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/sunrpc/bc_xprt.h>
  56 #include <linux/xattr.h>
  57 #include <linux/utsname.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65
  66 #define NFSDBG_FACILITY         NFSDBG_PROC
  67
  68 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  69 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  70
  71 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  72
  73 struct nfs4_opendata;
  74 static int _nfs4_proc_open(struct nfs4_opendata *data);
  75 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  76 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  77 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  78 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  79 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  80                             struct nfs_fattr *fattr, struct iattr *sattr,
  81                             struct nfs4_state *state);
  82 #ifdef CONFIG_NFS_V4_1
  83 static int nfs41_test_stateid(struct nfs_server *, struct nfs4_state *);
  84 static int nfs41_free_stateid(struct nfs_server *, struct nfs4_state *);
  85 #endif
  86 /* Prevent leaks of NFSv4 errors into userland */
  87 static int nfs4_map_errors(int err)
  88 {
  89         if (err >= -1000)
  90                 return err;
  91         switch (err) {
  92         case -NFS4ERR_RESOURCE:
  93                 return -EREMOTEIO;
  94         case -NFS4ERR_WRONGSEC:
  95                 return -EPERM;
  96         case -NFS4ERR_BADOWNER:
  97         case -NFS4ERR_BADNAME:
  98                 return -EINVAL;
  99         default:
 100                 dprintk("%s could not handle NFSv4 error %d\n",
 101                                 __func__, -err);
 102                 break;
 103         }
 104         return -EIO;
 105 }
 106
 107 /*
 108  * This is our standard bitmap for GETATTR requests.
 109  */
 110 const u32 nfs4_fattr_bitmap[2] = {
 111         FATTR4_WORD0_TYPE
 112         | FATTR4_WORD0_CHANGE
 113         | FATTR4_WORD0_SIZE
 114         | FATTR4_WORD0_FSID
 115         | FATTR4_WORD0_FILEID,
 116         FATTR4_WORD1_MODE
 117         | FATTR4_WORD1_NUMLINKS
 118         | FATTR4_WORD1_OWNER
 119         | FATTR4_WORD1_OWNER_GROUP
 120         | FATTR4_WORD1_RAWDEV
 121         | FATTR4_WORD1_SPACE_USED
 122         | FATTR4_WORD1_TIME_ACCESS
 123         | FATTR4_WORD1_TIME_METADATA
 124         | FATTR4_WORD1_TIME_MODIFY
 125 };
 126
 127 const u32 nfs4_statfs_bitmap[2] = {
 128         FATTR4_WORD0_FILES_AVAIL
 129         | FATTR4_WORD0_FILES_FREE
 130         | FATTR4_WORD0_FILES_TOTAL,
 131         FATTR4_WORD1_SPACE_AVAIL
 132         | FATTR4_WORD1_SPACE_FREE
 133         | FATTR4_WORD1_SPACE_TOTAL
 134 };
 135
 136 const u32 nfs4_pathconf_bitmap[2] = {
 137         FATTR4_WORD0_MAXLINK
 138         | FATTR4_WORD0_MAXNAME,
 139         0
 140 };
 141
 142 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 143                         | FATTR4_WORD0_MAXREAD
 144                         | FATTR4_WORD0_MAXWRITE
 145                         | FATTR4_WORD0_LEASE_TIME,
 146                         FATTR4_WORD1_TIME_DELTA
 147                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 148                         FATTR4_WORD2_LAYOUT_BLKSIZE
 149 };
 150
 151 const u32 nfs4_fs_locations_bitmap[2] = {
 152         FATTR4_WORD0_TYPE
 153         | FATTR4_WORD0_CHANGE
 154         | FATTR4_WORD0_SIZE
 155         | FATTR4_WORD0_FSID
 156         | FATTR4_WORD0_FILEID
 157         | FATTR4_WORD0_FS_LOCATIONS,
 158         FATTR4_WORD1_MODE
 159         | FATTR4_WORD1_NUMLINKS
 160         | FATTR4_WORD1_OWNER
 161         | FATTR4_WORD1_OWNER_GROUP
 162         | FATTR4_WORD1_RAWDEV
 163         | FATTR4_WORD1_SPACE_USED
 164         | FATTR4_WORD1_TIME_ACCESS
 165         | FATTR4_WORD1_TIME_METADATA
 166         | FATTR4_WORD1_TIME_MODIFY
 167         | FATTR4_WORD1_MOUNTED_ON_FILEID
 168 };
 169
 170 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 171                 struct nfs4_readdir_arg *readdir)
 172 {
 173         __be32 *start, *p;
 174
 175         BUG_ON(readdir->count < 80);
 176         if (cookie > 2) {
 177                 readdir->cookie = cookie;
 178                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 179                 return;
 180         }
 181
 182         readdir->cookie = 0;
 183         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 184         if (cookie == 2)
 185                 return;
 186
 187         /*
 188          * NFSv4 servers do not return entries for '.' and '..'
 189          * Therefore, we fake these entries here.  We let '.'
 190          * have cookie 0 and '..' have cookie 1.  Note that
 191          * when talking to the server, we always send cookie 0
 192          * instead of 1 or 2.
 193          */
 194         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 195
 196         if (cookie == 0) {
 197                 *p++ = xdr_one;                                  /* next */
 198                 *p++ = xdr_zero;                   /* cookie, first word */
 199                 *p++ = xdr_one;                   /* cookie, second word */
 200                 *p++ = xdr_one;                             /* entry len */
 201                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 202                 p++;
 203                 *p++ = xdr_one;                         /* bitmap length */
 204                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 205                 *p++ = htonl(8);              /* attribute buffer length */
 206                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 207         }
 208
 209         *p++ = xdr_one;                                  /* next */
 210         *p++ = xdr_zero;                   /* cookie, first word */
 211         *p++ = xdr_two;                   /* cookie, second word */
 212         *p++ = xdr_two;                             /* entry len */
 213         memcpy(p, "..\0\0", 4);                         /* entry */
 214         p++;
 215         *p++ = xdr_one;                         /* bitmap length */
 216         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 217         *p++ = htonl(8);              /* attribute buffer length */
 218         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 219
 220         readdir->pgbase = (char *)p - (char *)start;
 221         readdir->count -= readdir->pgbase;
 222         kunmap_atomic(start, KM_USER0);
 223 }
 224
 225 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 226 {
 227         int res;
 228
 229         might_sleep();
 230
 231         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 232                         nfs_wait_bit_killable, TASK_KILLABLE);
 233         return res;
 234 }
 235
 236 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 237 {
 238         int res = 0;
 239
 240         might_sleep();
 241
 242         if (*timeout <= 0)
 243                 *timeout = NFS4_POLL_RETRY_MIN;
 244         if (*timeout > NFS4_POLL_RETRY_MAX)
 245                 *timeout = NFS4_POLL_RETRY_MAX;
 246         schedule_timeout_killable(*timeout);
 247         if (fatal_signal_pending(current))
 248                 res = -ERESTARTSYS;
 249         *timeout <<= 1;
 250         return res;
 251 }
 252
 253 /* This is the error handling routine for processes that are allowed
 254  * to sleep.
 255  */
 256 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 257 {
 258         struct nfs_client *clp = server->nfs_client;
 259         struct nfs4_state *state = exception->state;
 260         int ret = errorcode;
 261
 262         exception->retry = 0;
 263         switch(errorcode) {
 264                 case 0:
 265                         return 0;
 266                 case -NFS4ERR_ADMIN_REVOKED:
 267                 case -NFS4ERR_BAD_STATEID:
 268                 case -NFS4ERR_OPENMODE:
 269                         if (state == NULL)
 270                                 break;
 271                         nfs4_schedule_stateid_recovery(server, state);
 272                         goto wait_on_recovery;
 273                 case -NFS4ERR_EXPIRED:
 274                         if (state != NULL)
 275                                 nfs4_schedule_stateid_recovery(server, state);
 276                 case -NFS4ERR_STALE_STATEID:
 277                 case -NFS4ERR_STALE_CLIENTID:
 278                         nfs4_schedule_lease_recovery(clp);
 279                         goto wait_on_recovery;
 280 #if defined(CONFIG_NFS_V4_1)
 281                 case -NFS4ERR_BADSESSION:
 282                 case -NFS4ERR_BADSLOT:
 283                 case -NFS4ERR_BAD_HIGH_SLOT:
 284                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 285                 case -NFS4ERR_DEADSESSION:
 286                 case -NFS4ERR_SEQ_FALSE_RETRY:
 287                 case -NFS4ERR_SEQ_MISORDERED:
 288                         dprintk("%s ERROR: %d Reset session\n", __func__,
 289                                 errorcode);
 290                         nfs4_schedule_session_recovery(clp->cl_session);
 291                         exception->retry = 1;
 292                         break;
 293 #endif /* defined(CONFIG_NFS_V4_1) */
 294                 case -NFS4ERR_FILE_OPEN:
 295                         if (exception->timeout > HZ) {
 296                                 /* We have retried a decent amount, time to
 297                                  * fail
 298                                  */
 299                                 ret = -EBUSY;
 300                                 break;
 301                         }
 302                 case -NFS4ERR_GRACE:
 303                 case -NFS4ERR_DELAY:
 304                 case -EKEYEXPIRED:
 305                         ret = nfs4_delay(server->client, &exception->timeout);
 306                         if (ret != 0)
 307                                 break;
 308                 case -NFS4ERR_RETRY_UNCACHED_REP:
 309                 case -NFS4ERR_OLD_STATEID:
 310                         exception->retry = 1;
 311                         break;
 312                 case -NFS4ERR_BADOWNER:
 313                         /* The following works around a Linux server bug! */
 314                 case -NFS4ERR_BADNAME:
 315                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 316                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 317                                 exception->retry = 1;
 318                                 printk(KERN_WARNING "NFS: v4 server %s "
 319                                                 "does not accept raw "
 320                                                 "uid/gids. "
 321                                                 "Reenabling the idmapper.\n",
 322                                                 server->nfs_client->cl_hostname);
 323                         }
 324         }
 325         /* We failed to handle the error */
 326         return nfs4_map_errors(ret);
 327 wait_on_recovery:
 328         ret = nfs4_wait_clnt_recover(clp);
 329         if (ret == 0)
 330                 exception->retry = 1;
 331         return ret;
 332 }
 333
 334
 335 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 336 {
 337         spin_lock(&clp->cl_lock);
 338         if (time_before(clp->cl_last_renewal,timestamp))
 339                 clp->cl_last_renewal = timestamp;
 340         spin_unlock(&clp->cl_lock);
 341 }
 342
 343 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 344 {
 345         do_renew_lease(server->nfs_client, timestamp);
 346 }
 347
 348 #if defined(CONFIG_NFS_V4_1)
 349
 350 /*
 351  * nfs4_free_slot - free a slot and efficiently update slot table.
 352  *
 353  * freeing a slot is trivially done by clearing its respective bit
 354  * in the bitmap.
 355  * If the freed slotid equals highest_used_slotid we want to update it
 356  * so that the server would be able to size down the slot table if needed,
 357  * otherwise we know that the highest_used_slotid is still in use.
 358  * When updating highest_used_slotid there may be "holes" in the bitmap
 359  * so we need to scan down from highest_used_slotid to 0 looking for the now
 360  * highest slotid in use.
 361  * If none found, highest_used_slotid is set to -1.
 362  *
 363  * Must be called while holding tbl->slot_tbl_lock
 364  */
 365 static void
 366 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
 367 {
 368         int slotid = free_slotid;
 369
 370         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 371         /* clear used bit in bitmap */
 372         __clear_bit(slotid, tbl->used_slots);
 373
 374         /* update highest_used_slotid when it is freed */
 375         if (slotid == tbl->highest_used_slotid) {
 376                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 377                 if (slotid < tbl->max_slots)
 378                         tbl->highest_used_slotid = slotid;
 379                 else
 380                         tbl->highest_used_slotid = -1;
 381         }
 382         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 383                 free_slotid, tbl->highest_used_slotid);
 384 }
 385
 386 /*
 387  * Signal state manager thread if session fore channel is drained
 388  */
 389 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 390 {
 391         struct rpc_task *task;
 392
 393         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 394                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 395                 if (task)
 396                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 397                 return;
 398         }
 399
 400         if (ses->fc_slot_table.highest_used_slotid != -1)
 401                 return;
 402
 403         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 404         complete(&ses->fc_slot_table.complete);
 405 }
 406
 407 /*
 408  * Signal state manager thread if session back channel is drained
 409  */
 410 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 411 {
 412         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 413             ses->bc_slot_table.highest_used_slotid != -1)
 414                 return;
 415         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 416         complete(&ses->bc_slot_table.complete);
 417 }
 418
 419 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 420 {
 421         struct nfs4_slot_table *tbl;
 422
 423         tbl = &res->sr_session->fc_slot_table;
 424         if (!res->sr_slot) {
 425                 /* just wake up the next guy waiting since
 426                  * we may have not consumed a slot after all */
 427                 dprintk("%s: No slot\n", __func__);
 428                 return;
 429         }
 430
 431         spin_lock(&tbl->slot_tbl_lock);
 432         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 433         nfs4_check_drain_fc_complete(res->sr_session);
 434         spin_unlock(&tbl->slot_tbl_lock);
 435         res->sr_slot = NULL;
 436 }
 437
 438 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 439 {
 440         unsigned long timestamp;
 441         struct nfs_client *clp;
 442
 443         /*
 444          * sr_status remains 1 if an RPC level error occurred. The server
 445          * may or may not have processed the sequence operation..
 446          * Proceed as if the server received and processed the sequence
 447          * operation.
 448          */
 449         if (res->sr_status == 1)
 450                 res->sr_status = NFS_OK;
 451
 452         /* don't increment the sequence number if the task wasn't sent */
 453         if (!RPC_WAS_SENT(task))
 454                 goto out;
 455
 456         /* Check the SEQUENCE operation status */
 457         switch (res->sr_status) {
 458         case 0:
 459                 /* Update the slot's sequence and clientid lease timer */
 460                 ++res->sr_slot->seq_nr;
 461                 timestamp = res->sr_renewal_time;
 462                 clp = res->sr_session->clp;
 463                 do_renew_lease(clp, timestamp);
 464                 /* Check sequence flags */
 465                 if (res->sr_status_flags != 0)
 466                         nfs4_schedule_lease_recovery(clp);
 467                 break;
 468         case -NFS4ERR_DELAY:
 469                 /* The server detected a resend of the RPC call and
 470                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 471                  * of RFC5661.
 472                  */
 473                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 474                         __func__,
 475                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 476                         res->sr_slot->seq_nr);
 477                 goto out_retry;
 478         default:
 479                 /* Just update the slot sequence no. */
 480                 ++res->sr_slot->seq_nr;
 481         }
 482 out:
 483         /* The session may be reset by one of the error handlers. */
 484         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 485         nfs41_sequence_free_slot(res);
 486         return 1;
 487 out_retry:
 488         if (!rpc_restart_call(task))
 489                 goto out;
 490         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 491         return 0;
 492 }
 493
 494 static int nfs4_sequence_done(struct rpc_task *task,
 495                                struct nfs4_sequence_res *res)
 496 {
 497         if (res->sr_session == NULL)
 498                 return 1;
 499         return nfs41_sequence_done(task, res);
 500 }
 501
 502 /*
 503  * nfs4_find_slot - efficiently look for a free slot
 504  *
 505  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 506  * If found, we mark the slot as used, update the highest_used_slotid,
 507  * and respectively set up the sequence operation args.
 508  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 509  *
 510  * Note: must be called with under the slot_tbl_lock.
 511  */
 512 static u8
 513 nfs4_find_slot(struct nfs4_slot_table *tbl)
 514 {
 515         int slotid;
 516         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 517         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 518
 519         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 520                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 521                 tbl->max_slots);
 522         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 523         if (slotid >= tbl->max_slots)
 524                 goto out;
 525         __set_bit(slotid, tbl->used_slots);
 526         if (slotid > tbl->highest_used_slotid)
 527                 tbl->highest_used_slotid = slotid;
 528         ret_id = slotid;
 529 out:
 530         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 531                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 532         return ret_id;
 533 }
 534
 535 int nfs41_setup_sequence(struct nfs4_session *session,
 536                                 struct nfs4_sequence_args *args,
 537                                 struct nfs4_sequence_res *res,
 538                                 int cache_reply,
 539                                 struct rpc_task *task)
 540 {
 541         struct nfs4_slot *slot;
 542         struct nfs4_slot_table *tbl;
 543         u8 slotid;
 544
 545         dprintk("--> %s\n", __func__);
 546         /* slot already allocated? */
 547         if (res->sr_slot != NULL)
 548                 return 0;
 549
 550         tbl = &session->fc_slot_table;
 551
 552         spin_lock(&tbl->slot_tbl_lock);
 553         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 554             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 555                 /* The state manager will wait until the slot table is empty */
 556                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 557                 spin_unlock(&tbl->slot_tbl_lock);
 558                 dprintk("%s session is draining\n", __func__);
 559                 return -EAGAIN;
 560         }
 561
 562         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 563             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 564                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 565                 spin_unlock(&tbl->slot_tbl_lock);
 566                 dprintk("%s enforce FIFO order\n", __func__);
 567                 return -EAGAIN;
 568         }
 569
 570         slotid = nfs4_find_slot(tbl);
 571         if (slotid == NFS4_MAX_SLOT_TABLE) {
 572                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 573                 spin_unlock(&tbl->slot_tbl_lock);
 574                 dprintk("<-- %s: no free slots\n", __func__);
 575                 return -EAGAIN;
 576         }
 577         spin_unlock(&tbl->slot_tbl_lock);
 578
 579         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 580         slot = tbl->slots + slotid;
 581         args->sa_session = session;
 582         args->sa_slotid = slotid;
 583         args->sa_cache_this = cache_reply;
 584
 585         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 586
 587         res->sr_session = session;
 588         res->sr_slot = slot;
 589         res->sr_renewal_time = jiffies;
 590         res->sr_status_flags = 0;
 591         /*
 592          * sr_status is only set in decode_sequence, and so will remain
 593          * set to 1 if an rpc level failure occurs.
 594          */
 595         res->sr_status = 1;
 596         return 0;
 597 }
 598 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 599
 600 int nfs4_setup_sequence(const struct nfs_server *server,
 601                         struct nfs4_sequence_args *args,
 602                         struct nfs4_sequence_res *res,
 603                         int cache_reply,
 604                         struct rpc_task *task)
 605 {
 606         struct nfs4_session *session = nfs4_get_session(server);
 607         int ret = 0;
 608
 609         if (session == NULL) {
 610                 args->sa_session = NULL;
 611                 res->sr_session = NULL;
 612                 goto out;
 613         }
 614
 615         dprintk("--> %s clp %p session %p sr_slot %td\n",
 616                 __func__, session->clp, session, res->sr_slot ?
