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