fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/sunrpc/gss_api.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/nfs_idmap.h>
  56 #include <linux/sunrpc/bc_xprt.h>
  57 #include <linux/xattr.h>
  58 #include <linux/utsname.h>
  59 #include <linux/freezer.h>
  60
  61 #include "nfs4_fs.h"
  62 #include "delegation.h"
  63 #include "internal.h"
  64 #include "iostat.h"
  65 #include "callback.h"
  66 #include "pnfs.h"
  67
  68 #define NFSDBG_FACILITY         NFSDBG_PROC
  69
  70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  72
  73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  74
  75 static unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state);
  87 #ifdef CONFIG_NFS_V4_1
  88 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  89 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  90 #endif
  91 /* Prevent leaks of NFSv4 errors into userland */
  92 static int nfs4_map_errors(int err)
  93 {
  94         if (err >= -1000)
  95                 return err;
  96         switch (err) {
  97         case -NFS4ERR_RESOURCE:
  98                 return -EREMOTEIO;
  99         case -NFS4ERR_WRONGSEC:
 100                 return -EPERM;
 101         case -NFS4ERR_BADOWNER:
 102         case -NFS4ERR_BADNAME:
 103                 return -EINVAL;
 104         default:
 105                 dprintk("%s could not handle NFSv4 error %d\n",
 106                                 __func__, -err);
 107                 break;
 108         }
 109         return -EIO;
 110 }
 111
 112 /*
 113  * This is our standard bitmap for GETATTR requests.
 114  */
 115 const u32 nfs4_fattr_bitmap[2] = {
 116         FATTR4_WORD0_TYPE
 117         | FATTR4_WORD0_CHANGE
 118         | FATTR4_WORD0_SIZE
 119         | FATTR4_WORD0_FSID
 120         | FATTR4_WORD0_FILEID,
 121         FATTR4_WORD1_MODE
 122         | FATTR4_WORD1_NUMLINKS
 123         | FATTR4_WORD1_OWNER
 124         | FATTR4_WORD1_OWNER_GROUP
 125         | FATTR4_WORD1_RAWDEV
 126         | FATTR4_WORD1_SPACE_USED
 127         | FATTR4_WORD1_TIME_ACCESS
 128         | FATTR4_WORD1_TIME_METADATA
 129         | FATTR4_WORD1_TIME_MODIFY
 130 };
 131
 132 const u32 nfs4_statfs_bitmap[2] = {
 133         FATTR4_WORD0_FILES_AVAIL
 134         | FATTR4_WORD0_FILES_FREE
 135         | FATTR4_WORD0_FILES_TOTAL,
 136         FATTR4_WORD1_SPACE_AVAIL
 137         | FATTR4_WORD1_SPACE_FREE
 138         | FATTR4_WORD1_SPACE_TOTAL
 139 };
 140
 141 const u32 nfs4_pathconf_bitmap[2] = {
 142         FATTR4_WORD0_MAXLINK
 143         | FATTR4_WORD0_MAXNAME,
 144         0
 145 };
 146
 147 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 148                         | FATTR4_WORD0_MAXREAD
 149                         | FATTR4_WORD0_MAXWRITE
 150                         | FATTR4_WORD0_LEASE_TIME,
 151                         FATTR4_WORD1_TIME_DELTA
 152                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 153                         FATTR4_WORD2_LAYOUT_BLKSIZE
 154 };
 155
 156 const u32 nfs4_fs_locations_bitmap[2] = {
 157         FATTR4_WORD0_TYPE
 158         | FATTR4_WORD0_CHANGE
 159         | FATTR4_WORD0_SIZE
 160         | FATTR4_WORD0_FSID
 161         | FATTR4_WORD0_FILEID
 162         | FATTR4_WORD0_FS_LOCATIONS,
 163         FATTR4_WORD1_MODE
 164         | FATTR4_WORD1_NUMLINKS
 165         | FATTR4_WORD1_OWNER
 166         | FATTR4_WORD1_OWNER_GROUP
 167         | FATTR4_WORD1_RAWDEV
 168         | FATTR4_WORD1_SPACE_USED
 169         | FATTR4_WORD1_TIME_ACCESS
 170         | FATTR4_WORD1_TIME_METADATA
 171         | FATTR4_WORD1_TIME_MODIFY
 172         | FATTR4_WORD1_MOUNTED_ON_FILEID
 173 };
 174
 175 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 176                 struct nfs4_readdir_arg *readdir)
 177 {
 178         __be32 *start, *p;
 179
 180         BUG_ON(readdir->count < 80);
 181         if (cookie > 2) {
 182                 readdir->cookie = cookie;
 183                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 184                 return;
 185         }
 186
 187         readdir->cookie = 0;
 188         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 189         if (cookie == 2)
 190                 return;
 191
 192         /*
 193          * NFSv4 servers do not return entries for '.' and '..'
 194          * Therefore, we fake these entries here.  We let '.'
 195          * have cookie 0 and '..' have cookie 1.  Note that
 196          * when talking to the server, we always send cookie 0
 197          * instead of 1 or 2.
 198          */
 199         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 200
 201         if (cookie == 0) {
 202                 *p++ = xdr_one;                                  /* next */
 203                 *p++ = xdr_zero;                   /* cookie, first word */
 204                 *p++ = xdr_one;                   /* cookie, second word */
 205                 *p++ = xdr_one;                             /* entry len */
 206                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 207                 p++;
 208                 *p++ = xdr_one;                         /* bitmap length */
 209                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 210                 *p++ = htonl(8);              /* attribute buffer length */
 211                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 212         }
 213
 214         *p++ = xdr_one;                                  /* next */
 215         *p++ = xdr_zero;                   /* cookie, first word */
 216         *p++ = xdr_two;                   /* cookie, second word */
 217         *p++ = xdr_two;                             /* entry len */
 218         memcpy(p, "..\0\0", 4);                         /* entry */
 219         p++;
 220         *p++ = xdr_one;                         /* bitmap length */
 221         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 222         *p++ = htonl(8);              /* attribute buffer length */
 223         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 224
 225         readdir->pgbase = (char *)p - (char *)start;
 226         readdir->count -= readdir->pgbase;
 227         kunmap_atomic(start, KM_USER0);
 228 }
 229
 230 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 231 {
 232         int res;
 233
 234         might_sleep();
 235
 236         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 237                         nfs_wait_bit_killable, TASK_KILLABLE);
 238         return res;
 239 }
 240
 241 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 242 {
 243         int res = 0;
 244
 245         might_sleep();
 246
 247         if (*timeout <= 0)
 248                 *timeout = NFS4_POLL_RETRY_MIN;
 249         if (*timeout > NFS4_POLL_RETRY_MAX)
 250                 *timeout = NFS4_POLL_RETRY_MAX;
 251         freezable_schedule_timeout_killable(*timeout);
 252         if (fatal_signal_pending(current))
 253                 res = -ERESTARTSYS;
 254         *timeout <<= 1;
 255         return res;
 256 }
 257
 258 /* This is the error handling routine for processes that are allowed
 259  * to sleep.
 260  */
 261 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 262 {
 263         struct nfs_client *clp = server->nfs_client;
 264         struct nfs4_state *state = exception->state;
 265         int ret = errorcode;
 266
 267         exception->retry = 0;
 268         switch(errorcode) {
 269                 case 0:
 270                         return 0;
 271                 case -NFS4ERR_DELEG_REVOKED:
 272                 case -NFS4ERR_ADMIN_REVOKED:
 273                 case -NFS4ERR_BAD_STATEID:
 274                         if (state != NULL)
 275                                 nfs_remove_bad_delegation(state->inode);
 276                 case -NFS4ERR_OPENMODE:
 277                         if (state == NULL)
 278                                 break;
 279                         nfs4_schedule_stateid_recovery(server, state);
 280                         goto wait_on_recovery;
 281                 case -NFS4ERR_EXPIRED:
 282                         if (state != NULL)
 283                                 nfs4_schedule_stateid_recovery(server, state);
 284                 case -NFS4ERR_STALE_STATEID:
 285                 case -NFS4ERR_STALE_CLIENTID:
 286                         nfs4_schedule_lease_recovery(clp);
 287                         goto wait_on_recovery;
 288 #if defined(CONFIG_NFS_V4_1)
 289                 case -NFS4ERR_BADSESSION:
 290                 case -NFS4ERR_BADSLOT:
 291                 case -NFS4ERR_BAD_HIGH_SLOT:
 292                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 293                 case -NFS4ERR_DEADSESSION:
 294                 case -NFS4ERR_SEQ_FALSE_RETRY:
 295                 case -NFS4ERR_SEQ_MISORDERED:
 296                         dprintk("%s ERROR: %d Reset session\n", __func__,
 297                                 errorcode);
 298                         nfs4_schedule_session_recovery(clp->cl_session);
 299                         exception->retry = 1;
 300                         break;
 301 #endif /* defined(CONFIG_NFS_V4_1) */
 302                 case -NFS4ERR_FILE_OPEN:
 303                         if (exception->timeout > HZ) {
 304                                 /* We have retried a decent amount, time to
 305                                  * fail
 306                                  */
 307                                 ret = -EBUSY;
 308                                 break;
 309                         }
 310                 case -NFS4ERR_GRACE:
 311                 case -NFS4ERR_DELAY:
 312                 case -EKEYEXPIRED:
 313                         ret = nfs4_delay(server->client, &exception->timeout);
 314                         if (ret != 0)
 315                                 break;
 316                 case -NFS4ERR_RETRY_UNCACHED_REP:
 317                 case -NFS4ERR_OLD_STATEID:
 318                         exception->retry = 1;
 319                         break;
 320                 case -NFS4ERR_BADOWNER:
 321                         /* The following works around a Linux server bug! */
 322                 case -NFS4ERR_BADNAME:
 323                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 324                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 325                                 exception->retry = 1;
 326                                 printk(KERN_WARNING "NFS: v4 server %s "
 327                                                 "does not accept raw "
 328                                                 "uid/gids. "
 329                                                 "Reenabling the idmapper.\n",
 330                                                 server->nfs_client->cl_hostname);
 331                         }
 332         }
 333         /* We failed to handle the error */
 334         return nfs4_map_errors(ret);
 335 wait_on_recovery:
 336         ret = nfs4_wait_clnt_recover(clp);
 337         if (ret == 0)
 338                 exception->retry = 1;
 339         return ret;
 340 }
 341
 342
 343 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 344 {
 345         spin_lock(&clp->cl_lock);
 346         if (time_before(clp->cl_last_renewal,timestamp))
 347                 clp->cl_last_renewal = timestamp;
 348         spin_unlock(&clp->cl_lock);
 349 }
 350
 351 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 352 {
 353         do_renew_lease(server->nfs_client, timestamp);
 354 }
 355
 356 #if defined(CONFIG_NFS_V4_1)
 357
 358 /*
 359  * nfs4_free_slot - free a slot and efficiently update slot table.
 360  *
 361  * freeing a slot is trivially done by clearing its respective bit
 362  * in the bitmap.
 363  * If the freed slotid equals highest_used_slotid we want to update it
 364  * so that the server would be able to size down the slot table if needed,
 365  * otherwise we know that the highest_used_slotid is still in use.
 366  * When updating highest_used_slotid there may be "holes" in the bitmap
 367  * so we need to scan down from highest_used_slotid to 0 looking for the now
 368  * highest slotid in use.
 369  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 370  *
 371  * Must be called while holding tbl->slot_tbl_lock
 372  */
 373 static void
 374 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 375 {
 376         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 377         /* clear used bit in bitmap */
 378         __clear_bit(slotid, tbl->used_slots);
 379
 380         /* update highest_used_slotid when it is freed */
 381         if (slotid == tbl->highest_used_slotid) {
 382                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 383                 if (slotid < tbl->max_slots)
 384                         tbl->highest_used_slotid = slotid;
 385                 else
 386                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 387         }
 388         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 389                 slotid, tbl->highest_used_slotid);
 390 }
 391
 392 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 393 {
 394         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 395         return true;
 396 }
 397
 398 /*
 399  * Signal state manager thread if session fore channel is drained
 400  */
 401 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 402 {
 403         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 404                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 405                                 nfs4_set_task_privileged, NULL);
 406                 return;
 407         }
 408
 409         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 410                 return;
 411
 412         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 413         complete(&ses->fc_slot_table.complete);
 414 }
 415
 416 /*
 417  * Signal state manager thread if session back channel is drained
 418  */
 419 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 420 {
 421         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 422             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 423                 return;
 424         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 425         complete(&ses->bc_slot_table.complete);
 426 }
 427
 428 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 429 {
 430         struct nfs4_slot_table *tbl;
 431
 432         tbl = &res->sr_session->fc_slot_table;
 433         if (!res->sr_slot) {
 434                 /* just wake up the next guy waiting since
 435                  * we may have not consumed a slot after all */
 436                 dprintk("%s: No slot\n", __func__);
 437                 return;
 438         }
 439
 440         spin_lock(&tbl->slot_tbl_lock);
 441         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 442         nfs4_check_drain_fc_complete(res->sr_session);
 443         spin_unlock(&tbl->slot_tbl_lock);
 444         res->sr_slot = NULL;
 445 }
 446
 447 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 448 {
 449         unsigned long timestamp;
 450         struct nfs_client *clp;
 451
 452         /*
 453          * sr_status remains 1 if an RPC level error occurred. The server
 454          * may or may not have processed the sequence operation..
 455          * Proceed as if the server received and processed the sequence
 456          * operation.
 457          */
 458         if (res->sr_status == 1)
 459                 res->sr_status = NFS_OK;
 460
 461         /* don't increment the sequence number if the task wasn't sent */
 462         if (!RPC_WAS_SENT(task))
 463                 goto out;
 464
 465         /* Check the SEQUENCE operation status */
 466         switch (res->sr_status) {
 467         case 0:
 468                 /* Update the slot's sequence and clientid lease timer */
 469                 ++res->sr_slot->seq_nr;
 470                 timestamp = res->sr_renewal_time;
 471                 clp = res->sr_session->clp;
 472                 do_renew_lease(clp, timestamp);
 473                 /* Check sequence flags */
 474                 if (res->sr_status_flags != 0)
 475                         nfs4_schedule_lease_recovery(clp);
 476                 break;
 477         case -NFS4ERR_DELAY:
 478                 /* The server detected a resend of the RPC call and
 479                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 480                  * of RFC5661.
 481                  */
 482                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 483                         __func__,
 484                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 485                         res->sr_slot->seq_nr);
 486                 goto out_retry;
 487         default:
 488                 /* Just update the slot sequence no. */
 489                 ++res->sr_slot->seq_nr;
 490         }
 491 out:
 492         /* The session may be reset by one of the error handlers. */
 493         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 494         nfs41_sequence_free_slot(res);
 495         return 1;
 496 out_retry:
 497         if (!rpc_restart_call(task))
 498                 goto out;
 499         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 500         return 0;
 501 }
 502
 503 static int nfs4_sequence_done(struct rpc_task *task,
 504                                struct nfs4_sequence_res *res)
 505 {
 506         if (res->sr_session == NULL)
 507                 return 1;
 508         return nfs41_sequence_done(task, res);
 509 }
 510
 511 /*
 512  * nfs4_find_slot - efficiently look for a free slot
 513  *
 514  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 515  * If found, we mark the slot as used, update the highest_used_slotid,
 516  * and respectively set up the sequence operation args.
 517  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 518  *
 519  * Note: must be called with under the slot_tbl_lock.
 520  */
 521 static u32
 522 nfs4_find_slot(struct nfs4_slot_table *tbl)
 523 {
 524         u32 slotid;
 525         u32 ret_id = NFS4_NO_SLOT;
 526
 527         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 528                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 529                 tbl->max_slots);
 530         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 531         if (slotid >= tbl->max_slots)
 532                 goto out;
 533         __set_bit(slotid, tbl->used_slots);
 534         if (slotid > tbl->highest_used_slotid ||
 535                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 536                 tbl->highest_used_slotid = slotid;
 537         ret_id = slotid;
 538 out:
 539         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 540                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 541         return ret_id;
 542 }
 543
 544 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 545                 struct nfs4_sequence_res *res, int cache_reply)
 546 {
 547         args->sa_session = NULL;
 548         args->sa_cache_this = 0;
 549         if (cache_reply)
 550                 args->sa_cache_this = 1;
 551         res->sr_session = NULL;
 552         res->sr_slot = NULL;
 553 }
 554
 555 int nfs41_setup_sequence(struct nfs4_session *session,
 556                                 struct nfs4_sequence_args *args,
 557                                 struct nfs4_sequence_res *res,
 558                                 struct rpc_task *task)
 559 {
 560         struct nfs4_slot *slot;
 561         struct nfs4_slot_table *tbl;
 562         u32 slotid;
 563
 564         dprintk("--> %s\n", __func__);
 565         /* slot already allocated? */
 566         if (res->sr_slot != NULL)
 567                 return 0;
 568
 569         tbl = &session->fc_slot_table;
 570
 571         spin_lock(&tbl->slot_tbl_lock);
 572         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 573             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 574                 /* The state manager will wait until the slot table is empty */
 575                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 576                 spin_unlock(&tbl->slot_tbl_lock);
 577                 dprintk("%s session is draining\n", __func__);
 578                 return -EAGAIN;
 579         }
 580
 581         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 582             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 583                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 584                 spin_unlock(&tbl->slot_tbl_lock);
 585                 dprintk("%s enforce FIFO order\n", __func__);
 586                 return -EAGAIN;
 587         }
 588
 589         slotid = nfs4_find_slot(tbl);
 590         if (slotid == NFS4_NO_SLOT) {
 591                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 592                 spin_unlock(&tbl->slot_tbl_lock);
 593                 dprintk("<-- %s: no free slots\n", __func__);
 594                 return -EAGAIN;
 595         }
 596         spin_unlock(&tbl->slot_tbl_lock);
 597
 598         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 599         slot = tbl->slots + slotid;
 600         args->sa_session = session;
 601         args->sa_slotid = slotid;
 602
 603         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 604
 605         res->sr_session = session;
 606         res->sr_slot = slot;
 607         res->sr_renewal_time = jiffies;
 608         res->sr_status_flags = 0;
 609         /*
 610          * sr_status is only set in decode_sequence, and so will remain
 611          * set to 1 if an rpc level failure occurs.
 612          */
 613         res->sr_status = 1;
 614         return 0;
 615 }
 616 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 617
 618 int nfs4_setup_sequence(const struct nfs_server *server,
 619                         struct nfs4_sequence_args *args,
 620                         struct nfs4_sequence_res *res,
 621                         struct rpc_task *task)
 622 {
 623         struct nfs4_session *session = nfs4_get_session(server);
 624         int ret = 0;
 625
 626         if (session == NULL)
 627                 goto out;
 628
 629         dprintk("--> %s clp %p session %p sr_slot %td\n",
 630                 __func__, session->clp, session, res->sr_slot ?