 617                         res->sr_slot - session->fc_slot_table.slots : -1);
 618
 619         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 620                                    task);
 621 out:
 622         dprintk("<-- %s status=%d\n", __func__, ret);
 623         return ret;
 624 }
 625
 626 struct nfs41_call_sync_data {
 627         const struct nfs_server *seq_server;
 628         struct nfs4_sequence_args *seq_args;
 629         struct nfs4_sequence_res *seq_res;
 630         int cache_reply;
 631 };
 632
 633 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 634 {
 635         struct nfs41_call_sync_data *data = calldata;
 636
 637         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 638
 639         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 640                                 data->seq_res, data->cache_reply, task))
 641                 return;
 642         rpc_call_start(task);
 643 }
 644
 645 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 646 {
 647         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 648         nfs41_call_sync_prepare(task, calldata);
 649 }
 650
 651 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 652 {
 653         struct nfs41_call_sync_data *data = calldata;
 654
 655         nfs41_sequence_done(task, data->seq_res);
 656 }
 657
 658 struct rpc_call_ops nfs41_call_sync_ops = {
 659         .rpc_call_prepare = nfs41_call_sync_prepare,
 660         .rpc_call_done = nfs41_call_sync_done,
 661 };
 662
 663 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 664         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 665         .rpc_call_done = nfs41_call_sync_done,
 666 };
 667
 668 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 669                                    struct nfs_server *server,
 670                                    struct rpc_message *msg,
 671                                    struct nfs4_sequence_args *args,
 672                                    struct nfs4_sequence_res *res,
 673                                    int cache_reply,
 674                                    int privileged)
 675 {
 676         int ret;
 677         struct rpc_task *task;
 678         struct nfs41_call_sync_data data = {
 679                 .seq_server = server,
 680                 .seq_args = args,
 681                 .seq_res = res,
 682                 .cache_reply = cache_reply,
 683         };
 684         struct rpc_task_setup task_setup = {
 685                 .rpc_client = clnt,
 686                 .rpc_message = msg,
 687                 .callback_ops = &nfs41_call_sync_ops,
 688                 .callback_data = &data
 689         };
 690
 691         res->sr_slot = NULL;
 692         if (privileged)
 693                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 694         task = rpc_run_task(&task_setup);
 695         if (IS_ERR(task))
 696                 ret = PTR_ERR(task);
 697         else {
 698                 ret = task->tk_status;
 699                 rpc_put_task(task);
 700         }
 701         return ret;
 702 }
 703
 704 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 705                             struct nfs_server *server,
 706                             struct rpc_message *msg,
 707                             struct nfs4_sequence_args *args,
 708                             struct nfs4_sequence_res *res,
 709                             int cache_reply)
 710 {
 711         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 712 }
 713
 714 #else
 715 static int nfs4_sequence_done(struct rpc_task *task,
 716                                struct nfs4_sequence_res *res)
 717 {
 718         return 1;
 719 }
 720 #endif /* CONFIG_NFS_V4_1 */
 721
 722 int _nfs4_call_sync(struct rpc_clnt *clnt,
 723                     struct nfs_server *server,
 724                     struct rpc_message *msg,
 725                     struct nfs4_sequence_args *args,
 726                     struct nfs4_sequence_res *res,
 727                     int cache_reply)
 728 {
 729         args->sa_session = res->sr_session = NULL;
 730         return rpc_call_sync(clnt, msg, 0);
 731 }
 732
 733 static inline
 734 int nfs4_call_sync(struct rpc_clnt *clnt,
 735                    struct nfs_server *server,
 736                    struct rpc_message *msg,
 737                    struct nfs4_sequence_args *args,
 738                    struct nfs4_sequence_res *res,
 739                    int cache_reply)
 740 {
 741         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 742                                                 args, res, cache_reply);
 743 }
 744
 745 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 746 {
 747         struct nfs_inode *nfsi = NFS_I(dir);
 748
 749         spin_lock(&dir->i_lock);
 750         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 751         if (!cinfo->atomic || cinfo->before != dir->i_version)
 752                 nfs_force_lookup_revalidate(dir);
 753         dir->i_version = cinfo->after;
 754         spin_unlock(&dir->i_lock);
 755 }
 756
 757 struct nfs4_opendata {
 758         struct kref kref;
 759         struct nfs_openargs o_arg;
 760         struct nfs_openres o_res;
 761         struct nfs_open_confirmargs c_arg;
 762         struct nfs_open_confirmres c_res;
 763         struct nfs_fattr f_attr;
 764         struct nfs_fattr dir_attr;
 765         struct dentry *dir;
 766         struct dentry *dentry;
 767         struct nfs4_state_owner *owner;
 768         struct nfs4_state *state;
 769         struct iattr attrs;
 770         unsigned long timestamp;
 771         unsigned int rpc_done : 1;
 772         int rpc_status;
 773         int cancelled;
 774 };
 775
 776
 777 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 778 {
 779         p->o_res.f_attr = &p->f_attr;
 780         p->o_res.dir_attr = &p->dir_attr;
 781         p->o_res.seqid = p->o_arg.seqid;
 782         p->c_res.seqid = p->c_arg.seqid;
 783         p->o_res.server = p->o_arg.server;
 784         nfs_fattr_init(&p->f_attr);
 785         nfs_fattr_init(&p->dir_attr);
 786 }
 787
 788 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 789                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 790                 const struct iattr *attrs,
 791                 gfp_t gfp_mask)
 792 {
 793         struct dentry *parent = dget_parent(dentry);
 794         struct inode *dir = parent->d_inode;
 795         struct nfs_server *server = NFS_SERVER(dir);
 796         struct nfs4_opendata *p;
 797
 798         p = kzalloc(sizeof(*p), gfp_mask);
 799         if (p == NULL)
 800                 goto err;
 801         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 802         if (p->o_arg.seqid == NULL)
 803                 goto err_free;
 804         nfs_sb_active(dentry->d_sb);
 805         p->dentry = dget(dentry);
 806         p->dir = parent;
 807         p->owner = sp;
 808         atomic_inc(&sp->so_count);
 809         p->o_arg.fh = NFS_FH(dir);
 810         p->o_arg.open_flags = flags;
 811         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 812         p->o_arg.clientid = server->nfs_client->cl_clientid;
 813         p->o_arg.id = sp->so_owner_id.id;
 814         p->o_arg.name = &dentry->d_name;
 815         p->o_arg.server = server;
 816         p->o_arg.bitmask = server->attr_bitmask;
 817         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 818         if (flags & O_CREAT) {
 819                 u32 *s;
 820
 821                 p->o_arg.u.attrs = &p->attrs;
 822                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 823                 s = (u32 *) p->o_arg.u.verifier.data;
 824                 s[0] = jiffies;
 825                 s[1] = current->pid;
 826         }
 827         p->c_arg.fh = &p->o_res.fh;
 828         p->c_arg.stateid = &p->o_res.stateid;
 829         p->c_arg.seqid = p->o_arg.seqid;
 830         nfs4_init_opendata_res(p);
 831         kref_init(&p->kref);
 832         return p;
 833 err_free:
 834         kfree(p);
 835 err:
 836         dput(parent);
 837         return NULL;
 838 }
 839
 840 static void nfs4_opendata_free(struct kref *kref)
 841 {
 842         struct nfs4_opendata *p = container_of(kref,
 843                         struct nfs4_opendata, kref);
 844         struct super_block *sb = p->dentry->d_sb;
 845
 846         nfs_free_seqid(p->o_arg.seqid);
 847         if (p->state != NULL)
 848                 nfs4_put_open_state(p->state);
 849         nfs4_put_state_owner(p->owner);
 850         dput(p->dir);
 851         dput(p->dentry);
 852         nfs_sb_deactive(sb);
 853         kfree(p);
 854 }
 855
 856 static void nfs4_opendata_put(struct nfs4_opendata *p)
 857 {
 858         if (p != NULL)
 859                 kref_put(&p->kref, nfs4_opendata_free);
 860 }
 861
 862 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 863 {
 864         int ret;
 865
 866         ret = rpc_wait_for_completion_task(task);
 867         return ret;
 868 }
 869
 870 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 871 {
 872         int ret = 0;
 873
 874         if (open_mode & O_EXCL)
 875                 goto out;
 876         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 877                 case FMODE_READ:
 878                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 879                                 && state->n_rdonly != 0;
 880                         break;
 881                 case FMODE_WRITE:
 882                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 883                                 && state->n_wronly != 0;
 884                         break;
 885                 case FMODE_READ|FMODE_WRITE:
 886                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 887                                 && state->n_rdwr != 0;
 888         }
 889 out:
 890         return ret;
 891 }
 892
 893 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 894 {
 895         if (delegation == NULL)
 896                 return 0;
 897         if ((delegation->type & fmode) != fmode)
 898                 return 0;
 899         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 900                 return 0;
 901         nfs_mark_delegation_referenced(delegation);
 902         return 1;
 903 }
 904
 905 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 906 {
 907         switch (fmode) {
 908                 case FMODE_WRITE:
 909                         state->n_wronly++;
 910                         break;
 911                 case FMODE_READ:
 912                         state->n_rdonly++;
 913                         break;
 914                 case FMODE_READ|FMODE_WRITE:
 915                         state->n_rdwr++;
 916         }
 917         nfs4_state_set_mode_locked(state, state->state | fmode);
 918 }
 919
 920 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 921 {
 922         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 923                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 924         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 925         switch (fmode) {
 926                 case FMODE_READ:
 927                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 928                         break;
 929                 case FMODE_WRITE:
 930                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 931                         break;
 932                 case FMODE_READ|FMODE_WRITE:
 933                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 934         }
 935 }
 936
 937 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 938 {
 939         write_seqlock(&state->seqlock);
 940         nfs_set_open_stateid_locked(state, stateid, fmode);
 941         write_sequnlock(&state->seqlock);
 942 }
 943
 944 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 945 {
 946         /*
 947          * Protect the call to nfs4_state_set_mode_locked and
 948          * serialise the stateid update
 949          */
 950         write_seqlock(&state->seqlock);
 951         if (deleg_stateid != NULL) {
 952                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 953                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 954         }
 955         if (open_stateid != NULL)
 956                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 957         write_sequnlock(&state->seqlock);
 958         spin_lock(&state->owner->so_lock);
 959         update_open_stateflags(state, fmode);
 960         spin_unlock(&state->owner->so_lock);
 961 }
 962
 963 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 964 {
 965         struct nfs_inode *nfsi = NFS_I(state->inode);
 966         struct nfs_delegation *deleg_cur;
 967         int ret = 0;
 968
 969         fmode &= (FMODE_READ|FMODE_WRITE);
 970
 971         rcu_read_lock();
 972         deleg_cur = rcu_dereference(nfsi->delegation);
 973         if (deleg_cur == NULL)
 974                 goto no_delegation;
 975
 976         spin_lock(&deleg_cur->lock);
 977         if (nfsi->delegation != deleg_cur ||
 978             (deleg_cur->type & fmode) != fmode)
 979                 goto no_delegation_unlock;
 980
 981         if (delegation == NULL)
 982                 delegation = &deleg_cur->stateid;
 983         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 984                 goto no_delegation_unlock;
 985
 986         nfs_mark_delegation_referenced(deleg_cur);
 987         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 988         ret = 1;
 989 no_delegation_unlock:
 990         spin_unlock(&deleg_cur->lock);
 991 no_delegation:
 992         rcu_read_unlock();
 993
 994         if (!ret && open_stateid != NULL) {
 995                 __update_open_stateid(state, open_stateid, NULL, fmode);
 996                 ret = 1;
 997         }
 998
 999         return ret;
1000 }
1001
1002
1003 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1004 {
1005         struct nfs_delegation *delegation;
1006
1007         rcu_read_lock();
1008         delegation = rcu_dereference(NFS_I(inode)->delegation);
1009         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1010                 rcu_read_unlock();
1011                 return;
1012         }
1013         rcu_read_unlock();
1014         nfs_inode_return_delegation(inode);
1015 }
1016
1017 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1018 {
1019         struct nfs4_state *state = opendata->state;
1020         struct nfs_inode *nfsi = NFS_I(state->inode);
1021         struct nfs_delegation *delegation;
1022         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1023         fmode_t fmode = opendata->o_arg.fmode;
1024         nfs4_stateid stateid;
1025         int ret = -EAGAIN;
1026
1027         for (;;) {
1028                 if (can_open_cached(state, fmode, open_mode)) {
1029                         spin_lock(&state->owner->so_lock);
1030                         if (can_open_cached(state, fmode, open_mode)) {
1031                                 update_open_stateflags(state, fmode);
1032                                 spin_unlock(&state->owner->so_lock);
1033                                 goto out_return_state;
1034                         }
1035                         spin_unlock(&state->owner->so_lock);
1036                 }
1037                 rcu_read_lock();
1038                 delegation = rcu_dereference(nfsi->delegation);
1039                 if (!can_open_delegated(delegation, fmode)) {
1040                         rcu_read_unlock();
1041                         break;
1042                 }
1043                 /* Save the delegation */
1044                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1045                 rcu_read_unlock();
1046                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1047                 if (ret != 0)
1048                         goto out;
1049                 ret = -EAGAIN;
1050
1051                 /* Try to update the stateid using the delegation */
1052                 if (update_open_stateid(state, NULL, &stateid, fmode))
1053                         goto out_return_state;
1054         }
1055 out:
1056         return ERR_PTR(ret);
1057 out_return_state:
1058         atomic_inc(&state->count);
1059         return state;
1060 }
1061
1062 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1063 {
1064         struct inode *inode;
1065         struct nfs4_state *state = NULL;
1066         struct nfs_delegation *delegation;
1067         int ret;
1068
1069         if (!data->rpc_done) {
1070                 state = nfs4_try_open_cached(data);
1071                 goto out;
1072         }
1073
1074         ret = -EAGAIN;
1075         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1076                 goto err;
1077         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1078         ret = PTR_ERR(inode);
1079         if (IS_ERR(inode))
1080                 goto err;
1081         ret = -ENOMEM;
1082         state = nfs4_get_open_state(inode, data->owner);
1083         if (state == NULL)
1084                 goto err_put_inode;
1085         if (data->o_res.delegation_type != 0) {
1086                 int delegation_flags = 0;
1087
1088                 rcu_read_lock();
1089                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1090                 if (delegation)
1091                         delegation_flags = delegation->flags;
1092                 rcu_read_unlock();
1093                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1094                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1095                                         "returning a delegation for "
1096                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1097                                         NFS_CLIENT(inode)->cl_server);
1098                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1099                         nfs_inode_set_delegation(state->inode,
1100                                         data->owner->so_cred,
1101                                         &data->o_res);
1102                 else
1103                         nfs_inode_reclaim_delegation(state->inode,
1104                                         data->owner->so_cred,
1105                                         &data->o_res);
1106         }
1107
1108         update_open_stateid(state, &data->o_res.stateid, NULL,
1109                         data->o_arg.fmode);
1110         iput(inode);
1111 out:
1112         return state;
1113 err_put_inode:
1114         iput(inode);
1115 err:
1116         return ERR_PTR(ret);
1117 }
1118
1119 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1120 {
1121         struct nfs_inode *nfsi = NFS_I(state->inode);
1122         struct nfs_open_context *ctx;
1123
1124         spin_lock(&state->inode->i_lock);
1125         list_for_each_entry(ctx, &nfsi->open_files, list) {
1126                 if (ctx->state != state)
1127                         continue;
1128                 get_nfs_open_context(ctx);
1129                 spin_unlock(&state->inode->i_lock);
1130                 return ctx;
1131         }
1132         spin_unlock(&state->inode->i_lock);
1133         return ERR_PTR(-ENOENT);
1134 }
1135
1136 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1137 {
1138         struct nfs4_opendata *opendata;
1139
1140         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1141         if (opendata == NULL)
1142                 return ERR_PTR(-ENOMEM);
1143         opendata->state = state;
1144         atomic_inc(&state->count);
1145         return opendata;
1146 }
1147
1148 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1149 {
1150         struct nfs4_state *newstate;
1151         int ret;
1152
1153         opendata->o_arg.open_flags = 0;
1154         opendata->o_arg.fmode = fmode;
1155         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1156         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1157         nfs4_init_opendata_res(opendata);
1158         ret = _nfs4_recover_proc_open(opendata);
1159         if (ret != 0)
1160                 return ret;
1161         newstate = nfs4_opendata_to_nfs4_state(opendata);
1162         if (IS_ERR(newstate))
1163                 return PTR_ERR(newstate);
1164         nfs4_close_state(newstate, fmode);
1165         *res = newstate;
1166         return 0;
1167 }
1168
1169 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1170 {
1171         struct nfs4_state *newstate;
1172         int ret;
1173
1174         /* memory barrier prior to reading state->n_* */
1175         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1176         smp_rmb();
1177         if (state->n_rdwr != 0) {
1178                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1179                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1180                 if (ret != 0)
1181                         return ret;
1182                 if (newstate != state)
1183                         return -ESTALE;
1184         }
1185         if (state->n_wronly != 0) {
1186                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1187                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1188                 if (ret != 0)
1189                         return ret;
1190                 if (newstate != state)
1191                         return -ESTALE;
1192         }
1193         if (state->n_rdonly != 0) {
1194                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1195                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1196                 if (ret != 0)
1197                         return ret;
1198                 if (newstate != state)
1199                         return -ESTALE;
1200         }
1201         /*
1202          * We may have performed cached opens for all three recoveries.
1203          * Check if we need to update the current stateid.
1204          */
1205         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1206             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1207                 write_seqlock(&state->seqlock);
1208                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1209                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1210                 write_sequnlock(&state->seqlock);
1211         }
1212         return 0;
1213 }
1214
1215 /*
1216  * OPEN_RECLAIM:
1217  *      reclaim state on the server after a reboot.
1218  */
1219 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1220 {
1221         struct nfs_delegation *delegation;
1222         struct nfs4_opendata *opendata;
1223         fmode_t delegation_type = 0;
1224         int status;
1225
1226         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1227         if (IS_ERR(opendata))
1228                 return PTR_ERR(opendata);
1229         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1230         opendata->o_arg.fh = NFS_FH(state->inode);
1231         rcu_read_lock();
1232         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1233         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1234                 delegation_type = delegation->type;
1235         rcu_read_unlock();
1236         opendata->o_arg.u.delegation_type = delegation_type;
1237         status = nfs4_open_recover(opendata, state);
1238         nfs4_opendata_put(opendata);
1239         return status;
1240 }
1241
1242 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1243 {
1244         struct nfs_server *server = NFS_SERVER(state->inode);
1245         struct nfs4_exception exception = { };
1246         int err;
1247         do {
1248                 err = _nfs4_do_open_reclaim(ctx, state);
1249                 if (err != -NFS4ERR_DELAY)
1250                         break;
1251                 nfs4_handle_exception(server, err, &exception);
1252         } while (exception.retry);
1253         return err;
1254 }
1255
1256 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1257 {
1258         struct nfs_open_context *ctx;
1259         int ret;
1260
1261         ctx = nfs4_state_find_open_context(state);
1262         if (IS_ERR(ctx))
1263                 return PTR_ERR(ctx);
1264         ret = nfs4_do_open_reclaim(ctx, state);
1265         put_nfs_open_context(ctx);
1266         return ret;
1267 }
1268
1269 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1270 {
1271         struct nfs4_opendata *opendata;
1272         int ret;
1273
1274         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1275         if (IS_ERR(opendata))
1276                 return PTR_ERR(opendata);
1277         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1278         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1279                         sizeof(opendata->o_arg.u.delegation.data));
1280         ret = nfs4_open_recover(opendata, state);
1281         nfs4_opendata_put(opendata);
1282         return ret;
1283 }
1284
1285 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1286 {
1287         struct nfs4_exception exception = { };
1288         struct nfs_server *server = NFS_SERVER(state->inode);
1289         int err;
1290         do {
1291                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1292                 switch (err) {
1293                         case 0:
1294                         case -ENOENT:
1295                         case -ESTALE:
1296                                 goto out;
1297                         case -NFS4ERR_BADSESSION:
1298                         case -NFS4ERR_BADSLOT:
1299                         case -NFS4ERR_BAD_HIGH_SLOT:
1300                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1301                         case -NFS4ERR_DEADSESSION:
1302                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1303                                 goto out;
1304                         case -NFS4ERR_STALE_CLIENTID:
1305                         case -NFS4ERR_STALE_STATEID:
1306                         case -NFS4ERR_EXPIRED:
1307                                 /* Don't recall a delegation if it was lost */
1308                                 nfs4_schedule_lease_recovery(server->nfs_client);
1309                                 goto out;
1310                         case -ERESTARTSYS:
1311                                 /*
1312                                  * The show must go on: exit, but mark the
1313                                  * stateid as needing recovery.
1314                                  */
1315                         case -NFS4ERR_ADMIN_REVOKED:
1316                         case -NFS4ERR_BAD_STATEID:
1317                                 nfs4_schedule_stateid_recovery(server, state);
1318                         case -EKEYEXPIRED:
1319                                 /*
1320                                  * User RPCSEC_GSS context has expired.
1321                                  * We cannot recover this stateid now, so
1322                                  * skip it and allow recovery thread to
1323                                  * proceed.
1324                                  */
1325                         case -ENOMEM:
1326                                 err = 0;
1327                                 goto out;
1328                 }
1329                 err = nfs4_handle_exception(server, err, &exception);
1330         } while (exception.retry);
1331 out:
1332         return err;
1333 }
1334
1335 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1336 {
1337         struct nfs4_opendata *data = calldata;
1338
1339         data->rpc_status = task->tk_status;
1340         if (data->rpc_status == 0) {
1341                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1342                                 sizeof(data->o_res.stateid.data));
1343                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1344                 renew_lease(data->o_res.server, data->timestamp);
1345                 data->rpc_done = 1;
1346         }
1347 }
1348
1349 static void nfs4_open_confirm_release(void *calldata)
1350 {
1351         struct nfs4_opendata *data = calldata;
1352         struct nfs4_state *state = NULL;
1353
1354         /* If this request hasn't been cancelled, do nothing */
1355         if (data->cancelled == 0)
1356                 goto out_free;
1357         /* In case of error, no cleanup! */
1358         if (!data->rpc_done)
1359                 goto out_free;
1360         state = nfs4_opendata_to_nfs4_state(data);
1361         if (!IS_ERR(state))
1362                 nfs4_close_state(state, data->o_arg.fmode);
1363 out_free:
1364         nfs4_opendata_put(data);
1365 }
1366
1367 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1368         .rpc_call_done = nfs4_open_confirm_done,
1369         .rpc_release = nfs4_open_confirm_release,
1370 };
1371
1372 /*
1373  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1374  */
1375 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1376 {
1377         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1378         struct rpc_task *task;
1379         struct  rpc_message msg = {
1380                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1381                 .rpc_argp = &data->c_arg,
1382                 .rpc_resp = &data->c_res,
1383                 .rpc_cred = data->owner->so_cred,
1384         };
1385         struct rpc_task_setup task_setup_data = {
1386                 .rpc_client = server->client,
1387                 .rpc_message = &msg,
1388                 .callback_ops = &nfs4_open_confirm_ops,
1389                 .callback_data = data,
1390                 .workqueue = nfsiod_workqueue,
1391                 .flags = RPC_TASK_ASYNC,
1392         };
1393         int status;
1394
1395         kref_get(&data->kref);
1396         data->rpc_done = 0;
1397         data->rpc_status = 0;
1398         data->timestamp = jiffies;
1399         task = rpc_run_task(&task_setup_data);
1400         if (IS_ERR(task))
1401                 return PTR_ERR(task);
1402         status = nfs4_wait_for_completion_rpc_task(task);
1403         if (status != 0) {
1404                 data->cancelled = 1;
1405                 smp_wmb();
1406         } else
1407                 status = data->rpc_status;
1408         rpc_put_task(task);
1409         return status;
1410 }
1411
1412 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1413 {
1414         struct nfs4_opendata *data = calldata;
1415         struct nfs4_state_owner *sp = data->owner;
1416
1417         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1418                 return;
1419         /*
1420          * Check if we still need to send an OPEN call, or if we can use
1421          * a delegation instead.
