nfs4_server_supports_acls(server))
3900                 return -EOPNOTSUPP;
3901         ret = nfs_revalidate_inode(server, inode);
3902         if (ret < 0)
3903                 return ret;
3904         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3905                 nfs_zap_acl_cache(inode);
3906         ret = nfs4_read_cached_acl(inode, buf, buflen);
3907         if (ret != -ENOENT)
3908                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3909                  * but no cached acl data, just the acl length */
3910                 return ret;
3911         return nfs4_get_acl_uncached(inode, buf, buflen);
3912 }
3913
3914 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3915 {
3916         struct nfs_server *server = NFS_SERVER(inode);
3917         struct page *pages[NFS4ACL_MAXPAGES];
3918         struct nfs_setaclargs arg = {
3919                 .fh             = NFS_FH(inode),
3920                 .acl_pages      = pages,
3921                 .acl_len        = buflen,
3922         };
3923         struct nfs_setaclres res;
3924         struct rpc_message msg = {
3925                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3926                 .rpc_argp       = &arg,
3927                 .rpc_resp       = &res,
3928         };
3929         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
3930         int ret, i;
3931
3932         if (!nfs4_server_supports_acls(server))
3933                 return -EOPNOTSUPP;
3934         if (npages > ARRAY_SIZE(pages))
3935                 return -ERANGE;
3936         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3937         if (i < 0)
3938                 return i;
3939         nfs4_inode_return_delegation(inode);
3940         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3941
3942         /*
3943          * Free each page after tx, so the only ref left is
3944          * held by the network stack
3945          */
3946         for (; i > 0; i--)
3947                 put_page(pages[i-1]);
3948
3949         /*
3950          * Acl update can result in inode attribute update.
3951          * so mark the attribute cache invalid.
3952          */
3953         spin_lock(&inode->i_lock);
3954         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3955         spin_unlock(&inode->i_lock);
3956         nfs_access_zap_cache(inode);
3957         nfs_zap_acl_cache(inode);
3958         return ret;
3959 }
3960
3961 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3962 {
3963         struct nfs4_exception exception = { };
3964         int err;
3965         do {
3966                 err = nfs4_handle_exception(NFS_SERVER(inode),
3967                                 __nfs4_proc_set_acl(inode, buf, buflen),
3968                                 &exception);
3969         } while (exception.retry);
3970         return err;
3971 }
3972
3973 static int
3974 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3975 {
3976         struct nfs_client *clp = server->nfs_client;
3977
3978         if (task->tk_status >= 0)
3979                 return 0;
3980         switch(task->tk_status) {
3981                 case -NFS4ERR_DELEG_REVOKED:
3982                 case -NFS4ERR_ADMIN_REVOKED:
3983                 case -NFS4ERR_BAD_STATEID:
3984                         if (state == NULL)
3985                                 break;
3986                         nfs_remove_bad_delegation(state->inode);
3987                 case -NFS4ERR_OPENMODE:
3988                         if (state == NULL)
3989                                 break;
3990                         nfs4_schedule_stateid_recovery(server, state);
3991                         goto wait_on_recovery;
3992                 case -NFS4ERR_EXPIRED:
3993                         if (state != NULL)
3994                                 nfs4_schedule_stateid_recovery(server, state);
3995                 case -NFS4ERR_STALE_STATEID:
3996                 case -NFS4ERR_STALE_CLIENTID:
3997                         nfs4_schedule_lease_recovery(clp);
3998                         goto wait_on_recovery;
3999 #if defined(CONFIG_NFS_V4_1)
4000                 case -NFS4ERR_BADSESSION:
4001                 case -NFS4ERR_BADSLOT:
4002                 case -NFS4ERR_BAD_HIGH_SLOT:
4003                 case -NFS4ERR_DEADSESSION:
4004                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4005                 case -NFS4ERR_SEQ_FALSE_RETRY:
4006                 case -NFS4ERR_SEQ_MISORDERED:
4007                         dprintk("%s ERROR %d, Reset session\n", __func__,
4008                                 task->tk_status);
4009                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4010                         task->tk_status = 0;
4011                         return -EAGAIN;
4012 #endif /* CONFIG_NFS_V4_1 */
4013                 case -NFS4ERR_DELAY:
4014                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4015                 case -NFS4ERR_GRACE:
4016                 case -EKEYEXPIRED:
4017                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4018                         task->tk_status = 0;
4019                         return -EAGAIN;
4020                 case -NFS4ERR_RETRY_UNCACHED_REP:
4021                 case -NFS4ERR_OLD_STATEID:
4022                         task->tk_status = 0;
4023                         return -EAGAIN;
4024         }
4025         task->tk_status = nfs4_map_errors(task->tk_status);
4026         return 0;
4027 wait_on_recovery:
4028         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4029         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4030                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4031         task->tk_status = 0;
4032         return -EAGAIN;
4033 }
4034
4035 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4036                                     nfs4_verifier *bootverf)
4037 {
4038         __be32 verf[2];
4039
4040         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4041                 /* An impossible timestamp guarantees this value
4042                  * will never match a generated boot time. */
4043                 verf[0] = 0;
4044                 verf[1] = (__be32)(NSEC_PER_SEC + 1);
4045         } else {
4046                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4047                 verf[0] = (__be32)nn->boot_time.tv_sec;
4048                 verf[1] = (__be32)nn->boot_time.tv_nsec;
4049         }
4050         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4051 }
4052
4053 static unsigned int
4054 nfs4_init_nonuniform_client_string(const struct nfs_client *clp,
4055                                    char *buf, size_t len)
4056 {
4057         unsigned int result;
4058
4059         rcu_read_lock();
4060         result = scnprintf(buf, len, "Linux NFSv4.0 %s/%s %s",
4061                                 clp->cl_ipaddr,
4062                                 rpc_peeraddr2str(clp->cl_rpcclient,
4063                                                         RPC_DISPLAY_ADDR),
4064                                 rpc_peeraddr2str(clp->cl_rpcclient,
4065                                                         RPC_DISPLAY_PROTO));
4066         rcu_read_unlock();
4067         return result;
4068 }
4069
4070 static unsigned int
4071 nfs4_init_uniform_client_string(const struct nfs_client *clp,
4072                                 char *buf, size_t len)
4073 {
4074         char *nodename = clp->cl_rpcclient->cl_nodename;
4075
4076         if (nfs4_client_id_uniquifier[0] != '\0')
4077                 nodename = nfs4_client_id_uniquifier;
4078         return scnprintf(buf, len, "Linux NFSv%u.%u %s",
4079                                 clp->rpc_ops->version, clp->cl_minorversion,
4080                                 nodename);
4081 }
4082
4083 /**
4084  * nfs4_proc_setclientid - Negotiate client ID
4085  * @clp: state data structure
4086  * @program: RPC program for NFSv4 callback service
4087  * @port: IP port number for NFS4 callback service
4088  * @cred: RPC credential to use for this call
4089  * @res: where to place the result
4090  *
4091  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4092  */
4093 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
4094                 unsigned short port, struct rpc_cred *cred,
4095                 struct nfs4_setclientid_res *res)
4096 {
4097         nfs4_verifier sc_verifier;
4098         struct nfs4_setclientid setclientid = {
4099                 .sc_verifier = &sc_verifier,
4100                 .sc_prog = program,
4101                 .sc_cb_ident = clp->cl_cb_ident,
4102         };
4103         struct rpc_message msg = {
4104                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
4105                 .rpc_argp = &setclientid,
4106                 .rpc_resp = res,
4107                 .rpc_cred = cred,
4108         };
4109         int status;
4110
4111         /* nfs_client_id4 */
4112         nfs4_init_boot_verifier(clp, &sc_verifier);
4113         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
4114                 setclientid.sc_name_len =
4115                                 nfs4_init_uniform_client_string(clp,
4116                                                 setclientid.sc_name,
4117                                                 sizeof(setclientid.sc_name));
4118         else
4119                 setclientid.sc_name_len =
4120                                 nfs4_init_nonuniform_client_string(clp,
4121                                                 setclientid.sc_name,
4122                                                 sizeof(setclientid.sc_name));
4123         /* cb_client4 */
4124         rcu_read_lock();
4125         setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
4126                                 sizeof(setclientid.sc_netid),
4127                                 rpc_peeraddr2str(clp->cl_rpcclient,
4128                                                         RPC_DISPLAY_NETID));
4129         rcu_read_unlock();
4130         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
4131                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
4132                                 clp->cl_ipaddr, port >> 8, port & 255);
4133
4134         dprintk("NFS call  setclientid auth=%s, '%.*s'\n",
4135                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4136                 setclientid.sc_name_len, setclientid.sc_name);
4137         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4138         dprintk("NFS reply setclientid: %d\n", status);
4139         return status;
4140 }
4141
4142 /**
4143  * nfs4_proc_setclientid_confirm - Confirm client ID
4144  * @clp: state data structure
4145  * @res: result of a previous SETCLIENTID
4146  * @cred: RPC credential to use for this call
4147  *
4148  * Returns zero, a negative errno, or a negative NFS4ERR status code.
4149  */
4150 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
4151                 struct nfs4_setclientid_res *arg,
4152                 struct rpc_cred *cred)
4153 {
4154         struct nfs_fsinfo fsinfo;
4155         struct rpc_message msg = {
4156                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
4157                 .rpc_argp = arg,
4158                 .rpc_resp = &fsinfo,
4159                 .rpc_cred = cred,
4160         };
4161         unsigned long now;
4162         int status;
4163
4164         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
4165                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
4166                 clp->cl_clientid);
4167         now = jiffies;
4168         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4169         if (status == 0) {
4170                 spin_lock(&clp->cl_lock);
4171                 clp->cl_lease_time = fsinfo.lease_time * HZ;
4172                 clp->cl_last_renewal = now;
4173                 spin_unlock(&clp->cl_lock);
4174         }
4175         dprintk("NFS reply setclientid_confirm: %d\n", status);
4176         return status;
4177 }
4178
4179 struct nfs4_delegreturndata {
4180         struct nfs4_delegreturnargs args;
4181         struct nfs4_delegreturnres res;
4182         struct nfs_fh fh;
4183         nfs4_stateid stateid;
4184         unsigned long timestamp;
4185         struct nfs_fattr fattr;
4186         int rpc_status;
4187 };
4188
4189 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
4190 {
4191         struct nfs4_delegreturndata *data = calldata;
4192
4193         if (!nfs4_sequence_done(task, &data->res.seq_res))
4194                 return;
4195
4196         switch (task->tk_status) {
4197         case -NFS4ERR_STALE_STATEID:
4198         case -NFS4ERR_EXPIRED:
4199         case 0:
4200                 renew_lease(data->res.server, data->timestamp);
4201                 break;
4202         default:
4203                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
4204                                 -EAGAIN) {
4205                         rpc_restart_call_prepare(task);
4206                         return;
4207                 }
4208         }
4209         data->rpc_status = task->tk_status;
4210 }
4211
4212 static void nfs4_delegreturn_release(void *calldata)
4213 {
4214         kfree(calldata);
4215 }
4216
4217 #if defined(CONFIG_NFS_V4_1)
4218 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
4219 {
4220         struct nfs4_delegreturndata *d_data;
4221
4222         d_data = (struct nfs4_delegreturndata *)data;
4223
4224         if (nfs4_setup_sequence(d_data->res.server,
4225                                 &d_data->args.seq_args,
4226                                 &d_data->res.seq_res, task))
4227                 return;
4228         rpc_call_start(task);
4229 }
4230 #endif /* CONFIG_NFS_V4_1 */
4231
4232 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4233 #if defined(CONFIG_NFS_V4_1)
4234         .rpc_call_prepare = nfs4_delegreturn_prepare,
4235 #endif /* CONFIG_NFS_V4_1 */
4236         .rpc_call_done = nfs4_delegreturn_done,
4237         .rpc_release = nfs4_delegreturn_release,
4238 };
4239
4240 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4241 {
4242         struct nfs4_delegreturndata *data;
4243         struct nfs_server *server = NFS_SERVER(inode);
4244         struct rpc_task *task;
4245         struct rpc_message msg = {
4246                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4247                 .rpc_cred = cred,
4248         };
4249         struct rpc_task_setup task_setup_data = {
4250                 .rpc_client = server->client,
4251                 .rpc_message = &msg,
4252                 .callback_ops = &nfs4_delegreturn_ops,
4253                 .flags = RPC_TASK_ASYNC,
4254         };
4255         int status = 0;
4256
4257         data = kzalloc(sizeof(*data), GFP_NOFS);
4258         if (data == NULL)
4259                 return -ENOMEM;
4260         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4261         data->args.fhandle = &data->fh;
4262         data->args.stateid = &data->stateid;
4263         data->args.bitmask = server->cache_consistency_bitmask;
4264         nfs_copy_fh(&data->fh, NFS_FH(inode));
4265         nfs4_stateid_copy(&data->stateid, stateid);
4266         data->res.fattr = &data->fattr;
4267         data->res.server = server;
4268         nfs_fattr_init(data->res.fattr);
4269         data->timestamp = jiffies;
4270         data->rpc_status = 0;
4271
4272         task_setup_data.callback_data = data;
4273         msg.rpc_argp = &data->args;
4274         msg.rpc_resp = &data->res;
4275         task = rpc_run_task(&task_setup_data);
4276         if (IS_ERR(task))
4277                 return PTR_ERR(task);
4278         if (!issync)
4279                 goto out;
4280         status = nfs4_wait_for_completion_rpc_task(task);
4281         if (status != 0)
4282                 goto out;
4283         status = data->rpc_status;
4284         if (status == 0)
4285                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
4286         else
4287                 nfs_refresh_inode(inode, &data->fattr);
4288 out:
4289         rpc_put_task(task);
4290         return status;
4291 }
4292
4293 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4294 {
4295         struct nfs_server *server = NFS_SERVER(inode);
4296         struct nfs4_exception exception = { };
4297         int err;
4298         do {
4299                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4300                 switch (err) {
4301                         case -NFS4ERR_STALE_STATEID:
4302                         case -NFS4ERR_EXPIRED:
4303                         case 0:
4304                                 return 0;
4305                 }
4306                 err = nfs4_handle_exception(server, err, &exception);
4307         } while (exception.retry);
4308         return err;
4309 }
4310
4311 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4312 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4313
4314 /*
4315  * sleep, with exponential backoff, and retry the LOCK operation.