 631                         res->sr_slot - session->fc_slot_table.slots : -1);
 632
 633         ret = nfs41_setup_sequence(session, args, res, task);
 634 out:
 635         dprintk("<-- %s status=%d\n", __func__, ret);
 636         return ret;
 637 }
 638
 639 struct nfs41_call_sync_data {
 640         const struct nfs_server *seq_server;
 641         struct nfs4_sequence_args *seq_args;
 642         struct nfs4_sequence_res *seq_res;
 643 };
 644
 645 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 646 {
 647         struct nfs41_call_sync_data *data = calldata;
 648
 649         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 650
 651         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 652                                 data->seq_res, task))
 653                 return;
 654         rpc_call_start(task);
 655 }
 656
 657 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 658 {
 659         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 660         nfs41_call_sync_prepare(task, calldata);
 661 }
 662
 663 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 664 {
 665         struct nfs41_call_sync_data *data = calldata;
 666
 667         nfs41_sequence_done(task, data->seq_res);
 668 }
 669
 670 struct rpc_call_ops nfs41_call_sync_ops = {
 671         .rpc_call_prepare = nfs41_call_sync_prepare,
 672         .rpc_call_done = nfs41_call_sync_done,
 673 };
 674
 675 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 676         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 677         .rpc_call_done = nfs41_call_sync_done,
 678 };
 679
 680 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 681                                    struct nfs_server *server,
 682                                    struct rpc_message *msg,
 683                                    struct nfs4_sequence_args *args,
 684                                    struct nfs4_sequence_res *res,
 685                                    int privileged)
 686 {
 687         int ret;
 688         struct rpc_task *task;
 689         struct nfs41_call_sync_data data = {
 690                 .seq_server = server,
 691                 .seq_args = args,
 692                 .seq_res = res,
 693         };
 694         struct rpc_task_setup task_setup = {
 695                 .rpc_client = clnt,
 696                 .rpc_message = msg,
 697                 .callback_ops = &nfs41_call_sync_ops,
 698                 .callback_data = &data
 699         };
 700
 701         if (privileged)
 702                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 703         task = rpc_run_task(&task_setup);
 704         if (IS_ERR(task))
 705                 ret = PTR_ERR(task);
 706         else {
 707                 ret = task->tk_status;
 708                 rpc_put_task(task);
 709         }
 710         return ret;
 711 }
 712
 713 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 714                             struct nfs_server *server,
 715                             struct rpc_message *msg,
 716                             struct nfs4_sequence_args *args,
 717                             struct nfs4_sequence_res *res,
 718                             int cache_reply)
 719 {
 720         nfs41_init_sequence(args, res, cache_reply);
 721         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 722 }
 723
 724 #else
 725 static inline
 726 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 727                 struct nfs4_sequence_res *res, int cache_reply)
 728 {
 729 }
 730
 731 static int nfs4_sequence_done(struct rpc_task *task,
 732                                struct nfs4_sequence_res *res)
 733 {
 734         return 1;
 735 }
 736 #endif /* CONFIG_NFS_V4_1 */
 737
 738 int _nfs4_call_sync(struct rpc_clnt *clnt,
 739                     struct nfs_server *server,
 740                     struct rpc_message *msg,
 741                     struct nfs4_sequence_args *args,
 742                     struct nfs4_sequence_res *res,
 743                     int cache_reply)
 744 {
 745         nfs41_init_sequence(args, res, cache_reply);
 746         return rpc_call_sync(clnt, msg, 0);
 747 }
 748
 749 static inline
 750 int nfs4_call_sync(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         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 758                                                 args, res, cache_reply);
 759 }
 760
 761 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 762 {
 763         struct nfs_inode *nfsi = NFS_I(dir);
 764
 765         spin_lock(&dir->i_lock);
 766         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 767         if (!cinfo->atomic || cinfo->before != dir->i_version)
 768                 nfs_force_lookup_revalidate(dir);
 769         dir->i_version = cinfo->after;
 770         spin_unlock(&dir->i_lock);
 771 }
 772
 773 struct nfs4_opendata {
 774         struct kref kref;
 775         struct nfs_openargs o_arg;
 776         struct nfs_openres o_res;
 777         struct nfs_open_confirmargs c_arg;
 778         struct nfs_open_confirmres c_res;
 779         struct nfs4_string owner_name;
 780         struct nfs4_string group_name;
 781         struct nfs_fattr f_attr;
 782         struct nfs_fattr dir_attr;
 783         struct dentry *dir;
 784         struct dentry *dentry;
 785         struct nfs4_state_owner *owner;
 786         struct nfs4_state *state;
 787         struct iattr attrs;
 788         unsigned long timestamp;
 789         unsigned int rpc_done : 1;
 790         int rpc_status;
 791         int cancelled;
 792 };
 793
 794
 795 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 796 {
 797         p->o_res.f_attr = &p->f_attr;
 798         p->o_res.dir_attr = &p->dir_attr;
 799         p->o_res.seqid = p->o_arg.seqid;
 800         p->c_res.seqid = p->c_arg.seqid;
 801         p->o_res.server = p->o_arg.server;
 802         nfs_fattr_init(&p->f_attr);
 803         nfs_fattr_init(&p->dir_attr);
 804         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 805 }
 806
 807 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 808                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 809                 const struct iattr *attrs,
 810                 gfp_t gfp_mask)
 811 {
 812         struct dentry *parent = dget_parent(dentry);
 813         struct inode *dir = parent->d_inode;
 814         struct nfs_server *server = NFS_SERVER(dir);
 815         struct nfs4_opendata *p;
 816
 817         p = kzalloc(sizeof(*p), gfp_mask);
 818         if (p == NULL)
 819                 goto err;
 820         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 821         if (p->o_arg.seqid == NULL)
 822                 goto err_free;
 823         nfs_sb_active(dentry->d_sb);
 824         p->dentry = dget(dentry);
 825         p->dir = parent;
 826         p->owner = sp;
 827         atomic_inc(&sp->so_count);
 828         p->o_arg.fh = NFS_FH(dir);
 829         p->o_arg.open_flags = flags;
 830         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 831         p->o_arg.clientid = server->nfs_client->cl_clientid;
 832         p->o_arg.id = sp->so_seqid.owner_id;
 833         p->o_arg.name = &dentry->d_name;
 834         p->o_arg.server = server;
 835         p->o_arg.bitmask = server->attr_bitmask;
 836         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 837         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 838         if (attrs != NULL && attrs->ia_valid != 0) {
 839                 __be32 verf[2];
 840
 841                 p->o_arg.u.attrs = &p->attrs;
 842                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 843
 844                 verf[0] = jiffies;
 845                 verf[1] = current->pid;
 846                 memcpy(p->o_arg.u.verifier.data, verf,
 847                                 sizeof(p->o_arg.u.verifier.data));
 848         }
 849         p->c_arg.fh = &p->o_res.fh;
 850         p->c_arg.stateid = &p->o_res.stateid;
 851         p->c_arg.seqid = p->o_arg.seqid;
 852         nfs4_init_opendata_res(p);
 853         kref_init(&p->kref);
 854         return p;
 855 err_free:
 856         kfree(p);
 857 err:
 858         dput(parent);
 859         return NULL;
 860 }
 861
 862 static void nfs4_opendata_free(struct kref *kref)
 863 {
 864         struct nfs4_opendata *p = container_of(kref,
 865                         struct nfs4_opendata, kref);
 866         struct super_block *sb = p->dentry->d_sb;
 867
 868         nfs_free_seqid(p->o_arg.seqid);
 869         if (p->state != NULL)
 870                 nfs4_put_open_state(p->state);
 871         nfs4_put_state_owner(p->owner);
 872         dput(p->dir);
 873         dput(p->dentry);
 874         nfs_sb_deactive(sb);
 875         nfs_fattr_free_names(&p->f_attr);
 876         kfree(p);
 877 }
 878
 879 static void nfs4_opendata_put(struct nfs4_opendata *p)
 880 {
 881         if (p != NULL)
 882                 kref_put(&p->kref, nfs4_opendata_free);
 883 }
 884
 885 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 886 {
 887         int ret;
 888
 889         ret = rpc_wait_for_completion_task(task);
 890         return ret;
 891 }
 892
 893 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 894 {
 895         int ret = 0;
 896
 897         if (open_mode & (O_EXCL|O_TRUNC))
 898                 goto out;
 899         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 900                 case FMODE_READ:
 901                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 902                                 && state->n_rdonly != 0;
 903                         break;
 904                 case FMODE_WRITE:
 905                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 906                                 && state->n_wronly != 0;
 907                         break;
 908                 case FMODE_READ|FMODE_WRITE:
 909                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 910                                 && state->n_rdwr != 0;
 911         }
 912 out:
 913         return ret;
 914 }
 915
 916 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 917 {
 918         if (delegation == NULL)
 919                 return 0;
 920         if ((delegation->type & fmode) != fmode)
 921                 return 0;
 922         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 923                 return 0;
 924         nfs_mark_delegation_referenced(delegation);
 925         return 1;
 926 }
 927
 928 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 929 {
 930         switch (fmode) {
 931                 case FMODE_WRITE:
 932                         state->n_wronly++;
 933                         break;
 934                 case FMODE_READ:
 935                         state->n_rdonly++;
 936                         break;
 937                 case FMODE_READ|FMODE_WRITE:
 938                         state->n_rdwr++;
 939         }
 940         nfs4_state_set_mode_locked(state, state->state | fmode);
 941 }
 942
 943 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 944 {
 945         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 946                 nfs4_stateid_copy(&state->stateid, stateid);
 947         nfs4_stateid_copy(&state->open_stateid, stateid);
 948         switch (fmode) {
 949                 case FMODE_READ:
 950                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 951                         break;
 952                 case FMODE_WRITE:
 953                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 954                         break;
 955                 case FMODE_READ|FMODE_WRITE:
 956                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 957         }
 958 }
 959
 960 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 961 {
 962         write_seqlock(&state->seqlock);
 963         nfs_set_open_stateid_locked(state, stateid, fmode);
 964         write_sequnlock(&state->seqlock);
 965 }
 966
 967 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 968 {
 969         /*
 970          * Protect the call to nfs4_state_set_mode_locked and
 971          * serialise the stateid update
 972          */
 973         write_seqlock(&state->seqlock);
 974         if (deleg_stateid != NULL) {
 975                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
 976                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 977         }
 978         if (open_stateid != NULL)
 979                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 980         write_sequnlock(&state->seqlock);
 981         spin_lock(&state->owner->so_lock);
 982         update_open_stateflags(state, fmode);
 983         spin_unlock(&state->owner->so_lock);
 984 }
 985
 986 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 987 {
 988         struct nfs_inode *nfsi = NFS_I(state->inode);
 989         struct nfs_delegation *deleg_cur;
 990         int ret = 0;
 991
 992         fmode &= (FMODE_READ|FMODE_WRITE);
 993
 994         rcu_read_lock();
 995         deleg_cur = rcu_dereference(nfsi->delegation);
 996         if (deleg_cur == NULL)
 997                 goto no_delegation;
 998
 999         spin_lock(&deleg_cur->lock);
1000         if (nfsi->delegation != deleg_cur ||
1001             (deleg_cur->type & fmode) != fmode)
1002                 goto no_delegation_unlock;
1003
1004         if (delegation == NULL)
1005                 delegation = &deleg_cur->stateid;
1006         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1007                 goto no_delegation_unlock;
1008
1009         nfs_mark_delegation_referenced(deleg_cur);
1010         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1011         ret = 1;
1012 no_delegation_unlock:
1013         spin_unlock(&deleg_cur->lock);
1014 no_delegation:
1015         rcu_read_unlock();
1016
1017         if (!ret && open_stateid != NULL) {
1018                 __update_open_stateid(state, open_stateid, NULL, fmode);
1019                 ret = 1;
1020         }
1021
1022         return ret;
1023 }
1024
1025
1026 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1027 {
1028         struct nfs_delegation *delegation;
1029
1030         rcu_read_lock();
1031         delegation = rcu_dereference(NFS_I(inode)->delegation);
1032         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1033                 rcu_read_unlock();
1034                 return;
1035         }
1036         rcu_read_unlock();
1037         nfs_inode_return_delegation(inode);
1038 }
1039
1040 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1041 {
1042         struct nfs4_state *state = opendata->state;
1043         struct nfs_inode *nfsi = NFS_I(state->inode);
1044         struct nfs_delegation *delegation;
1045         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1046         fmode_t fmode = opendata->o_arg.fmode;
1047         nfs4_stateid stateid;
1048         int ret = -EAGAIN;
1049
1050         for (;;) {
1051                 if (can_open_cached(state, fmode, open_mode)) {
1052                         spin_lock(&state->owner->so_lock);
1053                         if (can_open_cached(state, fmode, open_mode)) {
1054                                 update_open_stateflags(state, fmode);
1055                                 spin_unlock(&state->owner->so_lock);
1056                                 goto out_return_state;
1057                         }
1058                         spin_unlock(&state->owner->so_lock);
1059                 }
1060                 rcu_read_lock();
1061                 delegation = rcu_dereference(nfsi->delegation);
1062                 if (!can_open_delegated(delegation, fmode)) {
1063                         rcu_read_unlock();
1064                         break;
1065                 }
1066                 /* Save the delegation */
1067                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1068                 rcu_read_unlock();
1069                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1070                 if (ret != 0)
1071                         goto out;
1072                 ret = -EAGAIN;
1073
1074                 /* Try to update the stateid using the delegation */
1075                 if (update_open_stateid(state, NULL, &stateid, fmode))
1076                         goto out_return_state;
1077         }
1078 out:
1079         return ERR_PTR(ret);
1080 out_return_state:
1081         atomic_inc(&state->count);
1082         return state;
1083 }
1084
1085 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1086 {
1087         struct inode *inode;
1088         struct nfs4_state *state = NULL;
1089         struct nfs_delegation *delegation;
1090         int ret;
1091
1092         if (!data->rpc_done) {
1093                 state = nfs4_try_open_cached(data);
1094                 goto out;
1095         }
1096
1097         ret = -EAGAIN;
1098         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1099                 goto err;
1100         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1101         ret = PTR_ERR(inode);
1102         if (IS_ERR(inode))
1103                 goto err;
1104         ret = -ENOMEM;
1105         state = nfs4_get_open_state(inode, data->owner);
1106         if (state == NULL)
1107                 goto err_put_inode;
1108         if (data->o_res.delegation_type != 0) {
1109                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1110                 int delegation_flags = 0;
1111
1112                 rcu_read_lock();
1113                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1114                 if (delegation)
1115                         delegation_flags = delegation->flags;
1116                 rcu_read_unlock();
1117                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1118                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1119                                         "returning a delegation for "
1120                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1121                                         clp->cl_hostname);
1122                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1123                         nfs_inode_set_delegation(state->inode,
1124                                         data->owner->so_cred,
1125                                         &data->o_res);
1126                 else
1127                         nfs_inode_reclaim_delegation(state->inode,
1128                                         data->owner->so_cred,
1129                                         &data->o_res);
1130         }
1131
1132         update_open_stateid(state, &data->o_res.stateid, NULL,
1133                         data->o_arg.fmode);
1134         iput(inode);
1135 out:
1136         return state;
1137 err_put_inode:
1138         iput(inode);
1139 err:
1140         return ERR_PTR(ret);
1141 }
1142
1143 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1144 {
1145         struct nfs_inode *nfsi = NFS_I(state->inode);
1146         struct nfs_open_context *ctx;
1147
1148         spin_lock(&state->inode->i_lock);
1149         list_for_each_entry(ctx, &nfsi->open_files, list) {
1150                 if (ctx->state != state)
1151                         continue;
1152                 get_nfs_open_context(ctx);
1153                 spin_unlock(&state->inode->i_lock);
1154                 return ctx;
1155         }
1156         spin_unlock(&state->inode->i_lock);
1157         return ERR_PTR(-ENOENT);
1158 }
1159
1160 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1161 {
1162         struct nfs4_opendata *opendata;
1163
1164         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1165         if (opendata == NULL)
1166                 return ERR_PTR(-ENOMEM);
1167         opendata->state = state;
1168         atomic_inc(&state->count);
1169         return opendata;
1170 }
1171
1172 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1173 {
1174         struct nfs4_state *newstate;
1175         int ret;
1176
1177         opendata->o_arg.open_flags = 0;
1178         opendata->o_arg.fmode = fmode;
1179         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1180         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1181         nfs4_init_opendata_res(opendata);
1182         ret = _nfs4_recover_proc_open(opendata);
1183         if (ret != 0)
1184                 return ret;
1185         newstate = nfs4_opendata_to_nfs4_state(opendata);
1186         if (IS_ERR(newstate))
1187                 return PTR_ERR(newstate);
1188         nfs4_close_state(newstate, fmode);
1189         *res = newstate;
1190         return 0;
1191 }
1192
1193 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1194 {
1195         struct nfs4_state *newstate;
1196         int ret;
1197
1198         /* memory barrier prior to reading state->n_* */
1199         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1200         smp_rmb();
1201         if (state->n_rdwr != 0) {
1202                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1203                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1204                 if (ret != 0)
1205                         return ret;
1206                 if (newstate != state)
1207                         return -ESTALE;
1208         }
1209         if (state->n_wronly != 0) {
1210                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1211                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1212                 if (ret != 0)
1213                         return ret;
1214                 if (newstate != state)
1215                         return -ESTALE;
1216         }
1217         if (state->n_rdonly != 0) {
1218                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1219                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1220                 if (ret != 0)
1221                         return ret;
1222                 if (newstate != state)
1223                         return -ESTALE;
1224         }
1225         /*
1226          * We may have performed cached opens for all three recoveries.
1227          * Check if we need to update the current stateid.
1228          */
1229         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1230             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1231                 write_seqlock(&state->seqlock);
1232                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1233                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1234                 write_sequnlock(&state->seqlock);
1235         }
1236         return 0;
1237 }
1238
1239 /*
1240  * OPEN_RECLAIM:
1241  *      reclaim state on the server after a reboot.
1242  */
1243 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1244 {
1245         struct nfs_delegation *delegation;
1246         struct nfs4_opendata *opendata;
1247         fmode_t delegation_type = 0;
1248         int status;
1249
1250         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1251         if (IS_ERR(opendata))
1252                 return PTR_ERR(opendata);
1253         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1254         opendata->o_arg.fh = NFS_FH(state->inode);
1255         rcu_read_lock();
1256         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1257         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1258                 delegation_type = delegation->type;
1259         rcu_read_unlock();
1260         opendata->o_arg.u.delegation_type = delegation_type;
1261         status = nfs4_open_recover(opendata, state);
1262         nfs4_opendata_put(opendata);
1263         return status;
1264 }
1265
1266 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1267 {
1268         struct nfs_server *server = NFS_SERVER(state->inode);
1269         struct nfs4_exception exception = { };
1270         int err;
1271         do {
1272                 err = _nfs4_do_open_reclaim(ctx, state);
1273                 if (err != -NFS4ERR_DELAY)
1274                         break;
1275                 nfs4_handle_exception(server, err, &exception);
1276         } while (exception.retry);
1277         return err;
1278 }
1279
1280 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1281 {
1282         struct nfs_open_context *ctx;
1283         int ret;
1284
1285         ctx = nfs4_state_find_open_context(state);
1286         if (IS_ERR(ctx))
1287                 return PTR_ERR(ctx);
1288         ret = nfs4_do_open_reclaim(ctx, state);
1289         put_nfs_open_context(ctx);
1290         return ret;
1291 }
1292
1293 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1294 {
1295         struct nfs4_opendata *opendata;
1296         int ret;
1297
1298         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1299         if (IS_ERR(opendata))
1300                 return PTR_ERR(opendata);
1301         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1302         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1303         ret = nfs4_open_recover(opendata, state);
1304         nfs4_opendata_put(opendata);
1305         return ret;
1306 }
1307
1308 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1309 {
1310         struct nfs4_exception exception = { };
1311         struct nfs_server *server = NFS_SERVER(state->inode);
1312         int err;
1313         do {
1314                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1315                 switch (err) {
1316                         case 0:
1317                         case -ENOENT:
1318                         case -ESTALE:
1319                                 goto out;
1320                         case -NFS4ERR_BADSESSION:
1321                         case -NFS4ERR_BADSLOT:
1322                         case -NFS4ERR_BAD_HIGH_SLOT:
1323                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1324                         case -NFS4ERR_DEADSESSION:
1325                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1326                                 goto out;
1327                         case -NFS4ERR_STALE_CLIENTID:
1328                         case -NFS4ERR_STALE_STATEID:
1329                         case -NFS4ERR_EXPIRED:
1330                                 /* Don't recall a delegation if it was lost */
1331                                 nfs4_schedule_lease_recovery(server->nfs_client);
1332                                 goto out;
1333                         case -ERESTARTSYS:
1334                                 /*
1335                                  * The show must go on: exit, but mark the
1336                                  * stateid as needing recovery.
1337                                  */
1338                         case -NFS4ERR_DELEG_REVOKED:
1339                         case -NFS4ERR_ADMIN_REVOKED:
1340                         case -NFS4ERR_BAD_STATEID:
1341                                 nfs_inode_find_state_and_recover(state->inode,
1342                                                 stateid);
1343                                 nfs4_schedule_stateid_recovery(server, state);
1344                         case -EKEYEXPIRED:
1345                                 /*
1346                                  * User RPCSEC_GSS context has expired.
1347                                  * We cannot recover this stateid now, so
1348                                  * skip it and allow recovery thread to
1349                                  * proceed.
1350                                  */
1351                         case -ENOMEM:
1352                                 err = 0;
1353                                 goto out;
1354                 }
1355                 err = nfs4_handle_exception(server, err, &exception);
1356         } while (exception.retry);
1357 out:
1358         return err;
1359 }
1360
1361 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1362 {
1363         struct nfs4_opendata *data = calldata;
1364
1365         data->rpc_status = task->tk_status;
1366         if (data->rpc_status == 0) {
1367                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1368                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1369                 renew_lease(data->o_res.server, data->timestamp);
1370                 data->rpc_done = 1;
1371         }
1372 }
1373
1374 static void nfs4_open_confirm_release(void *calldata)
1375 {
1376         struct nfs4_opendata *data = calldata;
1377         struct nfs4_state *state = NULL;
1378
1379         /* If this request hasn't been cancelled, do nothing */
1380         if (data->cancelled == 0)
1381                 goto out_free;
1382         /* In case of error, no cleanup! */
1383         if (!data->rpc_done)
1384                 goto out_free;
1385         state = nfs4_opendata_to_nfs4_state(data);
1386         if (!IS_ERR(state))
1387                 nfs4_close_state(state, data->o_arg.fmode);
1388 out_free:
1389         nfs4_opendata_put(data);
1390 }
1391
1392 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1393         .rpc_call_done = nfs4_open_confirm_done,
1394         .rpc_release = nfs4_open_confirm_release,
1395 };
1396
1397 /*
1398  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1399  */
1400 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1401 {
1402         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1403         struct rpc_task *task;
1404         struct  rpc_message msg = {
1405                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1406                 .rpc_argp = &data->c_arg,
1407                 .rpc_resp = &data->c_res,
1408                 .rpc_cred = data->owner->so_cred,
1409         };
1410         struct rpc_task_setup task_setup_data = {
1411                 .rpc_client = server->client,
1412                 .rpc_message = &msg,
1413                 .callback_ops = &nfs4_open_confirm_ops,
1414                 .callback_data = data,
1415                 .workqueue = nfsiod_workqueue,
1416                 .flags = RPC_TASK_ASYNC,
1417         };
1418         int status;
1419
1420         kref_get(&data->kref);
1421         data->rpc_done = 0;
1422         data->rpc_status = 0;
1423         data->timestamp = jiffies;
1424         task = rpc_run_task(&task_setup_data);
1425         if (IS_ERR(task))
1426                 return PTR_ERR(task);
1427         status = nfs4_wait_for_completion_rpc_task(task);
1428         if (status != 0) {
1429                 data->cancelled = 1;
1430                 smp_wmb();
1431         } else
1432                 status = data->rpc_status;
1433         rpc_put_task(task);
1434         return status;
1435 }
1436
1437 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1438 {
1439         struct nfs4_opendata *data = calldata;
1440         struct nfs4_state_owner *sp = data->owner;
1441
1442         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1443                 return;
1444         /*
1445          * Check if we still need to send an OPEN call, or if we can use
1446          * a delegation instead.
1447          */
1448         if (data->state != NULL) {
1449                 struct nfs_delegation *delegation;
1450
1451                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1452                         goto out_no_action;
1453                 rcu_read_lock();
1454                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1455                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1456                     can_open_delegated(delegation, data->o_arg.fmode))
1457                         goto unlock_no_action;
1458                 rcu_read_unlock();
1459         }
1460         /* Update sequence id. */
1461         data->o_arg.id = sp->so_seqid.owner_id;
1462         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1463         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1464                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1465                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1466         }
1467         data->timestamp = jiffies;
1468         if (nfs4_setup_sequence(data->o_arg.server,
1469                                 &data->o_arg.seq_args,
1470                                 &data->o_res.seq_res, task))
1471                 return;
1472         rpc_call_start(task);
1473         return;
1474 unlock_no_action:
1475         rcu_read_unlock();
1476 out_no_action:
1477         task->tk_action = NULL;
1478
1479 }
1480
1481 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1482 {
1483         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1484         nfs4_open_prepare(task, calldata);
1485 }
1486
1487 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1488 {
1489         struct nfs4_opendata *data = calldata;
1490
1491         data->rpc_status = task->tk_status;
1492
1493         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1494                 return;
1495
1496         if (task->tk_status == 0) {
1497                 switch (data->o_res.f_attr->mode & S_IFMT) {
1498                         case S_IFREG:
1499                                 break;
1500                         case S_IFLNK:
1501                                 data->rpc_status = -ELOOP;
1502                                 break;
1503                         case S_IFDIR:
1504                                 data->rpc_status = -EISDIR;
1505                                 break;
1506                         default:
1507                                 data->rpc_status = -ENOTDIR;
1508                 }
1509                 renew_lease(data->o_res.server, data->timestamp);
1510                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1511                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1512         }
1513         data->rpc_done = 1;
1514 }
1515
1516 static void nfs4_open_release(void *calldata)
1517 {
1518         struct nfs4_opendata *data = calldata;
1519         struct nfs4_state *state = NULL;
1520
1521         /* If this request hasn't been cancelled, do nothing */
1522         if (data->cancelled == 0)
1523                 goto out_free;
1524         /* In case of error, no cleanup! */
1525         if (data->rpc_status != 0 || !data->rpc_done)
1526                 goto out_free;
1527         /* In case we need an open_confirm, no cleanup! */
1528         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1529                 goto out_free;
1530         state = nfs4_opendata_to_nfs4_state(data);
1531         if (!IS_ERR(state))
1532                 nfs4_close_state(state, data->o_arg.fmode);
1533 out_free:
1534         nfs4_opendata_put(data);
1535 }
1536
1537 static const struct rpc_call_ops nfs4_open_ops = {
1538         .rpc_call_prepare = nfs4_open_prepare,
1539         .rpc_call_done = nfs4_open_done,
1540         .rpc_release = nfs4_open_release,
1541 };
1542
1543 static const struct rpc_call_ops nfs4_recover_open_ops = {
1544         .rpc_call_prepare = nfs4_recover_open_prepare,
1545         .rpc_call_done = nfs4_open_done,
1546         .rpc_release = nfs4_open_release,
1547 };
1548
1549 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1550 {
1551         struct inode *dir = data->dir->d_inode;
1552         struct nfs_server *server = NFS_SERVER(dir);
1553         struct nfs_openargs *o_arg = &data->o_arg;
1554         struct nfs_openres *o_res = &data->o_res;
1555         struct rpc_task *task;
1556         struct rpc_message msg = {
1557                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1558                 .rpc_argp = o_arg,
1559                 .rpc_resp = o_res,
1560                 .rpc_cred = data->owner->so_cred,
1561         };
1562         struct rpc_task_setup task_setup_data = {
1563                 .rpc_client = server->client,
1564                 .rpc_message = &msg,
1565                 .callback_ops = &nfs4_open_ops,
1566                 .callback_data = data,
1567                 .workqueue = nfsiod_workqueue,
1568                 .flags = RPC_TASK_ASYNC,
1569         };
1570         int status;
1571
1572         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1573         kref_get(&data->kref);
1574         data->rpc_done = 0;
1575         data->rpc_status = 0;
1576         data->cancelled = 0;
1577         if (isrecover)
1578                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1579         task = rpc_run_task(&task_setup_data);
1580         if (IS_ERR(task))
1581                 return PTR_ERR(task);
1582         status = nfs4_wait_for_completion_rpc_task(task);
1583         if (status != 0) {
1584                 data->cancelled = 1;
1585                 smp_wmb();
1586         } else
1587                 status = data->rpc_status;
1588         rpc_put_task(task);
1589
1590         return status;
1591 }
1592
1593 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1594 {
1595         struct inode *dir = data->dir->d_inode;
1596         struct nfs_openres *o_res = &data->o_res;
1597         int status;
1598
1599         status = nfs4_run_open_task(data, 1);
1600         if (status != 0 || !data->rpc_done)
1601                 return status;
1602
1603         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1604
1605         nfs_refresh_inode(dir, o_res->dir_attr);
1606
1607         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1608                 status = _nfs4_proc_open_confirm(data);
1609                 if (status != 0)
1610                         return status;
1611         }
1612
1613         return status;
1614 }
1615
1616 /*
1617  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1618  */
1619 static int _nfs4_proc_open(struct nfs4_opendata *data)
1620 {
1621         struct inode *dir = data->dir->d_inode;
1622         struct nfs_server *server = NFS_SERVER(dir);
1623         struct nfs_openargs *o_arg = &data->o_arg;
1624         struct nfs_openres *o_res = &data->o_res;
1625         int status;
1626
1627         status = nfs4_run_open_task(data, 0);
1628         if (!data->rpc_done)
1629                 return status;
1630         if (status != 0) {
1631                 if (status == -NFS4ERR_BADNAME &&
1632                                 !(o_arg->open_flags & O_CREAT))
1633                         return -ENOENT;
1634                 return status;
1635         }
1636
1637         nfs_fattr_map_and_free_names(server, &data->f_attr);
1638
1639         if (o_arg->open_flags & O_CREAT) {
1640                 update_changeattr(dir, &o_res->cinfo);
1641                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1642         } else
1643                 nfs_refresh_inode(dir, o_res->dir_attr);
1644         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1645                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1646         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1647                 status = _nfs4_proc_open_confirm(data);
1648                 if (status != 0)
1649                         return status;
1650         }
1651         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1652                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1653         return 0;
1654 }
1655
1656 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1657 {
1658         unsigned int loop;
1659         int ret;
1660
1661         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1662                 ret = nfs4_wait_clnt_recover(clp);
1663                 if (ret != 0)
1664                         break;
1665                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1666                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1667                         break;
1668                 nfs4_schedule_state_manager(clp);
1669                 ret = -EIO;
1670         }
1671         return ret;
1672 }
1673
1674 static int nfs4_recover_expired_lease(struct nfs_server *server)
1675 {
1676         return nfs4_client_recover_expired_lease(server->nfs_client);
1677 }
1678
1679 /*
1680  * OPEN_EXPIRED:
1681  *      reclaim state on the server after a network partition.