1422          */
1423         if (data->state != NULL) {
1424                 struct nfs_delegation *delegation;
1425
1426                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1427                         goto out_no_action;
1428                 rcu_read_lock();
1429                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1430                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1431                     can_open_delegated(delegation, data->o_arg.fmode))
1432                         goto unlock_no_action;
1433                 rcu_read_unlock();
1434         }
1435         /* Update sequence id. */
1436         data->o_arg.id = sp->so_owner_id.id;
1437         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1438         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1439                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1440                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1441         }
1442         data->timestamp = jiffies;
1443         if (nfs4_setup_sequence(data->o_arg.server,
1444                                 &data->o_arg.seq_args,
1445                                 &data->o_res.seq_res, 1, task))
1446                 return;
1447         rpc_call_start(task);
1448         return;
1449 unlock_no_action:
1450         rcu_read_unlock();
1451 out_no_action:
1452         task->tk_action = NULL;
1453
1454 }
1455
1456 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1457 {
1458         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1459         nfs4_open_prepare(task, calldata);
1460 }
1461
1462 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1463 {
1464         struct nfs4_opendata *data = calldata;
1465
1466         data->rpc_status = task->tk_status;
1467
1468         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1469                 return;
1470
1471         if (task->tk_status == 0) {
1472                 switch (data->o_res.f_attr->mode & S_IFMT) {
1473                         case S_IFREG:
1474                                 break;
1475                         case S_IFLNK:
1476                                 data->rpc_status = -ELOOP;
1477                                 break;
1478                         case S_IFDIR:
1479                                 data->rpc_status = -EISDIR;
1480                                 break;
1481                         default:
1482                                 data->rpc_status = -ENOTDIR;
1483                 }
1484                 renew_lease(data->o_res.server, data->timestamp);
1485                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1486                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1487         }
1488         data->rpc_done = 1;
1489 }
1490
1491 static void nfs4_open_release(void *calldata)
1492 {
1493         struct nfs4_opendata *data = calldata;
1494         struct nfs4_state *state = NULL;
1495
1496         /* If this request hasn't been cancelled, do nothing */
1497         if (data->cancelled == 0)
1498                 goto out_free;
1499         /* In case of error, no cleanup! */
1500         if (data->rpc_status != 0 || !data->rpc_done)
1501                 goto out_free;
1502         /* In case we need an open_confirm, no cleanup! */
1503         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1504                 goto out_free;
1505         state = nfs4_opendata_to_nfs4_state(data);
1506         if (!IS_ERR(state))
1507                 nfs4_close_state(state, data->o_arg.fmode);
1508 out_free:
1509         nfs4_opendata_put(data);
1510 }
1511
1512 static const struct rpc_call_ops nfs4_open_ops = {
1513         .rpc_call_prepare = nfs4_open_prepare,
1514         .rpc_call_done = nfs4_open_done,
1515         .rpc_release = nfs4_open_release,
1516 };
1517
1518 static const struct rpc_call_ops nfs4_recover_open_ops = {
1519         .rpc_call_prepare = nfs4_recover_open_prepare,
1520         .rpc_call_done = nfs4_open_done,
1521         .rpc_release = nfs4_open_release,
1522 };
1523
1524 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1525 {
1526         struct inode *dir = data->dir->d_inode;
1527         struct nfs_server *server = NFS_SERVER(dir);
1528         struct nfs_openargs *o_arg = &data->o_arg;
1529         struct nfs_openres *o_res = &data->o_res;
1530         struct rpc_task *task;
1531         struct rpc_message msg = {
1532                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1533                 .rpc_argp = o_arg,
1534                 .rpc_resp = o_res,
1535                 .rpc_cred = data->owner->so_cred,
1536         };
1537         struct rpc_task_setup task_setup_data = {
1538                 .rpc_client = server->client,
1539                 .rpc_message = &msg,
1540                 .callback_ops = &nfs4_open_ops,
1541                 .callback_data = data,
1542                 .workqueue = nfsiod_workqueue,
1543                 .flags = RPC_TASK_ASYNC,
1544         };
1545         int status;
1546
1547         kref_get(&data->kref);
1548         data->rpc_done = 0;
1549         data->rpc_status = 0;
1550         data->cancelled = 0;
1551         if (isrecover)
1552                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1553         task = rpc_run_task(&task_setup_data);
1554         if (IS_ERR(task))
1555                 return PTR_ERR(task);
1556         status = nfs4_wait_for_completion_rpc_task(task);
1557         if (status != 0) {
1558                 data->cancelled = 1;
1559                 smp_wmb();
1560         } else
1561                 status = data->rpc_status;
1562         rpc_put_task(task);
1563
1564         return status;
1565 }
1566
1567 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1568 {
1569         struct inode *dir = data->dir->d_inode;
1570         struct nfs_openres *o_res = &data->o_res;
1571         int status;
1572
1573         status = nfs4_run_open_task(data, 1);
1574         if (status != 0 || !data->rpc_done)
1575                 return status;
1576
1577         nfs_refresh_inode(dir, o_res->dir_attr);
1578
1579         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1580                 status = _nfs4_proc_open_confirm(data);
1581                 if (status != 0)
1582                         return status;
1583         }
1584
1585         return status;
1586 }
1587
1588 /*
1589  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1590  */
1591 static int _nfs4_proc_open(struct nfs4_opendata *data)
1592 {
1593         struct inode *dir = data->dir->d_inode;
1594         struct nfs_server *server = NFS_SERVER(dir);
1595         struct nfs_openargs *o_arg = &data->o_arg;
1596         struct nfs_openres *o_res = &data->o_res;
1597         int status;
1598
1599         status = nfs4_run_open_task(data, 0);
1600         if (!data->rpc_done)
1601                 return status;
1602         if (status != 0) {
1603                 if (status == -NFS4ERR_BADNAME &&
1604                                 !(o_arg->open_flags & O_CREAT))
1605                         return -ENOENT;
1606                 return status;
1607         }
1608
1609         if (o_arg->open_flags & O_CREAT) {
1610                 update_changeattr(dir, &o_res->cinfo);
1611                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1612         } else
1613                 nfs_refresh_inode(dir, o_res->dir_attr);
1614         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1615                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1616         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1617                 status = _nfs4_proc_open_confirm(data);
1618                 if (status != 0)
1619                         return status;
1620         }
1621         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1622                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1623         return 0;
1624 }
1625
1626 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1627 {
1628         unsigned int loop;
1629         int ret;
1630
1631         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1632                 ret = nfs4_wait_clnt_recover(clp);
1633                 if (ret != 0)
1634                         break;
1635                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1636                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1637                         break;
1638                 nfs4_schedule_state_manager(clp);
1639                 ret = -EIO;
1640         }
1641         return ret;
1642 }
1643
1644 static int nfs4_recover_expired_lease(struct nfs_server *server)
1645 {
1646         return nfs4_client_recover_expired_lease(server->nfs_client);
1647 }
1648
1649 /*
1650  * OPEN_EXPIRED:
1651  *      reclaim state on the server after a network partition.
1652  *      Assumes caller holds the appropriate lock
1653  */
1654 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1655 {
1656         struct nfs4_opendata *opendata;
1657         int ret;
1658
1659         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1660         if (IS_ERR(opendata))
1661                 return PTR_ERR(opendata);
1662         ret = nfs4_open_recover(opendata, state);
1663         if (ret == -ESTALE)
1664                 d_drop(ctx->dentry);
1665         nfs4_opendata_put(opendata);
1666         return ret;
1667 }
1668
1669 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1670 {
1671         struct nfs_server *server = NFS_SERVER(state->inode);
1672         struct nfs4_exception exception = { };
1673         int err;
1674
1675         do {
1676                 err = _nfs4_open_expired(ctx, state);
1677                 switch (err) {
1678                 default:
1679                         goto out;
1680                 case -NFS4ERR_GRACE:
1681                 case -NFS4ERR_DELAY:
1682                         nfs4_handle_exception(server, err, &exception);
1683                         err = 0;
1684                 }
1685         } while (exception.retry);
1686 out:
1687         return err;
1688 }
1689
1690 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1691 {
1692         struct nfs_open_context *ctx;
1693         int ret;
1694
1695         ctx = nfs4_state_find_open_context(state);
1696         if (IS_ERR(ctx))
1697                 return PTR_ERR(ctx);
1698         ret = nfs4_do_open_expired(ctx, state);
1699         put_nfs_open_context(ctx);
1700         return ret;
1701 }
1702
1703 #if defined(CONFIG_NFS_V4_1)
1704 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1705 {
1706         int status;
1707         struct nfs_server *server = NFS_SERVER(state->inode);
1708
1709         status = nfs41_test_stateid(server, state);
1710         if (status == NFS_OK)
1711                 return 0;
1712         nfs41_free_stateid(server, state);
1713         return nfs4_open_expired(sp, state);
1714 }
1715 #endif
1716
1717 /*
1718  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1719  * fields corresponding to attributes that were used to store the verifier.
1720  * Make sure we clobber those fields in the later setattr call
1721  */
1722 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1723 {
1724         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1725             !(sattr->ia_valid & ATTR_ATIME_SET))
1726                 sattr->ia_valid |= ATTR_ATIME;
1727
1728         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1729             !(sattr->ia_valid & ATTR_MTIME_SET))
1730                 sattr->ia_valid |= ATTR_MTIME;
1731 }
1732
1733 /*
1734  * Returns a referenced nfs4_state
1735  */
1736 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)
1737 {
1738         struct nfs4_state_owner  *sp;
1739         struct nfs4_state     *state = NULL;
1740         struct nfs_server       *server = NFS_SERVER(dir);
1741         struct nfs4_opendata *opendata;
1742         int status;
1743
1744         /* Protect against reboot recovery conflicts */
1745         status = -ENOMEM;
1746         if (!(sp = nfs4_get_state_owner(server, cred))) {
1747                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1748                 goto out_err;
1749         }
1750         status = nfs4_recover_expired_lease(server);
1751         if (status != 0)
1752                 goto err_put_state_owner;
1753         if (dentry->d_inode != NULL)
1754                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1755         status = -ENOMEM;
1756         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1757         if (opendata == NULL)
1758                 goto err_put_state_owner;
1759
1760         if (dentry->d_inode != NULL)
1761                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1762
1763         status = _nfs4_proc_open(opendata);
1764         if (status != 0)
1765                 goto err_opendata_put;
1766
1767         state = nfs4_opendata_to_nfs4_state(opendata);
1768         status = PTR_ERR(state);
1769         if (IS_ERR(state))
1770                 goto err_opendata_put;
1771         if (server->caps & NFS_CAP_POSIX_LOCK)
1772                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1773
1774         if (opendata->o_arg.open_flags & O_EXCL) {
1775                 nfs4_exclusive_attrset(opendata, sattr);
1776
1777                 nfs_fattr_init(opendata->o_res.f_attr);
1778                 status = nfs4_do_setattr(state->inode, cred,
1779                                 opendata->o_res.f_attr, sattr,
1780                                 state);
1781                 if (status == 0)
1782                         nfs_setattr_update_inode(state->inode, sattr);
1783                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1784         }
1785         nfs4_opendata_put(opendata);
1786         nfs4_put_state_owner(sp);
1787         *res = state;
1788         return 0;
1789 err_opendata_put:
1790         nfs4_opendata_put(opendata);
1791 err_put_state_owner:
1792         nfs4_put_state_owner(sp);
1793 out_err:
1794         *res = NULL;
1795         return status;
1796 }
1797
1798
1799 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)
1800 {
1801         struct nfs4_exception exception = { };
1802         struct nfs4_state *res;
1803         int status;
1804
1805         do {
1806                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1807                 if (status == 0)
1808                         break;
1809                 /* NOTE: BAD_SEQID means the server and client disagree about the
1810                  * book-keeping w.r.t. state-changing operations
1811                  * (OPEN/CLOSE/LOCK/LOCKU...)
1812                  * It is actually a sign of a bug on the client or on the server.
1813                  *
1814                  * If we receive a BAD_SEQID error in the particular case of
1815                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1816                  * have unhashed the old state_owner for us, and that we can
1817                  * therefore safely retry using a new one. We should still warn
1818                  * the user though...
1819                  */
1820                 if (status == -NFS4ERR_BAD_SEQID) {
1821                         printk(KERN_WARNING "NFS: v4 server %s "
1822                                         " returned a bad sequence-id error!\n",
1823                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1824                         exception.retry = 1;
1825                         continue;
1826                 }
1827                 /*
1828                  * BAD_STATEID on OPEN means that the server cancelled our
1829                  * state before it received the OPEN_CONFIRM.
1830                  * Recover by retrying the request as per the discussion
1831                  * on Page 181 of RFC3530.
1832                  */
1833                 if (status == -NFS4ERR_BAD_STATEID) {
1834                         exception.retry = 1;
1835                         continue;
1836                 }
1837                 if (status == -EAGAIN) {
1838                         /* We must have found a delegation */
1839                         exception.retry = 1;
1840                         continue;
1841                 }
1842                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1843                                         status, &exception));
1844         } while (exception.retry);
1845         return res;
1846 }
1847
1848 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1849                             struct nfs_fattr *fattr, struct iattr *sattr,
1850                             struct nfs4_state *state)
1851 {
1852         struct nfs_server *server = NFS_SERVER(inode);
1853         struct nfs_setattrargs  arg = {
1854                 .fh             = NFS_FH(inode),
1855                 .iap            = sattr,
1856                 .server         = server,
1857                 .bitmask = server->attr_bitmask,
1858         };
1859         struct nfs_setattrres  res = {
1860                 .fattr          = fattr,
1861                 .server         = server,
1862         };
1863         struct rpc_message msg = {
1864                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1865                 .rpc_argp       = &arg,
1866                 .rpc_resp       = &res,
1867                 .rpc_cred       = cred,
1868         };
1869         unsigned long timestamp = jiffies;
1870         int status;
1871
1872         nfs_fattr_init(fattr);
1873
1874         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1875                 /* Use that stateid */
1876         } else if (state != NULL) {
1877                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1878         } else
1879                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1880
1881         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1882         if (status == 0 && state != NULL)
1883                 renew_lease(server, timestamp);
1884         return status;
1885 }
1886
1887 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1888                            struct nfs_fattr *fattr, struct iattr *sattr,
1889                            struct nfs4_state *state)
1890 {
1891         struct nfs_server *server = NFS_SERVER(inode);
1892         struct nfs4_exception exception = { };
1893         int err;
1894         do {
1895                 err = nfs4_handle_exception(server,
1896                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1897                                 &exception);
1898         } while (exception.retry);
1899         return err;
1900 }
1901
1902 struct nfs4_closedata {
1903         struct inode *inode;
1904         struct nfs4_state *state;
1905         struct nfs_closeargs arg;
1906         struct nfs_closeres res;
1907         struct nfs_fattr fattr;
1908         unsigned long timestamp;
1909         bool roc;
1910         u32 roc_barrier;
1911 };
1912
1913 static void nfs4_free_closedata(void *data)
1914 {
1915         struct nfs4_closedata *calldata = data;
1916         struct nfs4_state_owner *sp = calldata->state->owner;
1917         struct super_block *sb = calldata->state->inode->i_sb;
1918
1919         if (calldata->roc)
1920                 pnfs_roc_release(calldata->state->inode);
1921         nfs4_put_open_state(calldata->state);
1922         nfs_free_seqid(calldata->arg.seqid);
1923         nfs4_put_state_owner(sp);
1924         nfs_sb_deactive(sb);
1925         kfree(calldata);
1926 }
1927
1928 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1929                 fmode_t fmode)
1930 {
1931         spin_lock(&state->owner->so_lock);
1932         if (!(fmode & FMODE_READ))
1933                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1934         if (!(fmode & FMODE_WRITE))
1935                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1936         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1937         spin_unlock(&state->owner->so_lock);
1938 }
1939
1940 static void nfs4_close_done(struct rpc_task *task, void *data)
1941 {
1942         struct nfs4_closedata *calldata = data;
1943         struct nfs4_state *state = calldata->state;
1944         struct nfs_server *server = NFS_SERVER(calldata->inode);
1945
1946         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1947                 return;
1948         /* hmm. we are done with the inode, and in the process of freeing
1949          * the state_owner. we keep this around to process errors
1950          */
1951         switch (task->tk_status) {
1952                 case 0:
1953                         if (calldata->roc)
1954                                 pnfs_roc_set_barrier(state->inode,
1955                                                      calldata->roc_barrier);
1956                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1957                         renew_lease(server, calldata->timestamp);
1958                         nfs4_close_clear_stateid_flags(state,
1959                                         calldata->arg.fmode);
1960                         break;
1961                 case -NFS4ERR_STALE_STATEID:
1962                 case -NFS4ERR_OLD_STATEID:
1963                 case -NFS4ERR_BAD_STATEID:
1964                 case -NFS4ERR_EXPIRED:
1965                         if (calldata->arg.fmode == 0)
1966                                 break;
1967                 default:
1968                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1969                                 rpc_restart_call_prepare(task);
1970         }
1971         nfs_release_seqid(calldata->arg.seqid);
1972         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1973 }
1974
1975 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1976 {
1977         struct nfs4_closedata *calldata = data;
1978         struct nfs4_state *state = calldata->state;
1979         int call_close = 0;
1980
1981         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1982                 return;
1983
1984         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1985         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1986         spin_lock(&state->owner->so_lock);
1987         /* Calculate the change in open mode */
1988         if (state->n_rdwr == 0) {
1989                 if (state->n_rdonly == 0) {
1990                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1991                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1992                         calldata->arg.fmode &= ~FMODE_READ;
1993                 }
1994                 if (state->n_wronly == 0) {
1995                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1996                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1997                         calldata->arg.fmode &= ~FMODE_WRITE;
1998                 }
1999         }
2000         spin_unlock(&state->owner->so_lock);
2001
2002         if (!call_close) {
2003                 /* Note: exit _without_ calling nfs4_close_done */
2004                 task->tk_action = NULL;
2005                 return;
2006         }
2007
2008         if (calldata->arg.fmode == 0) {
2009                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2010                 if (calldata->roc &&
2011                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2012                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2013                                      task, NULL);
2014                         return;
2015                 }
2016         }
2017
2018         nfs_fattr_init(calldata->res.fattr);
2019         calldata->timestamp = jiffies;
2020         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2021                                 &calldata->arg.seq_args, &calldata->res.seq_res,
2022                                 1, task))
2023                 return;
2024         rpc_call_start(task);
2025 }
2026
2027 static const struct rpc_call_ops nfs4_close_ops = {
2028         .rpc_call_prepare = nfs4_close_prepare,
2029         .rpc_call_done = nfs4_close_done,
2030         .rpc_release = nfs4_free_closedata,
2031 };
2032
2033 /*
2034  * It is possible for data to be read/written from a mem-mapped file
2035  * after the sys_close call (which hits the vfs layer as a flush).
2036  * This means that we can't safely call nfsv4 close on a file until
2037  * the inode is cleared. This in turn means that we are not good
2038  * NFSv4 citizens - we do not indicate to the server to update the file's
2039  * share state even when we are done with one of the three share
2040  * stateid's in the inode.
2041  *
2042  * NOTE: Caller must be holding the sp->so_owner semaphore!
2043  */
2044 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2045 {
2046         struct nfs_server *server = NFS_SERVER(state->inode);
2047         struct nfs4_closedata *calldata;
2048         struct nfs4_state_owner *sp = state->owner;
2049         struct rpc_task *task;
2050         struct rpc_message msg = {
2051                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2052                 .rpc_cred = state->owner->so_cred,
2053         };
2054         struct rpc_task_setup task_setup_data = {
2055                 .rpc_client = server->client,
2056                 .rpc_message = &msg,
2057                 .callback_ops = &nfs4_close_ops,
2058                 .workqueue = nfsiod_workqueue,
2059                 .flags = RPC_TASK_ASYNC,
2060         };
2061         int status = -ENOMEM;
2062
2063         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2064         if (calldata == NULL)
2065                 goto out;
2066         calldata->inode = state->inode;
2067         calldata->state = state;
2068         calldata->arg.fh = NFS_FH(state->inode);
2069         calldata->arg.stateid = &state->open_stateid;
2070         /* Serialization for the sequence id */
2071         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2072         if (calldata->arg.seqid == NULL)
2073                 goto out_free_calldata;
2074         calldata->arg.fmode = 0;
2075         calldata->arg.bitmask = server->cache_consistency_bitmask;
2076         calldata->res.fattr = &calldata->fattr;
2077         calldata->res.seqid = calldata->arg.seqid;
2078         calldata->res.server = server;
2079         calldata->roc = roc;
2080         nfs_sb_active(calldata->inode->i_sb);
2081
2082         msg.rpc_argp = &calldata->arg;
2083         msg.rpc_resp = &calldata->res;
2084         task_setup_data.callback_data = calldata;
2085         task = rpc_run_task(&task_setup_data);
2086         if (IS_ERR(task))
2087                 return PTR_ERR(task);
2088         status = 0;
2089         if (wait)
2090                 status = rpc_wait_for_completion_task(task);
2091         rpc_put_task(task);
2092         return status;
2093 out_free_calldata:
2094         kfree(calldata);
2095 out:
2096         if (roc)
2097                 pnfs_roc_release(state->inode);
2098         nfs4_put_open_state(state);
2099         nfs4_put_state_owner(sp);
2100         return status;
2101 }
2102
2103 static struct inode *
2104 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2105 {
2106         struct nfs4_state *state;
2107
2108         /* Protect against concurrent sillydeletes */
2109         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2110         if (IS_ERR(state))
2111                 return ERR_CAST(state);
2112         ctx->state = state;
2113         return igrab(state->inode);
2114 }
2115
2116 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2117 {
2118         if (ctx->state == NULL)
2119                 return;
2120         if (is_sync)
2121                 nfs4_close_sync(ctx->state, ctx->mode);
2122         else
2123                 nfs4_close_state(ctx->state, ctx->mode);
2124 }
2125
2126 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2127 {
2128         struct nfs4_server_caps_arg args = {
2129                 .fhandle = fhandle,
2130         };
2131         struct nfs4_server_caps_res res = {};
2132         struct rpc_message msg = {
2133                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2134                 .rpc_argp = &args,
2135                 .rpc_resp = &res,
2136         };
2137         int status;
2138
2139         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2140         if (status == 0) {
2141                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2142                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2143                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2144                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2145                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2146                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2147                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2148                         server->caps |= NFS_CAP_ACLS;
2149                 if (res.has_links != 0)
2150                         server->caps |= NFS_CAP_HARDLINKS;
2151                 if (res.has_symlinks != 0)
2152                         server->caps |= NFS_CAP_SYMLINKS;
2153                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2154                         server->caps |= NFS_CAP_FILEID;
2155                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2156                         server->caps |= NFS_CAP_MODE;
2157                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2158                         server->caps |= NFS_CAP_NLINK;
2159                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2160                         server->caps |= NFS_CAP_OWNER;
2161                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2162                         server->caps |= NFS_CAP_OWNER_GROUP;
2163                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2164                         server->caps |= NFS_CAP_ATIME;
2165                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2166                         server->caps |= NFS_CAP_CTIME;
2167                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2168                         server->caps |= NFS_CAP_MTIME;
2169
2170                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2171                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2172                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2173                 server->acl_bitmask = res.acl_bitmask;
2174         }
2175
2176         return status;
2177 }
2178
2179 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2180 {
2181         struct nfs4_exception exception = { };
2182         int err;
2183         do {
2184                 err = nfs4_handle_exception(server,
2185                                 _nfs4_server_capabilities(server, fhandle),
2186                                 &exception);
2187         } while (exception.retry);
2188         return err;
2189 }
2190
2191 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2192                 struct nfs_fsinfo *info)
2193 {
2194         struct nfs4_lookup_root_arg args = {
2195                 .bitmask = nfs4_fattr_bitmap,
2196         };
2197         struct nfs4_lookup_res res = {
2198                 .server = server,
2199                 .fattr = info->fattr,
2200                 .fh = fhandle,
2201         };
2202         struct rpc_message msg = {
2203                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2204                 .rpc_argp = &args,
2205                 .rpc_resp = &res,
2206         };
2207
2208         nfs_fattr_init(info->fattr);
2209         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2210 }
2211
2212 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2213                 struct nfs_fsinfo *info)
2214 {
2215         struct nfs4_exception exception = { };
2216         int err;
2217         do {
2218                 err = _nfs4_lookup_root(server, fhandle, info);
2219                 switch (err) {
2220                 case 0:
2221                 case -NFS4ERR_WRONGSEC:
2222                         break;
2223                 default:
2224                         err = nfs4_handle_exception(server, err, &exception);
2225                 }
2226         } while (exception.retry);
2227         return err;
2228 }
2229
2230 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2231                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2232 {
2233         struct rpc_auth *auth;
2234         int ret;
2235
2236         auth = rpcauth_create(flavor, server->client);
2237         if (!auth) {
2238                 ret = -EIO;
2239                 goto out;
2240         }
2241         ret = nfs4_lookup_root(server, fhandle, info);
2242 out:
2243         return ret;
2244 }
2245
2246 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2247                               struct nfs_fsinfo *info)
2248 {
2249         int i, len, status = 0;
2250         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2251
2252         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2253         flav_array[len] = RPC_AUTH_NULL;
2254         len += 1;
2255
2256         for (i = 0; i < len; i++) {
2257                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2258                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2259                         continue;
2260                 break;
2261         }
2262         /*
2263          * -EACCESS could mean that the user doesn't have correct permissions
2264          * to access the mount.  It could also mean that we tried to mount
2265          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2266          * existing mount programs don't handle -EACCES very well so it should
2267          * be mapped to -EPERM instead.