4316  */
4317 static unsigned long
4318 nfs4_set_lock_task_retry(unsigned long timeout)
4319 {
4320         freezable_schedule_timeout_killable(timeout);
4321         timeout <<= 1;
4322         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4323                 return NFS4_LOCK_MAXTIMEOUT;
4324         return timeout;
4325 }
4326
4327 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4328 {
4329         struct inode *inode = state->inode;
4330         struct nfs_server *server = NFS_SERVER(inode);
4331         struct nfs_client *clp = server->nfs_client;
4332         struct nfs_lockt_args arg = {
4333                 .fh = NFS_FH(inode),
4334                 .fl = request,
4335         };
4336         struct nfs_lockt_res res = {
4337                 .denied = request,
4338         };
4339         struct rpc_message msg = {
4340                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4341                 .rpc_argp       = &arg,
4342                 .rpc_resp       = &res,
4343                 .rpc_cred       = state->owner->so_cred,
4344         };
4345         struct nfs4_lock_state *lsp;
4346         int status;
4347
4348         arg.lock_owner.clientid = clp->cl_clientid;
4349         status = nfs4_set_lock_state(state, request);
4350         if (status != 0)
4351                 goto out;
4352         lsp = request->fl_u.nfs4_fl.owner;
4353         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4354         arg.lock_owner.s_dev = server->s_dev;
4355         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4356         switch (status) {
4357                 case 0:
4358                         request->fl_type = F_UNLCK;
4359                         break;
4360                 case -NFS4ERR_DENIED:
4361                         status = 0;
4362         }
4363         request->fl_ops->fl_release_private(request);
4364 out:
4365         return status;
4366 }
4367
4368 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4369 {
4370         struct nfs4_exception exception = { };
4371         int err;
4372
4373         do {
4374                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4375                                 _nfs4_proc_getlk(state, cmd, request),
4376                                 &exception);
4377         } while (exception.retry);
4378         return err;
4379 }
4380
4381 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4382 {
4383         int res = 0;
4384         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4385                 case FL_POSIX:
4386                         res = posix_lock_file_wait(file, fl);
4387                         break;
4388                 case FL_FLOCK:
4389                         res = flock_lock_file_wait(file, fl);
4390                         break;
4391                 default:
4392                         BUG();
4393         }
4394         return res;
4395 }
4396
4397 struct nfs4_unlockdata {
4398         struct nfs_locku_args arg;
4399         struct nfs_locku_res res;
4400         struct nfs4_lock_state *lsp;
4401         struct nfs_open_context *ctx;
4402         struct file_lock fl;
4403         const struct nfs_server *server;
4404         unsigned long timestamp;
4405 };
4406
4407 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4408                 struct nfs_open_context *ctx,
4409                 struct nfs4_lock_state *lsp,
4410                 struct nfs_seqid *seqid)
4411 {
4412         struct nfs4_unlockdata *p;
4413         struct inode *inode = lsp->ls_state->inode;
4414
4415         p = kzalloc(sizeof(*p), GFP_NOFS);
4416         if (p == NULL)
4417                 return NULL;
4418         p->arg.fh = NFS_FH(inode);
4419         p->arg.fl = &p->fl;
4420         p->arg.seqid = seqid;
4421         p->res.seqid = seqid;
4422         p->arg.stateid = &lsp->ls_stateid;
4423         p->lsp = lsp;
4424         atomic_inc(&lsp->ls_count);
4425         /* Ensure we don't close file until we're done freeing locks! */
4426         p->ctx = get_nfs_open_context(ctx);
4427         memcpy(&p->fl, fl, sizeof(p->fl));
4428         p->server = NFS_SERVER(inode);
4429         return p;
4430 }
4431
4432 static void nfs4_locku_release_calldata(void *data)
4433 {
4434         struct nfs4_unlockdata *calldata = data;
4435         nfs_free_seqid(calldata->arg.seqid);
4436         nfs4_put_lock_state(calldata->lsp);
4437         put_nfs_open_context(calldata->ctx);
4438         kfree(calldata);
4439 }
4440
4441 static void nfs4_locku_done(struct rpc_task *task, void *data)
4442 {
4443         struct nfs4_unlockdata *calldata = data;
4444
4445         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4446                 return;
4447         switch (task->tk_status) {
4448                 case 0:
4449                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4450                                         &calldata->res.stateid);
4451                         renew_lease(calldata->server, calldata->timestamp);
4452                         break;
4453                 case -NFS4ERR_BAD_STATEID:
4454                 case -NFS4ERR_OLD_STATEID:
4455                 case -NFS4ERR_STALE_STATEID:
4456                 case -NFS4ERR_EXPIRED:
4457                         break;
4458                 default:
4459                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4460                                 rpc_restart_call_prepare(task);
4461         }
4462 }
4463
4464 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4465 {
4466         struct nfs4_unlockdata *calldata = data;
4467
4468         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4469                 return;
4470         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
4471                 /* Note: exit _without_ running nfs4_locku_done */
4472                 task->tk_action = NULL;
4473                 return;
4474         }
4475         calldata->timestamp = jiffies;
4476         if (nfs4_setup_sequence(calldata->server,
4477                                 &calldata->arg.seq_args,
4478                                 &calldata->res.seq_res, task))
4479                 return;
4480         rpc_call_start(task);
4481 }
4482
4483 static const struct rpc_call_ops nfs4_locku_ops = {
4484         .rpc_call_prepare = nfs4_locku_prepare,
4485         .rpc_call_done = nfs4_locku_done,
4486         .rpc_release = nfs4_locku_release_calldata,
4487 };
4488
4489 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4490                 struct nfs_open_context *ctx,
4491                 struct nfs4_lock_state *lsp,
4492                 struct nfs_seqid *seqid)
4493 {
4494         struct nfs4_unlockdata *data;
4495         struct rpc_message msg = {
4496                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4497                 .rpc_cred = ctx->cred,
4498         };
4499         struct rpc_task_setup task_setup_data = {
4500                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4501                 .rpc_message = &msg,
4502                 .callback_ops = &nfs4_locku_ops,
4503                 .workqueue = nfsiod_workqueue,
4504                 .flags = RPC_TASK_ASYNC,
4505         };
4506
4507         /* Ensure this is an unlock - when canceling a lock, the
4508          * canceled lock is passed in, and it won't be an unlock.
4509          */
4510         fl->fl_type = F_UNLCK;
4511
4512         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4513         if (data == NULL) {
4514                 nfs_free_seqid(seqid);
4515                 return ERR_PTR(-ENOMEM);
4516         }
4517
4518         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4519         msg.rpc_argp = &data->arg;
4520         msg.rpc_resp = &data->res;
4521         task_setup_data.callback_data = data;
4522         return rpc_run_task(&task_setup_data);
4523 }
4524
4525 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4526 {
4527         struct nfs_inode *nfsi = NFS_I(state->inode);
4528         struct nfs_seqid *seqid;
4529         struct nfs4_lock_state *lsp;
4530         struct rpc_task *task;
4531         int status = 0;
4532         unsigned char fl_flags = request->fl_flags;
4533
4534         status = nfs4_set_lock_state(state, request);
4535         /* Unlock _before_ we do the RPC call */
4536         request->fl_flags |= FL_EXISTS;
4537         down_read(&nfsi->rwsem);
4538         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4539                 up_read(&nfsi->rwsem);
4540                 goto out;
4541         }
4542         up_read(&nfsi->rwsem);
4543         if (status != 0)
4544                 goto out;
4545         /* Is this a delegated lock? */
4546         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4547                 goto out;
4548         lsp = request->fl_u.nfs4_fl.owner;
4549         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4550         status = -ENOMEM;
4551         if (seqid == NULL)
4552                 goto out;
4553         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4554         status = PTR_ERR(task);
4555         if (IS_ERR(task))
4556                 goto out;
4557         status = nfs4_wait_for_completion_rpc_task(task);
4558         rpc_put_task(task);
4559 out:
4560         request->fl_flags = fl_flags;
4561         return status;
4562 }
4563
4564 struct nfs4_lockdata {
4565         struct nfs_lock_args arg;
4566         struct nfs_lock_res res;
4567         struct nfs4_lock_state *lsp;
4568         struct nfs_open_context *ctx;
4569         struct file_lock fl;
4570         unsigned long timestamp;
4571         int rpc_status;
4572         int cancelled;
4573         struct nfs_server *server;
4574 };
4575
4576 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4577                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4578                 gfp_t gfp_mask)
4579 {
4580         struct nfs4_lockdata *p;
4581         struct inode *inode = lsp->ls_state->inode;
4582         struct nfs_server *server = NFS_SERVER(inode);
4583
4584         p = kzalloc(sizeof(*p), gfp_mask);
4585         if (p == NULL)
4586                 return NULL;
4587
4588         p->arg.fh = NFS_FH(inode);
4589         p->arg.fl = &p->fl;
4590         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4591         if (p->arg.open_seqid == NULL)
4592                 goto out_free;
4593         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4594         if (p->arg.lock_seqid == NULL)
4595                 goto out_free_seqid;
4596         p->arg.lock_stateid = &lsp->ls_stateid;
4597         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4598         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4599         p->arg.lock_owner.s_dev = server->s_dev;
4600         p->res.lock_seqid = p->arg.lock_seqid;
4601         p->lsp = lsp;
4602         p->server = server;
4603         atomic_inc(&lsp->ls_count);
4604         p->ctx = get_nfs_open_context(ctx);
4605         memcpy(&p->fl, fl, sizeof(p->fl));
4606         return p;
4607 out_free_seqid:
4608         nfs_free_seqid(p->arg.open_seqid);
4609 out_free:
4610         kfree(p);
4611         return NULL;
4612 }
4613
4614 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4615 {
4616         struct nfs4_lockdata *data = calldata;
4617         struct nfs4_state *state = data->lsp->ls_state;
4618
4619         dprintk("%s: begin!\n", __func__);
4620         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4621                 return;
4622         /* Do we need to do an open_to_lock_owner? */
4623         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4624                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4625                         return;
4626                 data->arg.open_stateid = &state->stateid;
4627                 data->arg.new_lock_owner = 1;
4628                 data->res.open_seqid = data->arg.open_seqid;
4629         } else
4630                 data->arg.new_lock_owner = 0;
4631         data->timestamp = jiffies;
4632         if (nfs4_setup_sequence(data->server,
4633                                 &data->arg.seq_args,
4634                                 &data->res.seq_res, task))
4635                 return;
4636         rpc_call_start(task);
4637         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4638 }
4639
4640 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4641 {
4642         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4643         nfs4_lock_prepare(task, calldata);
4644 }
4645
4646 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4647 {
4648         struct nfs4_lockdata *data = calldata;
4649
4650         dprintk("%s: begin!\n", __func__);
4651
4652         if (!nfs4_sequence_done(task, &data->res.seq_res))
4653                 return;
4654
4655         data->rpc_status = task->tk_status;
4656         if (data->arg.new_lock_owner != 0) {
4657                 if (data->rpc_status == 0)
4658                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4659                 else
4660                         goto out;
4661         }
4662         if (data->rpc_status == 0) {
4663                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4664                 set_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags);
4665                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4666         }
4667 out:
4668         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4669 }
4670
4671 static void nfs4_lock_release(void *calldata)
4672 {
4673         struct nfs4_lockdata *data = calldata;
4674
4675         dprintk("%s: begin!\n", __func__);
4676         nfs_free_seqid(data->arg.open_seqid);
4677         if (data->cancelled != 0) {
4678                 struct rpc_task *task;
4679                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4680                                 data->arg.lock_seqid);
4681                 if (!IS_ERR(task))