1682  *      Assumes caller holds the appropriate lock
1683  */
1684 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1685 {
1686         struct nfs4_opendata *opendata;
1687         int ret;
1688
1689         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1690         if (IS_ERR(opendata))
1691                 return PTR_ERR(opendata);
1692         ret = nfs4_open_recover(opendata, state);
1693         if (ret == -ESTALE)
1694                 d_drop(ctx->dentry);
1695         nfs4_opendata_put(opendata);
1696         return ret;
1697 }
1698
1699 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1700 {
1701         struct nfs_server *server = NFS_SERVER(state->inode);
1702         struct nfs4_exception exception = { };
1703         int err;
1704
1705         do {
1706                 err = _nfs4_open_expired(ctx, state);
1707                 switch (err) {
1708                 default:
1709                         goto out;
1710                 case -NFS4ERR_GRACE:
1711                 case -NFS4ERR_DELAY:
1712                         nfs4_handle_exception(server, err, &exception);
1713                         err = 0;
1714                 }
1715         } while (exception.retry);
1716 out:
1717         return err;
1718 }
1719
1720 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1721 {
1722         struct nfs_open_context *ctx;
1723         int ret;
1724
1725         ctx = nfs4_state_find_open_context(state);
1726         if (IS_ERR(ctx))
1727                 return PTR_ERR(ctx);
1728         ret = nfs4_do_open_expired(ctx, state);
1729         put_nfs_open_context(ctx);
1730         return ret;
1731 }
1732
1733 #if defined(CONFIG_NFS_V4_1)
1734 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1735 {
1736         int status = NFS_OK;
1737         struct nfs_server *server = NFS_SERVER(state->inode);
1738
1739         if (state->flags & flags) {
1740                 status = nfs41_test_stateid(server, stateid);
1741                 if (status != NFS_OK) {
1742                         nfs41_free_stateid(server, stateid);
1743                         state->flags &= ~flags;
1744                 }
1745         }
1746         return status;
1747 }
1748
1749 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1750 {
1751         int deleg_status, open_status;
1752         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1753         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1754
1755         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1756         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1757
1758         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1759                 return NFS_OK;
1760         return nfs4_open_expired(sp, state);
1761 }
1762 #endif
1763
1764 /*
1765  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1766  * fields corresponding to attributes that were used to store the verifier.
1767  * Make sure we clobber those fields in the later setattr call
1768  */
1769 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1770 {
1771         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1772             !(sattr->ia_valid & ATTR_ATIME_SET))
1773                 sattr->ia_valid |= ATTR_ATIME;
1774
1775         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1776             !(sattr->ia_valid & ATTR_MTIME_SET))
1777                 sattr->ia_valid |= ATTR_MTIME;
1778 }
1779
1780 /*
1781  * Returns a referenced nfs4_state
1782  */
1783 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1784 {
1785         struct nfs4_state_owner  *sp;
1786         struct nfs4_state     *state = NULL;
1787         struct nfs_server       *server = NFS_SERVER(dir);
1788         struct nfs4_opendata *opendata;
1789         int status;
1790
1791         /* Protect against reboot recovery conflicts */
1792         status = -ENOMEM;
1793         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1794         if (sp == NULL) {
1795                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1796                 goto out_err;
1797         }
1798         status = nfs4_recover_expired_lease(server);
1799         if (status != 0)
1800                 goto err_put_state_owner;
1801         if (dentry->d_inode != NULL)
1802                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1803         status = -ENOMEM;
1804         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1805         if (opendata == NULL)
1806                 goto err_put_state_owner;
1807
1808         if (dentry->d_inode != NULL)
1809                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1810
1811         status = _nfs4_proc_open(opendata);
1812         if (status != 0)
1813                 goto err_opendata_put;
1814
1815         state = nfs4_opendata_to_nfs4_state(opendata);
1816         status = PTR_ERR(state);
1817         if (IS_ERR(state))
1818                 goto err_opendata_put;
1819         if (server->caps & NFS_CAP_POSIX_LOCK)
1820                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1821
1822         if (opendata->o_arg.open_flags & O_EXCL) {
1823                 nfs4_exclusive_attrset(opendata, sattr);
1824
1825                 nfs_fattr_init(opendata->o_res.f_attr);
1826                 status = nfs4_do_setattr(state->inode, cred,
1827                                 opendata->o_res.f_attr, sattr,
1828                                 state);
1829                 if (status == 0)
1830                         nfs_setattr_update_inode(state->inode, sattr);
1831                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1832         }
1833         nfs4_opendata_put(opendata);
1834         nfs4_put_state_owner(sp);
1835         *res = state;
1836         return 0;
1837 err_opendata_put:
1838         nfs4_opendata_put(opendata);
1839 err_put_state_owner:
1840         nfs4_put_state_owner(sp);
1841 out_err:
1842         *res = NULL;
1843         return status;
1844 }
1845
1846
1847 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1848 {
1849         struct nfs4_exception exception = { };
1850         struct nfs4_state *res;
1851         int status;
1852
1853         do {
1854                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1855                 if (status == 0)
1856                         break;
1857                 /* NOTE: BAD_SEQID means the server and client disagree about the
1858                  * book-keeping w.r.t. state-changing operations
1859                  * (OPEN/CLOSE/LOCK/LOCKU...)
1860                  * It is actually a sign of a bug on the client or on the server.
1861                  *
1862                  * If we receive a BAD_SEQID error in the particular case of
1863                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1864                  * have unhashed the old state_owner for us, and that we can
1865                  * therefore safely retry using a new one. We should still warn
1866                  * the user though...
1867                  */
1868                 if (status == -NFS4ERR_BAD_SEQID) {
1869                         printk(KERN_WARNING "NFS: v4 server %s "
1870                                         " returned a bad sequence-id error!\n",
1871                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1872                         exception.retry = 1;
1873                         continue;
1874                 }
1875                 /*
1876                  * BAD_STATEID on OPEN means that the server cancelled our
1877                  * state before it received the OPEN_CONFIRM.
1878                  * Recover by retrying the request as per the discussion
1879                  * on Page 181 of RFC3530.
1880                  */
1881                 if (status == -NFS4ERR_BAD_STATEID) {
1882                         exception.retry = 1;
1883                         continue;
1884                 }
1885                 if (status == -EAGAIN) {
1886                         /* We must have found a delegation */
1887                         exception.retry = 1;
1888                         continue;
1889                 }
1890                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1891                                         status, &exception));
1892         } while (exception.retry);
1893         return res;
1894 }
1895
1896 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1897                             struct nfs_fattr *fattr, struct iattr *sattr,
1898                             struct nfs4_state *state)
1899 {
1900         struct nfs_server *server = NFS_SERVER(inode);
1901         struct nfs_setattrargs  arg = {
1902                 .fh             = NFS_FH(inode),
1903                 .iap            = sattr,
1904                 .server         = server,
1905                 .bitmask = server->attr_bitmask,
1906         };
1907         struct nfs_setattrres  res = {
1908                 .fattr          = fattr,
1909                 .server         = server,
1910         };
1911         struct rpc_message msg = {
1912                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1913                 .rpc_argp       = &arg,
1914                 .rpc_resp       = &res,
1915                 .rpc_cred       = cred,
1916         };
1917         unsigned long timestamp = jiffies;
1918         int status;
1919
1920         nfs_fattr_init(fattr);
1921
1922         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1923                 /* Use that stateid */
1924         } else if (state != NULL) {
1925                 nfs4_select_rw_stateid(&arg.stateid, state, current->files, current->tgid);
1926         } else
1927                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1928
1929         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1930         if (status == 0 && state != NULL)
1931                 renew_lease(server, timestamp);
1932         return status;
1933 }
1934
1935 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1936                            struct nfs_fattr *fattr, struct iattr *sattr,
1937                            struct nfs4_state *state)
1938 {
1939         struct nfs_server *server = NFS_SERVER(inode);
1940         struct nfs4_exception exception = {
1941                 .state = state,
1942         };
1943         int err;
1944         do {
1945                 err = nfs4_handle_exception(server,
1946                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1947                                 &exception);
1948         } while (exception.retry);
1949         return err;
1950 }
1951
1952 struct nfs4_closedata {
1953         struct inode *inode;
1954         struct nfs4_state *state;
1955         struct nfs_closeargs arg;
1956         struct nfs_closeres res;
1957         struct nfs_fattr fattr;
1958         unsigned long timestamp;
1959         bool roc;
1960         u32 roc_barrier;
1961 };
1962
1963 static void nfs4_free_closedata(void *data)
1964 {
1965         struct nfs4_closedata *calldata = data;
1966         struct nfs4_state_owner *sp = calldata->state->owner;
1967         struct super_block *sb = calldata->state->inode->i_sb;
1968
1969         if (calldata->roc)
1970                 pnfs_roc_release(calldata->state->inode);
1971         nfs4_put_open_state(calldata->state);
1972         nfs_free_seqid(calldata->arg.seqid);
1973         nfs4_put_state_owner(sp);
1974         nfs_sb_deactive(sb);
1975         kfree(calldata);
1976 }
1977
1978 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1979                 fmode_t fmode)
1980 {
1981         spin_lock(&state->owner->so_lock);
1982         if (!(fmode & FMODE_READ))
1983                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1984         if (!(fmode & FMODE_WRITE))
1985                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1986         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1987         spin_unlock(&state->owner->so_lock);
1988 }
1989
1990 static void nfs4_close_done(struct rpc_task *task, void *data)
1991 {
1992         struct nfs4_closedata *calldata = data;
1993         struct nfs4_state *state = calldata->state;
1994         struct nfs_server *server = NFS_SERVER(calldata->inode);
1995
1996         dprintk("%s: begin!\n", __func__);
1997         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1998                 return;
1999         /* hmm. we are done with the inode, and in the process of freeing
2000          * the state_owner. we keep this around to process errors
2001          */
2002         switch (task->tk_status) {
2003                 case 0:
2004                         if (calldata->roc)
2005                                 pnfs_roc_set_barrier(state->inode,
2006                                                      calldata->roc_barrier);
2007                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2008                         renew_lease(server, calldata->timestamp);
2009                         nfs4_close_clear_stateid_flags(state,
2010                                         calldata->arg.fmode);
2011                         break;
2012                 case -NFS4ERR_STALE_STATEID:
2013                 case -NFS4ERR_OLD_STATEID:
2014                 case -NFS4ERR_BAD_STATEID:
2015                 case -NFS4ERR_EXPIRED:
2016                         if (calldata->arg.fmode == 0)
2017                                 break;
2018                 default:
2019                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2020                                 rpc_restart_call_prepare(task);
2021         }
2022         nfs_release_seqid(calldata->arg.seqid);
2023         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2024         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2025 }
2026
2027 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2028 {
2029         struct nfs4_closedata *calldata = data;
2030         struct nfs4_state *state = calldata->state;
2031         int call_close = 0;
2032
2033         dprintk("%s: begin!\n", __func__);
2034         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2035                 return;
2036
2037         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2038         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2039         spin_lock(&state->owner->so_lock);
2040         /* Calculate the change in open mode */
2041         if (state->n_rdwr == 0) {
2042                 if (state->n_rdonly == 0) {
2043                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2044                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2045                         calldata->arg.fmode &= ~FMODE_READ;
2046                 }
2047                 if (state->n_wronly == 0) {
2048                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2049                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2050                         calldata->arg.fmode &= ~FMODE_WRITE;
2051                 }
2052         }
2053         spin_unlock(&state->owner->so_lock);
2054
2055         if (!call_close) {
2056                 /* Note: exit _without_ calling nfs4_close_done */
2057                 task->tk_action = NULL;
2058                 goto out;
2059         }
2060
2061         if (calldata->arg.fmode == 0) {
2062                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2063                 if (calldata->roc &&
2064                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2065                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2066                                      task, NULL);
2067                         goto out;
2068                 }
2069         }
2070
2071         nfs_fattr_init(calldata->res.fattr);
2072         calldata->timestamp = jiffies;
2073         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2074                                 &calldata->arg.seq_args,
2075                                 &calldata->res.seq_res,
2076                                 task))
2077                 goto out;
2078         rpc_call_start(task);
2079 out:
2080         dprintk("%s: done!\n", __func__);
2081 }
2082
2083 static const struct rpc_call_ops nfs4_close_ops = {
2084         .rpc_call_prepare = nfs4_close_prepare,
2085         .rpc_call_done = nfs4_close_done,
2086         .rpc_release = nfs4_free_closedata,
2087 };
2088
2089 /*
2090  * It is possible for data to be read/written from a mem-mapped file
2091  * after the sys_close call (which hits the vfs layer as a flush).
2092  * This means that we can't safely call nfsv4 close on a file until
2093  * the inode is cleared. This in turn means that we are not good
2094  * NFSv4 citizens - we do not indicate to the server to update the file's
2095  * share state even when we are done with one of the three share
2096  * stateid's in the inode.
2097  *
2098  * NOTE: Caller must be holding the sp->so_owner semaphore!
2099  */
2100 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2101 {
2102         struct nfs_server *server = NFS_SERVER(state->inode);
2103         struct nfs4_closedata *calldata;
2104         struct nfs4_state_owner *sp = state->owner;
2105         struct rpc_task *task;
2106         struct rpc_message msg = {
2107                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2108                 .rpc_cred = state->owner->so_cred,
2109         };
2110         struct rpc_task_setup task_setup_data = {
2111                 .rpc_client = server->client,
2112                 .rpc_message = &msg,
2113                 .callback_ops = &nfs4_close_ops,
2114                 .workqueue = nfsiod_workqueue,
2115                 .flags = RPC_TASK_ASYNC,
2116         };
2117         int status = -ENOMEM;
2118
2119         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2120         if (calldata == NULL)
2121                 goto out;
2122         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2123         calldata->inode = state->inode;
2124         calldata->state = state;
2125         calldata->arg.fh = NFS_FH(state->inode);
2126         calldata->arg.stateid = &state->open_stateid;
2127         /* Serialization for the sequence id */
2128         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2129         if (calldata->arg.seqid == NULL)
2130                 goto out_free_calldata;
2131         calldata->arg.fmode = 0;
2132         calldata->arg.bitmask = server->cache_consistency_bitmask;
2133         calldata->res.fattr = &calldata->fattr;
2134         calldata->res.seqid = calldata->arg.seqid;
2135         calldata->res.server = server;
2136         calldata->roc = roc;
2137         nfs_sb_active(calldata->inode->i_sb);
2138
2139         msg.rpc_argp = &calldata->arg;
2140         msg.rpc_resp = &calldata->res;
2141         task_setup_data.callback_data = calldata;
2142         task = rpc_run_task(&task_setup_data);
2143         if (IS_ERR(task))
2144                 return PTR_ERR(task);
2145         status = 0;
2146         if (wait)
2147                 status = rpc_wait_for_completion_task(task);
2148         rpc_put_task(task);
2149         return status;
2150 out_free_calldata:
2151         kfree(calldata);
2152 out:
2153         if (roc)
2154                 pnfs_roc_release(state->inode);
2155         nfs4_put_open_state(state);
2156         nfs4_put_state_owner(sp);
2157         return status;
2158 }
2159
2160 static struct inode *
2161 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2162 {
2163         struct nfs4_state *state;
2164
2165         /* Protect against concurrent sillydeletes */
2166         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2167         if (IS_ERR(state))
2168                 return ERR_CAST(state);
2169         ctx->state = state;
2170         return igrab(state->inode);
2171 }
2172
2173 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2174 {
2175         if (ctx->state == NULL)
2176                 return;
2177         if (is_sync)
2178                 nfs4_close_sync(ctx->state, ctx->mode);
2179         else
2180                 nfs4_close_state(ctx->state, ctx->mode);
2181 }
2182
2183 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2184 {
2185         struct nfs4_server_caps_arg args = {
2186                 .fhandle = fhandle,
2187         };
2188         struct nfs4_server_caps_res res = {};
2189         struct rpc_message msg = {
2190                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2191                 .rpc_argp = &args,
2192                 .rpc_resp = &res,
2193         };
2194         int status;
2195
2196         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2197         if (status == 0) {
2198                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2199                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2200                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2201                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2202                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2203                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2204                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2205                         server->caps |= NFS_CAP_ACLS;
2206                 if (res.has_links != 0)
2207                         server->caps |= NFS_CAP_HARDLINKS;
2208                 if (res.has_symlinks != 0)
2209                         server->caps |= NFS_CAP_SYMLINKS;
2210                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2211                         server->caps |= NFS_CAP_FILEID;
2212                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2213                         server->caps |= NFS_CAP_MODE;
2214                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2215                         server->caps |= NFS_CAP_NLINK;
2216                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2217                         server->caps |= NFS_CAP_OWNER;
2218                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2219                         server->caps |= NFS_CAP_OWNER_GROUP;
2220                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2221                         server->caps |= NFS_CAP_ATIME;
2222                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2223                         server->caps |= NFS_CAP_CTIME;
2224                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2225                         server->caps |= NFS_CAP_MTIME;
2226
2227                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2228                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2229                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2230                 server->acl_bitmask = res.acl_bitmask;
2231                 server->fh_expire_type = res.fh_expire_type;
2232         }
2233
2234         return status;
2235 }
2236
2237 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2238 {
2239         struct nfs4_exception exception = { };
2240         int err;
2241         do {
2242                 err = nfs4_handle_exception(server,
2243                                 _nfs4_server_capabilities(server, fhandle),
2244                                 &exception);
2245         } while (exception.retry);
2246         return err;
2247 }
2248
2249 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2250                 struct nfs_fsinfo *info)
2251 {
2252         struct nfs4_lookup_root_arg args = {
2253                 .bitmask = nfs4_fattr_bitmap,
2254         };
2255         struct nfs4_lookup_res res = {
2256                 .server = server,
2257                 .fattr = info->fattr,
2258                 .fh = fhandle,
2259         };
2260         struct rpc_message msg = {
2261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2262                 .rpc_argp = &args,
2263                 .rpc_resp = &res,
2264         };
2265
2266         nfs_fattr_init(info->fattr);
2267         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2268 }
2269
2270 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2271                 struct nfs_fsinfo *info)
2272 {
2273         struct nfs4_exception exception = { };
2274         int err;
2275         do {
2276                 err = _nfs4_lookup_root(server, fhandle, info);
2277                 switch (err) {
2278                 case 0:
2279                 case -NFS4ERR_WRONGSEC:
2280                         break;
2281                 default:
2282                         err = nfs4_handle_exception(server, err, &exception);
2283                 }
2284         } while (exception.retry);
2285         return err;
2286 }
2287
2288 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2289                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2290 {
2291         struct rpc_auth *auth;
2292         int ret;
2293
2294         auth = rpcauth_create(flavor, server->client);
2295         if (!auth) {
2296                 ret = -EIO;
2297                 goto out;
2298         }
2299         ret = nfs4_lookup_root(server, fhandle, info);
2300 out:
2301         return ret;
2302 }
2303
2304 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2305                               struct nfs_fsinfo *info)
2306 {
2307         int i, len, status = 0;
2308         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2309
2310         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2311         flav_array[len] = RPC_AUTH_NULL;
2312         len += 1;
2313
2314         for (i = 0; i < len; i++) {
2315                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2316                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2317                         continue;
2318                 break;
2319         }
2320         /*
2321          * -EACCESS could mean that the user doesn't have correct permissions
2322          * to access the mount.  It could also mean that we tried to mount
2323          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2324          * existing mount programs don't handle -EACCES very well so it should
2325          * be mapped to -EPERM instead.
2326          */
2327         if (status == -EACCES)
2328                 status = -EPERM;
2329         return status;
2330 }
2331
2332 /*
2333  * get the file handle for the "/" directory on the server
2334  */
2335 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2336                               struct nfs_fsinfo *info)
2337 {
2338         int minor_version = server->nfs_client->cl_minorversion;
2339         int status = nfs4_lookup_root(server, fhandle, info);
2340         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2341                 /*
2342                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2343                  * by nfs4_map_errors() as this function exits.
2344                  */
2345                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2346         if (status == 0)
2347                 status = nfs4_server_capabilities(server, fhandle);
2348         if (status == 0)
2349                 status = nfs4_do_fsinfo(server, fhandle, info);
2350         return nfs4_map_errors(status);
2351 }
2352
2353 /*
2354  * Get locations and (maybe) other attributes of a referral.
2355  * Note that we'll actually follow the referral later when
2356  * we detect fsid mismatch in inode revalidation
2357  */
2358 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2359                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2360 {
2361         int status = -ENOMEM;
2362         struct page *page = NULL;
2363         struct nfs4_fs_locations *locations = NULL;
2364
2365         page = alloc_page(GFP_KERNEL);
2366         if (page == NULL)
2367                 goto out;
2368         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2369         if (locations == NULL)
2370                 goto out;
2371
2372         status = nfs4_proc_fs_locations(dir, name, locations, page);
2373         if (status != 0)
2374                 goto out;
2375         /* Make sure server returned a different fsid for the referral */
2376         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2377                 dprintk("%s: server did not return a different fsid for"
2378                         " a referral at %s\n", __func__, name->name);
2379                 status = -EIO;
2380                 goto out;
2381         }
2382         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2383         nfs_fixup_referral_attributes(&locations->fattr);
2384
2385         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2386         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2387         memset(fhandle, 0, sizeof(struct nfs_fh));
2388 out:
2389         if (page)
2390                 __free_page(page);
2391         kfree(locations);
2392         return status;
2393 }
2394
2395 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2396 {
2397         struct nfs4_getattr_arg args = {
2398                 .fh = fhandle,
2399                 .bitmask = server->attr_bitmask,
2400         };
2401         struct nfs4_getattr_res res = {
2402                 .fattr = fattr,
2403                 .server = server,
2404         };
2405         struct rpc_message msg = {
2406                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2407                 .rpc_argp = &args,
2408                 .rpc_resp = &res,
2409         };
2410
2411         nfs_fattr_init(fattr);
2412         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2413 }
2414
2415 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2416 {
2417         struct nfs4_exception exception = { };
2418         int err;
2419         do {
2420                 err = nfs4_handle_exception(server,
2421                                 _nfs4_proc_getattr(server, fhandle, fattr),
2422                                 &exception);
2423         } while (exception.retry);
2424         return err;
2425 }
2426
2427 /*
2428  * The file is not closed if it is opened due to the a request to change
2429  * the size of the file. The open call will not be needed once the
2430  * VFS layer lookup-intents are implemented.