2268          */
2269         if (status == -EACCES)
2270                 status = -EPERM;
2271         return status;
2272 }
2273
2274 /*
2275  * get the file handle for the "/" directory on the server
2276  */
2277 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2278                               struct nfs_fsinfo *info)
2279 {
2280         int minor_version = server->nfs_client->cl_minorversion;
2281         int status = nfs4_lookup_root(server, fhandle, info);
2282         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2283                 /*
2284                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2285                  * by nfs4_map_errors() as this function exits.
2286                  */
2287                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2288         if (status == 0)
2289                 status = nfs4_server_capabilities(server, fhandle);
2290         if (status == 0)
2291                 status = nfs4_do_fsinfo(server, fhandle, info);
2292         return nfs4_map_errors(status);
2293 }
2294
2295 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
2296 /*
2297  * Get locations and (maybe) other attributes of a referral.
2298  * Note that we'll actually follow the referral later when
2299  * we detect fsid mismatch in inode revalidation
2300  */
2301 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2302                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2303 {
2304         int status = -ENOMEM;
2305         struct page *page = NULL;
2306         struct nfs4_fs_locations *locations = NULL;
2307
2308         page = alloc_page(GFP_KERNEL);
2309         if (page == NULL)
2310                 goto out;
2311         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2312         if (locations == NULL)
2313                 goto out;
2314
2315         status = nfs4_proc_fs_locations(dir, name, locations, page);
2316         if (status != 0)
2317                 goto out;
2318         /* Make sure server returned a different fsid for the referral */
2319         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2320                 dprintk("%s: server did not return a different fsid for"
2321                         " a referral at %s\n", __func__, name->name);
2322                 status = -EIO;
2323                 goto out;
2324         }
2325         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2326         nfs_fixup_referral_attributes(&locations->fattr);
2327
2328         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2329         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2330         memset(fhandle, 0, sizeof(struct nfs_fh));
2331 out:
2332         if (page)
2333                 __free_page(page);
2334         kfree(locations);
2335         return status;
2336 }
2337
2338 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2339 {
2340         struct nfs4_getattr_arg args = {
2341                 .fh = fhandle,
2342                 .bitmask = server->attr_bitmask,
2343         };
2344         struct nfs4_getattr_res res = {
2345                 .fattr = fattr,
2346                 .server = server,
2347         };
2348         struct rpc_message msg = {
2349                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2350                 .rpc_argp = &args,
2351                 .rpc_resp = &res,
2352         };
2353
2354         nfs_fattr_init(fattr);
2355         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2356 }
2357
2358 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2359 {
2360         struct nfs4_exception exception = { };
2361         int err;
2362         do {
2363                 err = nfs4_handle_exception(server,
2364                                 _nfs4_proc_getattr(server, fhandle, fattr),
2365                                 &exception);
2366         } while (exception.retry);
2367         return err;
2368 }
2369
2370 /*
2371  * The file is not closed if it is opened due to the a request to change
2372  * the size of the file. The open call will not be needed once the
2373  * VFS layer lookup-intents are implemented.
2374  *
2375  * Close is called when the inode is destroyed.
2376  * If we haven't opened the file for O_WRONLY, we
2377  * need to in the size_change case to obtain a stateid.
2378  *
2379  * Got race?
2380  * Because OPEN is always done by name in nfsv4, it is
2381  * possible that we opened a different file by the same
2382  * name.  We can recognize this race condition, but we
2383  * can't do anything about it besides returning an error.
2384  *
2385  * This will be fixed with VFS changes (lookup-intent).
2386  */
2387 static int
2388 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2389                   struct iattr *sattr)
2390 {
2391         struct inode *inode = dentry->d_inode;
2392         struct rpc_cred *cred = NULL;
2393         struct nfs4_state *state = NULL;
2394         int status;
2395
2396         if (pnfs_ld_layoutret_on_setattr(inode))
2397                 pnfs_return_layout(inode);
2398
2399         nfs_fattr_init(fattr);
2400
2401         /* Search for an existing open(O_WRITE) file */
2402         if (sattr->ia_valid & ATTR_FILE) {
2403                 struct nfs_open_context *ctx;
2404
2405                 ctx = nfs_file_open_context(sattr->ia_file);
2406                 if (ctx) {
2407                         cred = ctx->cred;
2408                         state = ctx->state;
2409                 }
2410         }
2411
2412         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2413         if (status == 0)
2414                 nfs_setattr_update_inode(inode, sattr);
2415         return status;
2416 }
2417
2418 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2419                 const struct qstr *name, struct nfs_fh *fhandle,
2420                 struct nfs_fattr *fattr)
2421 {
2422         struct nfs_server *server = NFS_SERVER(dir);
2423         int                    status;
2424         struct nfs4_lookup_arg args = {
2425                 .bitmask = server->attr_bitmask,
2426                 .dir_fh = NFS_FH(dir),
2427                 .name = name,
2428         };
2429         struct nfs4_lookup_res res = {
2430                 .server = server,
2431                 .fattr = fattr,
2432                 .fh = fhandle,
2433         };
2434         struct rpc_message msg = {
2435                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2436                 .rpc_argp = &args,
2437                 .rpc_resp = &res,
2438         };
2439
2440         nfs_fattr_init(fattr);
2441
2442         dprintk("NFS call  lookup %s\n", name->name);
2443         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2444         dprintk("NFS reply lookup: %d\n", status);
2445         return status;
2446 }
2447
2448 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2449 {
2450         memset(fh, 0, sizeof(struct nfs_fh));
2451         fattr->fsid.major = 1;
2452         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2453                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2454         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2455         fattr->nlink = 2;
2456 }
2457
2458 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2459                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2460 {
2461         struct nfs4_exception exception = { };
2462         int err;
2463         do {
2464                 int status;
2465
2466                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2467                 switch (status) {
2468                 case -NFS4ERR_BADNAME:
2469                         return -ENOENT;
2470                 case -NFS4ERR_MOVED:
2471                         return nfs4_get_referral(dir, name, fattr, fhandle);
2472                 case -NFS4ERR_WRONGSEC:
2473                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2474                 }
2475                 err = nfs4_handle_exception(NFS_SERVER(dir),
2476                                 status, &exception);
2477         } while (exception.retry);
2478         return err;
2479 }
2480
2481 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2482 {
2483         struct nfs_server *server = NFS_SERVER(inode);
2484         struct nfs4_accessargs args = {
2485                 .fh = NFS_FH(inode),
2486                 .bitmask = server->attr_bitmask,
2487         };
2488         struct nfs4_accessres res = {
2489                 .server = server,
2490         };
2491         struct rpc_message msg = {
2492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2493                 .rpc_argp = &args,
2494                 .rpc_resp = &res,
2495                 .rpc_cred = entry->cred,
2496         };
2497         int mode = entry->mask;
2498         int status;
2499
2500         /*
2501          * Determine which access bits we want to ask for...
2502          */
2503         if (mode & MAY_READ)
2504                 args.access |= NFS4_ACCESS_READ;
2505         if (S_ISDIR(inode->i_mode)) {
2506                 if (mode & MAY_WRITE)
2507                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2508                 if (mode & MAY_EXEC)
2509                         args.access |= NFS4_ACCESS_LOOKUP;
2510         } else {
2511                 if (mode & MAY_WRITE)
2512                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2513                 if (mode & MAY_EXEC)
2514                         args.access |= NFS4_ACCESS_EXECUTE;
2515         }
2516
2517         res.fattr = nfs_alloc_fattr();
2518         if (res.fattr == NULL)
2519                 return -ENOMEM;
2520
2521         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2522         if (!status) {
2523                 entry->mask = 0;
2524                 if (res.access & NFS4_ACCESS_READ)
2525                         entry->mask |= MAY_READ;
2526                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2527                         entry->mask |= MAY_WRITE;
2528                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2529                         entry->mask |= MAY_EXEC;
2530                 nfs_refresh_inode(inode, res.fattr);
2531         }
2532         nfs_free_fattr(res.fattr);
2533         return status;
2534 }
2535
2536 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2537 {
2538         struct nfs4_exception exception = { };
2539         int err;
2540         do {
2541                 err = nfs4_handle_exception(NFS_SERVER(inode),
2542                                 _nfs4_proc_access(inode, entry),
2543                                 &exception);
2544         } while (exception.retry);
2545         return err;
2546 }
2547
2548 /*
2549  * TODO: For the time being, we don't try to get any attributes
2550  * along with any of the zero-copy operations READ, READDIR,
2551  * READLINK, WRITE.
2552  *
2553  * In the case of the first three, we want to put the GETATTR
2554  * after the read-type operation -- this is because it is hard
2555  * to predict the length of a GETATTR response in v4, and thus
2556  * align the READ data correctly.  This means that the GETATTR
2557  * may end up partially falling into the page cache, and we should
2558  * shift it into the 'tail' of the xdr_buf before processing.
2559  * To do this efficiently, we need to know the total length
2560  * of data received, which doesn't seem to be available outside
2561  * of the RPC layer.
2562  *
2563  * In the case of WRITE, we also want to put the GETATTR after
2564  * the operation -- in this case because we want to make sure
2565  * we get the post-operation mtime and size.  This means that
2566  * we can't use xdr_encode_pages() as written: we need a variant
2567  * of it which would leave room in the 'tail' iovec.
2568  *
2569  * Both of these changes to the XDR layer would in fact be quite
2570  * minor, but I decided to leave them for a subsequent patch.
2571  */
2572 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2573                 unsigned int pgbase, unsigned int pglen)
2574 {
2575         struct nfs4_readlink args = {
2576                 .fh       = NFS_FH(inode),
2577                 .pgbase   = pgbase,
2578                 .pglen    = pglen,
2579                 .pages    = &page,
2580         };
2581         struct nfs4_readlink_res res;
2582         struct rpc_message msg = {
2583                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2584                 .rpc_argp = &args,
2585                 .rpc_resp = &res,
2586         };
2587
2588         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2589 }
2590
2591 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2592                 unsigned int pgbase, unsigned int pglen)
2593 {
2594         struct nfs4_exception exception = { };
2595         int err;
2596         do {
2597                 err = nfs4_handle_exception(NFS_SERVER(inode),
2598                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2599                                 &exception);
2600         } while (exception.retry);
2601         return err;
2602 }
2603
2604 /*
2605  * Got race?
2606  * We will need to arrange for the VFS layer to provide an atomic open.
2607  * Until then, this create/open method is prone to inefficiency and race
2608  * conditions due to the lookup, create, and open VFS calls from sys_open()
2609  * placed on the wire.
2610  *
2611  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2612  * The file will be opened again in the subsequent VFS open call
2613  * (nfs4_proc_file_open).
2614  *
2615  * The open for read will just hang around to be used by any process that
2616  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2617  */
2618
2619 static int
2620 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2621                  int flags, struct nfs_open_context *ctx)
2622 {
2623         struct dentry *de = dentry;
2624         struct nfs4_state *state;
2625         struct rpc_cred *cred = NULL;
2626         fmode_t fmode = 0;
2627         int status = 0;
2628
2629         if (ctx != NULL) {
2630                 cred = ctx->cred;
2631                 de = ctx->dentry;
2632                 fmode = ctx->mode;
2633         }
2634         sattr->ia_mode &= ~current_umask();
2635         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2636         d_drop(dentry);
2637         if (IS_ERR(state)) {
2638                 status = PTR_ERR(state);
2639                 goto out;
2640         }
2641         d_add(dentry, igrab(state->inode));
2642         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2643         if (ctx != NULL)
2644                 ctx->state = state;
2645         else
2646                 nfs4_close_sync(state, fmode);
2647 out:
2648         return status;
2649 }
2650
2651 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2652 {
2653         struct nfs_server *server = NFS_SERVER(dir);
2654         struct nfs_removeargs args = {
2655                 .fh = NFS_FH(dir),
2656                 .name.len = name->len,
2657                 .name.name = name->name,
2658                 .bitmask = server->attr_bitmask,
2659         };
2660         struct nfs_removeres res = {
2661                 .server = server,
2662         };
2663         struct rpc_message msg = {
2664                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2665                 .rpc_argp = &args,
2666                 .rpc_resp = &res,
2667         };
2668         int status = -ENOMEM;
2669
2670         res.dir_attr = nfs_alloc_fattr();
2671         if (res.dir_attr == NULL)
2672                 goto out;
2673
2674         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2675         if (status == 0) {
2676                 update_changeattr(dir, &res.cinfo);
2677                 nfs_post_op_update_inode(dir, res.dir_attr);
2678         }
2679         nfs_free_fattr(res.dir_attr);
2680 out:
2681         return status;
2682 }
2683
2684 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2685 {
2686         struct nfs4_exception exception = { };
2687         int err;
2688         do {
2689                 err = nfs4_handle_exception(NFS_SERVER(dir),
2690                                 _nfs4_proc_remove(dir, name),
2691                                 &exception);
2692         } while (exception.retry);
2693         return err;
2694 }
2695
2696 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2697 {
2698         struct nfs_server *server = NFS_SERVER(dir);
2699         struct nfs_removeargs *args = msg->rpc_argp;
2700         struct nfs_removeres *res = msg->rpc_resp;
2701
2702         args->bitmask = server->cache_consistency_bitmask;
2703         res->server = server;
2704         res->seq_res.sr_slot = NULL;
2705         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2706 }
2707
2708 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2709 {
2710         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2711
2712         if (!nfs4_sequence_done(task, &res->seq_res))
2713                 return 0;
2714         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2715                 return 0;
2716         update_changeattr(dir, &res->cinfo);
2717         nfs_post_op_update_inode(dir, res->dir_attr);
2718         return 1;
2719 }
2720
2721 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2722 {
2723         struct nfs_server *server = NFS_SERVER(dir);
2724         struct nfs_renameargs *arg = msg->rpc_argp;
2725         struct nfs_renameres *res = msg->rpc_resp;
2726
2727         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2728         arg->bitmask = server->attr_bitmask;
2729         res->server = server;
2730 }
2731
2732 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2733                                  struct inode *new_dir)
2734 {
2735         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2736
2737         if (!nfs4_sequence_done(task, &res->seq_res))
2738                 return 0;
2739         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2740                 return 0;
2741
2742         update_changeattr(old_dir, &res->old_cinfo);
2743         nfs_post_op_update_inode(old_dir, res->old_fattr);
2744         update_changeattr(new_dir, &res->new_cinfo);
2745         nfs_post_op_update_inode(new_dir, res->new_fattr);
2746         return 1;
2747 }
2748
2749 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2750                 struct inode *new_dir, struct qstr *new_name)
2751 {
2752         struct nfs_server *server = NFS_SERVER(old_dir);
2753         struct nfs_renameargs arg = {
2754                 .old_dir = NFS_FH(old_dir),
2755                 .new_dir = NFS_FH(new_dir),
2756                 .old_name = old_name,
2757                 .new_name = new_name,
2758                 .bitmask = server->attr_bitmask,
2759         };
2760         struct nfs_renameres res = {
2761                 .server = server,
2762         };
2763         struct rpc_message msg = {
2764                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2765                 .rpc_argp = &arg,
2766                 .rpc_resp = &res,
2767         };
2768         int status = -ENOMEM;
2769
2770         res.old_fattr = nfs_alloc_fattr();
2771         res.new_fattr = nfs_alloc_fattr();
2772         if (res.old_fattr == NULL || res.new_fattr == NULL)
2773                 goto out;
2774
2775         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2776         if (!status) {
2777                 update_changeattr(old_dir, &res.old_cinfo);
2778                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2779                 update_changeattr(new_dir, &res.new_cinfo);
2780                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2781         }
2782 out:
2783         nfs_free_fattr(res.new_fattr);
2784         nfs_free_fattr(res.old_fattr);
2785         return status;
2786 }
2787
2788 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2789                 struct inode *new_dir, struct qstr *new_name)
2790 {
2791         struct nfs4_exception exception = { };
2792         int err;
2793         do {
2794                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2795                                 _nfs4_proc_rename(old_dir, old_name,
2796                                         new_dir, new_name),
2797                                 &exception);
2798         } while (exception.retry);
2799         return err;
2800 }
2801
2802 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2803 {
2804         struct nfs_server *server = NFS_SERVER(inode);
2805         struct nfs4_link_arg arg = {
2806                 .fh     = NFS_FH(inode),
2807                 .dir_fh = NFS_FH(dir),
2808                 .name   = name,
2809                 .bitmask = server->attr_bitmask,
2810         };
2811         struct nfs4_link_res res = {
2812                 .server = server,
2813         };
2814         struct rpc_message msg = {
2815                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2816                 .rpc_argp = &arg,
2817                 .rpc_resp = &res,
2818         };
2819         int status = -ENOMEM;
2820
2821         res.fattr = nfs_alloc_fattr();
2822         res.dir_attr = nfs_alloc_fattr();
2823         if (res.fattr == NULL || res.dir_attr == NULL)
2824                 goto out;
2825
2826         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2827         if (!status) {
2828                 update_changeattr(dir, &res.cinfo);
2829                 nfs_post_op_update_inode(dir, res.dir_attr);
2830                 nfs_post_op_update_inode(inode, res.fattr);
2831         }
2832 out:
2833         nfs_free_fattr(res.dir_attr);
2834         nfs_free_fattr(res.fattr);
2835         return status;
2836 }
2837
2838 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2839 {
2840         struct nfs4_exception exception = { };
2841         int err;
2842         do {
2843                 err = nfs4_handle_exception(NFS_SERVER(inode),
2844                                 _nfs4_proc_link(inode, dir, name),
2845                                 &exception);
2846         } while (exception.retry);
2847         return err;
2848 }
2849
2850 struct nfs4_createdata {
2851         struct rpc_message msg;
2852         struct nfs4_create_arg arg;
2853         struct nfs4_create_res res;
2854         struct nfs_fh fh;
2855         struct nfs_fattr fattr;
2856         struct nfs_fattr dir_fattr;
2857 };
2858
2859 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2860                 struct qstr *name, struct iattr *sattr, u32 ftype)
2861 {
2862         struct nfs4_createdata *data;
2863
2864         data = kzalloc(sizeof(*data), GFP_KERNEL);
2865         if (data != NULL) {
2866                 struct nfs_server *server = NFS_SERVER(dir);
2867
2868                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2869                 data->msg.rpc_argp = &data->arg;
2870                 data->msg.rpc_resp = &data->res;
2871                 data->arg.dir_fh = NFS_FH(dir);
2872                 data->arg.server = server;
2873                 data->arg.name = name;
2874                 data->arg.attrs = sattr;
2875                 data->arg.ftype = ftype;
2876                 data->arg.bitmask = server->attr_bitmask;
2877                 data->res.server = server;
2878                 data->res.fh = &data->fh;
2879                 data->res.fattr = &data->fattr;
2880                 data->res.dir_fattr = &data->dir_fattr;
2881                 nfs_fattr_init(data->res.fattr);
2882                 nfs_fattr_init(data->res.dir_fattr);
2883         }
2884         return data;
2885 }
2886
2887 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2888 {
2889         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2890                                     &data->arg.seq_args, &data->res.seq_res, 1);
2891         if (status == 0) {
2892                 update_changeattr(dir, &data->res.dir_cinfo);
2893                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2894                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2895         }
2896         return status;
2897 }
2898
2899 static void nfs4_free_createdata(struct nfs4_createdata *data)
2900 {
2901         kfree(data);
2902 }
2903
2904 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2905                 struct page *page, unsigned int len, struct iattr *sattr)
2906 {
2907         struct nfs4_createdata *data;
2908         int status = -ENAMETOOLONG;
2909
2910         if (len > NFS4_MAXPATHLEN)
2911                 goto out;
2912
2913         status = -ENOMEM;
2914         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2915         if (data == NULL)
2916                 goto out;
2917
2918         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2919         data->arg.u.symlink.pages = &page;
2920         data->arg.u.symlink.len = len;
2921
2922         status = nfs4_do_create(dir, dentry, data);
2923
2924         nfs4_free_createdata(data);
2925 out:
2926         return status;
2927 }
2928
2929 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2930                 struct page *page, unsigned int len, struct iattr *sattr)
2931 {
2932         struct nfs4_exception exception = { };
2933         int err;
2934         do {
2935                 err = nfs4_handle_exception(NFS_SERVER(dir),
2936                                 _nfs4_proc_symlink(dir, dentry, page,
2937                                                         len, sattr),
2938                                 &exception);
2939         } while (exception.retry);
2940         return err;
2941 }
2942
2943 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2944                 struct iattr *sattr)
2945 {
2946         struct nfs4_createdata *data;
2947         int status = -ENOMEM;
2948
2949         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2950         if (data == NULL)
2951                 goto out;
2952
2953         status = nfs4_do_create(dir, dentry, data);
2954
2955         nfs4_free_createdata(data);
2956 out:
2957         return status;
2958 }