4682                         rpc_put_task_async(task);
4683                 dprintk("%s: cancelling lock!\n", __func__);
4684         } else
4685                 nfs_free_seqid(data->arg.lock_seqid);
4686         nfs4_put_lock_state(data->lsp);
4687         put_nfs_open_context(data->ctx);
4688         kfree(data);
4689         dprintk("%s: done!\n", __func__);
4690 }
4691
4692 static const struct rpc_call_ops nfs4_lock_ops = {
4693         .rpc_call_prepare = nfs4_lock_prepare,
4694         .rpc_call_done = nfs4_lock_done,
4695         .rpc_release = nfs4_lock_release,
4696 };
4697
4698 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4699         .rpc_call_prepare = nfs4_recover_lock_prepare,
4700         .rpc_call_done = nfs4_lock_done,
4701         .rpc_release = nfs4_lock_release,
4702 };
4703
4704 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4705 {
4706         switch (error) {
4707         case -NFS4ERR_ADMIN_REVOKED:
4708         case -NFS4ERR_BAD_STATEID:
4709                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4710                 if (new_lock_owner != 0 ||
4711                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
4712                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4713                 break;
4714         case -NFS4ERR_STALE_STATEID:
4715                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4716         case -NFS4ERR_EXPIRED:
4717                 nfs4_schedule_lease_recovery(server->nfs_client);
4718         };
4719 }
4720
4721 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4722 {
4723         struct nfs4_lockdata *data;
4724         struct rpc_task *task;
4725         struct rpc_message msg = {
4726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4727                 .rpc_cred = state->owner->so_cred,
4728         };
4729         struct rpc_task_setup task_setup_data = {
4730                 .rpc_client = NFS_CLIENT(state->inode),
4731                 .rpc_message = &msg,
4732                 .callback_ops = &nfs4_lock_ops,
4733                 .workqueue = nfsiod_workqueue,
4734                 .flags = RPC_TASK_ASYNC,
4735         };
4736         int ret;
4737
4738         dprintk("%s: begin!\n", __func__);
4739         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4740                         fl->fl_u.nfs4_fl.owner,
4741                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4742         if (data == NULL)
4743                 return -ENOMEM;
4744         if (IS_SETLKW(cmd))
4745                 data->arg.block = 1;
4746         if (recovery_type > NFS_LOCK_NEW) {
4747                 if (recovery_type == NFS_LOCK_RECLAIM)
4748                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4749                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4750         }
4751         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4752         msg.rpc_argp = &data->arg;
4753         msg.rpc_resp = &data->res;
4754         task_setup_data.callback_data = data;
4755         task = rpc_run_task(&task_setup_data);
4756         if (IS_ERR(task))
4757                 return PTR_ERR(task);
4758         ret = nfs4_wait_for_completion_rpc_task(task);
4759         if (ret == 0) {
4760                 ret = data->rpc_status;
4761                 if (ret)
4762                         nfs4_handle_setlk_error(data->server, data->lsp,
4763                                         data->arg.new_lock_owner, ret);
4764         } else
4765                 data->cancelled = 1;
4766         rpc_put_task(task);
4767         dprintk("%s: done, ret = %d!\n", __func__, ret);
4768         return ret;
4769 }
4770
4771 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4772 {
4773         struct nfs_server *server = NFS_SERVER(state->inode);
4774         struct nfs4_exception exception = {
4775                 .inode = state->inode,
4776         };
4777         int err;
4778
4779         do {
4780                 /* Cache the lock if possible... */
4781                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4782                         return 0;
4783                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4784                 if (err != -NFS4ERR_DELAY)
4785                         break;
4786                 nfs4_handle_exception(server, err, &exception);
4787         } while (exception.retry);
4788         return err;
4789 }
4790
4791 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4792 {
4793         struct nfs_server *server = NFS_SERVER(state->inode);
4794         struct nfs4_exception exception = {
4795                 .inode = state->inode,
4796         };
4797         int err;
4798
4799         err = nfs4_set_lock_state(state, request);
4800         if (err != 0)
4801                 return err;
4802         do {
4803                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4804                         return 0;
4805                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4806                 switch (err) {
4807                 default:
4808                         goto out;
4809                 case -NFS4ERR_GRACE:
4810                 case -NFS4ERR_DELAY:
4811                         nfs4_handle_exception(server, err, &exception);
4812                         err = 0;
4813                 }
4814         } while (exception.retry);
4815 out:
4816         return err;
4817 }
4818
4819 #if defined(CONFIG_NFS_V4_1)
4820 /**
4821  * nfs41_check_expired_locks - possibly free a lock stateid
4822  *
4823  * @state: NFSv4 state for an inode
4824  *
4825  * Returns NFS_OK if recovery for this stateid is now finished.
4826  * Otherwise a negative NFS4ERR value is returned.
4827  */
4828 static int nfs41_check_expired_locks(struct nfs4_state *state)
4829 {
4830         int status, ret = -NFS4ERR_BAD_STATEID;
4831         struct nfs4_lock_state *lsp;
4832         struct nfs_server *server = NFS_SERVER(state->inode);
4833
4834         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4835                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
4836                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4837                         if (status != NFS_OK) {
4838                                 /* Free the stateid unless the server
4839                                  * informs us the stateid is unrecognized. */
4840                                 if (status != -NFS4ERR_BAD_STATEID)
4841                                         nfs41_free_stateid(server,
4842                                                         &lsp->ls_stateid);
4843                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
4844                                 ret = status;
4845                         }
4846                 }
4847         };
4848
4849         return ret;
4850 }
4851
4852 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4853 {
4854         int status = NFS_OK;
4855
4856         if (test_bit(LK_STATE_IN_USE, &state->flags))
4857                 status = nfs41_check_expired_locks(state);
4858         if (status != NFS_OK)
4859                 status = nfs4_lock_expired(state, request);
4860         return status;
4861 }
4862 #endif
4863
4864 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4865 {
4866         struct nfs_inode *nfsi = NFS_I(state->inode);
4867         unsigned char fl_flags = request->fl_flags;
4868         int status = -ENOLCK;
4869
4870         if ((fl_flags & FL_POSIX) &&
4871                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4872                 goto out;
4873         /* Is this a delegated open? */
4874         status = nfs4_set_lock_state(state, request);
4875         if (status != 0)
4876                 goto out;
4877         request->fl_flags |= FL_ACCESS;
4878         status = do_vfs_lock(request->fl_file, request);
4879         if (status < 0)
4880                 goto out;
4881         down_read(&nfsi->rwsem);
4882         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4883                 /* Yes: cache locks! */
4884                 /* ...but avoid races with delegation recall... */
4885                 request->fl_flags = fl_flags & ~FL_SLEEP;
4886                 status = do_vfs_lock(request->fl_file, request);
4887                 goto out_unlock;
4888         }
4889         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4890         if (status != 0)
4891                 goto out_unlock;
4892         /* Note: we always want to sleep here! */
4893         request->fl_flags = fl_flags | FL_SLEEP;
4894         if (do_vfs_lock(request->fl_file, request) < 0)
4895                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4896                         "manager!\n", __func__);
4897 out_unlock:
4898         up_read(&nfsi->rwsem);
4899 out:
4900         request->fl_flags = fl_flags;
4901         return status;
4902 }
4903
4904 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4905 {
4906         struct nfs4_exception exception = {
4907                 .state = state,
4908                 .inode = state->inode,
4909         };
4910         int err;
4911
4912         do {
4913                 err = _nfs4_proc_setlk(state, cmd, request);
4914                 if (err == -NFS4ERR_DENIED)
4915                         err = -EAGAIN;
4916                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4917                                 err, &exception);
4918         } while (exception.retry);
4919         return err;
4920 }
4921
4922 static int
4923 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4924 {
4925         struct nfs_open_context *ctx;
4926         struct nfs4_state *state;
4927         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4928         int status;
4929
4930         /* verify open state */
4931         ctx = nfs_file_open_context(filp);
4932         state = ctx->state;
4933
4934         if (request->fl_start < 0 || request->fl_end < 0)
4935                 return -EINVAL;
4936
4937         if (IS_GETLK(cmd)) {
4938                 if (state != NULL)
4939                         return nfs4_proc_getlk(state, F_GETLK, request);
4940                 return 0;
4941         }
4942
4943         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4944                 return -EINVAL;
4945
4946         if (request->fl_type == F_UNLCK) {
4947                 if (state != NULL)
4948                         return nfs4_proc_unlck(state, cmd, request);
4949                 return 0;
4950         }
4951
4952         if (state == NULL)
4953                 return -ENOLCK;
4954         /*
4955          * Don't rely on the VFS having checked the file open mode,
4956          * since it won't do this for flock() locks.
4957          */
4958         switch (request->fl_type) {
4959         case F_RDLCK:
4960                 if (!(filp->f_mode & FMODE_READ))
4961                         return -EBADF;
4962                 break;
4963         case F_WRLCK:
4964                 if (!(filp->f_mode & FMODE_WRITE))
4965                         return -EBADF;
4966         }
4967
4968         do {
4969                 status = nfs4_proc_setlk(state, cmd, request);
4970                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4971                         break;
4972                 timeout = nfs4_set_lock_task_retry(timeout);
4973                 status = -ERESTARTSYS;
4974                 if (signalled())
4975                         break;
4976         } while(status < 0);
4977         return status;
4978 }
4979
4980 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4981 {
4982         struct nfs_server *server = NFS_SERVER(state->inode);
4983         struct nfs4_exception exception = { };
4984         int err;
4985
4986         err = nfs4_set_lock_state(state, fl);
4987         if (err != 0)
4988                 goto out;
4989         do {
4990                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4991                 switch (err) {
4992                         default:
4993                                 printk(KERN_ERR "NFS: %s: unhandled error "
4994                                         "%d.\n", __func__, err);
4995                         case 0:
4996                         case -ESTALE:
4997                                 goto out;
4998                         case -NFS4ERR_EXPIRED:
4999                                 nfs4_schedule_stateid_recovery(server, state);
5000                         case -NFS4ERR_STALE_CLIENTID:
5001                         case -NFS4ERR_STALE_STATEID:
5002                                 nfs4_schedule_lease_recovery(server->nfs_client);
5003                                 goto out;
5004                         case -NFS4ERR_BADSESSION:
5005                         case -NFS4ERR_BADSLOT:
5006                         case -NFS4ERR_BAD_HIGH_SLOT:
5007                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
5008                         case -NFS4ERR_DEADSESSION:
5009                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
5010                                 goto out;
5011                         case -ERESTARTSYS:
5012                                 /*
5013                                  * The show must go on: exit, but mark the
5014                                  * stateid as needing recovery.
5015                                  */
5016                         case -NFS4ERR_DELEG_REVOKED:
5017                         case -NFS4ERR_ADMIN_REVOKED:
5018                         case -NFS4ERR_BAD_STATEID:
5019                         case -NFS4ERR_OPENMODE:
5020                                 nfs4_schedule_stateid_recovery(server, state);
5021                                 err = 0;
5022                                 goto out;
5023                         case -EKEYEXPIRED:
5024                                 /*
5025                                  * User RPCSEC_GSS context has expired.