2431  *
2432  * Close is called when the inode is destroyed.
2433  * If we haven't opened the file for O_WRONLY, we
2434  * need to in the size_change case to obtain a stateid.
2435  *
2436  * Got race?
2437  * Because OPEN is always done by name in nfsv4, it is
2438  * possible that we opened a different file by the same
2439  * name.  We can recognize this race condition, but we
2440  * can't do anything about it besides returning an error.
2441  *
2442  * This will be fixed with VFS changes (lookup-intent).
2443  */
2444 static int
2445 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2446                   struct iattr *sattr)
2447 {
2448         struct inode *inode = dentry->d_inode;
2449         struct rpc_cred *cred = NULL;
2450         struct nfs4_state *state = NULL;
2451         int status;
2452
2453         if (pnfs_ld_layoutret_on_setattr(inode))
2454                 pnfs_return_layout(inode);
2455
2456         nfs_fattr_init(fattr);
2457
2458         /* Search for an existing open(O_WRITE) file */
2459         if (sattr->ia_valid & ATTR_FILE) {
2460                 struct nfs_open_context *ctx;
2461
2462                 ctx = nfs_file_open_context(sattr->ia_file);
2463                 if (ctx) {
2464                         cred = ctx->cred;
2465                         state = ctx->state;
2466                 }
2467         }
2468
2469         /* Deal with open(O_TRUNC) */
2470         if (sattr->ia_valid & ATTR_OPEN)
2471                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2472
2473         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2474         if (status == 0)
2475                 nfs_setattr_update_inode(inode, sattr);
2476         return status;
2477 }
2478
2479 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2480                 const struct qstr *name, struct nfs_fh *fhandle,
2481                 struct nfs_fattr *fattr)
2482 {
2483         struct nfs_server *server = NFS_SERVER(dir);
2484         int                    status;
2485         struct nfs4_lookup_arg args = {
2486                 .bitmask = server->attr_bitmask,
2487                 .dir_fh = NFS_FH(dir),
2488                 .name = name,
2489         };
2490         struct nfs4_lookup_res res = {
2491                 .server = server,
2492                 .fattr = fattr,
2493                 .fh = fhandle,
2494         };
2495         struct rpc_message msg = {
2496                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2497                 .rpc_argp = &args,
2498                 .rpc_resp = &res,
2499         };
2500
2501         nfs_fattr_init(fattr);
2502
2503         dprintk("NFS call  lookup %s\n", name->name);
2504         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2505         dprintk("NFS reply lookup: %d\n", status);
2506         return status;
2507 }
2508
2509 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2510 {
2511         memset(fh, 0, sizeof(struct nfs_fh));
2512         fattr->fsid.major = 1;
2513         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2514                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2515         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2516         fattr->nlink = 2;
2517 }
2518
2519 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2520                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2521 {
2522         struct nfs4_exception exception = { };
2523         int err;
2524         do {
2525                 int status;
2526
2527                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2528                 switch (status) {
2529                 case -NFS4ERR_BADNAME:
2530                         return -ENOENT;
2531                 case -NFS4ERR_MOVED:
2532                         return nfs4_get_referral(dir, name, fattr, fhandle);
2533                 case -NFS4ERR_WRONGSEC:
2534                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2535                 }
2536                 err = nfs4_handle_exception(NFS_SERVER(dir),
2537                                 status, &exception);
2538         } while (exception.retry);
2539         return err;
2540 }
2541
2542 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2543 {
2544         struct nfs_server *server = NFS_SERVER(inode);
2545         struct nfs4_accessargs args = {
2546                 .fh = NFS_FH(inode),
2547                 .bitmask = server->cache_consistency_bitmask,
2548         };
2549         struct nfs4_accessres res = {
2550                 .server = server,
2551         };
2552         struct rpc_message msg = {
2553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2554                 .rpc_argp = &args,
2555                 .rpc_resp = &res,
2556                 .rpc_cred = entry->cred,
2557         };
2558         int mode = entry->mask;
2559         int status;
2560
2561         /*
2562          * Determine which access bits we want to ask for...
2563          */
2564         if (mode & MAY_READ)
2565                 args.access |= NFS4_ACCESS_READ;
2566         if (S_ISDIR(inode->i_mode)) {
2567                 if (mode & MAY_WRITE)
2568                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2569                 if (mode & MAY_EXEC)
2570                         args.access |= NFS4_ACCESS_LOOKUP;
2571         } else {
2572                 if (mode & MAY_WRITE)
2573                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2574                 if (mode & MAY_EXEC)
2575                         args.access |= NFS4_ACCESS_EXECUTE;
2576         }
2577
2578         res.fattr = nfs_alloc_fattr();
2579         if (res.fattr == NULL)
2580                 return -ENOMEM;
2581
2582         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2583         if (!status) {
2584                 entry->mask = 0;
2585                 if (res.access & NFS4_ACCESS_READ)
2586                         entry->mask |= MAY_READ;
2587                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2588                         entry->mask |= MAY_WRITE;
2589                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2590                         entry->mask |= MAY_EXEC;
2591                 nfs_refresh_inode(inode, res.fattr);
2592         }
2593         nfs_free_fattr(res.fattr);
2594         return status;
2595 }
2596
2597 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2598 {
2599         struct nfs4_exception exception = { };
2600         int err;
2601         do {
2602                 err = nfs4_handle_exception(NFS_SERVER(inode),
2603                                 _nfs4_proc_access(inode, entry),
2604                                 &exception);
2605         } while (exception.retry);
2606         return err;
2607 }
2608
2609 /*
2610  * TODO: For the time being, we don't try to get any attributes
2611  * along with any of the zero-copy operations READ, READDIR,
2612  * READLINK, WRITE.
2613  *
2614  * In the case of the first three, we want to put the GETATTR
2615  * after the read-type operation -- this is because it is hard
2616  * to predict the length of a GETATTR response in v4, and thus
2617  * align the READ data correctly.  This means that the GETATTR
2618  * may end up partially falling into the page cache, and we should
2619  * shift it into the 'tail' of the xdr_buf before processing.
2620  * To do this efficiently, we need to know the total length
2621  * of data received, which doesn't seem to be available outside
2622  * of the RPC layer.
2623  *
2624  * In the case of WRITE, we also want to put the GETATTR after
2625  * the operation -- in this case because we want to make sure
2626  * we get the post-operation mtime and size.  This means that
2627  * we can't use xdr_encode_pages() as written: we need a variant
2628  * of it which would leave room in the 'tail' iovec.
2629  *
2630  * Both of these changes to the XDR layer would in fact be quite
2631  * minor, but I decided to leave them for a subsequent patch.
2632  */
2633 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2634                 unsigned int pgbase, unsigned int pglen)
2635 {
2636         struct nfs4_readlink args = {
2637                 .fh       = NFS_FH(inode),
2638                 .pgbase   = pgbase,
2639                 .pglen    = pglen,
2640                 .pages    = &page,
2641         };
2642         struct nfs4_readlink_res res;
2643         struct rpc_message msg = {
2644                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2645                 .rpc_argp = &args,
2646                 .rpc_resp = &res,
2647         };
2648
2649         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2650 }
2651
2652 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2653                 unsigned int pgbase, unsigned int pglen)
2654 {
2655         struct nfs4_exception exception = { };
2656         int err;
2657         do {
2658                 err = nfs4_handle_exception(NFS_SERVER(inode),
2659                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2660                                 &exception);
2661         } while (exception.retry);
2662         return err;
2663 }
2664
2665 /*
2666  * Got race?
2667  * We will need to arrange for the VFS layer to provide an atomic open.
2668  * Until then, this create/open method is prone to inefficiency and race
2669  * conditions due to the lookup, create, and open VFS calls from sys_open()
2670  * placed on the wire.
2671  *
2672  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2673  * The file will be opened again in the subsequent VFS open call
2674  * (nfs4_proc_file_open).
2675  *
2676  * The open for read will just hang around to be used by any process that
2677  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2678  */
2679
2680 static int
2681 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2682                  int flags, struct nfs_open_context *ctx)
2683 {
2684         struct dentry *de = dentry;
2685         struct nfs4_state *state;
2686         struct rpc_cred *cred = NULL;
2687         fmode_t fmode = 0;
2688         int status = 0;
2689
2690         if (ctx != NULL) {
2691                 cred = ctx->cred;
2692                 de = ctx->dentry;
2693                 fmode = ctx->mode;
2694         }
2695         sattr->ia_mode &= ~current_umask();
2696         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2697         d_drop(dentry);
2698         if (IS_ERR(state)) {
2699                 status = PTR_ERR(state);
2700                 goto out;
2701         }
2702         d_add(dentry, igrab(state->inode));
2703         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2704         if (ctx != NULL)
2705                 ctx->state = state;
2706         else
2707                 nfs4_close_sync(state, fmode);
2708 out:
2709         return status;
2710 }
2711
2712 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2713 {
2714         struct nfs_server *server = NFS_SERVER(dir);
2715         struct nfs_removeargs args = {
2716                 .fh = NFS_FH(dir),
2717                 .name.len = name->len,
2718                 .name.name = name->name,
2719                 .bitmask = server->attr_bitmask,
2720         };
2721         struct nfs_removeres res = {
2722                 .server = server,
2723         };
2724         struct rpc_message msg = {
2725                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2726                 .rpc_argp = &args,
2727                 .rpc_resp = &res,
2728         };
2729         int status = -ENOMEM;
2730
2731         res.dir_attr = nfs_alloc_fattr();
2732         if (res.dir_attr == NULL)
2733                 goto out;
2734
2735         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2736         if (status == 0) {
2737                 update_changeattr(dir, &res.cinfo);
2738                 nfs_post_op_update_inode(dir, res.dir_attr);
2739         }
2740         nfs_free_fattr(res.dir_attr);
2741 out:
2742         return status;
2743 }
2744
2745 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2746 {
2747         struct nfs4_exception exception = { };
2748         int err;
2749         do {
2750                 err = nfs4_handle_exception(NFS_SERVER(dir),
2751                                 _nfs4_proc_remove(dir, name),
2752                                 &exception);
2753         } while (exception.retry);
2754         return err;
2755 }
2756
2757 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2758 {
2759         struct nfs_server *server = NFS_SERVER(dir);
2760         struct nfs_removeargs *args = msg->rpc_argp;
2761         struct nfs_removeres *res = msg->rpc_resp;
2762
2763         args->bitmask = server->cache_consistency_bitmask;
2764         res->server = server;
2765         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2766         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2767 }
2768
2769 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2770 {
2771         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2772
2773         if (!nfs4_sequence_done(task, &res->seq_res))
2774                 return 0;
2775         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2776                 return 0;
2777         update_changeattr(dir, &res->cinfo);
2778         nfs_post_op_update_inode(dir, res->dir_attr);
2779         return 1;
2780 }
2781
2782 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2783 {
2784         struct nfs_server *server = NFS_SERVER(dir);
2785         struct nfs_renameargs *arg = msg->rpc_argp;
2786         struct nfs_renameres *res = msg->rpc_resp;
2787
2788         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2789         arg->bitmask = server->attr_bitmask;
2790         res->server = server;
2791         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2792 }
2793
2794 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2795                                  struct inode *new_dir)
2796 {
2797         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2798
2799         if (!nfs4_sequence_done(task, &res->seq_res))
2800                 return 0;
2801         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2802                 return 0;
2803
2804         update_changeattr(old_dir, &res->old_cinfo);
2805         nfs_post_op_update_inode(old_dir, res->old_fattr);
2806         update_changeattr(new_dir, &res->new_cinfo);
2807         nfs_post_op_update_inode(new_dir, res->new_fattr);
2808         return 1;
2809 }
2810
2811 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2812                 struct inode *new_dir, struct qstr *new_name)
2813 {
2814         struct nfs_server *server = NFS_SERVER(old_dir);
2815         struct nfs_renameargs arg = {
2816                 .old_dir = NFS_FH(old_dir),
2817                 .new_dir = NFS_FH(new_dir),
2818                 .old_name = old_name,
2819                 .new_name = new_name,
2820                 .bitmask = server->attr_bitmask,
2821         };
2822         struct nfs_renameres res = {
2823                 .server = server,
2824         };
2825         struct rpc_message msg = {
2826                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2827                 .rpc_argp = &arg,
2828                 .rpc_resp = &res,
2829         };
2830         int status = -ENOMEM;
2831
2832         res.old_fattr = nfs_alloc_fattr();
2833         res.new_fattr = nfs_alloc_fattr();
2834         if (res.old_fattr == NULL || res.new_fattr == NULL)
2835                 goto out;
2836
2837         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2838         if (!status) {
2839                 update_changeattr(old_dir, &res.old_cinfo);
2840                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2841                 update_changeattr(new_dir, &res.new_cinfo);
2842                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2843         }
2844 out:
2845         nfs_free_fattr(res.new_fattr);
2846         nfs_free_fattr(res.old_fattr);
2847         return status;
2848 }
2849
2850 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2851                 struct inode *new_dir, struct qstr *new_name)
2852 {
2853         struct nfs4_exception exception = { };
2854         int err;
2855         do {
2856                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2857                                 _nfs4_proc_rename(old_dir, old_name,
2858                                         new_dir, new_name),
2859                                 &exception);
2860         } while (exception.retry);
2861         return err;
2862 }
2863
2864 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2865 {
2866         struct nfs_server *server = NFS_SERVER(inode);
2867         struct nfs4_link_arg arg = {
2868                 .fh     = NFS_FH(inode),
2869                 .dir_fh = NFS_FH(dir),
2870                 .name   = name,
2871                 .bitmask = server->attr_bitmask,
2872         };
2873         struct nfs4_link_res res = {
2874                 .server = server,
2875         };
2876         struct rpc_message msg = {
2877                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2878                 .rpc_argp = &arg,
2879                 .rpc_resp = &res,
2880         };
2881         int status = -ENOMEM;
2882
2883         res.fattr = nfs_alloc_fattr();
2884         res.dir_attr = nfs_alloc_fattr();
2885         if (res.fattr == NULL || res.dir_attr == NULL)
2886                 goto out;
2887
2888         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2889         if (!status) {
2890                 update_changeattr(dir, &res.cinfo);
2891                 nfs_post_op_update_inode(dir, res.dir_attr);
2892                 nfs_post_op_update_inode(inode, res.fattr);
2893         }
2894 out:
2895         nfs_free_fattr(res.dir_attr);
2896         nfs_free_fattr(res.fattr);
2897         return status;
2898 }
2899
2900 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2901 {
2902         struct nfs4_exception exception = { };
2903         int err;
2904         do {
2905                 err = nfs4_handle_exception(NFS_SERVER(inode),
2906                                 _nfs4_proc_link(inode, dir, name),
2907                                 &exception);
2908         } while (exception.retry);
2909         return err;
2910 }
2911
2912 struct nfs4_createdata {
2913         struct rpc_message msg;
2914         struct nfs4_create_arg arg;
2915         struct nfs4_create_res res;
2916         struct nfs_fh fh;
2917         struct nfs_fattr fattr;
2918         struct nfs_fattr dir_fattr;
2919 };
2920
2921 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2922                 struct qstr *name, struct iattr *sattr, u32 ftype)
2923 {
2924         struct nfs4_createdata *data;
2925
2926         data = kzalloc(sizeof(*data), GFP_KERNEL);
2927         if (data != NULL) {
2928                 struct nfs_server *server = NFS_SERVER(dir);
2929
2930                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2931                 data->msg.rpc_argp = &data->arg;
2932                 data->msg.rpc_resp = &data->res;
2933                 data->arg.dir_fh = NFS_FH(dir);
2934                 data->arg.server = server;
2935                 data->arg.name = name;
2936                 data->arg.attrs = sattr;
2937                 data->arg.ftype = ftype;
2938                 data->arg.bitmask = server->attr_bitmask;
2939                 data->res.server = server;
2940                 data->res.fh = &data->fh;
2941                 data->res.fattr = &data->fattr;
2942                 data->res.dir_fattr = &data->dir_fattr;
2943                 nfs_fattr_init(data->res.fattr);
2944                 nfs_fattr_init(data->res.dir_fattr);
2945         }
2946         return data;
2947 }
2948
2949 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2950 {
2951         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2952                                     &data->arg.seq_args, &data->res.seq_res, 1);
2953         if (status == 0) {
2954                 update_changeattr(dir, &data->res.dir_cinfo);
2955                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2956                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2957         }
2958         return status;
2959 }
2960
2961 static void nfs4_free_createdata(struct nfs4_createdata *data)
2962 {
2963         kfree(data);
2964 }
2965
2966 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2967                 struct page *page, unsigned int len, struct iattr *sattr)
2968 {
2969         struct nfs4_createdata *data;
2970         int status = -ENAMETOOLONG;
2971
2972         if (len > NFS4_MAXPATHLEN)
2973                 goto out;
2974
2975         status = -ENOMEM;
2976         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2977         if (data == NULL)
2978                 goto out;
2979
2980         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2981         data->arg.u.symlink.pages = &page;
2982         data->arg.u.symlink.len = len;
2983
2984         status = nfs4_do_create(dir, dentry, data);
2985
2986         nfs4_free_createdata(data);
2987 out:
2988         return status;
2989 }
2990
2991 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2992                 struct page *page, unsigned int len, struct iattr *sattr)
2993 {
2994         struct nfs4_exception exception = { };
2995         int err;
2996         do {
2997                 err = nfs4_handle_exception(NFS_SERVER(dir),
2998                                 _nfs4_proc_symlink(dir, dentry, page,
2999                                                         len, sattr),
3000                                 &exception);
3001         } while (exception.retry);
3002         return err;
3003 }
3004
3005 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3006                 struct iattr *sattr)
3007 {
3008         struct nfs4_createdata *data;
3009         int status = -ENOMEM;
3010
3011         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3012         if (data == NULL)
3013                 goto out;
3014
3015         status = nfs4_do_create(dir, dentry, data);
3016
3017         nfs4_free_createdata(data);
3018 out:
3019         return status;
3020 }
3021
3022 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3023                 struct iattr *sattr)
3024 {
3025         struct nfs4_exception exception = { };
3026         int err;
3027
3028         sattr->ia_mode &= ~current_umask();
3029         do {
3030                 err = nfs4_handle_exception(NFS_SERVER(dir),
3031                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3032                                 &exception);
3033         } while (exception.retry);
3034         return err;
3035 }
3036
3037 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3038                 u64 cookie, struct page **pages, unsigned int count, int plus)
3039 {
3040         struct inode            *dir = dentry->d_inode;
3041         struct nfs4_readdir_arg args = {
3042                 .fh = NFS_FH(dir),
3043                 .pages = pages,
3044                 .pgbase = 0,
3045                 .count = count,
3046                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3047                 .plus = plus,
3048         };
3049         struct nfs4_readdir_res res;
3050         struct rpc_message msg = {
3051                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3052                 .rpc_argp = &args,
3053                 .rpc_resp = &res,
3054                 .rpc_cred = cred,
3055         };
3056         int                     status;
3057
3058         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3059                         dentry->d_parent->d_name.name,
3060                         dentry->d_name.name,
3061                         (unsigned long long)cookie);
3062         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3063         res.pgbase = args.pgbase;
3064         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3065         if (status >= 0) {
3066                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3067                 status += args.pgbase;
3068         }
3069
3070         nfs_invalidate_atime(dir);
3071
3072         dprintk("%s: returns %d\n", __func__, status);
3073         return status;
3074 }
3075
3076 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3077                 u64 cookie, struct page **pages, unsigned int count, int plus)
3078 {
3079         struct nfs4_exception exception = { };
3080         int err;
3081         do {
3082                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3083                                 _nfs4_proc_readdir(dentry, cred, cookie,
3084                                         pages, count, plus),
3085                                 &exception);
3086         } while (exception.retry);
3087         return err;
3088 }
3089
3090 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3091                 struct iattr *sattr, dev_t rdev)
3092 {
3093         struct nfs4_createdata *data;
3094         int mode = sattr->ia_mode;
3095         int status = -ENOMEM;
3096
3097         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3098         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3099
3100         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3101         if (data == NULL)
3102                 goto out;
3103
3104         if (S_ISFIFO(mode))
3105                 data->arg.ftype = NF4FIFO;
3106         else if (S_ISBLK(mode)) {
3107                 data->arg.ftype = NF4BLK;
3108                 data->arg.u.device.specdata1 = MAJOR(rdev);
3109                 data->arg.u.device.specdata2 = MINOR(rdev);
3110         }
3111         else if (S_ISCHR(mode)) {
3112                 data->arg.ftype = NF4CHR;
3113                 data->arg.u.device.specdata1 = MAJOR(rdev);
3114                 data->arg.u.device.specdata2 = MINOR(rdev);
3115         }
3116
3117         status = nfs4_do_create(dir, dentry, data);
3118
3119         nfs4_free_createdata(data);
3120 out:
3121         return status;
3122 }
3123
3124 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3125                 struct iattr *sattr, dev_t rdev)
3126 {
3127         struct nfs4_exception exception = { };
3128         int err;
3129
3130         sattr->ia_mode &= ~current_umask();
3131         do {
3132                 err = nfs4_handle_exception(NFS_SERVER(dir),
3133                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3134                                 &exception);
3135         } while (exception.retry);
3136         return err;
3137 }
3138
3139 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3140                  struct nfs_fsstat *fsstat)
3141 {
3142         struct nfs4_statfs_arg args = {
3143                 .fh = fhandle,
3144                 .bitmask = server->attr_bitmask,
3145         };
3146         struct nfs4_statfs_res res = {
3147                 .fsstat = fsstat,
3148         };
3149         struct rpc_message msg = {
3150                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3151                 .rpc_argp = &args,
3152                 .rpc_resp = &res,
3153         };
3154
3155         nfs_fattr_init(fsstat->fattr);
3156         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3157 }
3158
3159 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3160 {
3161         struct nfs4_exception exception = { };
3162         int err;
3163         do {
3164                 err = nfs4_handle_exception(server,
3165                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3166                                 &exception);
3167         } while (exception.retry);
3168         return err;
3169 }
3170
3171 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3172                 struct nfs_fsinfo *fsinfo)
3173 {
3174         struct nfs4_fsinfo_arg args = {
3175                 .fh = fhandle,
3176                 .bitmask = server->attr_bitmask,
3177         };
3178         struct nfs4_fsinfo_res res = {
3179                 .fsinfo = fsinfo,
3180         };
3181         struct rpc_message msg = {
3182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3183                 .rpc_argp = &args,
3184                 .rpc_resp = &res,
3185         };
3186
3187         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3188 }
3189
3190 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3191 {
3192         struct nfs4_exception exception = { };
3193         int err;
3194
3195         do {
3196                 err = nfs4_handle_exception(server,
3197                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3198                                 &exception);
3199         } while (exception.retry);
3200         return err;
3201 }
3202
3203 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3204 {
3205         nfs_fattr_init(fsinfo->fattr);
3206         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3207 }
3208
3209 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3210                 struct nfs_pathconf *pathconf)
3211 {
3212         struct nfs4_pathconf_arg args = {
3213                 .fh = fhandle,
3214                 .bitmask = server->attr_bitmask,
3215         };
3216         struct nfs4_pathconf_res res = {
3217                 .pathconf = pathconf,
3218         };
3219         struct rpc_message msg = {
3220                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3221                 .rpc_argp = &args,
3222                 .rpc_resp = &res,
3223         };
3224
3225         /* None of the pathconf attributes are mandatory to implement */
3226         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3227                 memset(pathconf, 0, sizeof(*pathconf));
3228                 return 0;
3229         }
3230
3231         nfs_fattr_init(pathconf->fattr);
3232         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3233 }
3234
3235 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3236                 struct nfs_pathconf *pathconf)
3237 {
3238         struct nfs4_exception exception = { };
3239         int err;
3240
3241         do {
3242                 err = nfs4_handle_exception(server,
3243                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3244                                 &exception);
3245         } while (exception.retry);
3246         return err;
3247 }
3248
3249 void __nfs4_read_done_cb(struct nfs_read_data *data)
3250 {
3251         nfs_invalidate_atime(data->inode);
3252 }
3253
3254 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3255 {
3256         struct nfs_server *server = NFS_SERVER(data->inode);
3257
3258         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3259                 rpc_restart_call_prepare(task);
3260                 return -EAGAIN;
3261         }
3262
3263         __nfs4_read_done_cb(data);
3264         if (task->tk_status > 0)
3265                 renew_lease(server, data->timestamp);
3266         return 0;
3267 }
3268
3269 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3270 {
3271
3272         dprintk("--> %s\n", __func__);
3273
3274         if (!nfs4_sequence_done(task, &data->res.seq_res))
3275                 return -EAGAIN;
3276
3277         return data->read_done_cb ? data->read_done_cb(task, data) :
3278                                     nfs4_read_done_cb(task, data);
3279 }
3280
3281 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3282 {
3283         data->timestamp   = jiffies;
3284         data->read_done_cb = nfs4_read_done_cb;
3285         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3286         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3287 }
3288
3289 /* Reset the the nfs_read_data to send the read to the MDS. */
3290 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3291 {
3292         dprintk("%s Reset task for i/o through\n", __func__);
3293         put_lseg(data->lseg);
3294         data->lseg = NULL;
3295         /* offsets will differ in the dense stripe case */
3296         data->args.offset = data->mds_offset;
3297         data->ds_clp = NULL;
3298         data->args.fh     = NFS_FH(data->inode);
3299         data->read_done_cb = nfs4_read_done_cb;
3300         task->tk_ops = data->mds_ops;
3301         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3302 }
3303 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3304
3305 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3306 {
3307         struct inode *inode = data->inode;
3308
3309         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3310                 rpc_restart_call_prepare(task);
3311                 return -EAGAIN;
3312         }
3313         if (task->tk_status >= 0) {
3314                 renew_lease(NFS_SERVER(inode), data->timestamp);
3315                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3316         }
3317         return 0;
3318 }
3319
3320 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3321 {
3322         if (!nfs4_sequence_done(task, &data->res.seq_res))
3323                 return -EAGAIN;
3324         return data->write_done_cb ? data->write_done_cb(task, data) :
3325                 nfs4_write_done_cb(task, data);
3326 }
3327
3328 /* Reset the the nfs_write_data to send the write to the MDS. */
3329 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3330 {
3331         dprintk("%s Reset task for i/o through\n", __func__);
3332         put_lseg(data->lseg);
3333         data->lseg          = NULL;
3334         data->ds_clp        = NULL;
3335         data->write_done_cb = nfs4_write_done_cb;
3336         data->args.fh       = NFS_FH(data->inode);
3337         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3338         data->args.offset   = data->mds_offset;
3339         data->res.fattr     = &data->fattr;
3340         task->tk_ops        = data->mds_ops;
3341         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3342 }
3343 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3344
3345 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3346 {
3347         struct nfs_server *server = NFS_SERVER(data->inode);
3348
3349         if (data->lseg) {
3350                 data->args.bitmask = NULL;
3351                 data->res.fattr = NULL;
3352         } else
3353                 data->args.bitmask = server->cache_consistency_bitmask;
3354         if (!data->write_done_cb)
3355                 data->write_done_cb = nfs4_write_done_cb;
3356         data->res.server = server;
3357         data->timestamp   = jiffies;
3358
3359         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3360         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3361 }
3362
3363 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3364 {
3365         struct inode *inode = data->inode;
3366
3367         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3368                 rpc_restart_call_prepare(task);
3369                 return -EAGAIN;
3370         }
3371         nfs_refresh_inode(inode, data->res.fattr);
3372         return 0;
3373 }
3374
3375 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3376 {
3377         if (!nfs4_sequence_done(task, &data->res.seq_res))
3378                 return -EAGAIN;
3379         return data->write_done_cb(task, data);
3380 }
3381
3382 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3383 {
3384         struct nfs_server *server = NFS_SERVER(data->inode);
3385
3386         if (data->lseg) {
3387                 data->args.bitmask = NULL;
3388                 data->res.fattr = NULL;
3389         } else
3390                 data->args.bitmask = server->cache_consistency_bitmask;
3391         if (!data->write_done_cb)
3392                 data->write_done_cb = nfs4_commit_done_cb;
3393         data->res.server = server;
3394         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3395         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3396 }
3397
3398 struct nfs4_renewdata {
3399         struct nfs_client       *client;
3400         unsigned long           timestamp;
3401 };
3402
3403 /*
3404  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3405  * standalone procedure for queueing an asynchronous RENEW.