2959
2960 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2961                 struct iattr *sattr)
2962 {
2963         struct nfs4_exception exception = { };
2964         int err;
2965
2966         sattr->ia_mode &= ~current_umask();
2967         do {
2968                 err = nfs4_handle_exception(NFS_SERVER(dir),
2969                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2970                                 &exception);
2971         } while (exception.retry);
2972         return err;
2973 }
2974
2975 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2976                 u64 cookie, struct page **pages, unsigned int count, int plus)
2977 {
2978         struct inode            *dir = dentry->d_inode;
2979         struct nfs4_readdir_arg args = {
2980                 .fh = NFS_FH(dir),
2981                 .pages = pages,
2982                 .pgbase = 0,
2983                 .count = count,
2984                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2985                 .plus = plus,
2986         };
2987         struct nfs4_readdir_res res;
2988         struct rpc_message msg = {
2989                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2990                 .rpc_argp = &args,
2991                 .rpc_resp = &res,
2992                 .rpc_cred = cred,
2993         };
2994         int                     status;
2995
2996         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2997                         dentry->d_parent->d_name.name,
2998                         dentry->d_name.name,
2999                         (unsigned long long)cookie);
3000         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3001         res.pgbase = args.pgbase;
3002         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3003         if (status >= 0) {
3004                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3005                 status += args.pgbase;
3006         }
3007
3008         nfs_invalidate_atime(dir);
3009
3010         dprintk("%s: returns %d\n", __func__, status);
3011         return status;
3012 }
3013
3014 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3015                 u64 cookie, struct page **pages, unsigned int count, int plus)
3016 {
3017         struct nfs4_exception exception = { };
3018         int err;
3019         do {
3020                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3021                                 _nfs4_proc_readdir(dentry, cred, cookie,
3022                                         pages, count, plus),
3023                                 &exception);
3024         } while (exception.retry);
3025         return err;
3026 }
3027
3028 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3029                 struct iattr *sattr, dev_t rdev)
3030 {
3031         struct nfs4_createdata *data;
3032         int mode = sattr->ia_mode;
3033         int status = -ENOMEM;
3034
3035         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3036         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3037
3038         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3039         if (data == NULL)
3040                 goto out;
3041
3042         if (S_ISFIFO(mode))
3043                 data->arg.ftype = NF4FIFO;
3044         else if (S_ISBLK(mode)) {
3045                 data->arg.ftype = NF4BLK;
3046                 data->arg.u.device.specdata1 = MAJOR(rdev);
3047                 data->arg.u.device.specdata2 = MINOR(rdev);
3048         }
3049         else if (S_ISCHR(mode)) {
3050                 data->arg.ftype = NF4CHR;
3051                 data->arg.u.device.specdata1 = MAJOR(rdev);
3052                 data->arg.u.device.specdata2 = MINOR(rdev);
3053         }
3054
3055         status = nfs4_do_create(dir, dentry, data);
3056
3057         nfs4_free_createdata(data);
3058 out:
3059         return status;
3060 }
3061
3062 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3063                 struct iattr *sattr, dev_t rdev)
3064 {
3065         struct nfs4_exception exception = { };
3066         int err;
3067
3068         sattr->ia_mode &= ~current_umask();
3069         do {
3070                 err = nfs4_handle_exception(NFS_SERVER(dir),
3071                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3072                                 &exception);
3073         } while (exception.retry);
3074         return err;
3075 }
3076
3077 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3078                  struct nfs_fsstat *fsstat)
3079 {
3080         struct nfs4_statfs_arg args = {
3081                 .fh = fhandle,
3082                 .bitmask = server->attr_bitmask,
3083         };
3084         struct nfs4_statfs_res res = {
3085                 .fsstat = fsstat,
3086         };
3087         struct rpc_message msg = {
3088                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3089                 .rpc_argp = &args,
3090                 .rpc_resp = &res,
3091         };
3092
3093         nfs_fattr_init(fsstat->fattr);
3094         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3095 }
3096
3097 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3098 {
3099         struct nfs4_exception exception = { };
3100         int err;
3101         do {
3102                 err = nfs4_handle_exception(server,
3103                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3104                                 &exception);
3105         } while (exception.retry);
3106         return err;
3107 }
3108
3109 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3110                 struct nfs_fsinfo *fsinfo)
3111 {
3112         struct nfs4_fsinfo_arg args = {
3113                 .fh = fhandle,
3114                 .bitmask = server->attr_bitmask,
3115         };
3116         struct nfs4_fsinfo_res res = {
3117                 .fsinfo = fsinfo,
3118         };
3119         struct rpc_message msg = {
3120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3121                 .rpc_argp = &args,
3122                 .rpc_resp = &res,
3123         };
3124
3125         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3126 }
3127
3128 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3129 {
3130         struct nfs4_exception exception = { };
3131         int err;
3132
3133         do {
3134                 err = nfs4_handle_exception(server,
3135                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3136                                 &exception);
3137         } while (exception.retry);
3138         return err;
3139 }
3140
3141 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3142 {
3143         nfs_fattr_init(fsinfo->fattr);
3144         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3145 }
3146
3147 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3148                 struct nfs_pathconf *pathconf)
3149 {
3150         struct nfs4_pathconf_arg args = {
3151                 .fh = fhandle,
3152                 .bitmask = server->attr_bitmask,
3153         };
3154         struct nfs4_pathconf_res res = {
3155                 .pathconf = pathconf,
3156         };
3157         struct rpc_message msg = {
3158                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3159                 .rpc_argp = &args,
3160                 .rpc_resp = &res,
3161         };
3162
3163         /* None of the pathconf attributes are mandatory to implement */
3164         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3165                 memset(pathconf, 0, sizeof(*pathconf));
3166                 return 0;
3167         }
3168
3169         nfs_fattr_init(pathconf->fattr);
3170         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3171 }
3172
3173 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3174                 struct nfs_pathconf *pathconf)
3175 {
3176         struct nfs4_exception exception = { };
3177         int err;
3178
3179         do {
3180                 err = nfs4_handle_exception(server,
3181                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3182                                 &exception);
3183         } while (exception.retry);
3184         return err;
3185 }
3186
3187 void __nfs4_read_done_cb(struct nfs_read_data *data)
3188 {
3189         nfs_invalidate_atime(data->inode);
3190 }
3191
3192 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3193 {
3194         struct nfs_server *server = NFS_SERVER(data->inode);
3195
3196         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3197                 rpc_restart_call_prepare(task);
3198                 return -EAGAIN;
3199         }
3200
3201         __nfs4_read_done_cb(data);
3202         if (task->tk_status > 0)
3203                 renew_lease(server, data->timestamp);
3204         return 0;
3205 }
3206
3207 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3208 {
3209
3210         dprintk("--> %s\n", __func__);
3211
3212         if (!nfs4_sequence_done(task, &data->res.seq_res))
3213                 return -EAGAIN;
3214
3215         return data->read_done_cb ? data->read_done_cb(task, data) :
3216                                     nfs4_read_done_cb(task, data);
3217 }
3218
3219 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3220 {
3221         data->timestamp   = jiffies;
3222         data->read_done_cb = nfs4_read_done_cb;
3223         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3224 }
3225
3226 /* Reset the the nfs_read_data to send the read to the MDS. */
3227 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3228 {
3229         dprintk("%s Reset task for i/o through\n", __func__);
3230         put_lseg(data->lseg);
3231         data->lseg = NULL;
3232         /* offsets will differ in the dense stripe case */
3233         data->args.offset = data->mds_offset;
3234         data->ds_clp = NULL;
3235         data->args.fh     = NFS_FH(data->inode);
3236         data->read_done_cb = nfs4_read_done_cb;
3237         task->tk_ops = data->mds_ops;
3238         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3239 }
3240 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3241
3242 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3243 {
3244         struct inode *inode = data->inode;
3245
3246         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3247                 rpc_restart_call_prepare(task);
3248                 return -EAGAIN;
3249         }
3250         if (task->tk_status >= 0) {
3251                 renew_lease(NFS_SERVER(inode), data->timestamp);
3252                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3253         }
3254         return 0;
3255 }
3256
3257 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3258 {
3259         if (!nfs4_sequence_done(task, &data->res.seq_res))
3260                 return -EAGAIN;
3261         return data->write_done_cb ? data->write_done_cb(task, data) :
3262                 nfs4_write_done_cb(task, data);
3263 }
3264
3265 /* Reset the the nfs_write_data to send the write to the MDS. */
3266 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3267 {
3268         dprintk("%s Reset task for i/o through\n", __func__);
3269         put_lseg(data->lseg);
3270         data->lseg          = NULL;
3271         data->ds_clp        = NULL;
3272         data->write_done_cb = nfs4_write_done_cb;
3273         data->args.fh       = NFS_FH(data->inode);
3274         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3275         data->args.offset   = data->mds_offset;
3276         data->res.fattr     = &data->fattr;
3277         task->tk_ops        = data->mds_ops;
3278         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3279 }
3280 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3281
3282 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3283 {
3284         struct nfs_server *server = NFS_SERVER(data->inode);
3285
3286         if (data->lseg) {
3287                 data->args.bitmask = NULL;
3288                 data->res.fattr = NULL;
3289         } else
3290                 data->args.bitmask = server->cache_consistency_bitmask;
3291         if (!data->write_done_cb)
3292                 data->write_done_cb = nfs4_write_done_cb;
3293         data->res.server = server;
3294         data->timestamp   = jiffies;
3295
3296         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3297 }
3298
3299 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3300 {
3301         struct inode *inode = data->inode;
3302
3303         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3304                 rpc_restart_call_prepare(task);
3305                 return -EAGAIN;
3306         }
3307         nfs_refresh_inode(inode, data->res.fattr);
3308         return 0;
3309 }
3310
3311 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3312 {
3313         if (!nfs4_sequence_done(task, &data->res.seq_res))
3314                 return -EAGAIN;
3315         return data->write_done_cb(task, data);
3316 }
3317
3318 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3319 {
3320         struct nfs_server *server = NFS_SERVER(data->inode);
3321
3322         if (data->lseg) {
3323                 data->args.bitmask = NULL;
3324                 data->res.fattr = NULL;
3325         } else
3326                 data->args.bitmask = server->cache_consistency_bitmask;
3327         if (!data->write_done_cb)
3328                 data->write_done_cb = nfs4_commit_done_cb;
3329         data->res.server = server;
3330         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3331 }
3332
3333 struct nfs4_renewdata {
3334         struct nfs_client       *client;
3335         unsigned long           timestamp;
3336 };
3337
3338 /*
3339  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3340  * standalone procedure for queueing an asynchronous RENEW.
3341  */
3342 static void nfs4_renew_release(void *calldata)
3343 {
3344         struct nfs4_renewdata *data = calldata;
3345         struct nfs_client *clp = data->client;
3346
3347         if (atomic_read(&clp->cl_count) > 1)
3348                 nfs4_schedule_state_renewal(clp);
3349         nfs_put_client(clp);
3350         kfree(data);
3351 }
3352
3353 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3354 {
3355         struct nfs4_renewdata *data = calldata;
3356         struct nfs_client *clp = data->client;
3357         unsigned long timestamp = data->timestamp;
3358
3359         if (task->tk_status < 0) {
3360                 /* Unless we're shutting down, schedule state recovery! */
3361                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3362                         return;
3363                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3364                         nfs4_schedule_lease_recovery(clp);
3365                         return;
3366                 }
3367                 nfs4_schedule_path_down_recovery(clp);
3368         }
3369         do_renew_lease(clp, timestamp);
3370 }
3371
3372 static const struct rpc_call_ops nfs4_renew_ops = {
3373         .rpc_call_done = nfs4_renew_done,
3374         .rpc_release = nfs4_renew_release,
3375 };
3376
3377 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3378 {
3379         struct rpc_message msg = {
3380                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3381                 .rpc_argp       = clp,
3382                 .rpc_cred       = cred,
3383         };
3384         struct nfs4_renewdata *data;
3385
3386         if (renew_flags == 0)
3387                 return 0;
3388         if (!atomic_inc_not_zero(&clp->cl_count))
3389                 return -EIO;
3390         data = kmalloc(sizeof(*data), GFP_NOFS);
3391         if (data == NULL)
3392                 return -ENOMEM;
3393         data->client = clp;
3394         data->timestamp = jiffies;
3395         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3396                         &nfs4_renew_ops, data);
3397 }
3398
3399 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3400 {
3401         struct rpc_message msg = {
3402                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3403                 .rpc_argp       = clp,
3404                 .rpc_cred       = cred,
3405         };
3406         unsigned long now = jiffies;
3407         int status;
3408
3409         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3410         if (status < 0)
3411                 return status;
3412         do_renew_lease(clp, now);
3413         return 0;
3414 }
3415
3416 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3417 {
3418         return (server->caps & NFS_CAP_ACLS)
3419                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3420                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3421 }
3422
3423 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3424  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3425  * the stack.
3426  */
3427 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3428
3429 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3430                 struct page **pages, unsigned int *pgbase)
3431 {
3432         struct page *newpage, **spages;
3433         int rc = 0;
3434         size_t len;
3435         spages = pages;
3436
3437         do {
3438                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3439                 newpage = alloc_page(GFP_KERNEL);
3440
3441                 if (newpage == NULL)
3442                         goto unwind;
3443                 memcpy(page_address(newpage), buf, len);
3444                 buf += len;
3445                 buflen -= len;
3446                 *pages++ = newpage;
3447                 rc++;
3448         } while (buflen != 0);
3449
3450         return rc;
3451
3452 unwind:
3453         for(; rc > 0; rc--)
3454                 __free_page(spages[rc-1]);
3455         return -ENOMEM;
3456 }
3457
3458 struct nfs4_cached_acl {
3459         int cached;
3460         size_t len;
3461         char data[0];
3462 };
3463
3464 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3465 {
3466         struct nfs_inode *nfsi = NFS_I(inode);
3467
3468         spin_lock(&inode->i_lock);
3469         kfree(nfsi->nfs4_acl);
3470         nfsi->nfs4_acl = acl;
3471         spin_unlock(&inode->i_lock);
3472 }
3473
3474 static void nfs4_zap_acl_attr(struct inode *inode)
3475 {
3476         nfs4_set_cached_acl(inode, NULL);
3477 }
3478
3479 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3480 {
3481         struct nfs_inode *nfsi = NFS_I(inode);
3482         struct nfs4_cached_acl *acl;
3483         int ret = -ENOENT;
3484
3485         spin_lock(&inode->i_lock);
3486         acl = nfsi->nfs4_acl;
3487         if (acl == NULL)
3488                 goto out;
3489         if (buf == NULL) /* user is just asking for length */
3490                 goto out_len;
3491         if (acl->cached == 0)
3492                 goto out;
3493         ret = -ERANGE; /* see getxattr(2) man page */
3494         if (acl->len > buflen)
3495                 goto out;
3496         memcpy(buf, acl->data, acl->len);
3497 out_len:
3498         ret = acl->len;
3499 out:
3500         spin_unlock(&inode->i_lock);
3501         return ret;
3502 }
3503
3504 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3505 {
3506         struct nfs4_cached_acl *acl;
3507
3508         if (buf && acl_len <= PAGE_SIZE) {
3509                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3510                 if (acl == NULL)
3511                         goto out;
3512                 acl->cached = 1;
3513                 memcpy(acl->data, buf, acl_len);
3514         } else {
3515                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3516                 if (acl == NULL)
3517                         goto out;
3518                 acl->cached = 0;
3519         }
3520         acl->len = acl_len;
3521 out:
3522         nfs4_set_cached_acl(inode, acl);
3523 }
3524
3525 /*
3526  * The getxattr API returns the required buffer length when called with a
3527  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3528  * the required buf.  On a NULL buf, we send a page of data to the server
3529  * guessing that the ACL request can be serviced by a page. If so, we cache
3530  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3531  * the cache. If not so, we throw away the page, and cache the required
3532  * length. The next getxattr call will then produce another round trip to
3533  * the server, this time with the input buf of the required size.
3534  */
3535 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3536 {
3537         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3538         struct nfs_getaclargs args = {
3539                 .fh = NFS_FH(inode),
3540                 .acl_pages = pages,
3541                 .acl_len = buflen,
3542         };
3543         struct nfs_getaclres res = {
3544                 .acl_len = buflen,
3545         };
3546         void *resp_buf;
3547         struct rpc_message msg = {
3548                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3549                 .rpc_argp = &args,
3550                 .rpc_resp = &res,
3551         };
3552         int ret = -ENOMEM, npages, i, acl_len = 0;
3553
3554         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3555         /* As long as we're doing a round trip to the server anyway,
3556          * let's be prepared for a page of acl data. */
3557         if (npages == 0)
3558                 npages = 1;
3559
3560         for (i = 0; i < npages; i++) {
3561                 pages[i] = alloc_page(GFP_KERNEL);
3562                 if (!pages[i])
3563                         goto out_free;
3564         }
3565         if (npages > 1) {
3566                 /* for decoding across pages */
3567                 args.acl_scratch = alloc_page(GFP_KERNEL);
3568                 if (!args.acl_scratch)
3569                         goto out_free;
3570         }
3571         args.acl_len = npages * PAGE_SIZE;
3572         args.acl_pgbase = 0;
3573         /* Let decode_getfacl know not to fail if the ACL data is larger than
3574          * the page we send as a guess */
3575         if (buf == NULL)
3576                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3577         resp_buf = page_address(pages[0]);
3578
3579         dprintk("%s  buf %p buflen %ld npages %d args.acl_len %ld\n",
3580                 __func__, buf, buflen, npages, args.acl_len);
3581         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3582                              &msg, &args.seq_args, &res.seq_res, 0);
3583         if (ret)
3584                 goto out_free;
3585
3586         acl_len = res.acl_len - res.acl_data_offset;
3587         if (acl_len > args.acl_len)
3588                 nfs4_write_cached_acl(inode, NULL, acl_len);
3589         else
3590                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3591                                       acl_len);
3592         if (buf) {
3593                 ret = -ERANGE;
3594                 if (acl_len > buflen)
3595                         goto out_free;
3596                 _copy_from_pages(buf, pages, res.acl_data_offset,
3597                                 res.acl_len);
3598         }
3599         ret = acl_len;
3600 out_free:
3601         for (i = 0; i < npages; i++)
3602                 if (pages[i])
3603                         __free_page(pages[i]);
3604         if (args.acl_scratch)
3605                 __free_page(args.acl_scratch);
3606         return ret;
3607 }
3608
3609 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3610 {
3611         struct nfs4_exception exception = { };
3612         ssize_t ret;
3613         do {
3614                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3615                 if (ret >= 0)
3616                         break;
3617                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3618         } while (exception.retry);
3619         return ret;
3620 }
3621
3622 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3623 {
3624         struct nfs_server *server = NFS_SERVER(inode);
3625         int ret;
3626
3627         if (!nfs4_server_supports_acls(server))
3628                 return -EOPNOTSUPP;
3629         ret = nfs_revalidate_inode(server, inode);
3630         if (ret < 0)
3631                 return ret;
3632         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3633                 nfs_zap_acl_cache(inode);
3634         ret = nfs4_read_cached_acl(inode, buf, buflen);
3635         if (ret != -ENOENT)
3636                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3637                  * but no cached acl data, just the acl length */
3638                 return ret;
3639         return nfs4_get_acl_uncached(inode, buf, buflen);
3640 }
3641
3642 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3643 {
3644         struct nfs_server *server = NFS_SERVER(inode);
3645         struct page *pages[NFS4ACL_MAXPAGES];
3646         struct nfs_setaclargs arg = {
3647                 .fh             = NFS_FH(inode),
3648                 .acl_pages      = pages,
3649                 .acl_len        = buflen,
3650         };
3651         struct nfs_setaclres res;
3652         struct rpc_message msg = {
3653                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3654                 .rpc_argp       = &arg,
3655                 .rpc_resp       = &res,
3656         };
3657         int ret, i;
3658
3659         if (!nfs4_server_supports_acls(server))
3660                 return -EOPNOTSUPP;
3661         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3662         if (i < 0)
3663                 return i;
3664         nfs_inode_return_delegation(inode);
3665         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3666
3667         /*
3668          * Free each page after tx, so the only ref left is
3669          * held by the network stack
3670          */
3671         for (; i > 0; i--)
3672                 put_page(pages[i-1]);
3673
3674         /*
3675          * Acl update can result in inode attribute update.
3676          * so mark the attribute cache invalid.