5026                                  * We cannot recover this stateid now, so
5027                                  * skip it and allow recovery thread to
5028                                  * proceed.
5029                                  */
5030                                 err = 0;
5031                                 goto out;
5032                         case -ENOMEM:
5033                         case -NFS4ERR_DENIED:
5034                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
5035                                 err = 0;
5036                                 goto out;
5037                         case -NFS4ERR_DELAY:
5038                                 break;
5039                 }
5040                 err = nfs4_handle_exception(server, err, &exception);
5041         } while (exception.retry);
5042 out:
5043         return err;
5044 }
5045
5046 struct nfs_release_lockowner_data {
5047         struct nfs4_lock_state *lsp;
5048         struct nfs_server *server;
5049         struct nfs_release_lockowner_args args;
5050 };
5051
5052 static void nfs4_release_lockowner_release(void *calldata)
5053 {
5054         struct nfs_release_lockowner_data *data = calldata;
5055         nfs4_free_lock_state(data->server, data->lsp);
5056         kfree(calldata);
5057 }
5058
5059 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
5060         .rpc_release = nfs4_release_lockowner_release,
5061 };
5062
5063 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
5064 {
5065         struct nfs_server *server = lsp->ls_state->owner->so_server;
5066         struct nfs_release_lockowner_data *data;
5067         struct rpc_message msg = {
5068                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
5069         };
5070
5071         if (server->nfs_client->cl_mvops->minor_version != 0)
5072                 return -EINVAL;
5073         data = kmalloc(sizeof(*data), GFP_NOFS);
5074         if (!data)
5075                 return -ENOMEM;
5076         data->lsp = lsp;
5077         data->server = server;
5078         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
5079         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
5080         data->args.lock_owner.s_dev = server->s_dev;
5081         msg.rpc_argp = &data->args;
5082         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
5083         return 0;
5084 }
5085
5086 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
5087
5088 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
5089                                    const void *buf, size_t buflen,
5090                                    int flags, int type)
5091 {
5092         if (strcmp(key, "") != 0)
5093                 return -EINVAL;
5094
5095         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
5096 }
5097
5098 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
5099                                    void *buf, size_t buflen, int type)
5100 {
5101         if (strcmp(key, "") != 0)
5102                 return -EINVAL;
5103
5104         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
5105 }
5106
5107 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
5108                                        size_t list_len, const char *name,
5109                                        size_t name_len, int type)
5110 {
5111         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
5112
5113         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
5114                 return 0;
5115
5116         if (list && len <= list_len)
5117                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
5118         return len;
5119 }
5120
5121 /*
5122  * nfs_fhget will use either the mounted_on_fileid or the fileid
5123  */
5124 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
5125 {
5126         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
5127                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
5128               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
5129               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
5130                 return;
5131
5132         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
5133                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
5134         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
5135         fattr->nlink = 2;
5136 }
5137
5138 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5139                                    const struct qstr *name,
5140                                    struct nfs4_fs_locations *fs_locations,
5141                                    struct page *page)
5142 {
5143         struct nfs_server *server = NFS_SERVER(dir);
5144         u32 bitmask[2] = {
5145                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
5146         };
5147         struct nfs4_fs_locations_arg args = {
5148                 .dir_fh = NFS_FH(dir),
5149                 .name = name,
5150                 .page = page,
5151                 .bitmask = bitmask,
5152         };
5153         struct nfs4_fs_locations_res res = {
5154                 .fs_locations = fs_locations,
5155         };
5156         struct rpc_message msg = {
5157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
5158                 .rpc_argp = &args,
5159                 .rpc_resp = &res,
5160         };
5161         int status;
5162
5163         dprintk("%s: start\n", __func__);
5164
5165         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
5166          * is not supported */
5167         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
5168                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
5169         else
5170                 bitmask[0] |= FATTR4_WORD0_FILEID;
5171
5172         nfs_fattr_init(&fs_locations->fattr);
5173         fs_locations->server = server;
5174         fs_locations->nlocations = 0;
5175         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
5176         dprintk("%s: returned status = %d\n", __func__, status);
5177         return status;
5178 }
5179
5180 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
5181                            const struct qstr *name,
5182                            struct nfs4_fs_locations *fs_locations,
5183                            struct page *page)
5184 {
5185         struct nfs4_exception exception = { };
5186         int err;
5187         do {
5188                 err = nfs4_handle_exception(NFS_SERVER(dir),
5189                                 _nfs4_proc_fs_locations(client, dir, name, fs_locations, page),
5190                                 &exception);
5191         } while (exception.retry);
5192         return err;
5193 }
5194
5195 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
5196 {
5197         int status;
5198         struct nfs4_secinfo_arg args = {
5199                 .dir_fh = NFS_FH(dir),
5200                 .name   = name,
5201         };
5202         struct nfs4_secinfo_res res = {
5203                 .flavors     = flavors,
5204         };
5205         struct rpc_message msg = {
5206                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
5207                 .rpc_argp = &args,
5208                 .rpc_resp = &res,
5209         };
5210
5211         dprintk("NFS call  secinfo %s\n", name->name);
5212         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
5213         dprintk("NFS reply  secinfo: %d\n", status);
5214         return status;
5215 }
5216
5217 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
5218                       struct nfs4_secinfo_flavors *flavors)
5219 {
5220         struct nfs4_exception exception = { };
5221         int err;
5222         do {
5223                 err = nfs4_handle_exception(NFS_SERVER(dir),
5224                                 _nfs4_proc_secinfo(dir, name, flavors),
5225                                 &exception);
5226         } while (exception.retry);
5227         return err;
5228 }
5229
5230 #ifdef CONFIG_NFS_V4_1
5231 /*
5232  * Check the exchange flags returned by the server for invalid flags, having
5233  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
5234  * DS flags set.
5235  */
5236 static int nfs4_check_cl_exchange_flags(u32 flags)
5237 {
5238         if (flags & ~EXCHGID4_FLAG_MASK_R)
5239                 goto out_inval;
5240         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
5241             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
5242                 goto out_inval;
5243         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
5244                 goto out_inval;
5245         return NFS_OK;
5246 out_inval:
5247         return -NFS4ERR_INVAL;
5248 }
5249
5250 static bool
5251 nfs41_same_server_scope(struct nfs41_server_scope *a,
5252                         struct nfs41_server_scope *b)
5253 {
5254         if (a->server_scope_sz == b->server_scope_sz &&
5255             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
5256                 return true;
5257
5258         return false;
5259 }
5260
5261 /*
5262  * nfs4_proc_bind_conn_to_session()
5263  *
5264  * The 4.1 client currently uses the same TCP connection for the
5265  * fore and backchannel.
5266  */
5267 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
5268 {
5269         int status;
5270         struct nfs41_bind_conn_to_session_res res;
5271         struct rpc_message msg = {
5272                 .rpc_proc =
5273                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
5274                 .rpc_argp = clp,
5275                 .rpc_resp = &res,
5276                 .rpc_cred = cred,
5277         };
5278
5279         dprintk("--> %s\n", __func__);
5280         BUG_ON(clp == NULL);
5281
5282         res.session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5283         if (unlikely(res.session == NULL)) {
5284                 status = -ENOMEM;
5285                 goto out;
5286         }
5287
5288         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5289         if (status == 0) {
5290                 if (memcmp(res.session->sess_id.data,
5291                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
5292                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
5293                         status = -EIO;
5294                         goto out_session;
5295                 }
5296                 if (res.dir != NFS4_CDFS4_BOTH) {
5297                         dprintk("NFS: %s: Unexpected direction from server\n",
5298                                 __func__);
5299                         status = -EIO;
5300                         goto out_session;
5301                 }
5302                 if (res.use_conn_in_rdma_mode) {
5303                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
5304                                 __func__);
5305                         status = -EIO;
5306                         goto out_session;
5307                 }
5308         }
5309 out_session:
5310         kfree(res.session);
5311 out:
5312         dprintk("<-- %s status= %d\n", __func__, status);
5313         return status;
5314 }
5315
5316 /*
5317  * nfs4_proc_exchange_id()
5318  *
5319  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5320  *
5321  * Since the clientid has expired, all compounds using sessions
5322  * associated with the stale clientid will be returning
5323  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
5324  * be in some phase of session reset.
5325  */
5326 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
5327 {
5328         nfs4_verifier verifier;
5329         struct nfs41_exchange_id_args args = {
5330                 .verifier = &verifier,
5331                 .client = clp,
5332                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
5333         };
5334         struct nfs41_exchange_id_res res = {
5335                 0
5336         };
5337         int status;
5338         struct rpc_message msg = {
5339                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5340                 .rpc_argp = &args,
5341                 .rpc_resp = &res,
5342                 .rpc_cred = cred,
5343         };
5344
5345         nfs4_init_boot_verifier(clp, &verifier);
5346         args.id_len = nfs4_init_uniform_client_string(clp, args.id,
5347                                                         sizeof(args.id));
5348         dprintk("NFS call  exchange_id auth=%s, '%.*s'\n",
5349                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5350                 args.id_len, args.id);
5351
5352         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
5353                                         GFP_NOFS);
5354         if (unlikely(res.server_owner == NULL)) {
5355                 status = -ENOMEM;
5356                 goto out;
5357         }
5358
5359         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
5360                                         GFP_NOFS);
5361         if (unlikely(res.server_scope == NULL)) {
5362                 status = -ENOMEM;
5363                 goto out_server_owner;
5364         }
5365
5366         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
5367         if (unlikely(res.impl_id == NULL)) {
5368                 status = -ENOMEM;
5369                 goto out_server_scope;
5370         }
5371
5372         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5373         if (status == 0)
5374                 status = nfs4_check_cl_exchange_flags(res.flags);
5375
5376         if (status == 0) {
5377                 clp->cl_clientid = res.clientid;
5378                 clp->cl_exchange_flags = (res.flags & ~EXCHGID4_FLAG_CONFIRMED_R);
5379                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R))
5380                         clp->cl_seqid = res.seqid;
5381
5382                 kfree(clp->cl_serverowner);
5383                 clp->cl_serverowner = res.server_owner;
5384                 res.server_owner = NULL;
5385
5386                 /* use the most recent implementation id */
5387                 kfree(clp->cl_implid);
5388                 clp->cl_implid = res.impl_id;
5389
5390                 if (clp->cl_serverscope != NULL &&
5391                     !nfs41_same_server_scope(clp->cl_serverscope,
5392                                              res.server_scope)) {
5393                         dprintk("%s: server_scope mismatch detected\n",
5394                                 __func__);
5395                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5396                         kfree(clp->cl_serverscope);
5397                         clp->cl_serverscope = NULL;
5398                 }
5399
5400                 if (clp->cl_serverscope == NULL) {
5401                         clp->cl_serverscope = res.server_scope;
5402                         goto out;
5403                 }
5404         } else
5405                 kfree(res.impl_id);
5406
5407 out_server_owner:
5408         kfree(res.server_owner);
5409 out_server_scope:
5410         kfree(res.server_scope);
5411 out:
5412         if (clp->cl_implid != NULL)
5413                 dprintk("NFS reply exchange_id: Server Implementation ID: "
5414                         "domain: %s, name: %s, date: %llu,%u\n",
5415                         clp->cl_implid->domain, clp->cl_implid->name,
5416                         clp->cl_implid->date.seconds,
5417                         clp->cl_implid->date.nseconds);
5418         dprintk("NFS reply exchange_id: %d\n", status);
5419         return status;
5420 }
5421
5422 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
5423                 struct rpc_cred *cred)
5424 {
5425         struct rpc_message msg = {
5426                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
5427                 .rpc_argp = clp,
5428                 .rpc_cred = cred,
5429         };
5430         int status;
5431
5432         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5433         if (status)
5434                 dprintk("NFS: Got error %d from the server %s on "
5435                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
5436         return status;
5437 }
5438
5439 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
5440                 struct rpc_cred *cred)
5441 {
5442         unsigned int loop;
5443         int ret;
5444
5445         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
5446                 ret = _nfs4_proc_destroy_clientid(clp, cred);
5447                 switch (ret) {
5448                 case -NFS4ERR_DELAY:
5449                 case -NFS4ERR_CLIENTID_BUSY:
5450                         ssleep(1);
5451                         break;
5452                 default:
5453                         return ret;
5454                 }
5455         }
5456         return 0;
5457 }
5458
5459 int nfs4_destroy_clientid(struct nfs_client *clp)
5460 {
5461         struct rpc_cred *cred;
5462         int ret = 0;
5463
5464         if (clp->cl_mvops->minor_version < 1)
5465                 goto out;
5466         if (clp->cl_exchange_flags == 0)
5467                 goto out;
5468         if (clp->cl_preserve_clid)
5469                 goto out;
5470         cred = nfs4_get_exchange_id_cred(clp);
5471         ret = nfs4_proc_destroy_clientid(clp, cred);
5472         if (cred)
5473                 put_rpccred(cred);
5474         switch (ret) {
5475         case 0:
5476         case -NFS4ERR_STALE_CLIENTID:
5477                 clp->cl_exchange_flags = 0;
5478         }
5479 out:
5480         return ret;
5481 }
5482
5483 struct nfs4_get_lease_time_data {
5484         struct nfs4_get_lease_time_args *args;
5485         struct nfs4_get_lease_time_res *res;
5486         struct nfs_client *clp;
5487 };
5488
5489 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5490                                         void *calldata)
5491 {
5492         int ret;
5493         struct nfs4_get_lease_time_data *data =
5494                         (struct nfs4_get_lease_time_data *)calldata;
5495
5496         dprintk("--> %s\n", __func__);
5497         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5498         /* just setup sequence, do not trigger session recovery
5499            since we're invoked within one */
5500         ret = nfs41_setup_sequence(data->clp->cl_session,
5501                                    &data->args->la_seq_args,
5502                                    &data->res->lr_seq_res, task);
5503
5504         BUG_ON(ret == -EAGAIN);
5505         rpc_call_start(task);
5506         dprintk("<-- %s\n", __func__);
5507 }
5508
5509 /*
5510  * Called from nfs4_state_manager thread for session setup, so don't recover
5511  * from sequence operation or clientid errors.