3406  */
3407 static void nfs4_renew_release(void *calldata)
3408 {
3409         struct nfs4_renewdata *data = calldata;
3410         struct nfs_client *clp = data->client;
3411
3412         if (atomic_read(&clp->cl_count) > 1)
3413                 nfs4_schedule_state_renewal(clp);
3414         nfs_put_client(clp);
3415         kfree(data);
3416 }
3417
3418 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3419 {
3420         struct nfs4_renewdata *data = calldata;
3421         struct nfs_client *clp = data->client;
3422         unsigned long timestamp = data->timestamp;
3423
3424         if (task->tk_status < 0) {
3425                 /* Unless we're shutting down, schedule state recovery! */
3426                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3427                         return;
3428                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3429                         nfs4_schedule_lease_recovery(clp);
3430                         return;
3431                 }
3432                 nfs4_schedule_path_down_recovery(clp);
3433         }
3434         do_renew_lease(clp, timestamp);
3435 }
3436
3437 static const struct rpc_call_ops nfs4_renew_ops = {
3438         .rpc_call_done = nfs4_renew_done,
3439         .rpc_release = nfs4_renew_release,
3440 };
3441
3442 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3443 {
3444         struct rpc_message msg = {
3445                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3446                 .rpc_argp       = clp,
3447                 .rpc_cred       = cred,
3448         };
3449         struct nfs4_renewdata *data;
3450
3451         if (renew_flags == 0)
3452                 return 0;
3453         if (!atomic_inc_not_zero(&clp->cl_count))
3454                 return -EIO;
3455         data = kmalloc(sizeof(*data), GFP_NOFS);
3456         if (data == NULL)
3457                 return -ENOMEM;
3458         data->client = clp;
3459         data->timestamp = jiffies;
3460         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3461                         &nfs4_renew_ops, data);
3462 }
3463
3464 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3465 {
3466         struct rpc_message msg = {
3467                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3468                 .rpc_argp       = clp,
3469                 .rpc_cred       = cred,
3470         };
3471         unsigned long now = jiffies;
3472         int status;
3473
3474         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3475         if (status < 0)
3476                 return status;
3477         do_renew_lease(clp, now);
3478         return 0;
3479 }
3480
3481 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3482 {
3483         return (server->caps & NFS_CAP_ACLS)
3484                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3485                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3486 }
3487
3488 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3489  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3490  * the stack.
3491  */
3492 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3493
3494 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3495                 struct page **pages, unsigned int *pgbase)
3496 {
3497         struct page *newpage, **spages;
3498         int rc = 0;
3499         size_t len;
3500         spages = pages;
3501
3502         do {
3503                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3504                 newpage = alloc_page(GFP_KERNEL);
3505
3506                 if (newpage == NULL)
3507                         goto unwind;
3508                 memcpy(page_address(newpage), buf, len);
3509                 buf += len;
3510                 buflen -= len;
3511                 *pages++ = newpage;
3512                 rc++;
3513         } while (buflen != 0);
3514
3515         return rc;
3516
3517 unwind:
3518         for(; rc > 0; rc--)
3519                 __free_page(spages[rc-1]);
3520         return -ENOMEM;
3521 }
3522
3523 struct nfs4_cached_acl {
3524         int cached;
3525         size_t len;
3526         char data[0];
3527 };
3528
3529 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3530 {
3531         struct nfs_inode *nfsi = NFS_I(inode);
3532
3533         spin_lock(&inode->i_lock);
3534         kfree(nfsi->nfs4_acl);
3535         nfsi->nfs4_acl = acl;
3536         spin_unlock(&inode->i_lock);
3537 }
3538
3539 static void nfs4_zap_acl_attr(struct inode *inode)
3540 {
3541         nfs4_set_cached_acl(inode, NULL);
3542 }
3543
3544 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3545 {
3546         struct nfs_inode *nfsi = NFS_I(inode);
3547         struct nfs4_cached_acl *acl;
3548         int ret = -ENOENT;
3549
3550         spin_lock(&inode->i_lock);
3551         acl = nfsi->nfs4_acl;
3552         if (acl == NULL)
3553                 goto out;
3554         if (buf == NULL) /* user is just asking for length */
3555                 goto out_len;
3556         if (acl->cached == 0)
3557                 goto out;
3558         ret = -ERANGE; /* see getxattr(2) man page */
3559         if (acl->len > buflen)
3560                 goto out;
3561         memcpy(buf, acl->data, acl->len);
3562 out_len:
3563         ret = acl->len;
3564 out:
3565         spin_unlock(&inode->i_lock);
3566         return ret;
3567 }
3568
3569 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3570 {
3571         struct nfs4_cached_acl *acl;
3572
3573         if (buf && acl_len <= PAGE_SIZE) {
3574                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3575                 if (acl == NULL)
3576                         goto out;
3577                 acl->cached = 1;
3578                 memcpy(acl->data, buf, acl_len);
3579         } else {
3580                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3581                 if (acl == NULL)
3582                         goto out;
3583                 acl->cached = 0;
3584         }
3585         acl->len = acl_len;
3586 out:
3587         nfs4_set_cached_acl(inode, acl);
3588 }
3589
3590 /*
3591  * The getxattr API returns the required buffer length when called with a
3592  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3593  * the required buf.  On a NULL buf, we send a page of data to the server
3594  * guessing that the ACL request can be serviced by a page. If so, we cache
3595  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3596  * the cache. If not so, we throw away the page, and cache the required
3597  * length. The next getxattr call will then produce another round trip to
3598  * the server, this time with the input buf of the required size.
3599  */
3600 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3601 {
3602         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3603         struct nfs_getaclargs args = {
3604                 .fh = NFS_FH(inode),
3605                 .acl_pages = pages,
3606                 .acl_len = buflen,
3607         };
3608         struct nfs_getaclres res = {
3609                 .acl_len = buflen,
3610         };
3611         void *resp_buf;
3612         struct rpc_message msg = {
3613                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3614                 .rpc_argp = &args,
3615                 .rpc_resp = &res,
3616         };
3617         int ret = -ENOMEM, npages, i, acl_len = 0;
3618
3619         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3620         /* As long as we're doing a round trip to the server anyway,
3621          * let's be prepared for a page of acl data. */
3622         if (npages == 0)
3623                 npages = 1;
3624
3625         for (i = 0; i < npages; i++) {
3626                 pages[i] = alloc_page(GFP_KERNEL);
3627                 if (!pages[i])
3628                         goto out_free;
3629         }
3630         if (npages > 1) {
3631                 /* for decoding across pages */
3632                 res.acl_scratch = alloc_page(GFP_KERNEL);
3633                 if (!res.acl_scratch)
3634                         goto out_free;
3635         }
3636         args.acl_len = npages * PAGE_SIZE;
3637         args.acl_pgbase = 0;
3638         /* Let decode_getfacl know not to fail if the ACL data is larger than
3639          * the page we send as a guess */
3640         if (buf == NULL)
3641                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3642         resp_buf = page_address(pages[0]);
3643
3644         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3645                 __func__, buf, buflen, npages, args.acl_len);
3646         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3647                              &msg, &args.seq_args, &res.seq_res, 0);
3648         if (ret)
3649                 goto out_free;
3650
3651         acl_len = res.acl_len - res.acl_data_offset;
3652         if (acl_len > args.acl_len)
3653                 nfs4_write_cached_acl(inode, NULL, acl_len);
3654         else
3655                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3656                                       acl_len);
3657         if (buf) {
3658                 ret = -ERANGE;
3659                 if (acl_len > buflen)
3660                         goto out_free;
3661                 _copy_from_pages(buf, pages, res.acl_data_offset,
3662                                 res.acl_len);
3663         }
3664         ret = acl_len;
3665 out_free:
3666         for (i = 0; i < npages; i++)
3667                 if (pages[i])
3668                         __free_page(pages[i]);
3669         if (res.acl_scratch)
3670                 __free_page(res.acl_scratch);
3671         return ret;
3672 }
3673
3674 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3675 {
3676         struct nfs4_exception exception = { };
3677         ssize_t ret;
3678         do {
3679                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3680                 if (ret >= 0)
3681                         break;
3682                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3683         } while (exception.retry);
3684         return ret;
3685 }
3686
3687 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3688 {
3689         struct nfs_server *server = NFS_SERVER(inode);
3690         int ret;
3691
3692         if (!nfs4_server_supports_acls(server))
3693                 return -EOPNOTSUPP;
3694         ret = nfs_revalidate_inode(server, inode);
3695         if (ret < 0)
3696                 return ret;
3697         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3698                 nfs_zap_acl_cache(inode);
3699         ret = nfs4_read_cached_acl(inode, buf, buflen);
3700         if (ret != -ENOENT)
3701                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3702                  * but no cached acl data, just the acl length */
3703                 return ret;
3704         return nfs4_get_acl_uncached(inode, buf, buflen);
3705 }
3706
3707 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3708 {
3709         struct nfs_server *server = NFS_SERVER(inode);
3710         struct page *pages[NFS4ACL_MAXPAGES];
3711         struct nfs_setaclargs arg = {
3712                 .fh             = NFS_FH(inode),
3713                 .acl_pages      = pages,
3714                 .acl_len        = buflen,
3715         };
3716         struct nfs_setaclres res;
3717         struct rpc_message msg = {
3718                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3719                 .rpc_argp       = &arg,
3720                 .rpc_resp       = &res,
3721         };
3722         int ret, i;
3723
3724         if (!nfs4_server_supports_acls(server))
3725                 return -EOPNOTSUPP;
3726         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3727         if (i < 0)
3728                 return i;
3729         nfs_inode_return_delegation(inode);
3730         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3731
3732         /*
3733          * Free each page after tx, so the only ref left is
3734          * held by the network stack
3735          */
3736         for (; i > 0; i--)
3737                 put_page(pages[i-1]);
3738
3739         /*
3740          * Acl update can result in inode attribute update.
3741          * so mark the attribute cache invalid.
3742          */
3743         spin_lock(&inode->i_lock);
3744         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3745         spin_unlock(&inode->i_lock);
3746         nfs_access_zap_cache(inode);
3747         nfs_zap_acl_cache(inode);
3748         return ret;
3749 }
3750
3751 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3752 {
3753         struct nfs4_exception exception = { };
3754         int err;
3755         do {
3756                 err = nfs4_handle_exception(NFS_SERVER(inode),
3757                                 __nfs4_proc_set_acl(inode, buf, buflen),
3758                                 &exception);
3759         } while (exception.retry);
3760         return err;
3761 }
3762
3763 static int
3764 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3765 {
3766         struct nfs_client *clp = server->nfs_client;
3767
3768         if (task->tk_status >= 0)
3769                 return 0;
3770         switch(task->tk_status) {
3771                 case -NFS4ERR_DELEG_REVOKED:
3772                 case -NFS4ERR_ADMIN_REVOKED:
3773                 case -NFS4ERR_BAD_STATEID:
3774                         if (state != NULL)
3775                                 nfs_remove_bad_delegation(state->inode);
3776                 case -NFS4ERR_OPENMODE:
3777                         if (state == NULL)
3778                                 break;
3779                         nfs4_schedule_stateid_recovery(server, state);
3780                         goto wait_on_recovery;
3781                 case -NFS4ERR_EXPIRED:
3782                         if (state != NULL)
3783                                 nfs4_schedule_stateid_recovery(server, state);
3784                 case -NFS4ERR_STALE_STATEID:
3785                 case -NFS4ERR_STALE_CLIENTID:
3786                         nfs4_schedule_lease_recovery(clp);
3787                         goto wait_on_recovery;
3788 #if defined(CONFIG_NFS_V4_1)
3789                 case -NFS4ERR_BADSESSION:
3790                 case -NFS4ERR_BADSLOT:
3791                 case -NFS4ERR_BAD_HIGH_SLOT:
3792                 case -NFS4ERR_DEADSESSION:
3793                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3794                 case -NFS4ERR_SEQ_FALSE_RETRY:
3795                 case -NFS4ERR_SEQ_MISORDERED:
3796                         dprintk("%s ERROR %d, Reset session\n", __func__,
3797                                 task->tk_status);
3798                         nfs4_schedule_session_recovery(clp->cl_session);
3799                         task->tk_status = 0;
3800                         return -EAGAIN;
3801 #endif /* CONFIG_NFS_V4_1 */
3802                 case -NFS4ERR_DELAY:
3803                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3804                 case -NFS4ERR_GRACE:
3805                 case -EKEYEXPIRED:
3806                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3807                         task->tk_status = 0;
3808                         return -EAGAIN;
3809                 case -NFS4ERR_RETRY_UNCACHED_REP:
3810                 case -NFS4ERR_OLD_STATEID:
3811                         task->tk_status = 0;
3812                         return -EAGAIN;
3813         }
3814         task->tk_status = nfs4_map_errors(task->tk_status);
3815         return 0;
3816 wait_on_recovery:
3817         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3818         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3819                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3820         task->tk_status = 0;
3821         return -EAGAIN;
3822 }
3823
3824 static void nfs4_construct_boot_verifier(struct nfs_client *clp,
3825                                          nfs4_verifier *bootverf)
3826 {
3827         __be32 verf[2];
3828
3829         verf[0] = htonl((u32)clp->cl_boot_time.tv_sec);
3830         verf[1] = htonl((u32)clp->cl_boot_time.tv_nsec);
3831         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3832 }
3833
3834 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3835                 unsigned short port, struct rpc_cred *cred,
3836                 struct nfs4_setclientid_res *res)
3837 {
3838         nfs4_verifier sc_verifier;
3839         struct nfs4_setclientid setclientid = {
3840                 .sc_verifier = &sc_verifier,
3841                 .sc_prog = program,
3842                 .sc_cb_ident = clp->cl_cb_ident,
3843         };
3844         struct rpc_message msg = {
3845                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3846                 .rpc_argp = &setclientid,
3847                 .rpc_resp = res,
3848                 .rpc_cred = cred,
3849         };
3850         int loop = 0;
3851         int status;
3852
3853         nfs4_construct_boot_verifier(clp, &sc_verifier);
3854
3855         for(;;) {
3856                 rcu_read_lock();
3857                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3858                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3859                                 clp->cl_ipaddr,
3860                                 rpc_peeraddr2str(clp->cl_rpcclient,
3861                                                         RPC_DISPLAY_ADDR),
3862                                 rpc_peeraddr2str(clp->cl_rpcclient,
3863                                                         RPC_DISPLAY_PROTO),
3864                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3865                                 clp->cl_id_uniquifier);
3866                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3867                                 sizeof(setclientid.sc_netid),
3868                                 rpc_peeraddr2str(clp->cl_rpcclient,
3869                                                         RPC_DISPLAY_NETID));
3870                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3871                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3872                                 clp->cl_ipaddr, port >> 8, port & 255);
3873                 rcu_read_unlock();
3874
3875                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3876                 if (status != -NFS4ERR_CLID_INUSE)
3877                         break;
3878                 if (loop != 0) {
3879                         ++clp->cl_id_uniquifier;
3880                         break;
3881                 }
3882                 ++loop;
3883                 ssleep(clp->cl_lease_time / HZ + 1);
3884         }
3885         return status;
3886 }
3887
3888 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3889                 struct nfs4_setclientid_res *arg,
3890                 struct rpc_cred *cred)
3891 {
3892         struct nfs_fsinfo fsinfo;
3893         struct rpc_message msg = {
3894                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3895                 .rpc_argp = arg,
3896                 .rpc_resp = &fsinfo,
3897                 .rpc_cred = cred,
3898         };
3899         unsigned long now;
3900         int status;
3901
3902         now = jiffies;
3903         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3904         if (status == 0) {
3905                 spin_lock(&clp->cl_lock);
3906                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3907                 clp->cl_last_renewal = now;
3908                 spin_unlock(&clp->cl_lock);
3909         }
3910         return status;
3911 }
3912
3913 struct nfs4_delegreturndata {
3914         struct nfs4_delegreturnargs args;
3915         struct nfs4_delegreturnres res;
3916         struct nfs_fh fh;
3917         nfs4_stateid stateid;
3918         unsigned long timestamp;
3919         struct nfs_fattr fattr;
3920         int rpc_status;
3921 };
3922
3923 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3924 {
3925         struct nfs4_delegreturndata *data = calldata;
3926
3927         if (!nfs4_sequence_done(task, &data->res.seq_res))
3928                 return;
3929
3930         switch (task->tk_status) {
3931         case -NFS4ERR_STALE_STATEID:
3932         case -NFS4ERR_EXPIRED:
3933         case 0:
3934                 renew_lease(data->res.server, data->timestamp);
3935                 break;
3936         default:
3937                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3938                                 -EAGAIN) {
3939                         rpc_restart_call_prepare(task);
3940                         return;
3941                 }
3942         }
3943         data->rpc_status = task->tk_status;
3944 }
3945
3946 static void nfs4_delegreturn_release(void *calldata)
3947 {
3948         kfree(calldata);
3949 }
3950
3951 #if defined(CONFIG_NFS_V4_1)
3952 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3953 {
3954         struct nfs4_delegreturndata *d_data;
3955
3956         d_data = (struct nfs4_delegreturndata *)data;
3957
3958         if (nfs4_setup_sequence(d_data->res.server,
3959                                 &d_data->args.seq_args,
3960                                 &d_data->res.seq_res, task))
3961                 return;
3962         rpc_call_start(task);
3963 }
3964 #endif /* CONFIG_NFS_V4_1 */
3965
3966 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3967 #if defined(CONFIG_NFS_V4_1)
3968         .rpc_call_prepare = nfs4_delegreturn_prepare,
3969 #endif /* CONFIG_NFS_V4_1 */
3970         .rpc_call_done = nfs4_delegreturn_done,
3971         .rpc_release = nfs4_delegreturn_release,
3972 };
3973
3974 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3975 {
3976         struct nfs4_delegreturndata *data;
3977         struct nfs_server *server = NFS_SERVER(inode);
3978         struct rpc_task *task;
3979         struct rpc_message msg = {
3980                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3981                 .rpc_cred = cred,
3982         };
3983         struct rpc_task_setup task_setup_data = {
3984                 .rpc_client = server->client,
3985                 .rpc_message = &msg,
3986                 .callback_ops = &nfs4_delegreturn_ops,
3987                 .flags = RPC_TASK_ASYNC,
3988         };
3989         int status = 0;
3990
3991         data = kzalloc(sizeof(*data), GFP_NOFS);
3992         if (data == NULL)
3993                 return -ENOMEM;
3994         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3995         data->args.fhandle = &data->fh;
3996         data->args.stateid = &data->stateid;
3997         data->args.bitmask = server->attr_bitmask;
3998         nfs_copy_fh(&data->fh, NFS_FH(inode));
3999         nfs4_stateid_copy(&data->stateid, stateid);
4000         data->res.fattr = &data->fattr;
4001         data->res.server = server;
4002         nfs_fattr_init(data->res.fattr);
4003         data->timestamp = jiffies;
4004         data->rpc_status = 0;
4005
4006         task_setup_data.callback_data = data;
4007         msg.rpc_argp = &data->args;
4008         msg.rpc_resp = &data->res;
4009         task = rpc_run_task(&task_setup_data);
4010         if (IS_ERR(task))
4011                 return PTR_ERR(task);
4012         if (!issync)
4013                 goto out;
4014         status = nfs4_wait_for_completion_rpc_task(task);
4015         if (status != 0)
4016                 goto out;
4017         status = data->rpc_status;
4018         if (status != 0)
4019                 goto out;
4020         nfs_refresh_inode(inode, &data->fattr);
4021 out:
4022         rpc_put_task(task);
4023         return status;
4024 }
4025
4026 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4027 {
4028         struct nfs_server *server = NFS_SERVER(inode);
4029         struct nfs4_exception exception = { };
4030         int err;
4031         do {
4032                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4033                 switch (err) {
4034                         case -NFS4ERR_STALE_STATEID:
4035                         case -NFS4ERR_EXPIRED:
4036                         case 0:
4037                                 return 0;
4038                 }
4039                 err = nfs4_handle_exception(server, err, &exception);
4040         } while (exception.retry);
4041         return err;
4042 }
4043
4044 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4045 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4046
4047 /*
4048  * sleep, with exponential backoff, and retry the LOCK operation.