3677          */
3678         spin_lock(&inode->i_lock);
3679         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3680         spin_unlock(&inode->i_lock);
3681         nfs_access_zap_cache(inode);
3682         nfs_zap_acl_cache(inode);
3683         return ret;
3684 }
3685
3686 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3687 {
3688         struct nfs4_exception exception = { };
3689         int err;
3690         do {
3691                 err = nfs4_handle_exception(NFS_SERVER(inode),
3692                                 __nfs4_proc_set_acl(inode, buf, buflen),
3693                                 &exception);
3694         } while (exception.retry);
3695         return err;
3696 }
3697
3698 static int
3699 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3700 {
3701         struct nfs_client *clp = server->nfs_client;
3702
3703         if (task->tk_status >= 0)
3704                 return 0;
3705         switch(task->tk_status) {
3706                 case -NFS4ERR_ADMIN_REVOKED:
3707                 case -NFS4ERR_BAD_STATEID:
3708                 case -NFS4ERR_OPENMODE:
3709                         if (state == NULL)
3710                                 break;
3711                         nfs4_schedule_stateid_recovery(server, state);
3712                         goto wait_on_recovery;
3713                 case -NFS4ERR_EXPIRED:
3714                         if (state != NULL)
3715                                 nfs4_schedule_stateid_recovery(server, state);
3716                 case -NFS4ERR_STALE_STATEID:
3717                 case -NFS4ERR_STALE_CLIENTID:
3718                         nfs4_schedule_lease_recovery(clp);
3719                         goto wait_on_recovery;
3720 #if defined(CONFIG_NFS_V4_1)
3721                 case -NFS4ERR_BADSESSION:
3722                 case -NFS4ERR_BADSLOT:
3723                 case -NFS4ERR_BAD_HIGH_SLOT:
3724                 case -NFS4ERR_DEADSESSION:
3725                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3726                 case -NFS4ERR_SEQ_FALSE_RETRY:
3727                 case -NFS4ERR_SEQ_MISORDERED:
3728                         dprintk("%s ERROR %d, Reset session\n", __func__,
3729                                 task->tk_status);
3730                         nfs4_schedule_session_recovery(clp->cl_session);
3731                         task->tk_status = 0;
3732                         return -EAGAIN;
3733 #endif /* CONFIG_NFS_V4_1 */
3734                 case -NFS4ERR_DELAY:
3735                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3736                 case -NFS4ERR_GRACE:
3737                 case -EKEYEXPIRED:
3738                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3739                         task->tk_status = 0;
3740                         return -EAGAIN;
3741                 case -NFS4ERR_RETRY_UNCACHED_REP:
3742                 case -NFS4ERR_OLD_STATEID:
3743                         task->tk_status = 0;
3744                         return -EAGAIN;
3745         }
3746         task->tk_status = nfs4_map_errors(task->tk_status);
3747         return 0;
3748 wait_on_recovery:
3749         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3750         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3751                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3752         task->tk_status = 0;
3753         return -EAGAIN;
3754 }
3755
3756 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3757                 unsigned short port, struct rpc_cred *cred,
3758                 struct nfs4_setclientid_res *res)
3759 {
3760         nfs4_verifier sc_verifier;
3761         struct nfs4_setclientid setclientid = {
3762                 .sc_verifier = &sc_verifier,
3763                 .sc_prog = program,
3764                 .sc_cb_ident = clp->cl_cb_ident,
3765         };
3766         struct rpc_message msg = {
3767                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3768                 .rpc_argp = &setclientid,
3769                 .rpc_resp = res,
3770                 .rpc_cred = cred,
3771         };
3772         __be32 *p;
3773         int loop = 0;
3774         int status;
3775
3776         p = (__be32*)sc_verifier.data;
3777         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3778         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3779
3780         for(;;) {
3781                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3782                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3783                                 clp->cl_ipaddr,
3784                                 rpc_peeraddr2str(clp->cl_rpcclient,
3785                                                         RPC_DISPLAY_ADDR),
3786                                 rpc_peeraddr2str(clp->cl_rpcclient,
3787                                                         RPC_DISPLAY_PROTO),
3788                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3789                                 clp->cl_id_uniquifier);
3790                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3791                                 sizeof(setclientid.sc_netid),
3792                                 rpc_peeraddr2str(clp->cl_rpcclient,
3793                                                         RPC_DISPLAY_NETID));
3794                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3795                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3796                                 clp->cl_ipaddr, port >> 8, port & 255);
3797
3798                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3799                 if (status != -NFS4ERR_CLID_INUSE)
3800                         break;
3801                 if (loop != 0) {
3802                         ++clp->cl_id_uniquifier;
3803                         break;
3804                 }
3805                 ++loop;
3806                 ssleep(clp->cl_lease_time / HZ + 1);
3807         }
3808         return status;
3809 }
3810
3811 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3812                 struct nfs4_setclientid_res *arg,
3813                 struct rpc_cred *cred)
3814 {
3815         struct nfs_fsinfo fsinfo;
3816         struct rpc_message msg = {
3817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3818                 .rpc_argp = arg,
3819                 .rpc_resp = &fsinfo,
3820                 .rpc_cred = cred,
3821         };
3822         unsigned long now;
3823         int status;
3824
3825         now = jiffies;
3826         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3827         if (status == 0) {
3828                 spin_lock(&clp->cl_lock);
3829                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3830                 clp->cl_last_renewal = now;
3831                 spin_unlock(&clp->cl_lock);
3832         }
3833         return status;
3834 }
3835
3836 struct nfs4_delegreturndata {
3837         struct nfs4_delegreturnargs args;
3838         struct nfs4_delegreturnres res;
3839         struct nfs_fh fh;
3840         nfs4_stateid stateid;
3841         unsigned long timestamp;
3842         struct nfs_fattr fattr;
3843         int rpc_status;
3844 };
3845
3846 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3847 {
3848         struct nfs4_delegreturndata *data = calldata;
3849
3850         if (!nfs4_sequence_done(task, &data->res.seq_res))
3851                 return;
3852
3853         switch (task->tk_status) {
3854         case -NFS4ERR_STALE_STATEID:
3855         case -NFS4ERR_EXPIRED:
3856         case 0:
3857                 renew_lease(data->res.server, data->timestamp);
3858                 break;
3859         default:
3860                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3861                                 -EAGAIN) {
3862                         rpc_restart_call_prepare(task);
3863                         return;
3864                 }
3865         }
3866         data->rpc_status = task->tk_status;
3867 }
3868
3869 static void nfs4_delegreturn_release(void *calldata)
3870 {
3871         kfree(calldata);
3872 }
3873
3874 #if defined(CONFIG_NFS_V4_1)
3875 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3876 {
3877         struct nfs4_delegreturndata *d_data;
3878
3879         d_data = (struct nfs4_delegreturndata *)data;
3880
3881         if (nfs4_setup_sequence(d_data->res.server,
3882                                 &d_data->args.seq_args,
3883                                 &d_data->res.seq_res, 1, task))
3884                 return;
3885         rpc_call_start(task);
3886 }
3887 #endif /* CONFIG_NFS_V4_1 */
3888
3889 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3890 #if defined(CONFIG_NFS_V4_1)
3891         .rpc_call_prepare = nfs4_delegreturn_prepare,
3892 #endif /* CONFIG_NFS_V4_1 */
3893         .rpc_call_done = nfs4_delegreturn_done,
3894         .rpc_release = nfs4_delegreturn_release,
3895 };
3896
3897 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3898 {
3899         struct nfs4_delegreturndata *data;
3900         struct nfs_server *server = NFS_SERVER(inode);
3901         struct rpc_task *task;
3902         struct rpc_message msg = {
3903                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3904                 .rpc_cred = cred,
3905         };
3906         struct rpc_task_setup task_setup_data = {
3907                 .rpc_client = server->client,
3908                 .rpc_message = &msg,
3909                 .callback_ops = &nfs4_delegreturn_ops,
3910                 .flags = RPC_TASK_ASYNC,
3911         };
3912         int status = 0;
3913
3914         data = kzalloc(sizeof(*data), GFP_NOFS);
3915         if (data == NULL)
3916                 return -ENOMEM;
3917         data->args.fhandle = &data->fh;
3918         data->args.stateid = &data->stateid;
3919         data->args.bitmask = server->attr_bitmask;
3920         nfs_copy_fh(&data->fh, NFS_FH(inode));
3921         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3922         data->res.fattr = &data->fattr;
3923         data->res.server = server;
3924         nfs_fattr_init(data->res.fattr);
3925         data->timestamp = jiffies;
3926         data->rpc_status = 0;
3927
3928         task_setup_data.callback_data = data;
3929         msg.rpc_argp = &data->args;
3930         msg.rpc_resp = &data->res;
3931         task = rpc_run_task(&task_setup_data);
3932         if (IS_ERR(task))
3933                 return PTR_ERR(task);
3934         if (!issync)
3935                 goto out;
3936         status = nfs4_wait_for_completion_rpc_task(task);
3937         if (status != 0)
3938                 goto out;
3939         status = data->rpc_status;
3940         if (status != 0)
3941                 goto out;
3942         nfs_refresh_inode(inode, &data->fattr);
3943 out:
3944         rpc_put_task(task);
3945         return status;
3946 }
3947
3948 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3949 {
3950         struct nfs_server *server = NFS_SERVER(inode);
3951         struct nfs4_exception exception = { };
3952         int err;
3953         do {
3954                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3955                 switch (err) {
3956                         case -NFS4ERR_STALE_STATEID:
3957                         case -NFS4ERR_EXPIRED:
3958                         case 0:
3959                                 return 0;
3960                 }
3961                 err = nfs4_handle_exception(server, err, &exception);
3962         } while (exception.retry);
3963         return err;
3964 }
3965
3966 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3967 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3968
3969 /*
3970  * sleep, with exponential backoff, and retry the LOCK operation.
3971  */
3972 static unsigned long
3973 nfs4_set_lock_task_retry(unsigned long timeout)
3974 {
3975         schedule_timeout_killable(timeout);
3976         timeout <<= 1;
3977         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3978                 return NFS4_LOCK_MAXTIMEOUT;
3979         return timeout;
3980 }
3981
3982 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3983 {
3984         struct inode *inode = state->inode;
3985         struct nfs_server *server = NFS_SERVER(inode);
3986         struct nfs_client *clp = server->nfs_client;
3987         struct nfs_lockt_args arg = {
3988                 .fh = NFS_FH(inode),
3989                 .fl = request,
3990         };
3991         struct nfs_lockt_res res = {
3992                 .denied = request,
3993         };
3994         struct rpc_message msg = {
3995                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3996                 .rpc_argp       = &arg,
3997                 .rpc_resp       = &res,
3998                 .rpc_cred       = state->owner->so_cred,
3999         };
4000         struct nfs4_lock_state *lsp;
4001         int status;
4002
4003         arg.lock_owner.clientid = clp->cl_clientid;
4004         status = nfs4_set_lock_state(state, request);
4005         if (status != 0)
4006                 goto out;
4007         lsp = request->fl_u.nfs4_fl.owner;
4008         arg.lock_owner.id = lsp->ls_id.id;
4009         arg.lock_owner.s_dev = server->s_dev;
4010         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4011         switch (status) {
4012                 case 0:
4013                         request->fl_type = F_UNLCK;
4014                         break;
4015                 case -NFS4ERR_DENIED:
4016                         status = 0;
4017         }
4018         request->fl_ops->fl_release_private(request);
4019 out:
4020         return status;
4021 }
4022
4023 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4024 {
4025         struct nfs4_exception exception = { };
4026         int err;
4027
4028         do {
4029                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4030                                 _nfs4_proc_getlk(state, cmd, request),
4031                                 &exception);
4032         } while (exception.retry);
4033         return err;
4034 }
4035
4036 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4037 {
4038         int res = 0;
4039         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4040                 case FL_POSIX:
4041                         res = posix_lock_file_wait(file, fl);
4042                         break;
4043                 case FL_FLOCK:
4044                         res = flock_lock_file_wait(file, fl);
4045                         break;
4046                 default:
4047                         BUG();
4048         }
4049         return res;
4050 }
4051
4052 struct nfs4_unlockdata {
4053         struct nfs_locku_args arg;
4054         struct nfs_locku_res res;
4055         struct nfs4_lock_state *lsp;
4056         struct nfs_open_context *ctx;
4057         struct file_lock fl;
4058         const struct nfs_server *server;
4059         unsigned long timestamp;
4060 };
4061
4062 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4063                 struct nfs_open_context *ctx,
4064                 struct nfs4_lock_state *lsp,
4065                 struct nfs_seqid *seqid)
4066 {
4067         struct nfs4_unlockdata *p;
4068         struct inode *inode = lsp->ls_state->inode;
4069
4070         p = kzalloc(sizeof(*p), GFP_NOFS);
4071         if (p == NULL)
4072                 return NULL;
4073         p->arg.fh = NFS_FH(inode);
4074         p->arg.fl = &p->fl;
4075         p->arg.seqid = seqid;
4076         p->res.seqid = seqid;
4077         p->arg.stateid = &lsp->ls_stateid;
4078         p->lsp = lsp;
4079         atomic_inc(&lsp->ls_count);
4080         /* Ensure we don't close file until we're done freeing locks! */
4081         p->ctx = get_nfs_open_context(ctx);
4082         memcpy(&p->fl, fl, sizeof(p->fl));
4083         p->server = NFS_SERVER(inode);
4084         return p;
4085 }
4086
4087 static void nfs4_locku_release_calldata(void *data)
4088 {
4089         struct nfs4_unlockdata *calldata = data;
4090         nfs_free_seqid(calldata->arg.seqid);
4091         nfs4_put_lock_state(calldata->lsp);
4092         put_nfs_open_context(calldata->ctx);
4093         kfree(calldata);
4094 }
4095
4096 static void nfs4_locku_done(struct rpc_task *task, void *data)
4097 {
4098         struct nfs4_unlockdata *calldata = data;
4099
4100         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4101                 return;
4102         switch (task->tk_status) {
4103                 case 0:
4104                         memcpy(calldata->lsp->ls_stateid.data,
4105                                         calldata->res.stateid.data,
4106                                         sizeof(calldata->lsp->ls_stateid.data));
4107                         renew_lease(calldata->server, calldata->timestamp);
4108                         break;
4109                 case -NFS4ERR_BAD_STATEID:
4110                 case -NFS4ERR_OLD_STATEID:
4111                 case -NFS4ERR_STALE_STATEID:
4112                 case -NFS4ERR_EXPIRED:
4113                         break;
4114                 default:
4115                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4116                                 rpc_restart_call_prepare(task);
4117         }
4118 }
4119
4120 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4121 {
4122         struct nfs4_unlockdata *calldata = data;
4123
4124         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4125                 return;
4126         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4127                 /* Note: exit _without_ running nfs4_locku_done */
4128                 task->tk_action = NULL;
4129                 return;
4130         }
4131         calldata->timestamp = jiffies;
4132         if (nfs4_setup_sequence(calldata->server,
4133                                 &calldata->arg.seq_args,
4134                                 &calldata->res.seq_res, 1, task))
4135                 return;
4136         rpc_call_start(task);
4137 }
4138
4139 static const struct rpc_call_ops nfs4_locku_ops = {
4140         .rpc_call_prepare = nfs4_locku_prepare,
4141         .rpc_call_done = nfs4_locku_done,
4142         .rpc_release = nfs4_locku_release_calldata,
4143 };
4144
4145 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4146                 struct nfs_open_context *ctx,
4147                 struct nfs4_lock_state *lsp,
4148                 struct nfs_seqid *seqid)
4149 {
4150         struct nfs4_unlockdata *data;
4151         struct rpc_message msg = {
4152                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4153                 .rpc_cred = ctx->cred,
4154         };
4155         struct rpc_task_setup task_setup_data = {
4156                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4157                 .rpc_message = &msg,
4158                 .callback_ops = &nfs4_locku_ops,
4159                 .workqueue = nfsiod_workqueue,
4160                 .flags = RPC_TASK_ASYNC,
4161         };
4162
4163         /* Ensure this is an unlock - when canceling a lock, the
4164          * canceled lock is passed in, and it won't be an unlock.
4165          */
4166         fl->fl_type = F_UNLCK;
4167
4168         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4169         if (data == NULL) {
4170                 nfs_free_seqid(seqid);
4171                 return ERR_PTR(-ENOMEM);
4172         }
4173
4174         msg.rpc_argp = &data->arg;
4175         msg.rpc_resp = &data->res;
4176         task_setup_data.callback_data = data;
4177         return rpc_run_task(&task_setup_data);
4178 }
4179
4180 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4181 {
4182         struct nfs_inode *nfsi = NFS_I(state->inode);
4183         struct nfs_seqid *seqid;
4184         struct nfs4_lock_state *lsp;
4185         struct rpc_task *task;
4186         int status = 0;
4187         unsigned char fl_flags = request->fl_flags;
4188
4189         status = nfs4_set_lock_state(state, request);
4190         /* Unlock _before_ we do the RPC call */
4191         request->fl_flags |= FL_EXISTS;
4192         down_read(&nfsi->rwsem);
4193         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4194                 up_read(&nfsi->rwsem);
4195                 goto out;
4196         }
4197         up_read(&nfsi->rwsem);
4198         if (status != 0)
4199                 goto out;
4200         /* Is this a delegated lock? */
4201         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4202                 goto out;
4203         lsp = request->fl_u.nfs4_fl.owner;
4204         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4205         status = -ENOMEM;
4206         if (seqid == NULL)
4207                 goto out;
4208         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4209         status = PTR_ERR(task);
4210         if (IS_ERR(task))
4211                 goto out;
4212         status = nfs4_wait_for_completion_rpc_task(task);
4213         rpc_put_task(task);
4214 out:
4215         request->fl_flags = fl_flags;
4216         return status;
4217 }
4218
4219 struct nfs4_lockdata {
4220         struct nfs_lock_args arg;
4221         struct nfs_lock_res res;
4222         struct nfs4_lock_state *lsp;
4223         struct nfs_open_context *ctx;
4224         struct file_lock fl;
4225         unsigned long timestamp;
4226         int rpc_status;
4227         int cancelled;
4228         struct nfs_server *server;
4229 };
4230
4231 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4232                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4233                 gfp_t gfp_mask)
4234 {
4235         struct nfs4_lockdata *p;
4236         struct inode *inode = lsp->ls_state->inode;
4237         struct nfs_server *server = NFS_SERVER(inode);
4238
4239         p = kzalloc(sizeof(*p), gfp_mask);
4240         if (p == NULL)
4241                 return NULL;
4242
4243         p->arg.fh = NFS_FH(inode);
4244         p->arg.fl = &p->fl;
4245         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4246         if (p->arg.open_seqid == NULL)
4247                 goto out_free;
4248         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4249         if (p->arg.lock_seqid == NULL)
4250                 goto out_free_seqid;
4251         p->arg.lock_stateid = &lsp->ls_stateid;
4252         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4253         p->arg.lock_owner.id = lsp->ls_id.id;
4254         p->arg.lock_owner.s_dev = server->s_dev;
4255         p->res.lock_seqid = p->arg.lock_seqid;
4256         p->lsp = lsp;
4257         p->server = server;
4258         atomic_inc(&lsp->ls_count);
4259         p->ctx = get_nfs_open_context(ctx);
4260         memcpy(&p->fl, fl, sizeof(p->fl));
4261         return p;
4262 out_free_seqid:
4263         nfs_free_seqid(p->arg.open_seqid);
4264 out_free:
4265         kfree(p);
4266         return NULL;
4267 }
4268
4269 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4270 {
4271         struct nfs4_lockdata *data = calldata;
4272         struct nfs4_state *state = data->lsp->ls_state;
4273
4274         dprintk("%s: begin!\n", __func__);
4275         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4276                 return;
4277         /* Do we need to do an open_to_lock_owner? */
4278         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4279                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4280                         return;
4281                 data->arg.open_stateid = &state->stateid;
4282                 data->arg.new_lock_owner = 1;
4283                 data->res.open_seqid = data->arg.open_seqid;
4284         } else
4285                 data->arg.new_lock_owner = 0;
4286         data->timestamp = jiffies;
4287         if (nfs4_setup_sequence(data->server,
4288                                 &data->arg.seq_args,
4289                                 &data->res.seq_res, 1, task))
4290                 return;
4291         rpc_call_start(task);
4292         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4293 }
4294
4295 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4296 {
4297         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4298         nfs4_lock_prepare(task, calldata);
4299 }
4300
4301 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4302 {
4303         struct nfs4_lockdata *data = calldata;
4304
4305         dprintk("%s: begin!\n", __func__);
4306
4307         if (!nfs4_sequence_done(task, &data->res.seq_res))
4308                 return;
4309
4310         data->rpc_status = task->tk_status;
4311         if (data->arg.new_lock_owner != 0) {
4312                 if (data->rpc_status == 0)
4313                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4314                 else
4315                         goto out;
4316         }
4317         if (data->rpc_status == 0) {
4318                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4319                                         sizeof(data->lsp->ls_stateid.data));
4320                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4321                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4322         }
4323 out:
4324         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4325 }
4326
4327 static void nfs4_lock_release(void *calldata)
4328 {
4329         struct nfs4_lockdata *data = calldata;
4330
4331         dprintk("%s: begin!\n", __func__);
4332         nfs_free_seqid(data->arg.open_seqid);
4333         if (data->cancelled != 0) {
4334                 struct rpc_task *task;
4335                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4336                                 data->arg.lock_seqid);
4337                 if (!IS_ERR(task))
4338                         rpc_put_task_async(task);
4339                 dprintk("%s: cancelling lock!\n", __func__);
4340         } else
4341                 nfs_free_seqid(data->arg.lock_seqid);
4342         nfs4_put_lock_state(data->lsp);
4343         put_nfs_open_context(data->ctx);
4344         kfree(data);
4345         dprintk("%s: done!\n", __func__);
4346 }
4347
4348 static const struct rpc_call_ops nfs4_lock_ops = {
4349         .rpc_call_prepare = nfs4_lock_prepare,
4350         .rpc_call_done = nfs4_lock_done,
4351         .rpc_release = nfs4_lock_release,
4352 };
4353
4354 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4355         .rpc_call_prepare = nfs4_recover_lock_prepare,
4356         .rpc_call_done = nfs4_lock_done,
4357         .rpc_release = nfs4_lock_release,
4358 };
4359
4360 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4361 {
4362         switch (error) {
4363         case -NFS4ERR_ADMIN_REVOKED:
4364         case -NFS4ERR_BAD_STATEID:
4365                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4366                 if (new_lock_owner != 0 ||
4367                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4368                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4369                 break;
4370         case -NFS4ERR_STALE_STATEID:
4371                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4372         case -NFS4ERR_EXPIRED:
4373                 nfs4_schedule_lease_recovery(server->nfs_client);
4374         };
4375 }
4376
4377 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4378 {
4379         struct nfs4_lockdata *data;
4380         struct rpc_task *task;
4381         struct rpc_message msg = {
4382                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4383                 .rpc_cred = state->owner->so_cred,
4384         };
4385         struct rpc_task_setup task_setup_data = {
4386                 .rpc_client = NFS_CLIENT(state->inode),
4387                 .rpc_message = &msg,
4388                 .callback_ops = &nfs4_lock_ops,
4389                 .workqueue = nfsiod_workqueue,
4390                 .flags = RPC_TASK_ASYNC,
4391         };
4392         int ret;
4393
4394         dprintk("%s: begin!\n", __func__);
4395         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4396                         fl->fl_u.nfs4_fl.owner,
4397                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4398         if (data == NULL)
4399                 return -ENOMEM;
4400         if (IS_SETLKW(cmd))
4401                 data->arg.block = 1;
4402         if (recovery_type > NFS_LOCK_NEW) {
4403                 if (recovery_type == NFS_LOCK_RECLAIM)
4404                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4405                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4406         }
4407         msg.rpc_argp = &data->arg;
4408         msg.rpc_resp = &data->res;
4409         task_setup_data.callback_data = data;
4410         task = rpc_run_task(&task_setup_data);
4411         if (IS_ERR(task))
4412                 return PTR_ERR(task);
4413         ret = nfs4_wait_for_completion_rpc_task(task);
4414         if (ret == 0) {
4415                 ret = data->rpc_status;
4416                 if (ret)
4417                         nfs4_handle_setlk_error(data->server, data->lsp,
4418                                         data->arg.new_lock_owner, ret);
4419         } else
4420                 data->cancelled = 1;
4421         rpc_put_task(task);
4422         dprintk("%s: done, ret = %d!\n", __func__, ret);
4423         return ret;
4424 }
4425
4426 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4427 {
4428         struct nfs_server *server = NFS_SERVER(state->inode);
4429         struct nfs4_exception exception = { };
4430         int err;
4431
4432         do {
4433                 /* Cache the lock if possible... */
4434                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4435                         return 0;
4436                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4437                 if (err != -NFS4ERR_DELAY)
4438                         break;
4439                 nfs4_handle_exception(server, err, &exception);
4440         } while (exception.retry);
4441         return err;
4442 }
4443
4444 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4445 {
4446         struct nfs_server *server = NFS_SERVER(state->inode);
4447         struct nfs4_exception exception = { };
4448         int err;
4449
4450         err = nfs4_set_lock_state(state, request);
4451         if (err != 0)
4452                 return err;
4453         do {
4454                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4455                         return 0;
4456                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4457                 switch (err) {
4458                 default:
4459                         goto out;
4460                 case -NFS4ERR_GRACE:
4461                 case -NFS4ERR_DELAY:
4462                         nfs4_handle_exception(server, err, &exception);
4463                         err = 0;
4464                 }
4465         } while (exception.retry);
4466 out:
4467         return err;
4468 }
4469
4470 #if defined(CONFIG_NFS_V4_1)
4471 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4472 {
4473         int status;
4474         struct nfs_server *server = NFS_SERVER(state->inode);
4475
4476         status = nfs41_test_stateid(server, state);
4477         if (status == NFS_OK)
4478                 return 0;
4479         nfs41_free_stateid(server, state);
4480         return nfs4_lock_expired(state, request);
4481 }
4482 #endif
4483
4484 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4485 {
4486         struct nfs_inode *nfsi = NFS_I(state->inode);
4487         unsigned char fl_flags = request->fl_flags;
4488         int status = -ENOLCK;
4489
4490         if ((fl_flags & FL_POSIX) &&
4491                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4492                 goto out;
4493         /* Is this a delegated open? */
4494         status = nfs4_set_lock_state(state, request);
4495         if (status != 0)
4496                 goto out;
4497         request->fl_flags |= FL_ACCESS;
4498         status = do_vfs_lock(request->fl_file, request);
4499         if (status < 0)
4500                 goto out;
4501         down_read(&nfsi->rwsem);
4502         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4503                 /* Yes: cache locks! */
4504                 /* ...but avoid races with delegation recall... */
4505                 request->fl_flags = fl_flags & ~FL_SLEEP;
4506                 status = do_vfs_lock(request->fl_file, request);
4507                 goto out_unlock;
4508         }
4509         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4510         if (status != 0)
4511                 goto out_unlock;
4512         /* Note: we always want to sleep here! */
4513         request->fl_flags = fl_flags | FL_SLEEP;
4514         if (do_vfs_lock(request->fl_file, request) < 0)
4515                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4516 out_unlock:
4517         up_read(&nfsi->rwsem);
4518 out:
4519         request->fl_flags = fl_flags;
4520         return status;
4521 }
4522
4523 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4524 {
4525         struct nfs4_exception exception = { };
4526         int err;
4527
4528         do {
4529                 err = _nfs4_proc_setlk(state, cmd, request);
4530                 if (err == -NFS4ERR_DENIED)
4531                         err = -EAGAIN;
4532                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4533                                 err, &exception);
4534         } while (exception.retry);
4535         return err;
4536 }
4537
4538 static int
4539 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4540 {
4541         struct nfs_open_context *ctx;
4542         struct nfs4_state *state;
4543         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4544         int status;
4545
4546         /* verify open state */
4547         ctx = nfs_file_open_context(filp);
4548         state = ctx->state;
4549
4550         if (request->fl_start < 0 || request->fl_end < 0)
4551                 return -EINVAL;
4552
4553         if (IS_GETLK(cmd)) {
4554                 if (state != NULL)
4555                         return nfs4_proc_getlk(state, F_GETLK, request);
4556                 return 0;
4557         }
4558
4559         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4560                 return -EINVAL;
4561
4562         if (request->fl_type == F_UNLCK) {
4563                 if (state != NULL)
4564                         return nfs4_proc_unlck(state, cmd, request);
4565                 return 0;
4566         }
4567
4568         if (state == NULL)
4569                 return -ENOLCK;
4570         do {
4571                 status = nfs4_proc_setlk(state, cmd, request);
4572                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4573                         break;
4574                 timeout = nfs4_set_lock_task_retry(timeout);
4575                 status = -ERESTARTSYS;
4576                 if (signalled())
4577                         break;
4578         } while(status < 0);
4579         return status;
4580 }
4581
4582 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4583 {
4584         struct nfs_server *server = NFS_SERVER(state->inode);
4585         struct nfs4_exception exception = { };
4586         int err;
4587
4588         err = nfs4_set_lock_state(state, fl);
4589         if (err != 0)
4590                 goto out;
4591         do {
4592                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4593                 switch (err) {
4594                         default:
4595                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4596                                                 __func__, err);
4597                         case 0:
4598                         case -ESTALE:
4599                                 goto out;
4600                         case -NFS4ERR_EXPIRED:
4601                                 nfs4_schedule_stateid_recovery(server, state);
4602                         case -NFS4ERR_STALE_CLIENTID:
4603                         case -NFS4ERR_STALE_STATEID:
4604                                 nfs4_schedule_lease_recovery(server->nfs_client);
4605                                 goto out;
4606                         case -NFS4ERR_BADSESSION:
4607                         case -NFS4ERR_BADSLOT:
4608                         case -NFS4ERR_BAD_HIGH_SLOT:
4609                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4610                         case -NFS4ERR_DEADSESSION:
4611                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4612                                 goto out;
4613                         case -ERESTARTSYS:
4614                                 /*
4615                                  * The show must go on: exit, but mark the
4616                                  * stateid as needing recovery.