5512  */
5513 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5514 {
5515         struct nfs4_get_lease_time_data *data =
5516                         (struct nfs4_get_lease_time_data *)calldata;
5517
5518         dprintk("--> %s\n", __func__);
5519         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5520                 return;
5521         switch (task->tk_status) {
5522         case -NFS4ERR_DELAY:
5523         case -NFS4ERR_GRACE:
5524                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5525                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5526                 task->tk_status = 0;
5527                 /* fall through */
5528         case -NFS4ERR_RETRY_UNCACHED_REP:
5529                 rpc_restart_call_prepare(task);
5530                 return;
5531         }
5532         dprintk("<-- %s\n", __func__);
5533 }
5534
5535 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5536         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5537         .rpc_call_done = nfs4_get_lease_time_done,
5538 };
5539
5540 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5541 {
5542         struct rpc_task *task;
5543         struct nfs4_get_lease_time_args args;
5544         struct nfs4_get_lease_time_res res = {
5545                 .lr_fsinfo = fsinfo,
5546         };
5547         struct nfs4_get_lease_time_data data = {
5548                 .args = &args,
5549                 .res = &res,
5550                 .clp = clp,
5551         };
5552         struct rpc_message msg = {
5553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5554                 .rpc_argp = &args,
5555                 .rpc_resp = &res,
5556         };
5557         struct rpc_task_setup task_setup = {
5558                 .rpc_client = clp->cl_rpcclient,
5559                 .rpc_message = &msg,
5560                 .callback_ops = &nfs4_get_lease_time_ops,
5561                 .callback_data = &data,
5562                 .flags = RPC_TASK_TIMEOUT,
5563         };
5564         int status;
5565
5566         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5567         dprintk("--> %s\n", __func__);
5568         task = rpc_run_task(&task_setup);
5569
5570         if (IS_ERR(task))
5571                 status = PTR_ERR(task);
5572         else {
5573                 status = task->tk_status;
5574                 rpc_put_task(task);
5575         }
5576         dprintk("<-- %s return %d\n", __func__, status);
5577
5578         return status;
5579 }
5580
5581 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5582 {
5583         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5584 }
5585
5586 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5587                 struct nfs4_slot *new,
5588                 u32 max_slots,
5589                 u32 ivalue)
5590 {
5591         struct nfs4_slot *old = NULL;
5592         u32 i;
5593
5594         spin_lock(&tbl->slot_tbl_lock);
5595         if (new) {
5596                 old = tbl->slots;
5597                 tbl->slots = new;
5598                 tbl->max_slots = max_slots;
5599         }
5600         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5601         for (i = 0; i < tbl->max_slots; i++)
5602                 tbl->slots[i].seq_nr = ivalue;
5603         spin_unlock(&tbl->slot_tbl_lock);
5604         kfree(old);
5605 }
5606
5607 /*
5608  * (re)Initialise a slot table
5609  */
5610 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5611                                  u32 ivalue)
5612 {
5613         struct nfs4_slot *new = NULL;
5614         int ret = -ENOMEM;
5615
5616         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5617                 max_reqs, tbl->max_slots);
5618
5619         /* Does the newly negotiated max_reqs match the existing slot table? */
5620         if (max_reqs != tbl->max_slots) {
5621                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5622                 if (!new)
5623                         goto out;
5624         }
5625         ret = 0;
5626
5627         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5628         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5629                 tbl, tbl->slots, tbl->max_slots);
5630 out:
5631         dprintk("<-- %s: return %d\n", __func__, ret);
5632         return ret;
5633 }
5634
5635 /* Destroy the slot table */
5636 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5637 {
5638         if (session->fc_slot_table.slots != NULL) {
5639                 kfree(session->fc_slot_table.slots);
5640                 session->fc_slot_table.slots = NULL;
5641         }
5642         if (session->bc_slot_table.slots != NULL) {
5643                 kfree(session->bc_slot_table.slots);
5644                 session->bc_slot_table.slots = NULL;
5645         }
5646         return;
5647 }
5648
5649 /*
5650  * Initialize or reset the forechannel and backchannel tables
5651  */
5652 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5653 {
5654         struct nfs4_slot_table *tbl;
5655         int status;
5656
5657         dprintk("--> %s\n", __func__);
5658         /* Fore channel */
5659         tbl = &ses->fc_slot_table;
5660         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5661         if (status) /* -ENOMEM */
5662                 return status;
5663         /* Back channel */
5664         tbl = &ses->bc_slot_table;
5665         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5666         if (status && tbl->slots == NULL)
5667                 /* Fore and back channel share a connection so get
5668                  * both slot tables or neither */
5669                 nfs4_destroy_slot_tables(ses);
5670         return status;
5671 }
5672
5673 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5674 {
5675         struct nfs4_session *session;
5676         struct nfs4_slot_table *tbl;
5677
5678         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5679         if (!session)
5680                 return NULL;
5681
5682         tbl = &session->fc_slot_table;
5683         tbl->highest_used_slotid = NFS4_NO_SLOT;
5684         spin_lock_init(&tbl->slot_tbl_lock);
5685         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5686         init_completion(&tbl->complete);
5687
5688         tbl = &session->bc_slot_table;
5689         tbl->highest_used_slotid = NFS4_NO_SLOT;
5690         spin_lock_init(&tbl->slot_tbl_lock);
5691         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5692         init_completion(&tbl->complete);
5693
5694         session->session_state = 1<<NFS4_SESSION_INITING;
5695
5696         session->clp = clp;
5697         return session;
5698 }
5699
5700 void nfs4_destroy_session(struct nfs4_session *session)
5701 {
5702         struct rpc_xprt *xprt;
5703         struct rpc_cred *cred;
5704
5705         cred = nfs4_get_exchange_id_cred(session->clp);
5706         nfs4_proc_destroy_session(session, cred);
5707         if (cred)
5708                 put_rpccred(cred);
5709
5710         rcu_read_lock();
5711         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5712         rcu_read_unlock();
5713         dprintk("%s Destroy backchannel for xprt %p\n",
5714                 __func__, xprt);
5715         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5716         nfs4_destroy_slot_tables(session);
5717         kfree(session);
5718 }
5719
5720 /*
5721  * Initialize the values to be used by the client in CREATE_SESSION
5722  * If nfs4_init_session set the fore channel request and response sizes,
5723  * use them.
5724  *
5725  * Set the back channel max_resp_sz_cached to zero to force the client to
5726  * always set csa_cachethis to FALSE because the current implementation
5727  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5728  */
5729 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5730 {
5731         struct nfs4_session *session = args->client->cl_session;
5732         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5733                      mxresp_sz = session->fc_attrs.max_resp_sz;
5734
5735         if (mxrqst_sz == 0)
5736                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5737         if (mxresp_sz == 0)
5738                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5739         /* Fore channel attributes */
5740         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5741         args->fc_attrs.max_resp_sz = mxresp_sz;
5742         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5743         args->fc_attrs.max_reqs = max_session_slots;
5744
5745         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5746                 "max_ops=%u max_reqs=%u\n",
5747                 __func__,
5748                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5749                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5750
5751         /* Back channel attributes */
5752         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5753         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5754         args->bc_attrs.max_resp_sz_cached = 0;
5755         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5756         args->bc_attrs.max_reqs = 1;
5757
5758         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5759                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5760                 __func__,
5761                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5762                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5763                 args->bc_attrs.max_reqs);
5764 }
5765
5766 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5767 {
5768         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5769         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5770
5771         if (rcvd->max_resp_sz > sent->max_resp_sz)
5772                 return -EINVAL;
5773         /*
5774          * Our requested max_ops is the minimum we need; we're not
5775          * prepared to break up compounds into smaller pieces than that.
5776          * So, no point even trying to continue if the server won't
5777          * cooperate:
5778          */
5779         if (rcvd->max_ops < sent->max_ops)
5780                 return -EINVAL;
5781         if (rcvd->max_reqs == 0)
5782                 return -EINVAL;
5783         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5784                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5785         return 0;
5786 }
5787
5788 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5789 {
5790         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5791         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5792
5793         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5794                 return -EINVAL;
5795         if (rcvd->max_resp_sz < sent->max_resp_sz)
5796                 return -EINVAL;
5797         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5798                 return -EINVAL;
5799         /* These would render the backchannel useless: */
5800         if (rcvd->max_ops != sent->max_ops)
5801                 return -EINVAL;
5802         if (rcvd->max_reqs != sent->max_reqs)
5803                 return -EINVAL;
5804         return 0;
5805 }
5806
5807 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5808                                      struct nfs4_session *session)
5809 {
5810         int ret;
5811
5812         ret = nfs4_verify_fore_channel_attrs(args, session);
5813         if (ret)
5814                 return ret;
5815         return nfs4_verify_back_channel_attrs(args, session);
5816 }
5817
5818 static int _nfs4_proc_create_session(struct nfs_client *clp,
5819                 struct rpc_cred *cred)
5820 {
5821         struct nfs4_session *session = clp->cl_session;
5822         struct nfs41_create_session_args args = {
5823                 .client = clp,
5824                 .cb_program = NFS4_CALLBACK,
5825         };
5826         struct nfs41_create_session_res res = {
5827                 .client = clp,
5828         };
5829         struct rpc_message msg = {
5830                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5831                 .rpc_argp = &args,
5832                 .rpc_resp = &res,
5833                 .rpc_cred = cred,
5834         };
5835         int status;
5836
5837         nfs4_init_channel_attrs(&args);
5838         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5839
5840         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5841
5842         if (!status)
5843                 /* Verify the session's negotiated channel_attrs values */
5844                 status = nfs4_verify_channel_attrs(&args, session);
5845         if (!status) {
5846                 /* Increment the clientid slot sequence id */
5847                 clp->cl_seqid++;
5848         }
5849
5850         return status;
5851 }
5852
5853 /*
5854  * Issues a CREATE_SESSION operation to the server.
5855  * It is the responsibility of the caller to verify the session is
5856  * expired before calling this routine.
5857  */
5858 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
5859 {
5860         int status;
5861         unsigned *ptr;
5862         struct nfs4_session *session = clp->cl_session;
5863
5864         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5865
5866         status = _nfs4_proc_create_session(clp, cred);
5867         if (status)
5868                 goto out;
5869
5870         /* Init or reset the session slot tables */
5871         status = nfs4_setup_session_slot_tables(session);
5872         dprintk("slot table setup returned %d\n", status);
5873         if (status)
5874                 goto out;
5875
5876         ptr = (unsigned *)&session->sess_id.data[0];
5877         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5878                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5879 out:
5880         dprintk("<-- %s\n", __func__);
5881         return status;
5882 }
5883
5884 /*
5885  * Issue the over-the-wire RPC DESTROY_SESSION.