4049  */
4050 static unsigned long
4051 nfs4_set_lock_task_retry(unsigned long timeout)
4052 {
4053         freezable_schedule_timeout_killable(timeout);
4054         timeout <<= 1;
4055         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4056                 return NFS4_LOCK_MAXTIMEOUT;
4057         return timeout;
4058 }
4059
4060 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4061 {
4062         struct inode *inode = state->inode;
4063         struct nfs_server *server = NFS_SERVER(inode);
4064         struct nfs_client *clp = server->nfs_client;
4065         struct nfs_lockt_args arg = {
4066                 .fh = NFS_FH(inode),
4067                 .fl = request,
4068         };
4069         struct nfs_lockt_res res = {
4070                 .denied = request,
4071         };
4072         struct rpc_message msg = {
4073                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4074                 .rpc_argp       = &arg,
4075                 .rpc_resp       = &res,
4076                 .rpc_cred       = state->owner->so_cred,
4077         };
4078         struct nfs4_lock_state *lsp;
4079         int status;
4080
4081         arg.lock_owner.clientid = clp->cl_clientid;
4082         status = nfs4_set_lock_state(state, request);
4083         if (status != 0)
4084                 goto out;
4085         lsp = request->fl_u.nfs4_fl.owner;
4086         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4087         arg.lock_owner.s_dev = server->s_dev;
4088         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4089         switch (status) {
4090                 case 0:
4091                         request->fl_type = F_UNLCK;
4092                         break;
4093                 case -NFS4ERR_DENIED:
4094                         status = 0;
4095         }
4096         request->fl_ops->fl_release_private(request);
4097 out:
4098         return status;
4099 }
4100
4101 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4102 {
4103         struct nfs4_exception exception = { };
4104         int err;
4105
4106         do {
4107                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4108                                 _nfs4_proc_getlk(state, cmd, request),
4109                                 &exception);
4110         } while (exception.retry);
4111         return err;
4112 }
4113
4114 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4115 {
4116         int res = 0;
4117         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4118                 case FL_POSIX:
4119                         res = posix_lock_file_wait(file, fl);
4120                         break;
4121                 case FL_FLOCK:
4122                         res = flock_lock_file_wait(file, fl);
4123                         break;
4124                 default:
4125                         BUG();
4126         }
4127         return res;
4128 }
4129
4130 struct nfs4_unlockdata {
4131         struct nfs_locku_args arg;
4132         struct nfs_locku_res res;
4133         struct nfs4_lock_state *lsp;
4134         struct nfs_open_context *ctx;
4135         struct file_lock fl;
4136         const struct nfs_server *server;
4137         unsigned long timestamp;
4138 };
4139
4140 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4141                 struct nfs_open_context *ctx,
4142                 struct nfs4_lock_state *lsp,
4143                 struct nfs_seqid *seqid)
4144 {
4145         struct nfs4_unlockdata *p;
4146         struct inode *inode = lsp->ls_state->inode;
4147
4148         p = kzalloc(sizeof(*p), GFP_NOFS);
4149         if (p == NULL)
4150                 return NULL;
4151         p->arg.fh = NFS_FH(inode);
4152         p->arg.fl = &p->fl;
4153         p->arg.seqid = seqid;
4154         p->res.seqid = seqid;
4155         p->arg.stateid = &lsp->ls_stateid;
4156         p->lsp = lsp;
4157         atomic_inc(&lsp->ls_count);
4158         /* Ensure we don't close file until we're done freeing locks! */
4159         p->ctx = get_nfs_open_context(ctx);
4160         memcpy(&p->fl, fl, sizeof(p->fl));
4161         p->server = NFS_SERVER(inode);
4162         return p;
4163 }
4164
4165 static void nfs4_locku_release_calldata(void *data)
4166 {
4167         struct nfs4_unlockdata *calldata = data;
4168         nfs_free_seqid(calldata->arg.seqid);
4169         nfs4_put_lock_state(calldata->lsp);
4170         put_nfs_open_context(calldata->ctx);
4171         kfree(calldata);
4172 }
4173
4174 static void nfs4_locku_done(struct rpc_task *task, void *data)
4175 {
4176         struct nfs4_unlockdata *calldata = data;
4177
4178         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4179                 return;
4180         switch (task->tk_status) {
4181                 case 0:
4182                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4183                                         &calldata->res.stateid);
4184                         renew_lease(calldata->server, calldata->timestamp);
4185                         break;
4186                 case -NFS4ERR_BAD_STATEID:
4187                 case -NFS4ERR_OLD_STATEID:
4188                 case -NFS4ERR_STALE_STATEID:
4189                 case -NFS4ERR_EXPIRED:
4190                         break;
4191                 default:
4192                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4193                                 rpc_restart_call_prepare(task);
4194         }
4195 }
4196
4197 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4198 {
4199         struct nfs4_unlockdata *calldata = data;
4200
4201         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4202                 return;
4203         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4204                 /* Note: exit _without_ running nfs4_locku_done */
4205                 task->tk_action = NULL;
4206                 return;
4207         }
4208         calldata->timestamp = jiffies;
4209         if (nfs4_setup_sequence(calldata->server,
4210                                 &calldata->arg.seq_args,
4211                                 &calldata->res.seq_res, task))
4212                 return;
4213         rpc_call_start(task);
4214 }
4215
4216 static const struct rpc_call_ops nfs4_locku_ops = {
4217         .rpc_call_prepare = nfs4_locku_prepare,
4218         .rpc_call_done = nfs4_locku_done,
4219         .rpc_release = nfs4_locku_release_calldata,
4220 };
4221
4222 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4223                 struct nfs_open_context *ctx,
4224                 struct nfs4_lock_state *lsp,
4225                 struct nfs_seqid *seqid)
4226 {
4227         struct nfs4_unlockdata *data;
4228         struct rpc_message msg = {
4229                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4230                 .rpc_cred = ctx->cred,
4231         };
4232         struct rpc_task_setup task_setup_data = {
4233                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4234                 .rpc_message = &msg,
4235                 .callback_ops = &nfs4_locku_ops,
4236                 .workqueue = nfsiod_workqueue,
4237                 .flags = RPC_TASK_ASYNC,
4238         };
4239
4240         /* Ensure this is an unlock - when canceling a lock, the
4241          * canceled lock is passed in, and it won't be an unlock.
4242          */
4243         fl->fl_type = F_UNLCK;
4244
4245         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4246         if (data == NULL) {
4247                 nfs_free_seqid(seqid);
4248                 return ERR_PTR(-ENOMEM);
4249         }
4250
4251         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4252         msg.rpc_argp = &data->arg;
4253         msg.rpc_resp = &data->res;
4254         task_setup_data.callback_data = data;
4255         return rpc_run_task(&task_setup_data);
4256 }
4257
4258 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4259 {
4260         struct nfs_inode *nfsi = NFS_I(state->inode);
4261         struct nfs_seqid *seqid;
4262         struct nfs4_lock_state *lsp;
4263         struct rpc_task *task;
4264         int status = 0;
4265         unsigned char fl_flags = request->fl_flags;
4266
4267         status = nfs4_set_lock_state(state, request);
4268         /* Unlock _before_ we do the RPC call */
4269         request->fl_flags |= FL_EXISTS;
4270         down_read(&nfsi->rwsem);
4271         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4272                 up_read(&nfsi->rwsem);
4273                 goto out;
4274         }
4275         up_read(&nfsi->rwsem);
4276         if (status != 0)
4277                 goto out;
4278         /* Is this a delegated lock? */
4279         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4280                 goto out;
4281         lsp = request->fl_u.nfs4_fl.owner;
4282         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4283         status = -ENOMEM;
4284         if (seqid == NULL)
4285                 goto out;
4286         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4287         status = PTR_ERR(task);
4288         if (IS_ERR(task))
4289                 goto out;
4290         status = nfs4_wait_for_completion_rpc_task(task);
4291         rpc_put_task(task);
4292 out:
4293         request->fl_flags = fl_flags;
4294         return status;
4295 }
4296
4297 struct nfs4_lockdata {
4298         struct nfs_lock_args arg;
4299         struct nfs_lock_res res;
4300         struct nfs4_lock_state *lsp;
4301         struct nfs_open_context *ctx;
4302         struct file_lock fl;
4303         unsigned long timestamp;
4304         int rpc_status;
4305         int cancelled;
4306         struct nfs_server *server;
4307 };
4308
4309 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4310                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4311                 gfp_t gfp_mask)
4312 {
4313         struct nfs4_lockdata *p;
4314         struct inode *inode = lsp->ls_state->inode;
4315         struct nfs_server *server = NFS_SERVER(inode);
4316
4317         p = kzalloc(sizeof(*p), gfp_mask);
4318         if (p == NULL)
4319                 return NULL;
4320
4321         p->arg.fh = NFS_FH(inode);
4322         p->arg.fl = &p->fl;
4323         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4324         if (p->arg.open_seqid == NULL)
4325                 goto out_free;
4326         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4327         if (p->arg.lock_seqid == NULL)
4328                 goto out_free_seqid;
4329         p->arg.lock_stateid = &lsp->ls_stateid;
4330         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4331         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4332         p->arg.lock_owner.s_dev = server->s_dev;
4333         p->res.lock_seqid = p->arg.lock_seqid;
4334         p->lsp = lsp;
4335         p->server = server;
4336         atomic_inc(&lsp->ls_count);
4337         p->ctx = get_nfs_open_context(ctx);
4338         memcpy(&p->fl, fl, sizeof(p->fl));
4339         return p;
4340 out_free_seqid:
4341         nfs_free_seqid(p->arg.open_seqid);
4342 out_free:
4343         kfree(p);
4344         return NULL;
4345 }
4346
4347 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4348 {
4349         struct nfs4_lockdata *data = calldata;
4350         struct nfs4_state *state = data->lsp->ls_state;
4351
4352         dprintk("%s: begin!\n", __func__);
4353         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4354                 return;
4355         /* Do we need to do an open_to_lock_owner? */
4356         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4357                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4358                         return;
4359                 data->arg.open_stateid = &state->stateid;
4360                 data->arg.new_lock_owner = 1;
4361                 data->res.open_seqid = data->arg.open_seqid;
4362         } else
4363                 data->arg.new_lock_owner = 0;
4364         data->timestamp = jiffies;
4365         if (nfs4_setup_sequence(data->server,
4366                                 &data->arg.seq_args,
4367                                 &data->res.seq_res, task))
4368                 return;
4369         rpc_call_start(task);
4370         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4371 }
4372
4373 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4374 {
4375         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4376         nfs4_lock_prepare(task, calldata);
4377 }
4378
4379 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4380 {
4381         struct nfs4_lockdata *data = calldata;
4382
4383         dprintk("%s: begin!\n", __func__);
4384
4385         if (!nfs4_sequence_done(task, &data->res.seq_res))
4386                 return;
4387
4388         data->rpc_status = task->tk_status;
4389         if (data->arg.new_lock_owner != 0) {
4390                 if (data->rpc_status == 0)
4391                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4392                 else
4393                         goto out;
4394         }
4395         if (data->rpc_status == 0) {
4396                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4397                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4398                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4399         }
4400 out:
4401         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4402 }
4403
4404 static void nfs4_lock_release(void *calldata)
4405 {
4406         struct nfs4_lockdata *data = calldata;
4407
4408         dprintk("%s: begin!\n", __func__);
4409         nfs_free_seqid(data->arg.open_seqid);
4410         if (data->cancelled != 0) {
4411                 struct rpc_task *task;
4412                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4413                                 data->arg.lock_seqid);
4414                 if (!IS_ERR(task))
4415                         rpc_put_task_async(task);
4416                 dprintk("%s: cancelling lock!\n", __func__);
4417         } else
4418                 nfs_free_seqid(data->arg.lock_seqid);
4419         nfs4_put_lock_state(data->lsp);
4420         put_nfs_open_context(data->ctx);
4421         kfree(data);
4422         dprintk("%s: done!\n", __func__);
4423 }
4424
4425 static const struct rpc_call_ops nfs4_lock_ops = {
4426         .rpc_call_prepare = nfs4_lock_prepare,
4427         .rpc_call_done = nfs4_lock_done,
4428         .rpc_release = nfs4_lock_release,
4429 };
4430
4431 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4432         .rpc_call_prepare = nfs4_recover_lock_prepare,
4433         .rpc_call_done = nfs4_lock_done,
4434         .rpc_release = nfs4_lock_release,
4435 };
4436
4437 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4438 {
4439         switch (error) {
4440         case -NFS4ERR_ADMIN_REVOKED:
4441         case -NFS4ERR_BAD_STATEID:
4442                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4443                 if (new_lock_owner != 0 ||
4444                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4445                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4446                 break;
4447         case -NFS4ERR_STALE_STATEID:
4448                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4449         case -NFS4ERR_EXPIRED:
4450                 nfs4_schedule_lease_recovery(server->nfs_client);
4451         };
4452 }
4453
4454 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4455 {
4456         struct nfs4_lockdata *data;
4457         struct rpc_task *task;
4458         struct rpc_message msg = {
4459                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4460                 .rpc_cred = state->owner->so_cred,
4461         };
4462         struct rpc_task_setup task_setup_data = {
4463                 .rpc_client = NFS_CLIENT(state->inode),
4464                 .rpc_message = &msg,
4465                 .callback_ops = &nfs4_lock_ops,
4466                 .workqueue = nfsiod_workqueue,
4467                 .flags = RPC_TASK_ASYNC,
4468         };
4469         int ret;
4470
4471         dprintk("%s: begin!\n", __func__);
4472         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4473                         fl->fl_u.nfs4_fl.owner,
4474                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4475         if (data == NULL)
4476                 return -ENOMEM;
4477         if (IS_SETLKW(cmd))
4478                 data->arg.block = 1;
4479         if (recovery_type > NFS_LOCK_NEW) {
4480                 if (recovery_type == NFS_LOCK_RECLAIM)
4481                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4482                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4483         }
4484         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4485         msg.rpc_argp = &data->arg;
4486         msg.rpc_resp = &data->res;
4487         task_setup_data.callback_data = data;
4488         task = rpc_run_task(&task_setup_data);
4489         if (IS_ERR(task))
4490                 return PTR_ERR(task);
4491         ret = nfs4_wait_for_completion_rpc_task(task);
4492         if (ret == 0) {
4493                 ret = data->rpc_status;
4494                 if (ret)
4495                         nfs4_handle_setlk_error(data->server, data->lsp,
4496                                         data->arg.new_lock_owner, ret);
4497         } else
4498                 data->cancelled = 1;
4499         rpc_put_task(task);
4500         dprintk("%s: done, ret = %d!\n", __func__, ret);
4501         return ret;
4502 }
4503
4504 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4505 {
4506         struct nfs_server *server = NFS_SERVER(state->inode);
4507         struct nfs4_exception exception = { };
4508         int err;
4509
4510         do {
4511                 /* Cache the lock if possible... */
4512                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4513                         return 0;
4514                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4515                 if (err != -NFS4ERR_DELAY)
4516                         break;
4517                 nfs4_handle_exception(server, err, &exception);
4518         } while (exception.retry);
4519         return err;
4520 }
4521
4522 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4523 {
4524         struct nfs_server *server = NFS_SERVER(state->inode);
4525         struct nfs4_exception exception = { };
4526         int err;
4527
4528         err = nfs4_set_lock_state(state, request);
4529         if (err != 0)
4530                 return err;
4531         do {
4532                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4533                         return 0;
4534                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4535                 switch (err) {
4536                 default:
4537                         goto out;
4538                 case -NFS4ERR_GRACE:
4539                 case -NFS4ERR_DELAY:
4540                         nfs4_handle_exception(server, err, &exception);
4541                         err = 0;
4542                 }
4543         } while (exception.retry);
4544 out:
4545         return err;
4546 }
4547
4548 #if defined(CONFIG_NFS_V4_1)
4549 static int nfs41_check_expired_locks(struct nfs4_state *state)
4550 {
4551         int status, ret = NFS_OK;
4552         struct nfs4_lock_state *lsp;
4553         struct nfs_server *server = NFS_SERVER(state->inode);
4554
4555         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4556                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4557                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4558                         if (status != NFS_OK) {
4559                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4560                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4561                                 ret = status;
4562                         }
4563                 }
4564         };
4565
4566         return ret;
4567 }
4568
4569 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4570 {
4571         int status = NFS_OK;
4572
4573         if (test_bit(LK_STATE_IN_USE, &state->flags))
4574                 status = nfs41_check_expired_locks(state);
4575         if (status == NFS_OK)
4576                 return status;
4577         return nfs4_lock_expired(state, request);
4578 }
4579 #endif
4580
4581 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4582 {
4583         struct nfs_inode *nfsi = NFS_I(state->inode);
4584         unsigned char fl_flags = request->fl_flags;
4585         int status = -ENOLCK;
4586
4587         if ((fl_flags & FL_POSIX) &&
4588                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4589                 goto out;
4590         /* Is this a delegated open? */
4591         status = nfs4_set_lock_state(state, request);
4592         if (status != 0)
4593                 goto out;
4594         request->fl_flags |= FL_ACCESS;
4595         status = do_vfs_lock(request->fl_file, request);
4596         if (status < 0)
4597                 goto out;
4598         down_read(&nfsi->rwsem);
4599         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4600                 /* Yes: cache locks! */
4601                 /* ...but avoid races with delegation recall... */
4602                 request->fl_flags = fl_flags & ~FL_SLEEP;
4603                 status = do_vfs_lock(request->fl_file, request);
4604                 goto out_unlock;
4605         }
4606         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4607         if (status != 0)
4608                 goto out_unlock;
4609         /* Note: we always want to sleep here! */
4610         request->fl_flags = fl_flags | FL_SLEEP;
4611         if (do_vfs_lock(request->fl_file, request) < 0)
4612                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4613                         "manager!\n", __func__);
4614 out_unlock:
4615         up_read(&nfsi->rwsem);
4616 out:
4617         request->fl_flags = fl_flags;
4618         return status;
4619 }
4620
4621 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4622 {
4623         struct nfs4_exception exception = {
4624                 .state = state,
4625         };
4626         int err;
4627
4628         do {
4629                 err = _nfs4_proc_setlk(state, cmd, request);
4630                 if (err == -NFS4ERR_DENIED)
4631                         err = -EAGAIN;
4632                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4633                                 err, &exception);
4634         } while (exception.retry);
4635         return err;
4636 }
4637
4638 static int
4639 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4640 {
4641         struct nfs_open_context *ctx;
4642         struct nfs4_state *state;
4643         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4644         int status;
4645
4646         /* verify open state */
4647         ctx = nfs_file_open_context(filp);
4648         state = ctx->state;
4649
4650         if (request->fl_start < 0 || request->fl_end < 0)
4651                 return -EINVAL;
4652
4653         if (IS_GETLK(cmd)) {
4654                 if (state != NULL)
4655                         return nfs4_proc_getlk(state, F_GETLK, request);
4656                 return 0;
4657         }
4658
4659         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4660                 return -EINVAL;
4661
4662         if (request->fl_type == F_UNLCK) {
4663                 if (state != NULL)
4664                         return nfs4_proc_unlck(state, cmd, request);
4665                 return 0;
4666         }
4667
4668         if (state == NULL)
4669                 return -ENOLCK;
4670         do {
4671                 status = nfs4_proc_setlk(state, cmd, request);
4672                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4673                         break;
4674                 timeout = nfs4_set_lock_task_retry(timeout);
4675                 status = -ERESTARTSYS;
4676                 if (signalled())
4677                         break;
4678         } while(status < 0);
4679         return status;
4680 }
4681
4682 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4683 {
4684         struct nfs_server *server = NFS_SERVER(state->inode);
4685         struct nfs4_exception exception = { };
4686         int err;
4687
4688         err = nfs4_set_lock_state(state, fl);
4689         if (err != 0)
4690                 goto out;
4691         do {
4692                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4693                 switch (err) {
4694                         default:
4695                                 printk(KERN_ERR "NFS: %s: unhandled error "
4696                                         "%d.\n", __func__, err);
4697                         case 0:
4698                         case -ESTALE:
4699                                 goto out;
4700                         case -NFS4ERR_EXPIRED:
4701                                 nfs4_schedule_stateid_recovery(server, state);
4702                         case -NFS4ERR_STALE_CLIENTID:
4703                         case -NFS4ERR_STALE_STATEID:
4704                                 nfs4_schedule_lease_recovery(server->nfs_client);
4705                                 goto out;
4706                         case -NFS4ERR_BADSESSION:
4707                         case -NFS4ERR_BADSLOT:
4708                         case -NFS4ERR_BAD_HIGH_SLOT:
4709                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4710                         case -NFS4ERR_DEADSESSION:
4711                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4712                                 goto out;
4713                         case -ERESTARTSYS:
4714                                 /*
4715                                  * The show must go on: exit, but mark the
4716                                  * stateid as needing recovery.
4717                                  */
4718                         case -NFS4ERR_DELEG_REVOKED:
4719                         case -NFS4ERR_ADMIN_REVOKED:
4720                         case -NFS4ERR_BAD_STATEID:
4721                         case -NFS4ERR_OPENMODE:
4722                                 nfs4_schedule_stateid_recovery(server, state);
4723                                 err = 0;
4724                                 goto out;
4725                         case -EKEYEXPIRED:
4726                                 /*
4727                                  * User RPCSEC_GSS context has expired.
4728                                  * We cannot recover this stateid now, so
4729                                  * skip it and allow recovery thread to
4730                                  * proceed.