4617                                  */
4618                         case -NFS4ERR_ADMIN_REVOKED:
4619                         case -NFS4ERR_BAD_STATEID:
4620                         case -NFS4ERR_OPENMODE:
4621                                 nfs4_schedule_stateid_recovery(server, state);
4622                                 err = 0;
4623                                 goto out;
4624                         case -EKEYEXPIRED:
4625                                 /*
4626                                  * User RPCSEC_GSS context has expired.
4627                                  * We cannot recover this stateid now, so
4628                                  * skip it and allow recovery thread to
4629                                  * proceed.
4630                                  */
4631                                 err = 0;
4632                                 goto out;
4633                         case -ENOMEM:
4634                         case -NFS4ERR_DENIED:
4635                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4636                                 err = 0;
4637                                 goto out;
4638                         case -NFS4ERR_DELAY:
4639                                 break;
4640                 }
4641                 err = nfs4_handle_exception(server, err, &exception);
4642         } while (exception.retry);
4643 out:
4644         return err;
4645 }
4646
4647 static void nfs4_release_lockowner_release(void *calldata)
4648 {
4649         kfree(calldata);
4650 }
4651
4652 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4653         .rpc_release = nfs4_release_lockowner_release,
4654 };
4655
4656 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4657 {
4658         struct nfs_server *server = lsp->ls_state->owner->so_server;
4659         struct nfs_release_lockowner_args *args;
4660         struct rpc_message msg = {
4661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4662         };
4663
4664         if (server->nfs_client->cl_mvops->minor_version != 0)
4665                 return;
4666         args = kmalloc(sizeof(*args), GFP_NOFS);
4667         if (!args)
4668                 return;
4669         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4670         args->lock_owner.id = lsp->ls_id.id;
4671         args->lock_owner.s_dev = server->s_dev;
4672         msg.rpc_argp = args;
4673         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4674 }
4675
4676 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4677
4678 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4679                                    const void *buf, size_t buflen,
4680                                    int flags, int type)
4681 {
4682         if (strcmp(key, "") != 0)
4683                 return -EINVAL;
4684
4685         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4686 }
4687
4688 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4689                                    void *buf, size_t buflen, int type)
4690 {
4691         if (strcmp(key, "") != 0)
4692                 return -EINVAL;
4693
4694         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4695 }
4696
4697 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4698                                        size_t list_len, const char *name,
4699                                        size_t name_len, int type)
4700 {
4701         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4702
4703         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4704                 return 0;
4705
4706         if (list && len <= list_len)
4707                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4708         return len;
4709 }
4710
4711 /*
4712  * nfs_fhget will use either the mounted_on_fileid or the fileid
4713  */
4714 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4715 {
4716         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4717                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4718               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4719               (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4720                 return;
4721
4722         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4723                 NFS_ATTR_FATTR_NLINK;
4724         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4725         fattr->nlink = 2;
4726 }
4727
4728 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4729                 struct nfs4_fs_locations *fs_locations, struct page *page)
4730 {
4731         struct nfs_server *server = NFS_SERVER(dir);
4732         u32 bitmask[2] = {
4733                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4734         };
4735         struct nfs4_fs_locations_arg args = {
4736                 .dir_fh = NFS_FH(dir),
4737                 .name = name,
4738                 .page = page,
4739                 .bitmask = bitmask,
4740         };
4741         struct nfs4_fs_locations_res res = {
4742                 .fs_locations = fs_locations,
4743         };
4744         struct rpc_message msg = {
4745                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4746                 .rpc_argp = &args,
4747                 .rpc_resp = &res,
4748         };
4749         int status;
4750
4751         dprintk("%s: start\n", __func__);
4752
4753         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4754          * is not supported */
4755         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4756                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4757         else
4758                 bitmask[0] |= FATTR4_WORD0_FILEID;
4759
4760         nfs_fattr_init(&fs_locations->fattr);
4761         fs_locations->server = server;
4762         fs_locations->nlocations = 0;
4763         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4764         dprintk("%s: returned status = %d\n", __func__, status);
4765         return status;
4766 }
4767
4768 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4769 {
4770         int status;
4771         struct nfs4_secinfo_arg args = {
4772                 .dir_fh = NFS_FH(dir),
4773                 .name   = name,
4774         };
4775         struct nfs4_secinfo_res res = {
4776                 .flavors     = flavors,
4777         };
4778         struct rpc_message msg = {
4779                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4780                 .rpc_argp = &args,
4781                 .rpc_resp = &res,
4782         };
4783
4784         dprintk("NFS call  secinfo %s\n", name->name);
4785         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4786         dprintk("NFS reply  secinfo: %d\n", status);
4787         return status;
4788 }
4789
4790 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4791 {
4792         struct nfs4_exception exception = { };
4793         int err;
4794         do {
4795                 err = nfs4_handle_exception(NFS_SERVER(dir),
4796                                 _nfs4_proc_secinfo(dir, name, flavors),
4797                                 &exception);
4798         } while (exception.retry);
4799         return err;
4800 }
4801
4802 #ifdef CONFIG_NFS_V4_1
4803 /*
4804  * Check the exchange flags returned by the server for invalid flags, having
4805  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4806  * DS flags set.
4807  */
4808 static int nfs4_check_cl_exchange_flags(u32 flags)
4809 {
4810         if (flags & ~EXCHGID4_FLAG_MASK_R)
4811                 goto out_inval;
4812         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4813             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4814                 goto out_inval;
4815         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4816                 goto out_inval;
4817         return NFS_OK;
4818 out_inval:
4819         return -NFS4ERR_INVAL;
4820 }
4821
4822 static bool
4823 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4824 {
4825         if (a->server_scope_sz == b->server_scope_sz &&
4826             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4827                 return true;
4828
4829         return false;
4830 }
4831
4832 /*
4833  * nfs4_proc_exchange_id()
4834  *
4835  * Since the clientid has expired, all compounds using sessions
4836  * associated with the stale clientid will be returning
4837  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4838  * be in some phase of session reset.
4839  */
4840 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4841 {
4842         nfs4_verifier verifier;
4843         struct nfs41_exchange_id_args args = {
4844                 .client = clp,
4845                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4846         };
4847         struct nfs41_exchange_id_res res = {
4848                 .client = clp,
4849         };
4850         int status;
4851         struct rpc_message msg = {
4852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4853                 .rpc_argp = &args,
4854                 .rpc_resp = &res,
4855                 .rpc_cred = cred,
4856         };
4857         __be32 *p;
4858
4859         dprintk("--> %s\n", __func__);
4860         BUG_ON(clp == NULL);
4861
4862         p = (u32 *)verifier.data;
4863         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4864         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4865         args.verifier = &verifier;
4866
4867         args.id_len = scnprintf(args.id, sizeof(args.id),
4868                                 "%s/%s.%s/%u",
4869                                 clp->cl_ipaddr,
4870                                 init_utsname()->nodename,
4871                                 init_utsname()->domainname,
4872                                 clp->cl_rpcclient->cl_auth->au_flavor);
4873
4874         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4875         if (unlikely(!res.server_scope))
4876                 return -ENOMEM;
4877
4878         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4879         if (!status)
4880                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4881
4882         if (!status) {
4883                 if (clp->server_scope &&
4884                     !nfs41_same_server_scope(clp->server_scope,
4885                                              res.server_scope)) {
4886                         dprintk("%s: server_scope mismatch detected\n",
4887                                 __func__);
4888                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4889                         kfree(clp->server_scope);
4890                         clp->server_scope = NULL;
4891                 }
4892
4893                 if (!clp->server_scope)
4894                         clp->server_scope = res.server_scope;
4895                 else
4896                         kfree(res.server_scope);
4897         }
4898
4899         dprintk("<-- %s status= %d\n", __func__, status);
4900         return status;
4901 }
4902
4903 struct nfs4_get_lease_time_data {
4904         struct nfs4_get_lease_time_args *args;
4905         struct nfs4_get_lease_time_res *res;
4906         struct nfs_client *clp;
4907 };
4908
4909 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4910                                         void *calldata)
4911 {
4912         int ret;
4913         struct nfs4_get_lease_time_data *data =
4914                         (struct nfs4_get_lease_time_data *)calldata;
4915
4916         dprintk("--> %s\n", __func__);
4917         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4918         /* just setup sequence, do not trigger session recovery
4919            since we're invoked within one */
4920         ret = nfs41_setup_sequence(data->clp->cl_session,
4921                                    &data->args->la_seq_args,
4922                                    &data->res->lr_seq_res, 0, task);
4923
4924         BUG_ON(ret == -EAGAIN);
4925         rpc_call_start(task);
4926         dprintk("<-- %s\n", __func__);
4927 }
4928
4929 /*
4930  * Called from nfs4_state_manager thread for session setup, so don't recover
4931  * from sequence operation or clientid errors.
4932  */
4933 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4934 {
4935         struct nfs4_get_lease_time_data *data =
4936                         (struct nfs4_get_lease_time_data *)calldata;
4937
4938         dprintk("--> %s\n", __func__);
4939         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4940                 return;
4941         switch (task->tk_status) {
4942         case -NFS4ERR_DELAY:
4943         case -NFS4ERR_GRACE:
4944                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4945                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4946                 task->tk_status = 0;
4947                 /* fall through */
4948         case -NFS4ERR_RETRY_UNCACHED_REP:
4949                 rpc_restart_call_prepare(task);
4950                 return;
4951         }
4952         dprintk("<-- %s\n", __func__);
4953 }
4954
4955 struct rpc_call_ops nfs4_get_lease_time_ops = {
4956         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4957         .rpc_call_done = nfs4_get_lease_time_done,
4958 };
4959
4960 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4961 {
4962         struct rpc_task *task;
4963         struct nfs4_get_lease_time_args args;
4964         struct nfs4_get_lease_time_res res = {
4965                 .lr_fsinfo = fsinfo,
4966         };
4967         struct nfs4_get_lease_time_data data = {
4968                 .args = &args,
4969                 .res = &res,
4970                 .clp = clp,
4971         };
4972         struct rpc_message msg = {
4973                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4974                 .rpc_argp = &args,
4975                 .rpc_resp = &res,
4976         };
4977         struct rpc_task_setup task_setup = {
4978                 .rpc_client = clp->cl_rpcclient,
4979                 .rpc_message = &msg,
4980                 .callback_ops = &nfs4_get_lease_time_ops,
4981                 .callback_data = &data,
4982                 .flags = RPC_TASK_TIMEOUT,
4983         };
4984         int status;
4985
4986         dprintk("--> %s\n", __func__);
4987         task = rpc_run_task(&task_setup);
4988
4989         if (IS_ERR(task))
4990                 status = PTR_ERR(task);
4991         else {
4992                 status = task->tk_status;
4993                 rpc_put_task(task);
4994         }
4995         dprintk("<-- %s return %d\n", __func__, status);
4996
4997         return status;
4998 }
4999
5000 /*
5001  * Reset a slot table
5002  */
5003 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5004                                  int ivalue)
5005 {
5006         struct nfs4_slot *new = NULL;
5007         int i;
5008         int ret = 0;
5009
5010         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5011                 max_reqs, tbl->max_slots);
5012
5013         /* Does the newly negotiated max_reqs match the existing slot table? */
5014         if (max_reqs != tbl->max_slots) {
5015                 ret = -ENOMEM;
5016                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
5017                               GFP_NOFS);
5018                 if (!new)
5019                         goto out;
5020                 ret = 0;
5021                 kfree(tbl->slots);
5022         }
5023         spin_lock(&tbl->slot_tbl_lock);
5024         if (new) {
5025                 tbl->slots = new;
5026                 tbl->max_slots = max_reqs;
5027         }
5028         for (i = 0; i < tbl->max_slots; ++i)
5029                 tbl->slots[i].seq_nr = ivalue;
5030         spin_unlock(&tbl->slot_tbl_lock);
5031         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5032                 tbl, tbl->slots, tbl->max_slots);
5033 out:
5034         dprintk("<-- %s: return %d\n", __func__, ret);
5035         return ret;
5036 }
5037
5038 /* Destroy the slot table */
5039 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5040 {
5041         if (session->fc_slot_table.slots != NULL) {
5042                 kfree(session->fc_slot_table.slots);
5043                 session->fc_slot_table.slots = NULL;
5044         }
5045         if (session->bc_slot_table.slots != NULL) {
5046                 kfree(session->bc_slot_table.slots);
5047                 session->bc_slot_table.slots = NULL;
5048         }
5049         return;
5050 }
5051
5052 /*
5053  * Initialize slot table
5054  */
5055 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
5056                 int max_slots, int ivalue)
5057 {
5058         struct nfs4_slot *slot;
5059         int ret = -ENOMEM;
5060
5061         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
5062
5063         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
5064
5065         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
5066         if (!slot)
5067                 goto out;
5068         ret = 0;
5069
5070         spin_lock(&tbl->slot_tbl_lock);
5071         tbl->max_slots = max_slots;
5072         tbl->slots = slot;
5073         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5074         spin_unlock(&tbl->slot_tbl_lock);
5075         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5076                 tbl, tbl->slots, tbl->max_slots);
5077 out:
5078         dprintk("<-- %s: return %d\n", __func__, ret);
5079         return ret;
5080 }
5081
5082 /*
5083  * Initialize or reset the forechannel and backchannel tables
5084  */
5085 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5086 {
5087         struct nfs4_slot_table *tbl;
5088         int status;
5089
5090         dprintk("--> %s\n", __func__);
5091         /* Fore channel */
5092         tbl = &ses->fc_slot_table;
5093         if (tbl->slots == NULL) {
5094                 status = nfs4_init_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5095                 if (status) /* -ENOMEM */
5096                         return status;
5097         } else {
5098                 status = nfs4_reset_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5099                 if (status)
5100                         return status;
5101         }
5102         /* Back channel */
5103         tbl = &ses->bc_slot_table;
5104         if (tbl->slots == NULL) {
5105                 status = nfs4_init_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5106                 if (status)
5107                         /* Fore and back channel share a connection so get
5108                          * both slot tables or neither */
5109                         nfs4_destroy_slot_tables(ses);
5110         } else
5111                 status = nfs4_reset_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5112         return status;
5113 }
5114
5115 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5116 {
5117         struct nfs4_session *session;
5118         struct nfs4_slot_table *tbl;
5119
5120         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5121         if (!session)
5122                 return NULL;
5123
5124         tbl = &session->fc_slot_table;
5125         tbl->highest_used_slotid = -1;
5126         spin_lock_init(&tbl->slot_tbl_lock);
5127         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5128         init_completion(&tbl->complete);
5129
5130         tbl = &session->bc_slot_table;
5131         tbl->highest_used_slotid = -1;
5132         spin_lock_init(&tbl->slot_tbl_lock);
5133         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5134         init_completion(&tbl->complete);
5135
5136         session->session_state = 1<<NFS4_SESSION_INITING;
5137
5138         session->clp = clp;
5139         return session;
5140 }
5141
5142 void nfs4_destroy_session(struct nfs4_session *session)
5143 {
5144         nfs4_proc_destroy_session(session);
5145         dprintk("%s Destroy backchannel for xprt %p\n",
5146                 __func__, session->clp->cl_rpcclient->cl_xprt);
5147         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5148                                 NFS41_BC_MIN_CALLBACKS);
5149         nfs4_destroy_slot_tables(session);
5150         kfree(session);
5151 }
5152
5153 /*
5154  * Initialize the values to be used by the client in CREATE_SESSION
5155  * If nfs4_init_session set the fore channel request and response sizes,
5156  * use them.
5157  *
5158  * Set the back channel max_resp_sz_cached to zero to force the client to
5159  * always set csa_cachethis to FALSE because the current implementation
5160  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5161  */
5162 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5163 {
5164         struct nfs4_session *session = args->client->cl_session;
5165         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5166                      mxresp_sz = session->fc_attrs.max_resp_sz;
5167
5168         if (mxrqst_sz == 0)
5169                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5170         if (mxresp_sz == 0)
5171                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5172         /* Fore channel attributes */
5173         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5174         args->fc_attrs.max_resp_sz = mxresp_sz;
5175         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5176         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5177
5178         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5179                 "max_ops=%u max_reqs=%u\n",
5180                 __func__,
5181                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5182                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5183
5184         /* Back channel attributes */
5185         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5186         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5187         args->bc_attrs.max_resp_sz_cached = 0;
5188         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5189         args->bc_attrs.max_reqs = 1;
5190
5191         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5192                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5193                 __func__,
5194                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5195                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5196                 args->bc_attrs.max_reqs);
5197 }
5198
5199 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5200 {
5201         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5202         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5203
5204         if (rcvd->max_resp_sz > sent->max_resp_sz)
5205                 return -EINVAL;
5206         /*
5207          * Our requested max_ops is the minimum we need; we're not
5208          * prepared to break up compounds into smaller pieces than that.
5209          * So, no point even trying to continue if the server won't
5210          * cooperate:
5211          */
5212         if (rcvd->max_ops < sent->max_ops)
5213                 return -EINVAL;
5214         if (rcvd->max_reqs == 0)
5215                 return -EINVAL;
5216         return 0;
5217 }
5218
5219 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5220 {
5221         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5222         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5223
5224         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5225                 return -EINVAL;
5226         if (rcvd->max_resp_sz < sent->max_resp_sz)
5227                 return -EINVAL;
5228         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5229                 return -EINVAL;
5230         /* These would render the backchannel useless: */
5231         if (rcvd->max_ops  == 0)
5232                 return -EINVAL;
5233         if (rcvd->max_reqs == 0)
5234                 return -EINVAL;
5235         return 0;
5236 }
5237
5238 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5239                                      struct nfs4_session *session)
5240 {
5241         int ret;
5242
5243         ret = nfs4_verify_fore_channel_attrs(args, session);
5244         if (ret)
5245                 return ret;
5246         return nfs4_verify_back_channel_attrs(args, session);
5247 }
5248
5249 static int _nfs4_proc_create_session(struct nfs_client *clp)
5250 {
5251         struct nfs4_session *session = clp->cl_session;
5252         struct nfs41_create_session_args args = {
5253                 .client = clp,
5254                 .cb_program = NFS4_CALLBACK,
5255         };
5256         struct nfs41_create_session_res res = {
5257                 .client = clp,
5258         };
5259         struct rpc_message msg = {
5260                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5261                 .rpc_argp = &args,
5262                 .rpc_resp = &res,
5263         };
5264         int status;
5265
5266         nfs4_init_channel_attrs(&args);
5267         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5268
5269         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5270
5271         if (!status)
5272                 /* Verify the session's negotiated channel_attrs values */
5273                 status = nfs4_verify_channel_attrs(&args, session);
5274         if (!status) {
5275                 /* Increment the clientid slot sequence id */
5276                 clp->cl_seqid++;
5277         }
5278
5279         return status;
5280 }
5281
5282 /*
5283  * Issues a CREATE_SESSION operation to the server.
5284  * It is the responsibility of the caller to verify the session is
5285  * expired before calling this routine.
5286  */
5287 int nfs4_proc_create_session(struct nfs_client *clp)
5288 {
5289         int status;
5290         unsigned *ptr;
5291         struct nfs4_session *session = clp->cl_session;
5292
5293         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5294
5295         status = _nfs4_proc_create_session(clp);
5296         if (status)
5297                 goto out;
5298
5299         /* Init or reset the session slot tables */
5300         status = nfs4_setup_session_slot_tables(session);
5301         dprintk("slot table setup returned %d\n", status);
5302         if (status)
5303                 goto out;
5304
5305         ptr = (unsigned *)&session->sess_id.data[0];
5306         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5307                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5308 out:
5309         dprintk("<-- %s\n", __func__);
5310         return status;
5311 }
5312
5313 /*
5314  * Issue the over-the-wire RPC DESTROY_SESSION.
5315  * The caller must serialize access to this routine.