5886  * The caller must serialize access to this routine.
5887  */
5888 int nfs4_proc_destroy_session(struct nfs4_session *session,
5889                 struct rpc_cred *cred)
5890 {
5891         struct rpc_message msg = {
5892                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
5893                 .rpc_argp = session,
5894                 .rpc_cred = cred,
5895         };
5896         int status = 0;
5897
5898         dprintk("--> nfs4_proc_destroy_session\n");
5899
5900         /* session is still being setup */
5901         if (session->clp->cl_cons_state != NFS_CS_READY)
5902                 return status;
5903
5904         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5905
5906         if (status)
5907                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
5908                         "Session has been destroyed regardless...\n", status);
5909
5910         dprintk("<-- nfs4_proc_destroy_session\n");
5911         return status;
5912 }
5913
5914 /*
5915  * With sessions, the client is not marked ready until after a
5916  * successful EXCHANGE_ID and CREATE_SESSION.
5917  *
5918  * Map errors cl_cons_state errors to EPROTONOSUPPORT to indicate
5919  * other versions of NFS can be tried.
5920  */
5921 static int nfs41_check_session_ready(struct nfs_client *clp)
5922 {
5923         int ret;
5924
5925         if (clp->cl_cons_state == NFS_CS_SESSION_INITING) {
5926                 ret = nfs4_client_recover_expired_lease(clp);
5927                 if (ret)
5928                         return ret;
5929         }
5930         if (clp->cl_cons_state < NFS_CS_READY)
5931                 return -EPROTONOSUPPORT;
5932         smp_rmb();
5933         return 0;
5934 }
5935
5936 int nfs4_init_session(struct nfs_server *server)
5937 {
5938         struct nfs_client *clp = server->nfs_client;
5939         struct nfs4_session *session;
5940         unsigned int rsize, wsize;
5941
5942         if (!nfs4_has_session(clp))
5943                 return 0;
5944
5945         session = clp->cl_session;
5946         spin_lock(&clp->cl_lock);
5947         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5948
5949                 rsize = server->rsize;
5950                 if (rsize == 0)
5951                         rsize = NFS_MAX_FILE_IO_SIZE;
5952                 wsize = server->wsize;
5953                 if (wsize == 0)
5954                         wsize = NFS_MAX_FILE_IO_SIZE;
5955
5956                 session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5957                 session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5958         }
5959         spin_unlock(&clp->cl_lock);
5960
5961         return nfs41_check_session_ready(clp);
5962 }
5963
5964 int nfs4_init_ds_session(struct nfs_client *clp, unsigned long lease_time)
5965 {
5966         struct nfs4_session *session = clp->cl_session;
5967         int ret;
5968
5969         spin_lock(&clp->cl_lock);
5970         if (test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) {
5971                 /*
5972                  * Do not set NFS_CS_CHECK_LEASE_TIME instead set the
5973                  * DS lease to be equal to the MDS lease.
5974                  */
5975                 clp->cl_lease_time = lease_time;
5976                 clp->cl_last_renewal = jiffies;
5977         }
5978         spin_unlock(&clp->cl_lock);
5979
5980         ret = nfs41_check_session_ready(clp);
5981         if (ret)
5982                 return ret;
5983         /* Test for the DS role */
5984         if (!is_ds_client(clp))
5985                 return -ENODEV;
5986         return 0;
5987 }
5988 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5989
5990
5991 /*
5992  * Renew the cl_session lease.
5993  */
5994 struct nfs4_sequence_data {
5995         struct nfs_client *clp;
5996         struct nfs4_sequence_args args;
5997         struct nfs4_sequence_res res;
5998 };
5999
6000 static void nfs41_sequence_release(void *data)
6001 {
6002         struct nfs4_sequence_data *calldata = data;
6003         struct nfs_client *clp = calldata->clp;
6004
6005         if (atomic_read(&clp->cl_count) > 1)
6006                 nfs4_schedule_state_renewal(clp);
6007         nfs_put_client(clp);
6008         kfree(calldata);
6009 }
6010
6011 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6012 {
6013         switch(task->tk_status) {
6014         case -NFS4ERR_DELAY:
6015                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6016                 return -EAGAIN;
6017         default:
6018                 nfs4_schedule_lease_recovery(clp);
6019         }
6020         return 0;
6021 }
6022
6023 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
6024 {
6025         struct nfs4_sequence_data *calldata = data;
6026         struct nfs_client *clp = calldata->clp;
6027
6028         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
6029                 return;
6030
6031         if (task->tk_status < 0) {
6032                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
6033                 if (atomic_read(&clp->cl_count) == 1)
6034                         goto out;
6035
6036                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
6037                         rpc_restart_call_prepare(task);
6038                         return;
6039                 }
6040         }
6041         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
6042 out:
6043         dprintk("<-- %s\n", __func__);
6044 }
6045
6046 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
6047 {
6048         struct nfs4_sequence_data *calldata = data;
6049         struct nfs_client *clp = calldata->clp;
6050         struct nfs4_sequence_args *args;
6051         struct nfs4_sequence_res *res;
6052
6053         args = task->tk_msg.rpc_argp;
6054         res = task->tk_msg.rpc_resp;
6055
6056         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
6057                 return;
6058         rpc_call_start(task);
6059 }
6060
6061 static const struct rpc_call_ops nfs41_sequence_ops = {
6062         .rpc_call_done = nfs41_sequence_call_done,
6063         .rpc_call_prepare = nfs41_sequence_prepare,
6064         .rpc_release = nfs41_sequence_release,
6065 };
6066
6067 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6068 {
6069         struct nfs4_sequence_data *calldata;
6070         struct rpc_message msg = {
6071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
6072                 .rpc_cred = cred,
6073         };
6074         struct rpc_task_setup task_setup_data = {
6075                 .rpc_client = clp->cl_rpcclient,
6076                 .rpc_message = &msg,
6077                 .callback_ops = &nfs41_sequence_ops,
6078                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
6079         };
6080
6081         if (!atomic_inc_not_zero(&clp->cl_count))
6082                 return ERR_PTR(-EIO);
6083         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6084         if (calldata == NULL) {
6085                 nfs_put_client(clp);
6086                 return ERR_PTR(-ENOMEM);
6087         }
6088         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
6089         msg.rpc_argp = &calldata->args;
6090         msg.rpc_resp = &calldata->res;
6091         calldata->clp = clp;
6092         task_setup_data.callback_data = calldata;
6093
6094         return rpc_run_task(&task_setup_data);
6095 }
6096
6097 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
6098 {
6099         struct rpc_task *task;
6100         int ret = 0;
6101
6102         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
6103                 return 0;
6104         task = _nfs41_proc_sequence(clp, cred);
6105         if (IS_ERR(task))
6106                 ret = PTR_ERR(task);
6107         else
6108                 rpc_put_task_async(task);
6109         dprintk("<-- %s status=%d\n", __func__, ret);
6110         return ret;
6111 }
6112
6113 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
6114 {
6115         struct rpc_task *task;
6116         int ret;
6117
6118         task = _nfs41_proc_sequence(clp, cred);
6119         if (IS_ERR(task)) {
6120                 ret = PTR_ERR(task);
6121                 goto out;
6122         }
6123         ret = rpc_wait_for_completion_task(task);
6124         if (!ret) {
6125                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
6126
6127                 if (task->tk_status == 0)
6128                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
6129                 ret = task->tk_status;
6130         }
6131         rpc_put_task(task);
6132 out:
6133         dprintk("<-- %s status=%d\n", __func__, ret);
6134         return ret;
6135 }
6136
6137 struct nfs4_reclaim_complete_data {
6138         struct nfs_client *clp;
6139         struct nfs41_reclaim_complete_args arg;
6140         struct nfs41_reclaim_complete_res res;
6141 };
6142
6143 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
6144 {
6145         struct nfs4_reclaim_complete_data *calldata = data;
6146
6147         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
6148         if (nfs41_setup_sequence(calldata->clp->cl_session,
6149                                 &calldata->arg.seq_args,
6150                                 &calldata->res.seq_res, task))
6151                 return;
6152
6153         rpc_call_start(task);
6154 }
6155
6156 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
6157 {
6158         switch(task->tk_status) {
6159         case 0:
6160         case -NFS4ERR_COMPLETE_ALREADY:
6161         case -NFS4ERR_WRONG_CRED: /* What to do here? */
6162                 break;
6163         case -NFS4ERR_DELAY:
6164                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
6165                 /* fall through */
6166         case -NFS4ERR_RETRY_UNCACHED_REP:
6167                 return -EAGAIN;
6168         default:
6169                 nfs4_schedule_lease_recovery(clp);
6170         }
6171         return 0;
6172 }
6173
6174 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
6175 {
6176         struct nfs4_reclaim_complete_data *calldata = data;
6177         struct nfs_client *clp = calldata->clp;
6178         struct nfs4_sequence_res *res = &calldata->res.seq_res;
6179
6180         dprintk("--> %s\n", __func__);
6181         if (!nfs41_sequence_done(task, res))
6182                 return;
6183
6184         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
6185                 rpc_restart_call_prepare(task);
6186                 return;
6187         }
6188         dprintk("<-- %s\n", __func__);
6189 }
6190
6191 static void nfs4_free_reclaim_complete_data(void *data)
6192 {
6193         struct nfs4_reclaim_complete_data *calldata = data;
6194
6195         kfree(calldata);
6196 }
6197
6198 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
6199         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
6200         .rpc_call_done = nfs4_reclaim_complete_done,
6201         .rpc_release = nfs4_free_reclaim_complete_data,
6202 };
6203
6204 /*
6205  * Issue a global reclaim complete.
6206  */
6207 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
6208 {
6209         struct nfs4_reclaim_complete_data *calldata;
6210         struct rpc_task *task;
6211         struct rpc_message msg = {
6212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
6213         };
6214         struct rpc_task_setup task_setup_data = {
6215                 .rpc_client = clp->cl_rpcclient,
6216                 .rpc_message = &msg,
6217                 .callback_ops = &nfs4_reclaim_complete_call_ops,
6218                 .flags = RPC_TASK_ASYNC,
6219         };
6220         int status = -ENOMEM;
6221
6222         dprintk("--> %s\n", __func__);
6223         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
6224         if (calldata == NULL)
6225                 goto out;
6226         calldata->clp = clp;
6227         calldata->arg.one_fs = 0;
6228
6229         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
6230         msg.rpc_argp = &calldata->arg;
6231         msg.rpc_resp = &calldata->res;
6232         task_setup_data.callback_data = calldata;
6233         task = rpc_run_task(&task_setup_data);
6234         if (IS_ERR(task)) {
6235                 status = PTR_ERR(task);
6236                 goto out;
6237         }
6238         status = nfs4_wait_for_completion_rpc_task(task);
6239         if (status == 0)
6240                 status = task->tk_status;
6241         rpc_put_task(task);
6242         return 0;
6243 out:
6244         dprintk("<-- %s status=%d\n", __func__, status);
6245         return status;
6246 }
6247
6248 static void
6249 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
6250 {
6251         struct nfs4_layoutget *lgp = calldata;
6252         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6253
6254         dprintk("--> %s\n", __func__);
6255         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
6256          * right now covering the LAYOUTGET we are about to send.
6257          * However, that is not so catastrophic, and there seems
6258          * to be no way to prevent it completely.