4731                                  */
4732                                 err = 0;
4733                                 goto out;
4734                         case -ENOMEM:
4735                         case -NFS4ERR_DENIED:
4736                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4737                                 err = 0;
4738                                 goto out;
4739                         case -NFS4ERR_DELAY:
4740                                 break;
4741                 }
4742                 err = nfs4_handle_exception(server, err, &exception);
4743         } while (exception.retry);
4744 out:
4745         return err;
4746 }
4747
4748 static void nfs4_release_lockowner_release(void *calldata)
4749 {
4750         kfree(calldata);
4751 }
4752
4753 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4754         .rpc_release = nfs4_release_lockowner_release,
4755 };
4756
4757 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4758 {
4759         struct nfs_server *server = lsp->ls_state->owner->so_server;
4760         struct nfs_release_lockowner_args *args;
4761         struct rpc_message msg = {
4762                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4763         };
4764
4765         if (server->nfs_client->cl_mvops->minor_version != 0)
4766                 return;
4767         args = kmalloc(sizeof(*args), GFP_NOFS);
4768         if (!args)
4769                 return;
4770         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4771         args->lock_owner.id = lsp->ls_seqid.owner_id;
4772         args->lock_owner.s_dev = server->s_dev;
4773         msg.rpc_argp = args;
4774         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4775 }
4776
4777 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4778
4779 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4780                                    const void *buf, size_t buflen,
4781                                    int flags, int type)
4782 {
4783         if (strcmp(key, "") != 0)
4784                 return -EINVAL;
4785
4786         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4787 }
4788
4789 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4790                                    void *buf, size_t buflen, int type)
4791 {
4792         if (strcmp(key, "") != 0)
4793                 return -EINVAL;
4794
4795         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4796 }
4797
4798 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4799                                        size_t list_len, const char *name,
4800                                        size_t name_len, int type)
4801 {
4802         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4803
4804         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4805                 return 0;
4806
4807         if (list && len <= list_len)
4808                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4809         return len;
4810 }
4811
4812 /*
4813  * nfs_fhget will use either the mounted_on_fileid or the fileid
4814  */
4815 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4816 {
4817         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4818                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4819               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4820               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4821                 return;
4822
4823         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4824                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4825         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4826         fattr->nlink = 2;
4827 }
4828
4829 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4830                 struct nfs4_fs_locations *fs_locations, struct page *page)
4831 {
4832         struct nfs_server *server = NFS_SERVER(dir);
4833         u32 bitmask[2] = {
4834                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4835         };
4836         struct nfs4_fs_locations_arg args = {
4837                 .dir_fh = NFS_FH(dir),
4838                 .name = name,
4839                 .page = page,
4840                 .bitmask = bitmask,
4841         };
4842         struct nfs4_fs_locations_res res = {
4843                 .fs_locations = fs_locations,
4844         };
4845         struct rpc_message msg = {
4846                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4847                 .rpc_argp = &args,
4848                 .rpc_resp = &res,
4849         };
4850         int status;
4851
4852         dprintk("%s: start\n", __func__);
4853
4854         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4855          * is not supported */
4856         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4857                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4858         else
4859                 bitmask[0] |= FATTR4_WORD0_FILEID;
4860
4861         nfs_fattr_init(&fs_locations->fattr);
4862         fs_locations->server = server;
4863         fs_locations->nlocations = 0;
4864         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4865         dprintk("%s: returned status = %d\n", __func__, status);
4866         return status;
4867 }
4868
4869 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4870 {
4871         int status;
4872         struct nfs4_secinfo_arg args = {
4873                 .dir_fh = NFS_FH(dir),
4874                 .name   = name,
4875         };
4876         struct nfs4_secinfo_res res = {
4877                 .flavors     = flavors,
4878         };
4879         struct rpc_message msg = {
4880                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4881                 .rpc_argp = &args,
4882                 .rpc_resp = &res,
4883         };
4884
4885         dprintk("NFS call  secinfo %s\n", name->name);
4886         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4887         dprintk("NFS reply  secinfo: %d\n", status);
4888         return status;
4889 }
4890
4891 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4892 {
4893         struct nfs4_exception exception = { };
4894         int err;
4895         do {
4896                 err = nfs4_handle_exception(NFS_SERVER(dir),
4897                                 _nfs4_proc_secinfo(dir, name, flavors),
4898                                 &exception);
4899         } while (exception.retry);
4900         return err;
4901 }
4902
4903 #ifdef CONFIG_NFS_V4_1
4904 /*
4905  * Check the exchange flags returned by the server for invalid flags, having
4906  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4907  * DS flags set.
4908  */
4909 static int nfs4_check_cl_exchange_flags(u32 flags)
4910 {
4911         if (flags & ~EXCHGID4_FLAG_MASK_R)
4912                 goto out_inval;
4913         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4914             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4915                 goto out_inval;
4916         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4917                 goto out_inval;
4918         return NFS_OK;
4919 out_inval:
4920         return -NFS4ERR_INVAL;
4921 }
4922
4923 static bool
4924 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4925 {
4926         if (a->server_scope_sz == b->server_scope_sz &&
4927             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4928                 return true;
4929
4930         return false;
4931 }
4932
4933 /*
4934  * nfs4_proc_exchange_id()
4935  *
4936  * Since the clientid has expired, all compounds using sessions
4937  * associated with the stale clientid will be returning
4938  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4939  * be in some phase of session reset.
4940  */
4941 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4942 {
4943         nfs4_verifier verifier;
4944         struct nfs41_exchange_id_args args = {
4945                 .verifier = &verifier,
4946                 .client = clp,
4947                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4948         };
4949         struct nfs41_exchange_id_res res = {
4950                 .client = clp,
4951         };
4952         int status;
4953         struct rpc_message msg = {
4954                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4955                 .rpc_argp = &args,
4956                 .rpc_resp = &res,
4957                 .rpc_cred = cred,
4958         };
4959
4960         dprintk("--> %s\n", __func__);
4961         BUG_ON(clp == NULL);
4962
4963         nfs4_construct_boot_verifier(clp, &verifier);
4964
4965         args.id_len = scnprintf(args.id, sizeof(args.id),
4966                                 "%s/%s.%s/%u",
4967                                 clp->cl_ipaddr,
4968                                 init_utsname()->nodename,
4969                                 init_utsname()->domainname,
4970                                 clp->cl_rpcclient->cl_auth->au_flavor);
4971
4972         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4973         if (unlikely(!res.server_scope)) {
4974                 status = -ENOMEM;
4975                 goto out;
4976         }
4977
4978         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
4979         if (unlikely(!res.impl_id)) {
4980                 status = -ENOMEM;
4981                 goto out_server_scope;
4982         }
4983
4984         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4985         if (!status)
4986                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4987
4988         if (!status) {
4989                 /* use the most recent implementation id */
4990                 kfree(clp->impl_id);
4991                 clp->impl_id = res.impl_id;
4992         } else
4993                 kfree(res.impl_id);
4994
4995         if (!status) {
4996                 if (clp->server_scope &&
4997                     !nfs41_same_server_scope(clp->server_scope,
4998                                              res.server_scope)) {
4999                         dprintk("%s: server_scope mismatch detected\n",
5000                                 __func__);
5001                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5002                         kfree(clp->server_scope);
5003                         clp->server_scope = NULL;
5004                 }
5005
5006                 if (!clp->server_scope) {
5007                         clp->server_scope = res.server_scope;
5008                         goto out;
5009                 }
5010         }
5011
5012 out_server_scope:
5013         kfree(res.server_scope);
5014 out:
5015         if (clp->impl_id)
5016                 dprintk("%s: Server Implementation ID: "
5017                         "domain: %s, name: %s, date: %llu,%u\n",
5018                         __func__, clp->impl_id->domain, clp->impl_id->name,
5019                         clp->impl_id->date.seconds,
5020                         clp->impl_id->date.nseconds);
5021         dprintk("<-- %s status= %d\n", __func__, status);
5022         return status;
5023 }
5024
5025 struct nfs4_get_lease_time_data {
5026         struct nfs4_get_lease_time_args *args;
5027         struct nfs4_get_lease_time_res *res;
5028         struct nfs_client *clp;
5029 };
5030
5031 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5032                                         void *calldata)
5033 {
5034         int ret;
5035         struct nfs4_get_lease_time_data *data =
5036                         (struct nfs4_get_lease_time_data *)calldata;
5037
5038         dprintk("--> %s\n", __func__);
5039         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5040         /* just setup sequence, do not trigger session recovery
5041            since we're invoked within one */
5042         ret = nfs41_setup_sequence(data->clp->cl_session,
5043                                    &data->args->la_seq_args,
5044                                    &data->res->lr_seq_res, task);
5045
5046         BUG_ON(ret == -EAGAIN);
5047         rpc_call_start(task);
5048         dprintk("<-- %s\n", __func__);
5049 }
5050
5051 /*
5052  * Called from nfs4_state_manager thread for session setup, so don't recover
5053  * from sequence operation or clientid errors.
5054  */
5055 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5056 {
5057         struct nfs4_get_lease_time_data *data =
5058                         (struct nfs4_get_lease_time_data *)calldata;
5059
5060         dprintk("--> %s\n", __func__);
5061         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5062                 return;
5063         switch (task->tk_status) {
5064         case -NFS4ERR_DELAY:
5065         case -NFS4ERR_GRACE:
5066                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5067                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5068                 task->tk_status = 0;
5069                 /* fall through */
5070         case -NFS4ERR_RETRY_UNCACHED_REP:
5071                 rpc_restart_call_prepare(task);
5072                 return;
5073         }
5074         dprintk("<-- %s\n", __func__);
5075 }
5076
5077 struct rpc_call_ops nfs4_get_lease_time_ops = {
5078         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5079         .rpc_call_done = nfs4_get_lease_time_done,
5080 };
5081
5082 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5083 {
5084         struct rpc_task *task;
5085         struct nfs4_get_lease_time_args args;
5086         struct nfs4_get_lease_time_res res = {
5087                 .lr_fsinfo = fsinfo,
5088         };
5089         struct nfs4_get_lease_time_data data = {
5090                 .args = &args,
5091                 .res = &res,
5092                 .clp = clp,
5093         };
5094         struct rpc_message msg = {
5095                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5096                 .rpc_argp = &args,
5097                 .rpc_resp = &res,
5098         };
5099         struct rpc_task_setup task_setup = {
5100                 .rpc_client = clp->cl_rpcclient,
5101                 .rpc_message = &msg,
5102                 .callback_ops = &nfs4_get_lease_time_ops,
5103                 .callback_data = &data,
5104                 .flags = RPC_TASK_TIMEOUT,
5105         };
5106         int status;
5107
5108         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5109         dprintk("--> %s\n", __func__);
5110         task = rpc_run_task(&task_setup);
5111
5112         if (IS_ERR(task))
5113                 status = PTR_ERR(task);
5114         else {
5115                 status = task->tk_status;
5116                 rpc_put_task(task);
5117         }
5118         dprintk("<-- %s return %d\n", __func__, status);
5119
5120         return status;
5121 }
5122
5123 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5124 {
5125         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5126 }
5127
5128 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5129                 struct nfs4_slot *new,
5130                 u32 max_slots,
5131                 u32 ivalue)
5132 {
5133         struct nfs4_slot *old = NULL;
5134         u32 i;
5135
5136         spin_lock(&tbl->slot_tbl_lock);
5137         if (new) {
5138                 old = tbl->slots;
5139                 tbl->slots = new;
5140                 tbl->max_slots = max_slots;
5141         }
5142         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5143         for (i = 0; i < tbl->max_slots; i++)
5144                 tbl->slots[i].seq_nr = ivalue;
5145         spin_unlock(&tbl->slot_tbl_lock);
5146         kfree(old);
5147 }
5148
5149 /*
5150  * (re)Initialise a slot table
5151  */
5152 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5153                                  u32 ivalue)
5154 {
5155         struct nfs4_slot *new = NULL;
5156         int ret = -ENOMEM;
5157
5158         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5159                 max_reqs, tbl->max_slots);
5160
5161         /* Does the newly negotiated max_reqs match the existing slot table? */
5162         if (max_reqs != tbl->max_slots) {
5163                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5164                 if (!new)
5165                         goto out;
5166         }
5167         ret = 0;
5168
5169         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5170         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5171                 tbl, tbl->slots, tbl->max_slots);
5172 out:
5173         dprintk("<-- %s: return %d\n", __func__, ret);
5174         return ret;
5175 }
5176
5177 /* Destroy the slot table */
5178 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5179 {
5180         if (session->fc_slot_table.slots != NULL) {
5181                 kfree(session->fc_slot_table.slots);
5182                 session->fc_slot_table.slots = NULL;
5183         }
5184         if (session->bc_slot_table.slots != NULL) {
5185                 kfree(session->bc_slot_table.slots);
5186                 session->bc_slot_table.slots = NULL;
5187         }
5188         return;
5189 }
5190
5191 /*
5192  * Initialize or reset the forechannel and backchannel tables
5193  */
5194 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5195 {
5196         struct nfs4_slot_table *tbl;
5197         int status;
5198
5199         dprintk("--> %s\n", __func__);
5200         /* Fore channel */
5201         tbl = &ses->fc_slot_table;
5202         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5203         if (status) /* -ENOMEM */
5204                 return status;
5205         /* Back channel */
5206         tbl = &ses->bc_slot_table;
5207         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5208         if (status && tbl->slots == NULL)
5209                 /* Fore and back channel share a connection so get
5210                  * both slot tables or neither */
5211                 nfs4_destroy_slot_tables(ses);
5212         return status;
5213 }
5214
5215 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5216 {
5217         struct nfs4_session *session;
5218         struct nfs4_slot_table *tbl;
5219
5220         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5221         if (!session)
5222                 return NULL;
5223
5224         tbl = &session->fc_slot_table;
5225         tbl->highest_used_slotid = NFS4_NO_SLOT;
5226         spin_lock_init(&tbl->slot_tbl_lock);
5227         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5228         init_completion(&tbl->complete);
5229
5230         tbl = &session->bc_slot_table;
5231         tbl->highest_used_slotid = NFS4_NO_SLOT;
5232         spin_lock_init(&tbl->slot_tbl_lock);
5233         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5234         init_completion(&tbl->complete);
5235
5236         session->session_state = 1<<NFS4_SESSION_INITING;
5237
5238         session->clp = clp;
5239         return session;
5240 }
5241
5242 void nfs4_destroy_session(struct nfs4_session *session)
5243 {
5244         struct rpc_xprt *xprt;
5245
5246         nfs4_proc_destroy_session(session);
5247
5248         rcu_read_lock();
5249         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5250         rcu_read_unlock();
5251         dprintk("%s Destroy backchannel for xprt %p\n",
5252                 __func__, xprt);
5253         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5254         nfs4_destroy_slot_tables(session);
5255         kfree(session);
5256 }
5257
5258 /*
5259  * Initialize the values to be used by the client in CREATE_SESSION
5260  * If nfs4_init_session set the fore channel request and response sizes,
5261  * use them.
5262  *
5263  * Set the back channel max_resp_sz_cached to zero to force the client to
5264  * always set csa_cachethis to FALSE because the current implementation
5265  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5266  */
5267 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5268 {
5269         struct nfs4_session *session = args->client->cl_session;
5270         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5271                      mxresp_sz = session->fc_attrs.max_resp_sz;
5272
5273         if (mxrqst_sz == 0)
5274                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5275         if (mxresp_sz == 0)
5276                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5277         /* Fore channel attributes */
5278         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5279         args->fc_attrs.max_resp_sz = mxresp_sz;
5280         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5281         args->fc_attrs.max_reqs = max_session_slots;
5282
5283         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5284                 "max_ops=%u max_reqs=%u\n",
5285                 __func__,
5286                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5287                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5288
5289         /* Back channel attributes */
5290         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5291         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5292         args->bc_attrs.max_resp_sz_cached = 0;
5293         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5294         args->bc_attrs.max_reqs = 1;
5295
5296         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5297                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5298                 __func__,
5299                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5300                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5301                 args->bc_attrs.max_reqs);
5302 }
5303
5304 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5305 {
5306         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5307         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5308
5309         if (rcvd->max_resp_sz > sent->max_resp_sz)
5310                 return -EINVAL;
5311         /*
5312          * Our requested max_ops is the minimum we need; we're not
5313          * prepared to break up compounds into smaller pieces than that.
5314          * So, no point even trying to continue if the server won't
5315          * cooperate:
5316          */
5317         if (rcvd->max_ops < sent->max_ops)
5318                 return -EINVAL;
5319         if (rcvd->max_reqs == 0)
5320                 return -EINVAL;
5321         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5322                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5323         return 0;
5324 }
5325
5326 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5327 {
5328         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5329         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5330
5331         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5332                 return -EINVAL;
5333         if (rcvd->max_resp_sz < sent->max_resp_sz)
5334                 return -EINVAL;
5335         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5336                 return -EINVAL;
5337         /* These would render the backchannel useless: */
5338         if (rcvd->max_ops != sent->max_ops)
5339                 return -EINVAL;
5340         if (rcvd->max_reqs != sent->max_reqs)
5341                 return -EINVAL;
5342         return 0;
5343 }
5344
5345 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5346                                      struct nfs4_session *session)
5347 {
5348         int ret;
5349
5350         ret = nfs4_verify_fore_channel_attrs(args, session);
5351         if (ret)
5352                 return ret;
5353         return nfs4_verify_back_channel_attrs(args, session);
5354 }
5355
5356 static int _nfs4_proc_create_session(struct nfs_client *clp)
5357 {
5358         struct nfs4_session *session = clp->cl_session;
5359         struct nfs41_create_session_args args = {
5360                 .client = clp,
5361                 .cb_program = NFS4_CALLBACK,
5362         };
5363         struct nfs41_create_session_res res = {
5364                 .client = clp,
5365         };
5366         struct rpc_message msg = {
5367                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5368                 .rpc_argp = &args,
5369                 .rpc_resp = &res,
5370         };
5371         int status;
5372
5373         nfs4_init_channel_attrs(&args);
5374         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5375
5376         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5377
5378         if (!status)
5379                 /* Verify the session's negotiated channel_attrs values */
5380                 status = nfs4_verify_channel_attrs(&args, session);
5381         if (!status) {
5382                 /* Increment the clientid slot sequence id */
5383                 clp->cl_seqid++;
5384         }
5385
5386         return status;
5387 }
5388
5389 /*
5390  * Issues a CREATE_SESSION operation to the server.
5391  * It is the responsibility of the caller to verify the session is
5392  * expired before calling this routine.
5393  */
5394 int nfs4_proc_create_session(struct nfs_client *clp)
5395 {
5396         int status;
5397         unsigned *ptr;
5398         struct nfs4_session *session = clp->cl_session;
5399
5400         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5401
5402         status = _nfs4_proc_create_session(clp);
5403         if (status)
5404                 goto out;
5405
5406         /* Init or reset the session slot tables */
5407         status = nfs4_setup_session_slot_tables(session);
5408         dprintk("slot table setup returned %d\n", status);
5409         if (status)
5410                 goto out;
5411
5412         ptr = (unsigned *)&session->sess_id.data[0];
5413         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5414                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5415 out:
5416         dprintk("<-- %s\n", __func__);
5417         return status;
5418 }
5419
5420 /*
5421  * Issue the over-the-wire RPC DESTROY_SESSION.
5422  * The caller must serialize access to this routine.
5423  */
5424 int nfs4_proc_destroy_session(struct nfs4_session *session)
5425 {
5426         int status = 0;
5427         struct rpc_message msg;
5428
5429         dprintk("--> nfs4_proc_destroy_session\n");
5430
5431         /* session is still being setup */
5432         if (session->clp->cl_cons_state != NFS_CS_READY)
5433                 return status;
5434
5435         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5436         msg.rpc_argp = session;
5437         msg.rpc_resp = NULL;
5438         msg.rpc_cred = NULL;
5439         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5440
5441         if (status)
5442                 printk(KERN_WARNING
5443                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5444                         "Session has been destroyed regardless...\n", status);
5445
5446         dprintk("<-- nfs4_proc_destroy_session\n");
5447         return status;
5448 }
5449
5450 int nfs4_init_session(struct nfs_server *server)
5451 {
5452         struct nfs_client *clp = server->nfs_client;
5453         struct nfs4_session *session;
5454         unsigned int rsize, wsize;
5455         int ret;
5456
5457         if (!nfs4_has_session(clp))
5458                 return 0;
5459
5460         session = clp->cl_session;
5461         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5462                 return 0;
5463
5464         rsize = server->rsize;
5465         if (rsize == 0)
5466                 rsize = NFS_MAX_FILE_IO_SIZE;
5467         wsize = server->wsize;
5468         if (wsize == 0)
5469                 wsize = NFS_MAX_FILE_IO_SIZE;
5470
5471         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5472         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5473
5474         ret = nfs4_recover_expired_lease(server);
5475         if (!ret)
5476                 ret = nfs4_check_client_ready(clp);
5477         return ret;
5478 }
5479
5480 int nfs4_init_ds_session(struct nfs_client *clp)
5481 {
5482         struct nfs4_session *session = clp->cl_session;
5483         int ret;
5484
5485         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5486                 return 0;
5487
5488         ret = nfs4_client_recover_expired_lease(clp);
5489         if (!ret)
5490                 /* Test for the DS role */
5491                 if (!is_ds_client(clp))
5492                         ret = -ENODEV;
5493         if (!ret)
5494                 ret = nfs4_check_client_ready(clp);
5495         return ret;
5496
5497 }
5498 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5499
5500
5501 /*
5502  * Renew the cl_session lease.
5503  */
5504 struct nfs4_sequence_data {
5505         struct nfs_client *clp;
5506         struct nfs4_sequence_args args;
5507         struct nfs4_sequence_res res;
5508 };
5509
5510 static void nfs41_sequence_release(void *data)
5511 {
5512         struct nfs4_sequence_data *calldata = data;
5513         struct nfs_client *clp = calldata->clp;
5514
5515         if (atomic_read(&clp->cl_count) > 1)
5516                 nfs4_schedule_state_renewal(clp);
5517         nfs_put_client(clp);
5518         kfree(calldata);
5519 }
5520
5521 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5522 {
5523         switch(task->tk_status) {
5524         case -NFS4ERR_DELAY:
5525                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5526                 return -EAGAIN;
5527         default:
5528                 nfs4_schedule_lease_recovery(clp);
5529         }
5530         return 0;
5531 }
5532
5533 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5534 {
5535         struct nfs4_sequence_data *calldata = data;
5536         struct nfs_client *clp = calldata->clp;
5537
5538         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5539                 return;
5540
5541         if (task->tk_status < 0) {
5542                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5543                 if (atomic_read(&clp->cl_count) == 1)
5544                         goto out;
5545
5546                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5547                         rpc_restart_call_prepare(task);
5548                         return;
5549                 }
5550         }
5551         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5552 out:
5553         dprintk("<-- %s\n", __func__);
5554 }
5555
5556 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5557 {
5558         struct nfs4_sequence_data *calldata = data;
5559         struct nfs_client *clp = calldata->clp;
5560         struct nfs4_sequence_args *args;
5561         struct nfs4_sequence_res *res;
5562
5563         args = task->tk_msg.rpc_argp;
5564         res = task->tk_msg.rpc_resp;
5565
5566         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5567                 return;
5568         rpc_call_start(task);
5569 }
5570
5571 static const struct rpc_call_ops nfs41_sequence_ops = {
5572         .rpc_call_done = nfs41_sequence_call_done,
5573         .rpc_call_prepare = nfs41_sequence_prepare,
5574         .rpc_release = nfs41_sequence_release,
5575 };
5576
5577 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5578 {
5579         struct nfs4_sequence_data *calldata;
5580         struct rpc_message msg = {
5581                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5582                 .rpc_cred = cred,
5583         };
5584         struct rpc_task_setup task_setup_data = {
5585                 .rpc_client = clp->cl_rpcclient,
5586                 .rpc_message = &msg,
5587                 .callback_ops = &nfs41_sequence_ops,
5588                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5589         };
5590
5591         if (!atomic_inc_not_zero(&clp->cl_count))
5592                 return ERR_PTR(-EIO);
5593         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5594         if (calldata == NULL) {
5595                 nfs_put_client(clp);
5596                 return ERR_PTR(-ENOMEM);
5597         }
5598         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5599         msg.rpc_argp = &calldata->args;
5600         msg.rpc_resp = &calldata->res;
5601         calldata->clp = clp;
5602         task_setup_data.callback_data = calldata;
5603
5604         return rpc_run_task(&task_setup_data);
5605 }
5606
5607 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5608 {
5609         struct rpc_task *task;
5610         int ret = 0;
5611
5612         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5613                 return 0;
5614         task = _nfs41_proc_sequence(clp, cred);
5615         if (IS_ERR(task))
5616                 ret = PTR_ERR(task);
5617         else
5618                 rpc_put_task_async(task);
5619         dprintk("<-- %s status=%d\n", __func__, ret);
5620         return ret;
5621 }
5622
5623 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5624 {
5625         struct rpc_task *task;
5626         int ret;
5627
5628         task = _nfs41_proc_sequence(clp, cred);
5629         if (IS_ERR(task)) {
5630                 ret = PTR_ERR(task);
5631                 goto out;
5632         }
5633         ret = rpc_wait_for_completion_task(task);
5634         if (!ret) {
5635                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5636
5637                 if (task->tk_status == 0)
5638                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5639                 ret = task->tk_status;
5640         }
5641         rpc_put_task(task);
5642 out:
5643         dprintk("<-- %s status=%d\n", __func__, ret);
5644         return ret;
5645 }
5646
5647 struct nfs4_reclaim_complete_data {
5648         struct nfs_client *clp;
5649         struct nfs41_reclaim_complete_args arg;
5650         struct nfs41_reclaim_complete_res res;
5651 };
5652
5653 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5654 {
5655         struct nfs4_reclaim_complete_data *calldata = data;
5656
5657         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5658         if (nfs41_setup_sequence(calldata->clp->cl_session,
5659                                 &calldata->arg.seq_args,
5660                                 &calldata->res.seq_res, task))
5661                 return;
5662
5663         rpc_call_start(task);
5664 }
5665
5666 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5667 {
5668         switch(task->tk_status) {
5669         case 0:
5670         case -NFS4ERR_COMPLETE_ALREADY:
5671         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5672                 break;
5673         case -NFS4ERR_DELAY:
5674                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5675                 /* fall through */
5676         case -NFS4ERR_RETRY_UNCACHED_REP:
5677                 return -EAGAIN;
5678         default:
5679                 nfs4_schedule_lease_recovery(clp);
5680         }
5681         return 0;
5682 }
5683
5684 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5685 {
5686         struct nfs4_reclaim_complete_data *calldata = data;
5687         struct nfs_client *clp = calldata->clp;
5688         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5689
5690         dprintk("--> %s\n", __func__);
5691         if (!nfs41_sequence_done(task, res))
5692                 return;
5693
5694         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5695                 rpc_restart_call_prepare(task);
5696                 return;
5697         }
5698         dprintk("<-- %s\n", __func__);
5699 }
5700
5701 static void nfs4_free_reclaim_complete_data(void *data)
5702 {
5703         struct nfs4_reclaim_complete_data *calldata = data;
5704
5705         kfree(calldata);
5706 }
5707
5708 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5709         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5710         .rpc_call_done = nfs4_reclaim_complete_done,
5711         .rpc_release = nfs4_free_reclaim_complete_data,
5712 };
5713
5714 /*
5715  * Issue a global reclaim complete.
5716  */
5717 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5718 {
5719         struct nfs4_reclaim_complete_data *calldata;
5720         struct rpc_task *task;
5721         struct rpc_message msg = {
5722                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5723         };
5724         struct rpc_task_setup task_setup_data = {
5725                 .rpc_client = clp->cl_rpcclient,
5726                 .rpc_message = &msg,
5727                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5728                 .flags = RPC_TASK_ASYNC,
5729         };
5730         int status = -ENOMEM;
5731
5732         dprintk("--> %s\n", __func__);
5733         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5734         if (calldata == NULL)
5735                 goto out;
5736         calldata->clp = clp;
5737         calldata->arg.one_fs = 0;
5738
5739         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5740         msg.rpc_argp = &calldata->arg;
5741         msg.rpc_resp = &calldata->res;
5742         task_setup_data.callback_data = calldata;
5743         task = rpc_run_task(&task_setup_data);
5744         if (IS_ERR(task)) {
5745                 status = PTR_ERR(task);
5746                 goto out;
5747         }
5748         status = nfs4_wait_for_completion_rpc_task(task);
5749         if (status == 0)
5750                 status = task->tk_status;
5751         rpc_put_task(task);
5752         return 0;
5753 out:
5754         dprintk("<-- %s status=%d\n", __func__, status);
5755         return status;
5756 }
5757
5758 static void
5759 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5760 {
5761         struct nfs4_layoutget *lgp = calldata;
5762         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5763
5764         dprintk("--> %s\n", __func__);
5765         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5766          * right now covering the LAYOUTGET we are about to send.
5767          * However, that is not so catastrophic, and there seems
5768          * to be no way to prevent it completely.
5769          */
5770         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5771                                 &lgp->res.seq_res, task))
5772                 return;
5773         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5774                                           NFS_I(lgp->args.inode)->layout,
5775                                           lgp->args.ctx->state)) {
5776                 rpc_exit(task, NFS4_OK);
5777                 return;
5778         }
5779         rpc_call_start(task);
5780 }
5781
5782 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5783 {
5784         struct nfs4_layoutget *lgp = calldata;
5785         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5786
5787         dprintk("--> %s\n", __func__);
5788
5789         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5790                 return;
5791
5792         switch (task->tk_status) {
5793         case 0:
5794                 break;
5795         case -NFS4ERR_LAYOUTTRYLATER:
5796         case -NFS4ERR_RECALLCONFLICT:
5797                 task->tk_status = -NFS4ERR_DELAY;
5798                 /* Fall through */
5799         default:
5800                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5801                         rpc_restart_call_prepare(task);
5802                         return;
5803                 }
5804         }
5805         dprintk("<-- %s\n", __func__);
5806 }
5807
5808 static void nfs4_layoutget_release(void *calldata)
5809 {
5810         struct nfs4_layoutget *lgp = calldata;
5811
5812         dprintk("--> %s\n", __func__);
5813         put_nfs_open_context(lgp->args.ctx);
5814         kfree(calldata);
5815         dprintk("<-- %s\n", __func__);
5816 }
5817
5818 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5819         .rpc_call_prepare = nfs4_layoutget_prepare,
5820         .rpc_call_done = nfs4_layoutget_done,
5821         .rpc_release = nfs4_layoutget_release,
5822 };
5823
5824 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5825 {
5826         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5827         struct rpc_task *task;
5828         struct rpc_message msg = {
5829                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5830                 .rpc_argp = &lgp->args,
5831                 .rpc_resp = &lgp->res,
5832         };
5833         struct rpc_task_setup task_setup_data = {
5834                 .rpc_client = server->client,
5835                 .rpc_message = &msg,
5836                 .callback_ops = &nfs4_layoutget_call_ops,
5837                 .callback_data = lgp,
5838                 .flags = RPC_TASK_ASYNC,
5839         };
5840         int status = 0;
5841
5842         dprintk("--> %s\n", __func__);
5843
5844         lgp->res.layoutp = &lgp->args.layout;
5845         lgp->res.seq_res.sr_slot = NULL;
5846         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5847         task = rpc_run_task(&task_setup_data);
5848         if (IS_ERR(task))
5849                 return PTR_ERR(task);
5850         status = nfs4_wait_for_completion_rpc_task(task);
5851         if (status == 0)
5852                 status = task->tk_status;
5853         if (status == 0)
5854                 status = pnfs_layout_process(lgp);
5855         rpc_put_task(task);
5856         dprintk("<-- %s status=%d\n", __func__, status);
5857         return status;
5858 }
5859
5860 static void
5861 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5862 {
5863         struct nfs4_layoutreturn *lrp = calldata;
5864
5865         dprintk("--> %s\n", __func__);
5866         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5867                                 &lrp->res.seq_res, task))
5868                 return;
5869         rpc_call_start(task);
5870 }
5871
5872 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5873 {
5874         struct nfs4_layoutreturn *lrp = calldata;
5875         struct nfs_server *server;
5876         struct pnfs_layout_hdr *lo = lrp->args.layout;
5877
5878         dprintk("--> %s\n", __func__);
5879
5880         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5881                 return;
5882
5883         server = NFS_SERVER(lrp->args.inode);
5884         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5885                 rpc_restart_call_prepare(task);
5886                 return;
5887         }
5888         spin_lock(&lo->plh_inode->i_lock);
5889         if (task->tk_status == 0) {
5890                 if (lrp->res.lrs_present) {
5891                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5892                 } else
5893                         BUG_ON(!list_empty(&lo->plh_segs));
5894         }
5895         lo->plh_block_lgets--;
5896         spin_unlock(&lo->plh_inode->i_lock);
5897         dprintk("<-- %s\n", __func__);
5898 }
5899
5900 static void nfs4_layoutreturn_release(void *calldata)
5901 {
5902         struct nfs4_layoutreturn *lrp = calldata;
5903
5904         dprintk("--> %s\n", __func__);
5905         put_layout_hdr(lrp->args.layout);
5906         kfree(calldata);
5907         dprintk("<-- %s\n", __func__);
5908 }
5909
5910 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5911         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5912         .rpc_call_done = nfs4_layoutreturn_done,
5913         .rpc_release = nfs4_layoutreturn_release,
5914 };
5915
5916 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5917 {
5918         struct rpc_task *task;
5919         struct rpc_message msg = {
5920                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5921                 .rpc_argp = &lrp->args,
5922                 .rpc_resp = &lrp->res,
5923         };
5924         struct rpc_task_setup task_setup_data = {
5925                 .rpc_client = lrp->clp->cl_rpcclient,
5926                 .rpc_message = &msg,
5927                 .callback_ops = &nfs4_layoutreturn_call_ops,
5928                 .callback_data = lrp,
5929         };
5930         int status;
5931
5932         dprintk("--> %s\n", __func__);
5933         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5934         task = rpc_run_task(&task_setup_data);
5935         if (IS_ERR(task))
5936                 return PTR_ERR(task);
5937         status = task->tk_status;
5938         dprintk("<-- %s status=%d\n", __func__, status);
5939         rpc_put_task(task);
5940         return status;
5941 }
5942
5943 /*
5944  * Retrieve the list of Data Server devices from the MDS.
5945  */
5946 static int _nfs4_getdevicelist(struct nfs_server *server,
5947                                     const struct nfs_fh *fh,
5948                                     struct pnfs_devicelist *devlist)
5949 {
5950         struct nfs4_getdevicelist_args args = {
5951                 .fh = fh,
5952                 .layoutclass = server->pnfs_curr_ld->id,
5953         };
5954         struct nfs4_getdevicelist_res res = {
5955                 .devlist = devlist,
5956         };
5957         struct rpc_message msg = {
5958                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5959                 .rpc_argp = &args,
5960                 .rpc_resp = &res,
5961         };
5962         int status;
5963
5964         dprintk("--> %s\n", __func__);
5965         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5966                                 &res.seq_res, 0);
5967         dprintk("<-- %s status=%d\n", __func__, status);
5968         return status;
5969 }
5970
5971 int nfs4_proc_getdevicelist(struct nfs_server *server,
5972                             const struct nfs_fh *fh,
5973                             struct pnfs_devicelist *devlist)
5974 {
5975         struct nfs4_exception exception = { };
5976         int err;
5977
5978         do {
5979                 err = nfs4_handle_exception(server,
5980                                 _nfs4_getdevicelist(server, fh, devlist),
5981                                 &exception);
5982         } while (exception.retry);
5983
5984         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5985                 err, devlist->num_devs);
5986
5987         return err;
5988 }
5989 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5990
5991 static int
5992 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5993 {
5994         struct nfs4_getdeviceinfo_args args = {
5995                 .pdev = pdev,
5996         };
5997         struct nfs4_getdeviceinfo_res res = {
5998                 .pdev = pdev,
5999         };
6000         struct rpc_message msg = {
6001                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6002                 .rpc_argp = &args,
6003                 .rpc_resp = &res,
6004         };
6005         int status;
6006
6007         dprintk("--> %s\n", __func__);
6008         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6009         dprintk("<-- %s status=%d\n", __func__, status);
6010
6011         return status;
6012 }
6013
6014 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6015 {
6016         struct nfs4_exception exception = { };
6017         int err;
6018
6019         do {
6020                 err = nfs4_handle_exception(server,
6021                                         _nfs4_proc_getdeviceinfo(server, pdev),
6022                                         &exception);
6023         } while (exception.retry);
6024         return err;
6025 }
6026 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6027
6028 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6029 {
6030         struct nfs4_layoutcommit_data *data = calldata;
6031         struct nfs_server *server = NFS_SERVER(data->args.inode);
6032
6033         if (nfs4_setup_sequence(server, &data->args.seq_args,
6034                                 &data->res.seq_res, task))
6035                 return;
6036         rpc_call_start(task);
6037 }
6038
6039 static void
6040 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6041 {
6042         struct nfs4_layoutcommit_data *data = calldata;
6043         struct nfs_server *server = NFS_SERVER(data->args.inode);
6044
6045         if (!nfs4_sequence_done(task, &data->res.seq_res))
6046                 return;
6047
6048         switch (task->tk_status) { /* Just ignore these failures */
6049         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6050         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6051         case NFS4ERR_BADLAYOUT:     /* no layout */
6052         case NFS4ERR_GRACE:         /* loca_recalim always false */
6053                 task->tk_status = 0;
6054         }
6055
6056         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6057                 rpc_restart_call_prepare(task);
6058                 return;
6059         }
6060
6061         if (task->tk_status == 0)
6062                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6063                                                    data->res.fattr);
6064 }
6065
6066 static void nfs4_layoutcommit_release(void *calldata)
6067 {
6068         struct nfs4_layoutcommit_data *data = calldata;
6069         struct pnfs_layout_segment *lseg, *tmp;
6070         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6071
6072         pnfs_cleanup_layoutcommit(data);
6073         /* Matched by references in pnfs_set_layoutcommit */
6074         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6075                 list_del_init(&lseg->pls_lc_list);
6076                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6077                                        &lseg->pls_flags))
6078                         put_lseg(lseg);
6079         }
6080
6081         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6082         smp_mb__after_clear_bit();
6083         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6084
6085         put_rpccred(data->cred);
6086         kfree(data);
6087 }
6088
6089 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6090         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6091         .rpc_call_done = nfs4_layoutcommit_done,
6092         .rpc_release = nfs4_layoutcommit_release,
6093 };
6094
6095 int
6096 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6097 {
6098         struct rpc_message msg = {
6099                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6100                 .rpc_argp = &data->args,
6101                 .rpc_resp = &data->res,
6102                 .rpc_cred = data->cred,
6103         };
6104         struct rpc_task_setup task_setup_data = {
6105                 .task = &data->task,
6106                 .rpc_client = NFS_CLIENT(data->args.inode),
6107                 .rpc_message = &msg,
6108                 .callback_ops = &nfs4_layoutcommit_ops,
6109                 .callback_data = data,
6110                 .flags = RPC_TASK_ASYNC,
6111         };
6112         struct rpc_task *task;
6113         int status = 0;
6114
6115         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6116                 "lbw: %llu inode %lu\n",
6117                 data->task.tk_pid, sync,
6118                 data->args.lastbytewritten,
6119                 data->args.inode->i_ino);
6120
6121         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6122         task = rpc_run_task(&task_setup_data);
6123         if (IS_ERR(task))
6124                 return PTR_ERR(task);
6125         if (sync == false)
6126                 goto out;
6127         status = nfs4_wait_for_completion_rpc_task(task);
6128         if (status != 0)
6129                 goto out;
6130         status = task->tk_status;
6131 out:
6132         dprintk("%s: status %d\n", __func__, status);
6133         rpc_put_task(task);
6134         return status;
6135 }
6136
6137 static int
6138 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6139                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6140 {
6141         struct nfs41_secinfo_no_name_args args = {
6142                 .style = SECINFO_STYLE_CURRENT_FH,
6143         };
6144         struct nfs4_secinfo_res res = {
6145                 .flavors = flavors,
6146         };
6147         struct rpc_message msg = {
6148                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6149                 .rpc_argp = &args,
6150                 .rpc_resp = &res,
6151         };
6152         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6153 }
6154
6155 static int
6156 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6157                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6158 {
6159         struct nfs4_exception exception = { };
6160         int err;
6161         do {
6162                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6163                 switch (err) {
6164                 case 0:
6165                 case -NFS4ERR_WRONGSEC:
6166                 case -NFS4ERR_NOTSUPP:
6167                         break;
6168                 default:
6169                         err = nfs4_handle_exception(server, err, &exception);
6170                 }
6171         } while (exception.retry);
6172         return err;
6173 }
6174
6175 static int
6176 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6177                     struct nfs_fsinfo *info)
6178 {
6179         int err;
6180         struct page *page;
6181         rpc_authflavor_t flavor;
6182         struct nfs4_secinfo_flavors *flavors;
6183
6184         page = alloc_page(GFP_KERNEL);
6185         if (!page) {
6186                 err = -ENOMEM;
6187                 goto out;
6188         }
6189
6190         flavors = page_address(page);
6191         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6192
6193         /*
6194          * Fall back on "guess and check" method if
6195          * the server doesn't support SECINFO_NO_NAME
6196          */
6197         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6198                 err = nfs4_find_root_sec(server, fhandle, info);
6199                 goto out_freepage;
6200         }
6201         if (err)
6202                 goto out_freepage;
6203
6204         flavor = nfs_find_best_sec(flavors);
6205         if (err == 0)
6206                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6207
6208 out_freepage:
6209         put_page(page);
6210         if (err == -EACCES)
6211                 return -EPERM;
6212 out:
6213         return err;
6214 }
6215
6216 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6217 {
6218         int status;
6219         struct nfs41_test_stateid_args args = {
6220                 .stateid = stateid,
6221         };
6222         struct nfs41_test_stateid_res res;
6223         struct rpc_message msg = {
6224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6225                 .rpc_argp = &args,
6226                 .rpc_resp = &res,
6227         };
6228
6229         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6230         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6231
6232         if (status == NFS_OK)
6233                 return res.status;
6234         return status;
6235 }
6236
6237 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6238 {
6239         struct nfs4_exception exception = { };
6240         int err;
6241         do {
6242                 err = nfs4_handle_exception(server,
6243                                 _nfs41_test_stateid(server, stateid),
6244                                 &exception);
6245         } while (exception.retry);
6246         return err;
6247 }
6248
6249 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6250 {
6251         struct nfs41_free_stateid_args args = {
6252                 .stateid = stateid,
6253         };
6254         struct nfs41_free_stateid_res res;
6255         struct rpc_message msg = {
6256                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6257                 .rpc_argp = &args,
6258                 .rpc_resp = &res,
6259         };
6260
6261         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6262         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6263 }
6264
6265 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6266 {
6267         struct nfs4_exception exception = { };
6268         int err;
6269         do {
6270                 err = nfs4_handle_exception(server,
6271                                 _nfs4_free_stateid(server, stateid),
6272                                 &exception);
6273         } while (exception.retry);
6274         return err;
6275 }
6276
6277 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6278                 const nfs4_stateid *s2)
6279 {
6280         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6281                 return false;
6282
6283         if (s1->seqid == s2->seqid)
6284                 return true;
6285         if (s1->seqid == 0 || s2->seqid == 0)
6286                 return true;
6287
6288         return false;
6289 }
6290
6291 #endif /* CONFIG_NFS_V4_1 */
6292
6293 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6294                 const nfs4_stateid *s2)
6295 {
6296         return nfs4_stateid_match(s1, s2);
6297 }
6298
6299
6300 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6301         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6302         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6303         .recover_open   = nfs4_open_reclaim,
6304         .recover_lock   = nfs4_lock_reclaim,
6305         .establish_clid = nfs4_init_clientid,
6306         .get_clid_cred  = nfs4_get_setclientid_cred,
6307 };
6308
6309 #if defined(CONFIG_NFS_V4_1)
6310 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6311         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6312         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6313         .recover_open   = nfs4_open_reclaim,
6314         .recover_lock   = nfs4_lock_reclaim,
6315         .establish_clid = nfs41_init_clientid,
6316         .get_clid_cred  = nfs4_get_exchange_id_cred,
6317         .reclaim_complete = nfs41_proc_reclaim_complete,
6318 };
6319 #endif /* CONFIG_NFS_V4_1 */
6320
6321 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6322         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6323         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6324         .recover_open   = nfs4_open_expired,
6325         .recover_lock   = nfs4_lock_expired,
6326         .establish_clid = nfs4_init_clientid,
6327         .get_clid_cred  = nfs4_get_setclientid_cred,
6328 };
6329
6330 #if defined(CONFIG_NFS_V4_1)
6331 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6332         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6333         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6334         .recover_open   = nfs41_open_expired,
6335         .recover_lock   = nfs41_lock_expired,
6336         .establish_clid = nfs41_init_clientid,
6337         .get_clid_cred  = nfs4_get_exchange_id_cred,
6338 };
6339 #endif /* CONFIG_NFS_V4_1 */
6340
6341 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6342         .sched_state_renewal = nfs4_proc_async_renew,
6343         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6344         .renew_lease = nfs4_proc_renew,
6345 };
6346
6347 #if defined(CONFIG_NFS_V4_1)
6348 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6349         .sched_state_renewal = nfs41_proc_async_sequence,
6350         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6351         .renew_lease = nfs4_proc_sequence,
6352 };
6353 #endif
6354
6355 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6356         .minor_version = 0,
6357         .call_sync = _nfs4_call_sync,
6358         .match_stateid = nfs4_match_stateid,
6359         .find_root_sec = nfs4_find_root_sec,
6360         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6361         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6362         .state_renewal_ops = &nfs40_state_renewal_ops,
6363 };
6364
6365 #if defined(CONFIG_NFS_V4_1)
6366 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6367         .minor_version = 1,
6368         .call_sync = _nfs4_call_sync_session,
6369         .match_stateid = nfs41_match_stateid,
6370         .find_root_sec = nfs41_find_root_sec,
6371         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6372         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6373         .state_renewal_ops = &nfs41_state_renewal_ops,
6374 };
6375 #endif
6376
6377 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6378         [0] = &nfs_v4_0_minor_ops,
6379 #if defined(CONFIG_NFS_V4_1)
6380         [1] = &nfs_v4_1_minor_ops,
6381 #endif
6382 };
6383
6384 static const struct inode_operations nfs4_file_inode_operations = {
6385         .permission     = nfs_permission,
6386         .getattr        = nfs_getattr,
6387         .setattr        = nfs_setattr,
6388         .getxattr       = generic_getxattr,
6389         .setxattr       = generic_setxattr,
6390         .listxattr      = generic_listxattr,
6391         .removexattr    = generic_removexattr,
6392 };
6393
6394 const struct nfs_rpc_ops nfs_v4_clientops = {
6395         .version        = 4,                    /* protocol version */
6396         .dentry_ops     = &nfs4_dentry_operations,
6397         .dir_inode_ops  = &nfs4_dir_inode_operations,
6398         .file_inode_ops = &nfs4_file_inode_operations,
6399         .file_ops       = &nfs4_file_operations,
6400         .getroot        = nfs4_proc_get_root,
6401         .getattr        = nfs4_proc_getattr,
6402         .setattr        = nfs4_proc_setattr,
6403         .lookup         = nfs4_proc_lookup,
6404         .access         = nfs4_proc_access,
6405         .readlink       = nfs4_proc_readlink,
6406         .create         = nfs4_proc_create,
6407         .remove         = nfs4_proc_remove,
6408         .unlink_setup   = nfs4_proc_unlink_setup,
6409         .unlink_done    = nfs4_proc_unlink_done,
6410         .rename         = nfs4_proc_rename,
6411         .rename_setup   = nfs4_proc_rename_setup,
6412         .rename_done    = nfs4_proc_rename_done,
6413         .link           = nfs4_proc_link,
6414         .symlink        = nfs4_proc_symlink,
6415         .mkdir          = nfs4_proc_mkdir,
6416         .rmdir          = nfs4_proc_remove,
6417         .readdir        = nfs4_proc_readdir,
6418         .mknod          = nfs4_proc_mknod,
6419         .statfs         = nfs4_proc_statfs,
6420         .fsinfo         = nfs4_proc_fsinfo,
6421         .pathconf       = nfs4_proc_pathconf,
6422         .set_capabilities = nfs4_server_capabilities,
6423         .decode_dirent  = nfs4_decode_dirent,
6424         .read_setup     = nfs4_proc_read_setup,
6425         .read_done      = nfs4_read_done,
6426         .write_setup    = nfs4_proc_write_setup,
6427         .write_done     = nfs4_write_done,
6428         .commit_setup   = nfs4_proc_commit_setup,
6429         .commit_done    = nfs4_commit_done,
6430         .lock           = nfs4_proc_lock,
6431         .clear_acl_cache = nfs4_zap_acl_attr,
6432         .close_context  = nfs4_close_context,
6433         .open_context   = nfs4_atomic_open,
6434         .init_client    = nfs4_init_client,
6435         .secinfo        = nfs4_proc_secinfo,
6436 };
6437
6438 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6439         .prefix = XATTR_NAME_NFSV4_ACL,
6440         .list   = nfs4_xattr_list_nfs4_acl,
6441         .get    = nfs4_xattr_get_nfs4_acl,
6442         .set    = nfs4_xattr_set_nfs4_acl,
6443 };
6444
6445 const struct xattr_handler *nfs4_xattr_handlers[] = {
6446         &nfs4_xattr_nfs4_acl_handler,
6447         NULL
6448 };
6449
6450 module_param(max_session_slots, ushort, 0644);
6451 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6452                 "requests the client will negotiate");
6453
6454 /*
6455  * Local variables:
6456  *  c-basic-offset: 8
6457  * End:
6458  */