6259          */
6260         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
6261                                 &lgp->res.seq_res, task))
6262                 return;
6263         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
6264                                           NFS_I(lgp->args.inode)->layout,
6265                                           lgp->args.ctx->state)) {
6266                 rpc_exit(task, NFS4_OK);
6267                 return;
6268         }
6269         rpc_call_start(task);
6270 }
6271
6272 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
6273 {
6274         struct nfs4_layoutget *lgp = calldata;
6275         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6276
6277         dprintk("--> %s\n", __func__);
6278
6279         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
6280                 return;
6281
6282         switch (task->tk_status) {
6283         case 0:
6284                 break;
6285         case -NFS4ERR_LAYOUTTRYLATER:
6286         case -NFS4ERR_RECALLCONFLICT:
6287                 task->tk_status = -NFS4ERR_DELAY;
6288                 /* Fall through */
6289         default:
6290                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6291                         rpc_restart_call_prepare(task);
6292                         return;
6293                 }
6294         }
6295         dprintk("<-- %s\n", __func__);
6296 }
6297
6298 static size_t max_response_pages(struct nfs_server *server)
6299 {
6300         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
6301         return nfs_page_array_len(0, max_resp_sz);
6302 }
6303
6304 static void nfs4_free_pages(struct page **pages, size_t size)
6305 {
6306         int i;
6307
6308         if (!pages)
6309                 return;
6310
6311         for (i = 0; i < size; i++) {
6312                 if (!pages[i])
6313                         break;
6314                 __free_page(pages[i]);
6315         }
6316         kfree(pages);
6317 }
6318
6319 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
6320 {
6321         struct page **pages;
6322         int i;
6323
6324         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
6325         if (!pages) {
6326                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
6327                 return NULL;
6328         }
6329
6330         for (i = 0; i < size; i++) {
6331                 pages[i] = alloc_page(gfp_flags);
6332                 if (!pages[i]) {
6333                         dprintk("%s: failed to allocate page\n", __func__);
6334                         nfs4_free_pages(pages, size);
6335                         return NULL;
6336                 }
6337         }
6338
6339         return pages;
6340 }
6341
6342 static void nfs4_layoutget_release(void *calldata)
6343 {
6344         struct nfs4_layoutget *lgp = calldata;
6345         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6346         size_t max_pages = max_response_pages(server);
6347
6348         dprintk("--> %s\n", __func__);
6349         nfs4_free_pages(lgp->args.layout.pages, max_pages);
6350         put_nfs_open_context(lgp->args.ctx);
6351         kfree(calldata);
6352         dprintk("<-- %s\n", __func__);
6353 }
6354
6355 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
6356         .rpc_call_prepare = nfs4_layoutget_prepare,
6357         .rpc_call_done = nfs4_layoutget_done,
6358         .rpc_release = nfs4_layoutget_release,
6359 };
6360
6361 struct pnfs_layout_segment *
6362 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
6363 {
6364         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
6365         size_t max_pages = max_response_pages(server);
6366         struct rpc_task *task;
6367         struct rpc_message msg = {
6368                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
6369                 .rpc_argp = &lgp->args,
6370                 .rpc_resp = &lgp->res,
6371         };
6372         struct rpc_task_setup task_setup_data = {
6373                 .rpc_client = server->client,
6374                 .rpc_message = &msg,
6375                 .callback_ops = &nfs4_layoutget_call_ops,
6376                 .callback_data = lgp,
6377                 .flags = RPC_TASK_ASYNC,
6378         };
6379         struct pnfs_layout_segment *lseg = NULL;
6380         int status = 0;
6381
6382         dprintk("--> %s\n", __func__);
6383
6384         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
6385         if (!lgp->args.layout.pages) {
6386                 nfs4_layoutget_release(lgp);
6387                 return ERR_PTR(-ENOMEM);
6388         }
6389         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
6390
6391         lgp->res.layoutp = &lgp->args.layout;
6392         lgp->res.seq_res.sr_slot = NULL;
6393         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
6394         task = rpc_run_task(&task_setup_data);
6395         if (IS_ERR(task))
6396                 return ERR_CAST(task);
6397         status = nfs4_wait_for_completion_rpc_task(task);
6398         if (status == 0)
6399                 status = task->tk_status;
6400         if (status == 0)
6401                 lseg = pnfs_layout_process(lgp);
6402         rpc_put_task(task);
6403         dprintk("<-- %s status=%d\n", __func__, status);
6404         if (status)
6405                 return ERR_PTR(status);
6406         return lseg;
6407 }
6408
6409 static void
6410 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
6411 {
6412         struct nfs4_layoutreturn *lrp = calldata;
6413
6414         dprintk("--> %s\n", __func__);
6415         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
6416                                 &lrp->res.seq_res, task))
6417                 return;
6418         rpc_call_start(task);
6419 }
6420
6421 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
6422 {
6423         struct nfs4_layoutreturn *lrp = calldata;
6424         struct nfs_server *server;
6425
6426         dprintk("--> %s\n", __func__);
6427
6428         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
6429                 return;
6430
6431         server = NFS_SERVER(lrp->args.inode);
6432         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6433                 rpc_restart_call_prepare(task);
6434                 return;
6435         }
6436         dprintk("<-- %s\n", __func__);
6437 }
6438
6439 static void nfs4_layoutreturn_release(void *calldata)
6440 {
6441         struct nfs4_layoutreturn *lrp = calldata;
6442         struct pnfs_layout_hdr *lo = lrp->args.layout;
6443
6444         dprintk("--> %s\n", __func__);
6445         spin_lock(&lo->plh_inode->i_lock);
6446         if (lrp->res.lrs_present)
6447                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
6448         lo->plh_block_lgets--;
6449         spin_unlock(&lo->plh_inode->i_lock);
6450         pnfs_put_layout_hdr(lrp->args.layout);
6451         kfree(calldata);
6452         dprintk("<-- %s\n", __func__);
6453 }
6454
6455 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
6456         .rpc_call_prepare = nfs4_layoutreturn_prepare,
6457         .rpc_call_done = nfs4_layoutreturn_done,
6458         .rpc_release = nfs4_layoutreturn_release,
6459 };
6460
6461 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
6462 {
6463         struct rpc_task *task;
6464         struct rpc_message msg = {
6465                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
6466                 .rpc_argp = &lrp->args,
6467                 .rpc_resp = &lrp->res,
6468         };
6469         struct rpc_task_setup task_setup_data = {
6470                 .rpc_client = lrp->clp->cl_rpcclient,
6471                 .rpc_message = &msg,
6472                 .callback_ops = &nfs4_layoutreturn_call_ops,
6473                 .callback_data = lrp,
6474         };
6475         int status;
6476
6477         dprintk("--> %s\n", __func__);
6478         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
6479         task = rpc_run_task(&task_setup_data);
6480         if (IS_ERR(task))
6481                 return PTR_ERR(task);
6482         status = task->tk_status;
6483         dprintk("<-- %s status=%d\n", __func__, status);
6484         rpc_put_task(task);
6485         return status;
6486 }
6487
6488 /*
6489  * Retrieve the list of Data Server devices from the MDS.
6490  */
6491 static int _nfs4_getdevicelist(struct nfs_server *server,
6492                                     const struct nfs_fh *fh,
6493                                     struct pnfs_devicelist *devlist)
6494 {
6495         struct nfs4_getdevicelist_args args = {
6496                 .fh = fh,
6497                 .layoutclass = server->pnfs_curr_ld->id,
6498         };
6499         struct nfs4_getdevicelist_res res = {
6500                 .devlist = devlist,
6501         };
6502         struct rpc_message msg = {
6503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6504                 .rpc_argp = &args,
6505                 .rpc_resp = &res,
6506         };
6507         int status;
6508
6509         dprintk("--> %s\n", __func__);
6510         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6511                                 &res.seq_res, 0);
6512         dprintk("<-- %s status=%d\n", __func__, status);
6513         return status;
6514 }
6515
6516 int nfs4_proc_getdevicelist(struct nfs_server *server,
6517                             const struct nfs_fh *fh,
6518                             struct pnfs_devicelist *devlist)
6519 {
6520         struct nfs4_exception exception = { };
6521         int err;
6522
6523         do {
6524                 err = nfs4_handle_exception(server,
6525                                 _nfs4_getdevicelist(server, fh, devlist),
6526                                 &exception);
6527         } while (exception.retry);
6528
6529         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6530                 err, devlist->num_devs);
6531
6532         return err;
6533 }
6534 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6535
6536 static int
6537 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6538 {
6539         struct nfs4_getdeviceinfo_args args = {
6540                 .pdev = pdev,
6541         };
6542         struct nfs4_getdeviceinfo_res res = {
6543                 .pdev = pdev,
6544         };
6545         struct rpc_message msg = {
6546                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6547                 .rpc_argp = &args,
6548                 .rpc_resp = &res,
6549         };
6550         int status;
6551
6552         dprintk("--> %s\n", __func__);
6553         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6554         dprintk("<-- %s status=%d\n", __func__, status);
6555
6556         return status;
6557 }
6558
6559 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6560 {
6561         struct nfs4_exception exception = { };
6562         int err;
6563
6564         do {
6565                 err = nfs4_handle_exception(server,
6566                                         _nfs4_proc_getdeviceinfo(server, pdev),
6567                                         &exception);
6568         } while (exception.retry);
6569         return err;
6570 }
6571 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6572
6573 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6574 {
6575         struct nfs4_layoutcommit_data *data = calldata;
6576         struct nfs_server *server = NFS_SERVER(data->args.inode);
6577
6578         if (nfs4_setup_sequence(server, &data->args.seq_args,
6579                                 &data->res.seq_res, task))
6580                 return;
6581         rpc_call_start(task);
6582 }
6583
6584 static void
6585 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6586 {
6587         struct nfs4_layoutcommit_data *data = calldata;
6588         struct nfs_server *server = NFS_SERVER(data->args.inode);
6589
6590         if (!nfs4_sequence_done(task, &data->res.seq_res))
6591                 return;
6592
6593         switch (task->tk_status) { /* Just ignore these failures */
6594         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6595         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6596         case -NFS4ERR_BADLAYOUT:     /* no layout */
6597         case -NFS4ERR_GRACE:        /* loca_recalim always false */
6598                 task->tk_status = 0;
6599                 break;
6600         case 0:
6601                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6602                                                    data->res.fattr);
6603                 break;
6604         default:
6605                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6606                         rpc_restart_call_prepare(task);
6607                         return;
6608                 }
6609         }
6610 }
6611
6612 static void nfs4_layoutcommit_release(void *calldata)
6613 {
6614         struct nfs4_layoutcommit_data *data = calldata;
6615         struct pnfs_layout_segment *lseg, *tmp;
6616         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6617
6618         pnfs_cleanup_layoutcommit(data);
6619         /* Matched by references in pnfs_set_layoutcommit */
6620         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6621                 list_del_init(&lseg->pls_lc_list);
6622                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6623                                        &lseg->pls_flags))
6624                         pnfs_put_lseg(lseg);
6625         }
6626
6627         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6628         smp_mb__after_clear_bit();
6629         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6630
6631         put_rpccred(data->cred);
6632         kfree(data);
6633 }
6634
6635 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6636         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6637         .rpc_call_done = nfs4_layoutcommit_done,
6638         .rpc_release = nfs4_layoutcommit_release,
6639 };
6640
6641 int
6642 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6643 {
6644         struct rpc_message msg = {
6645                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6646                 .rpc_argp = &data->args,
6647                 .rpc_resp = &data->res,
6648                 .rpc_cred = data->cred,
6649         };
6650         struct rpc_task_setup task_setup_data = {
6651                 .task = &data->task,
6652                 .rpc_client = NFS_CLIENT(data->args.inode),
6653                 .rpc_message = &msg,
6654                 .callback_ops = &nfs4_layoutcommit_ops,
6655                 .callback_data = data,
6656                 .flags = RPC_TASK_ASYNC,
6657         };
6658         struct rpc_task *task;
6659         int status = 0;
6660
6661         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6662                 "lbw: %llu inode %lu\n",
6663                 data->task.tk_pid, sync,
6664                 data->args.lastbytewritten,
6665                 data->args.inode->i_ino);
6666
6667         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6668         task = rpc_run_task(&task_setup_data);
6669         if (IS_ERR(task))
6670                 return PTR_ERR(task);
6671         if (sync == false)
6672                 goto out;
6673         status = nfs4_wait_for_completion_rpc_task(task);
6674         if (status != 0)
6675                 goto out;
6676         status = task->tk_status;
6677 out:
6678         dprintk("%s: status %d\n", __func__, status);
6679         rpc_put_task(task);
6680         return status;
6681 }
6682
6683 static int
6684 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6685                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6686 {
6687         struct nfs41_secinfo_no_name_args args = {
6688                 .style = SECINFO_STYLE_CURRENT_FH,
6689         };
6690         struct nfs4_secinfo_res res = {
6691                 .flavors = flavors,
6692         };
6693         struct rpc_message msg = {
6694                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6695                 .rpc_argp = &args,
6696                 .rpc_resp = &res,
6697         };
6698         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6699 }
6700
6701 static int
6702 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6703                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6704 {
6705         struct nfs4_exception exception = { };
6706         int err;
6707         do {
6708                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6709                 switch (err) {
6710                 case 0:
6711                 case -NFS4ERR_WRONGSEC:
6712                 case -NFS4ERR_NOTSUPP:
6713                         goto out;
6714                 default:
6715                         err = nfs4_handle_exception(server, err, &exception);
6716                 }
6717         } while (exception.retry);
6718 out:
6719         return err;
6720 }
6721
6722 static int
6723 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6724                     struct nfs_fsinfo *info)
6725 {
6726         int err;
6727         struct page *page;
6728         rpc_authflavor_t flavor;
6729         struct nfs4_secinfo_flavors *flavors;
6730
6731         page = alloc_page(GFP_KERNEL);
6732         if (!page) {
6733                 err = -ENOMEM;
6734                 goto out;
6735         }
6736
6737         flavors = page_address(page);
6738         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6739
6740         /*
6741          * Fall back on "guess and check" method if
6742          * the server doesn't support SECINFO_NO_NAME
6743          */
6744         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6745                 err = nfs4_find_root_sec(server, fhandle, info);
6746                 goto out_freepage;
6747         }
6748         if (err)
6749                 goto out_freepage;
6750
6751         flavor = nfs_find_best_sec(flavors);
6752         if (err == 0)
6753                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6754
6755 out_freepage:
6756         put_page(page);
6757         if (err == -EACCES)
6758                 return -EPERM;
6759 out:
6760         return err;
6761 }
6762
6763 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6764 {
6765         int status;
6766         struct nfs41_test_stateid_args args = {
6767                 .stateid = stateid,
6768         };
6769         struct nfs41_test_stateid_res res;
6770         struct rpc_message msg = {
6771                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6772                 .rpc_argp = &args,
6773                 .rpc_resp = &res,
6774         };
6775
6776         dprintk("NFS call  test_stateid %p\n", stateid);
6777         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6778         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6779         if (status != NFS_OK) {
6780                 dprintk("NFS reply test_stateid: failed, %d\n", status);
6781                 return status;
6782         }
6783         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
6784         return -res.status;
6785 }
6786
6787 /**
6788  * nfs41_test_stateid - perform a TEST_STATEID operation
6789  *
6790  * @server: server / transport on which to perform the operation
6791  * @stateid: state ID to test
6792  *
6793  * Returns NFS_OK if the server recognizes that "stateid" is valid.
6794  * Otherwise a negative NFS4ERR value is returned if the operation
6795  * failed or the state ID is not currently valid.
6796  */
6797 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6798 {
6799         struct nfs4_exception exception = { };
6800         int err;
6801         do {
6802                 err = _nfs41_test_stateid(server, stateid);
6803                 if (err != -NFS4ERR_DELAY)
6804                         break;
6805                 nfs4_handle_exception(server, err, &exception);
6806         } while (exception.retry);
6807         return err;
6808 }
6809
6810 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6811 {
6812         struct nfs41_free_stateid_args args = {
6813                 .stateid = stateid,
6814         };
6815         struct nfs41_free_stateid_res res;
6816         struct rpc_message msg = {
6817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6818                 .rpc_argp = &args,
6819                 .rpc_resp = &res,
6820         };
6821         int status;
6822
6823         dprintk("NFS call  free_stateid %p\n", stateid);
6824         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6825         status = nfs4_call_sync_sequence(server->client, server, &msg,
6826                                          &args.seq_args, &res.seq_res, 1);
6827         dprintk("NFS reply free_stateid: %d\n", status);
6828         return status;
6829 }
6830
6831 /**
6832  * nfs41_free_stateid - perform a FREE_STATEID operation
6833  *
6834  * @server: server / transport on which to perform the operation
6835  * @stateid: state ID to release
6836  *
6837  * Returns NFS_OK if the server freed "stateid".  Otherwise a
6838  * negative NFS4ERR value is returned.
6839  */
6840 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6841 {
6842         struct nfs4_exception exception = { };
6843         int err;
6844         do {
6845                 err = _nfs4_free_stateid(server, stateid);
6846                 if (err != -NFS4ERR_DELAY)
6847                         break;
6848                 nfs4_handle_exception(server, err, &exception);
6849         } while (exception.retry);
6850         return err;
6851 }
6852
6853 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6854                 const nfs4_stateid *s2)
6855 {
6856         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6857                 return false;
6858
6859         if (s1->seqid == s2->seqid)
6860                 return true;
6861         if (s1->seqid == 0 || s2->seqid == 0)
6862                 return true;
6863
6864         return false;
6865 }
6866
6867 #endif /* CONFIG_NFS_V4_1 */
6868
6869 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6870                 const nfs4_stateid *s2)
6871 {
6872         return nfs4_stateid_match(s1, s2);
6873 }
6874
6875
6876 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6877         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6878         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6879         .recover_open   = nfs4_open_reclaim,
6880         .recover_lock   = nfs4_lock_reclaim,
6881         .establish_clid = nfs4_init_clientid,
6882         .get_clid_cred  = nfs4_get_setclientid_cred,
6883         .detect_trunking = nfs40_discover_server_trunking,
6884 };
6885
6886 #if defined(CONFIG_NFS_V4_1)
6887 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6888         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6889         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6890         .recover_open   = nfs4_open_reclaim,
6891         .recover_lock   = nfs4_lock_reclaim,
6892         .establish_clid = nfs41_init_clientid,
6893         .get_clid_cred  = nfs4_get_exchange_id_cred,
6894         .reclaim_complete = nfs41_proc_reclaim_complete,
6895         .detect_trunking = nfs41_discover_server_trunking,
6896 };
6897 #endif /* CONFIG_NFS_V4_1 */
6898
6899 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6900         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6901         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6902         .recover_open   = nfs4_open_expired,
6903         .recover_lock   = nfs4_lock_expired,
6904         .establish_clid = nfs4_init_clientid,
6905         .get_clid_cred  = nfs4_get_setclientid_cred,
6906 };
6907
6908 #if defined(CONFIG_NFS_V4_1)
6909 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6910         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6911         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6912         .recover_open   = nfs41_open_expired,
6913         .recover_lock   = nfs41_lock_expired,
6914         .establish_clid = nfs41_init_clientid,
6915         .get_clid_cred  = nfs4_get_exchange_id_cred,
6916 };
6917 #endif /* CONFIG_NFS_V4_1 */
6918
6919 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6920         .sched_state_renewal = nfs4_proc_async_renew,
6921         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6922         .renew_lease = nfs4_proc_renew,
6923 };
6924
6925 #if defined(CONFIG_NFS_V4_1)
6926 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6927         .sched_state_renewal = nfs41_proc_async_sequence,
6928         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6929         .renew_lease = nfs4_proc_sequence,
6930 };
6931 #endif
6932
6933 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6934         .minor_version = 0,
6935         .call_sync = _nfs4_call_sync,
6936         .match_stateid = nfs4_match_stateid,
6937         .find_root_sec = nfs4_find_root_sec,
6938         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6939         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6940         .state_renewal_ops = &nfs40_state_renewal_ops,
6941 };
6942
6943 #if defined(CONFIG_NFS_V4_1)
6944 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6945         .minor_version = 1,
6946         .call_sync = _nfs4_call_sync_session,
6947         .match_stateid = nfs41_match_stateid,
6948         .find_root_sec = nfs41_find_root_sec,
6949         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6950         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6951         .state_renewal_ops = &nfs41_state_renewal_ops,
6952 };
6953 #endif
6954
6955 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6956         [0] = &nfs_v4_0_minor_ops,
6957 #if defined(CONFIG_NFS_V4_1)
6958         [1] = &nfs_v4_1_minor_ops,
6959 #endif
6960 };
6961
6962 const struct inode_operations nfs4_dir_inode_operations = {
6963         .create         = nfs_create,
6964         .lookup         = nfs_lookup,
6965         .atomic_open    = nfs_atomic_open,
6966         .link           = nfs_link,
6967         .unlink         = nfs_unlink,
6968         .symlink        = nfs_symlink,
6969         .mkdir          = nfs_mkdir,
6970         .rmdir          = nfs_rmdir,
6971         .mknod          = nfs_mknod,
6972         .rename         = nfs_rename,
6973         .permission     = nfs_permission,
6974         .getattr        = nfs_getattr,
6975         .setattr        = nfs_setattr,
6976         .getxattr       = generic_getxattr,
6977         .setxattr       = generic_setxattr,
6978         .listxattr      = generic_listxattr,
6979         .removexattr    = generic_removexattr,
6980 };
6981
6982 static const struct inode_operations nfs4_file_inode_operations = {
6983         .permission     = nfs_permission,
6984         .getattr        = nfs_getattr,
6985         .setattr        = nfs_setattr,
6986         .getxattr       = generic_getxattr,
6987         .setxattr       = generic_setxattr,
6988         .listxattr      = generic_listxattr,
6989         .removexattr    = generic_removexattr,
6990 };
6991
6992 const struct nfs_rpc_ops nfs_v4_clientops = {
6993         .version        = 4,                    /* protocol version */
6994         .dentry_ops     = &nfs4_dentry_operations,
6995         .dir_inode_ops  = &nfs4_dir_inode_operations,
6996         .file_inode_ops = &nfs4_file_inode_operations,
6997         .file_ops       = &nfs4_file_operations,
6998         .getroot        = nfs4_proc_get_root,
6999         .submount       = nfs4_submount,
7000         .try_mount      = nfs4_try_mount,
7001         .getattr        = nfs4_proc_getattr,
7002         .setattr        = nfs4_proc_setattr,
7003         .lookup         = nfs4_proc_lookup,
7004         .access         = nfs4_proc_access,
7005         .readlink       = nfs4_proc_readlink,
7006         .create         = nfs4_proc_create,
7007         .remove         = nfs4_proc_remove,
7008         .unlink_setup   = nfs4_proc_unlink_setup,
7009         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
7010         .unlink_done    = nfs4_proc_unlink_done,
7011         .rename         = nfs4_proc_rename,
7012         .rename_setup   = nfs4_proc_rename_setup,
7013         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
7014         .rename_done    = nfs4_proc_rename_done,
7015         .link           = nfs4_proc_link,
7016         .symlink        = nfs4_proc_symlink,
7017         .mkdir          = nfs4_proc_mkdir,
7018         .rmdir          = nfs4_proc_remove,
7019         .readdir        = nfs4_proc_readdir,
7020         .mknod          = nfs4_proc_mknod,
7021         .statfs         = nfs4_proc_statfs,
7022         .fsinfo         = nfs4_proc_fsinfo,
7023         .pathconf       = nfs4_proc_pathconf,
7024         .set_capabilities = nfs4_server_capabilities,
7025         .decode_dirent  = nfs4_decode_dirent,
7026         .read_setup     = nfs4_proc_read_setup,
7027         .read_pageio_init = pnfs_pageio_init_read,
7028         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
7029         .read_done      = nfs4_read_done,
7030         .write_setup    = nfs4_proc_write_setup,
7031         .write_pageio_init = pnfs_pageio_init_write,
7032         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
7033         .write_done     = nfs4_write_done,
7034         .commit_setup   = nfs4_proc_commit_setup,
7035         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
7036         .commit_done    = nfs4_commit_done,
7037         .lock           = nfs4_proc_lock,
7038         .clear_acl_cache = nfs4_zap_acl_attr,
7039         .close_context  = nfs4_close_context,
7040         .open_context   = nfs4_atomic_open,
7041         .have_delegation = nfs4_have_delegation,
7042         .return_delegation = nfs4_inode_return_delegation,
7043         .alloc_client   = nfs4_alloc_client,
7044         .init_client    = nfs4_init_client,
7045         .free_client    = nfs4_free_client,
7046         .create_server  = nfs4_create_server,
7047         .clone_server   = nfs_clone_server,
7048 };
7049
7050 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
7051         .prefix = XATTR_NAME_NFSV4_ACL,
7052         .list   = nfs4_xattr_list_nfs4_acl,
7053         .get    = nfs4_xattr_get_nfs4_acl,
7054         .set    = nfs4_xattr_set_nfs4_acl,
7055 };
7056
7057 const struct xattr_handler *nfs4_xattr_handlers[] = {
7058         &nfs4_xattr_nfs4_acl_handler,
7059         NULL
7060 };
7061
7062 /*
7063  * Local variables:
7064  *  c-basic-offset: 8
7065  * End:
7066  */