5316  */
5317 int nfs4_proc_destroy_session(struct nfs4_session *session)
5318 {
5319         int status = 0;
5320         struct rpc_message msg;
5321
5322         dprintk("--> nfs4_proc_destroy_session\n");
5323
5324         /* session is still being setup */
5325         if (session->clp->cl_cons_state != NFS_CS_READY)
5326                 return status;
5327
5328         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5329         msg.rpc_argp = session;
5330         msg.rpc_resp = NULL;
5331         msg.rpc_cred = NULL;
5332         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5333
5334         if (status)
5335                 printk(KERN_WARNING
5336                         "Got error %d from the server on DESTROY_SESSION. "
5337                         "Session has been destroyed regardless...\n", status);
5338
5339         dprintk("<-- nfs4_proc_destroy_session\n");
5340         return status;
5341 }
5342
5343 int nfs4_init_session(struct nfs_server *server)
5344 {
5345         struct nfs_client *clp = server->nfs_client;
5346         struct nfs4_session *session;
5347         unsigned int rsize, wsize;
5348         int ret;
5349
5350         if (!nfs4_has_session(clp))
5351                 return 0;
5352
5353         session = clp->cl_session;
5354         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5355                 return 0;
5356
5357         rsize = server->rsize;
5358         if (rsize == 0)
5359                 rsize = NFS_MAX_FILE_IO_SIZE;
5360         wsize = server->wsize;
5361         if (wsize == 0)
5362                 wsize = NFS_MAX_FILE_IO_SIZE;
5363
5364         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5365         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5366
5367         ret = nfs4_recover_expired_lease(server);
5368         if (!ret)
5369                 ret = nfs4_check_client_ready(clp);
5370         return ret;
5371 }
5372
5373 int nfs4_init_ds_session(struct nfs_client *clp)
5374 {
5375         struct nfs4_session *session = clp->cl_session;
5376         int ret;
5377
5378         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5379                 return 0;
5380
5381         ret = nfs4_client_recover_expired_lease(clp);
5382         if (!ret)
5383                 /* Test for the DS role */
5384                 if (!is_ds_client(clp))
5385                         ret = -ENODEV;
5386         if (!ret)
5387                 ret = nfs4_check_client_ready(clp);
5388         return ret;
5389
5390 }
5391 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5392
5393
5394 /*
5395  * Renew the cl_session lease.
5396  */
5397 struct nfs4_sequence_data {
5398         struct nfs_client *clp;
5399         struct nfs4_sequence_args args;
5400         struct nfs4_sequence_res res;
5401 };
5402
5403 static void nfs41_sequence_release(void *data)
5404 {
5405         struct nfs4_sequence_data *calldata = data;
5406         struct nfs_client *clp = calldata->clp;
5407
5408         if (atomic_read(&clp->cl_count) > 1)
5409                 nfs4_schedule_state_renewal(clp);
5410         nfs_put_client(clp);
5411         kfree(calldata);
5412 }
5413
5414 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5415 {
5416         switch(task->tk_status) {
5417         case -NFS4ERR_DELAY:
5418                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5419                 return -EAGAIN;
5420         default:
5421                 nfs4_schedule_lease_recovery(clp);
5422         }
5423         return 0;
5424 }
5425
5426 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5427 {
5428         struct nfs4_sequence_data *calldata = data;
5429         struct nfs_client *clp = calldata->clp;
5430
5431         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5432                 return;
5433
5434         if (task->tk_status < 0) {
5435                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5436                 if (atomic_read(&clp->cl_count) == 1)
5437                         goto out;
5438
5439                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5440                         rpc_restart_call_prepare(task);
5441                         return;
5442                 }
5443         }
5444         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5445 out:
5446         dprintk("<-- %s\n", __func__);
5447 }
5448
5449 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5450 {
5451         struct nfs4_sequence_data *calldata = data;
5452         struct nfs_client *clp = calldata->clp;
5453         struct nfs4_sequence_args *args;
5454         struct nfs4_sequence_res *res;
5455
5456         args = task->tk_msg.rpc_argp;
5457         res = task->tk_msg.rpc_resp;
5458
5459         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5460                 return;
5461         rpc_call_start(task);
5462 }
5463
5464 static const struct rpc_call_ops nfs41_sequence_ops = {
5465         .rpc_call_done = nfs41_sequence_call_done,
5466         .rpc_call_prepare = nfs41_sequence_prepare,
5467         .rpc_release = nfs41_sequence_release,
5468 };
5469
5470 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5471 {
5472         struct nfs4_sequence_data *calldata;
5473         struct rpc_message msg = {
5474                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5475                 .rpc_cred = cred,
5476         };
5477         struct rpc_task_setup task_setup_data = {
5478                 .rpc_client = clp->cl_rpcclient,
5479                 .rpc_message = &msg,
5480                 .callback_ops = &nfs41_sequence_ops,
5481                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5482         };
5483
5484         if (!atomic_inc_not_zero(&clp->cl_count))
5485                 return ERR_PTR(-EIO);
5486         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5487         if (calldata == NULL) {
5488                 nfs_put_client(clp);
5489                 return ERR_PTR(-ENOMEM);
5490         }
5491         msg.rpc_argp = &calldata->args;
5492         msg.rpc_resp = &calldata->res;
5493         calldata->clp = clp;
5494         task_setup_data.callback_data = calldata;
5495
5496         return rpc_run_task(&task_setup_data);
5497 }
5498
5499 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5500 {
5501         struct rpc_task *task;
5502         int ret = 0;
5503
5504         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5505                 return 0;
5506         task = _nfs41_proc_sequence(clp, cred);
5507         if (IS_ERR(task))
5508                 ret = PTR_ERR(task);
5509         else
5510                 rpc_put_task_async(task);
5511         dprintk("<-- %s status=%d\n", __func__, ret);
5512         return ret;
5513 }
5514
5515 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5516 {
5517         struct rpc_task *task;
5518         int ret;
5519
5520         task = _nfs41_proc_sequence(clp, cred);
5521         if (IS_ERR(task)) {
5522                 ret = PTR_ERR(task);
5523                 goto out;
5524         }
5525         ret = rpc_wait_for_completion_task(task);
5526         if (!ret) {
5527                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5528
5529                 if (task->tk_status == 0)
5530                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5531                 ret = task->tk_status;
5532         }
5533         rpc_put_task(task);
5534 out:
5535         dprintk("<-- %s status=%d\n", __func__, ret);
5536         return ret;
5537 }
5538
5539 struct nfs4_reclaim_complete_data {
5540         struct nfs_client *clp;
5541         struct nfs41_reclaim_complete_args arg;
5542         struct nfs41_reclaim_complete_res res;
5543 };
5544
5545 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5546 {
5547         struct nfs4_reclaim_complete_data *calldata = data;
5548
5549         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5550         if (nfs41_setup_sequence(calldata->clp->cl_session,
5551                                 &calldata->arg.seq_args,
5552                                 &calldata->res.seq_res, 0, task))
5553                 return;
5554
5555         rpc_call_start(task);
5556 }
5557
5558 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5559 {
5560         switch(task->tk_status) {
5561         case 0:
5562         case -NFS4ERR_COMPLETE_ALREADY:
5563         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5564                 break;
5565         case -NFS4ERR_DELAY:
5566                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5567                 /* fall through */
5568         case -NFS4ERR_RETRY_UNCACHED_REP:
5569                 return -EAGAIN;
5570         default:
5571                 nfs4_schedule_lease_recovery(clp);
5572         }
5573         return 0;
5574 }
5575
5576 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5577 {
5578         struct nfs4_reclaim_complete_data *calldata = data;
5579         struct nfs_client *clp = calldata->clp;
5580         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5581
5582         dprintk("--> %s\n", __func__);
5583         if (!nfs41_sequence_done(task, res))
5584                 return;
5585
5586         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5587                 rpc_restart_call_prepare(task);
5588                 return;
5589         }
5590         dprintk("<-- %s\n", __func__);
5591 }
5592
5593 static void nfs4_free_reclaim_complete_data(void *data)
5594 {
5595         struct nfs4_reclaim_complete_data *calldata = data;
5596
5597         kfree(calldata);
5598 }
5599
5600 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5601         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5602         .rpc_call_done = nfs4_reclaim_complete_done,
5603         .rpc_release = nfs4_free_reclaim_complete_data,
5604 };
5605
5606 /*
5607  * Issue a global reclaim complete.
5608  */
5609 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5610 {
5611         struct nfs4_reclaim_complete_data *calldata;
5612         struct rpc_task *task;
5613         struct rpc_message msg = {
5614                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5615         };
5616         struct rpc_task_setup task_setup_data = {
5617                 .rpc_client = clp->cl_rpcclient,
5618                 .rpc_message = &msg,
5619                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5620                 .flags = RPC_TASK_ASYNC,
5621         };
5622         int status = -ENOMEM;
5623
5624         dprintk("--> %s\n", __func__);
5625         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5626         if (calldata == NULL)
5627                 goto out;
5628         calldata->clp = clp;
5629         calldata->arg.one_fs = 0;
5630
5631         msg.rpc_argp = &calldata->arg;
5632         msg.rpc_resp = &calldata->res;
5633         task_setup_data.callback_data = calldata;
5634         task = rpc_run_task(&task_setup_data);
5635         if (IS_ERR(task)) {
5636                 status = PTR_ERR(task);
5637                 goto out;
5638         }
5639         status = nfs4_wait_for_completion_rpc_task(task);
5640         if (status == 0)
5641                 status = task->tk_status;
5642         rpc_put_task(task);
5643         return 0;
5644 out:
5645         dprintk("<-- %s status=%d\n", __func__, status);
5646         return status;
5647 }
5648
5649 static void
5650 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5651 {
5652         struct nfs4_layoutget *lgp = calldata;
5653         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5654
5655         dprintk("--> %s\n", __func__);
5656         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5657          * right now covering the LAYOUTGET we are about to send.
5658          * However, that is not so catastrophic, and there seems
5659          * to be no way to prevent it completely.
5660          */
5661         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5662                                 &lgp->res.seq_res, 0, task))
5663                 return;
5664         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5665                                           NFS_I(lgp->args.inode)->layout,
5666                                           lgp->args.ctx->state)) {
5667                 rpc_exit(task, NFS4_OK);
5668                 return;
5669         }
5670         rpc_call_start(task);
5671 }
5672
5673 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5674 {
5675         struct nfs4_layoutget *lgp = calldata;
5676         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5677
5678         dprintk("--> %s\n", __func__);
5679
5680         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5681                 return;
5682
5683         switch (task->tk_status) {
5684         case 0:
5685                 break;
5686         case -NFS4ERR_LAYOUTTRYLATER:
5687         case -NFS4ERR_RECALLCONFLICT:
5688                 task->tk_status = -NFS4ERR_DELAY;
5689                 /* Fall through */
5690         default:
5691                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5692                         rpc_restart_call_prepare(task);
5693                         return;
5694                 }
5695         }
5696         dprintk("<-- %s\n", __func__);
5697 }
5698
5699 static void nfs4_layoutget_release(void *calldata)
5700 {
5701         struct nfs4_layoutget *lgp = calldata;
5702
5703         dprintk("--> %s\n", __func__);
5704         put_nfs_open_context(lgp->args.ctx);
5705         kfree(calldata);
5706         dprintk("<-- %s\n", __func__);
5707 }
5708
5709 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5710         .rpc_call_prepare = nfs4_layoutget_prepare,
5711         .rpc_call_done = nfs4_layoutget_done,
5712         .rpc_release = nfs4_layoutget_release,
5713 };
5714
5715 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5716 {
5717         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5718         struct rpc_task *task;
5719         struct rpc_message msg = {
5720                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5721                 .rpc_argp = &lgp->args,
5722                 .rpc_resp = &lgp->res,
5723         };
5724         struct rpc_task_setup task_setup_data = {
5725                 .rpc_client = server->client,
5726                 .rpc_message = &msg,
5727                 .callback_ops = &nfs4_layoutget_call_ops,
5728                 .callback_data = lgp,
5729                 .flags = RPC_TASK_ASYNC,
5730         };
5731         int status = 0;
5732
5733         dprintk("--> %s\n", __func__);
5734
5735         lgp->res.layoutp = &lgp->args.layout;
5736         lgp->res.seq_res.sr_slot = NULL;
5737         task = rpc_run_task(&task_setup_data);
5738         if (IS_ERR(task))
5739                 return PTR_ERR(task);
5740         status = nfs4_wait_for_completion_rpc_task(task);
5741         if (status == 0)
5742                 status = task->tk_status;
5743         if (status == 0)
5744                 status = pnfs_layout_process(lgp);
5745         rpc_put_task(task);
5746         dprintk("<-- %s status=%d\n", __func__, status);
5747         return status;
5748 }
5749
5750 static void
5751 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5752 {
5753         struct nfs4_layoutreturn *lrp = calldata;
5754
5755         dprintk("--> %s\n", __func__);
5756         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5757                                 &lrp->res.seq_res, 0, task))
5758                 return;
5759         rpc_call_start(task);
5760 }
5761
5762 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5763 {
5764         struct nfs4_layoutreturn *lrp = calldata;
5765         struct nfs_server *server;
5766         struct pnfs_layout_hdr *lo = lrp->args.layout;
5767
5768         dprintk("--> %s\n", __func__);
5769
5770         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5771                 return;
5772
5773         server = NFS_SERVER(lrp->args.inode);
5774         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5775                 rpc_restart_call_prepare(task);
5776                 return;
5777         }
5778         spin_lock(&lo->plh_inode->i_lock);
5779         if (task->tk_status == 0) {
5780                 if (lrp->res.lrs_present) {
5781                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5782                 } else
5783                         BUG_ON(!list_empty(&lo->plh_segs));
5784         }
5785         lo->plh_block_lgets--;
5786         spin_unlock(&lo->plh_inode->i_lock);
5787         dprintk("<-- %s\n", __func__);
5788 }
5789
5790 static void nfs4_layoutreturn_release(void *calldata)
5791 {
5792         struct nfs4_layoutreturn *lrp = calldata;
5793
5794         dprintk("--> %s\n", __func__);
5795         put_layout_hdr(lrp->args.layout);
5796         kfree(calldata);
5797         dprintk("<-- %s\n", __func__);
5798 }
5799
5800 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5801         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5802         .rpc_call_done = nfs4_layoutreturn_done,
5803         .rpc_release = nfs4_layoutreturn_release,
5804 };
5805
5806 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5807 {
5808         struct rpc_task *task;
5809         struct rpc_message msg = {
5810                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5811                 .rpc_argp = &lrp->args,
5812                 .rpc_resp = &lrp->res,
5813         };
5814         struct rpc_task_setup task_setup_data = {
5815                 .rpc_client = lrp->clp->cl_rpcclient,
5816                 .rpc_message = &msg,
5817                 .callback_ops = &nfs4_layoutreturn_call_ops,
5818                 .callback_data = lrp,
5819         };
5820         int status;
5821
5822         dprintk("--> %s\n", __func__);
5823         task = rpc_run_task(&task_setup_data);
5824         if (IS_ERR(task))
5825                 return PTR_ERR(task);
5826         status = task->tk_status;
5827         dprintk("<-- %s status=%d\n", __func__, status);
5828         rpc_put_task(task);
5829         return status;
5830 }
5831
5832 /*
5833  * Retrieve the list of Data Server devices from the MDS.
5834  */
5835 static int _nfs4_getdevicelist(struct nfs_server *server,
5836                                     const struct nfs_fh *fh,
5837                                     struct pnfs_devicelist *devlist)
5838 {
5839         struct nfs4_getdevicelist_args args = {
5840                 .fh = fh,
5841                 .layoutclass = server->pnfs_curr_ld->id,
5842         };
5843         struct nfs4_getdevicelist_res res = {
5844                 .devlist = devlist,
5845         };
5846         struct rpc_message msg = {
5847                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5848                 .rpc_argp = &args,
5849                 .rpc_resp = &res,
5850         };
5851         int status;
5852
5853         dprintk("--> %s\n", __func__);
5854         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5855                                 &res.seq_res, 0);
5856         dprintk("<-- %s status=%d\n", __func__, status);
5857         return status;
5858 }
5859
5860 int nfs4_proc_getdevicelist(struct nfs_server *server,
5861                             const struct nfs_fh *fh,
5862                             struct pnfs_devicelist *devlist)
5863 {
5864         struct nfs4_exception exception = { };
5865         int err;
5866
5867         do {
5868                 err = nfs4_handle_exception(server,
5869                                 _nfs4_getdevicelist(server, fh, devlist),
5870                                 &exception);
5871         } while (exception.retry);
5872
5873         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5874                 err, devlist->num_devs);
5875
5876         return err;
5877 }
5878 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5879
5880 static int
5881 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5882 {
5883         struct nfs4_getdeviceinfo_args args = {
5884                 .pdev = pdev,
5885         };
5886         struct nfs4_getdeviceinfo_res res = {
5887                 .pdev = pdev,
5888         };
5889         struct rpc_message msg = {
5890                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5891                 .rpc_argp = &args,
5892                 .rpc_resp = &res,
5893         };
5894         int status;
5895
5896         dprintk("--> %s\n", __func__);
5897         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5898         dprintk("<-- %s status=%d\n", __func__, status);
5899
5900         return status;
5901 }
5902
5903 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5904 {
5905         struct nfs4_exception exception = { };
5906         int err;
5907
5908         do {
5909                 err = nfs4_handle_exception(server,
5910                                         _nfs4_proc_getdeviceinfo(server, pdev),
5911                                         &exception);
5912         } while (exception.retry);
5913         return err;
5914 }
5915 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5916
5917 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5918 {
5919         struct nfs4_layoutcommit_data *data = calldata;
5920         struct nfs_server *server = NFS_SERVER(data->args.inode);
5921
5922         if (nfs4_setup_sequence(server, &data->args.seq_args,
5923                                 &data->res.seq_res, 1, task))
5924                 return;
5925         rpc_call_start(task);
5926 }
5927
5928 static void
5929 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5930 {
5931         struct nfs4_layoutcommit_data *data = calldata;
5932         struct nfs_server *server = NFS_SERVER(data->args.inode);
5933
5934         if (!nfs4_sequence_done(task, &data->res.seq_res))
5935                 return;
5936
5937         switch (task->tk_status) { /* Just ignore these failures */
5938         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5939         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5940         case NFS4ERR_BADLAYOUT:     /* no layout */
5941         case NFS4ERR_GRACE:         /* loca_recalim always false */
5942                 task->tk_status = 0;
5943         }
5944
5945         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5946                 rpc_restart_call_prepare(task);
5947                 return;
5948         }
5949
5950         if (task->tk_status == 0)
5951                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5952                                                    data->res.fattr);
5953 }
5954
5955 static void nfs4_layoutcommit_release(void *calldata)
5956 {
5957         struct nfs4_layoutcommit_data *data = calldata;
5958         struct pnfs_layout_segment *lseg, *tmp;
5959         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
5960
5961         pnfs_cleanup_layoutcommit(data);
5962         /* Matched by references in pnfs_set_layoutcommit */
5963         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
5964                 list_del_init(&lseg->pls_lc_list);
5965                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
5966                                        &lseg->pls_flags))
5967                         put_lseg(lseg);
5968         }
5969
5970         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
5971         smp_mb__after_clear_bit();
5972         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
5973
5974         put_rpccred(data->cred);
5975         kfree(data);
5976 }
5977
5978 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5979         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5980         .rpc_call_done = nfs4_layoutcommit_done,
5981         .rpc_release = nfs4_layoutcommit_release,
5982 };
5983
5984 int
5985 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5986 {
5987         struct rpc_message msg = {
5988                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5989                 .rpc_argp = &data->args,
5990                 .rpc_resp = &data->res,
5991                 .rpc_cred = data->cred,
5992         };
5993         struct rpc_task_setup task_setup_data = {
5994                 .task = &data->task,
5995                 .rpc_client = NFS_CLIENT(data->args.inode),
5996                 .rpc_message = &msg,
5997                 .callback_ops = &nfs4_layoutcommit_ops,
5998                 .callback_data = data,
5999                 .flags = RPC_TASK_ASYNC,
6000         };
6001         struct rpc_task *task;
6002         int status = 0;
6003
6004         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6005                 "lbw: %llu inode %lu\n",
6006                 data->task.tk_pid, sync,
6007                 data->args.lastbytewritten,
6008                 data->args.inode->i_ino);
6009
6010         task = rpc_run_task(&task_setup_data);
6011         if (IS_ERR(task))
6012                 return PTR_ERR(task);
6013         if (sync == false)
6014                 goto out;
6015         status = nfs4_wait_for_completion_rpc_task(task);
6016         if (status != 0)
6017                 goto out;
6018         status = task->tk_status;
6019 out:
6020         dprintk("%s: status %d\n", __func__, status);
6021         rpc_put_task(task);
6022         return status;
6023 }
6024
6025 static int
6026 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6027                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6028 {
6029         struct nfs41_secinfo_no_name_args args = {
6030                 .style = SECINFO_STYLE_CURRENT_FH,
6031         };
6032         struct nfs4_secinfo_res res = {
6033                 .flavors = flavors,
6034         };
6035         struct rpc_message msg = {
6036                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6037                 .rpc_argp = &args,
6038                 .rpc_resp = &res,
6039         };
6040         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6041 }
6042
6043 static int
6044 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6045                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6046 {
6047         struct nfs4_exception exception = { };
6048         int err;
6049         do {
6050                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6051                 switch (err) {
6052                 case 0:
6053                 case -NFS4ERR_WRONGSEC:
6054                 case -NFS4ERR_NOTSUPP:
6055                         break;
6056                 default:
6057                         err = nfs4_handle_exception(server, err, &exception);
6058                 }
6059         } while (exception.retry);
6060         return err;
6061 }
6062
6063 static int
6064 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6065                     struct nfs_fsinfo *info)
6066 {
6067         int err;
6068         struct page *page;
6069         rpc_authflavor_t flavor;
6070         struct nfs4_secinfo_flavors *flavors;
6071
6072         page = alloc_page(GFP_KERNEL);
6073         if (!page) {
6074                 err = -ENOMEM;
6075                 goto out;
6076         }
6077
6078         flavors = page_address(page);
6079         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6080
6081         /*
6082          * Fall back on "guess and check" method if
6083          * the server doesn't support SECINFO_NO_NAME
6084          */
6085         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6086                 err = nfs4_find_root_sec(server, fhandle, info);
6087                 goto out_freepage;
6088         }
6089         if (err)
6090                 goto out_freepage;
6091
6092         flavor = nfs_find_best_sec(flavors);
6093         if (err == 0)
6094                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6095
6096 out_freepage:
6097         put_page(page);
6098         if (err == -EACCES)
6099                 return -EPERM;
6100 out:
6101         return err;
6102 }
6103 static int _nfs41_test_stateid(struct nfs_server *server, struct nfs4_state *state)
6104 {
6105         int status;
6106         struct nfs41_test_stateid_args args = {
6107                 .stateid = &state->stateid,
6108         };
6109         struct nfs41_test_stateid_res res;
6110         struct rpc_message msg = {
6111                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6112                 .rpc_argp = &args,
6113                 .rpc_resp = &res,
6114         };
6115         args.seq_args.sa_session = res.seq_res.sr_session = NULL;
6116         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 0, 1);
6117         return status;
6118 }
6119
6120 static int nfs41_test_stateid(struct nfs_server *server, struct nfs4_state *state)
6121 {
6122         struct nfs4_exception exception = { };
6123         int err;
6124         do {
6125                 err = nfs4_handle_exception(server,
6126                                 _nfs41_test_stateid(server, state),
6127                                 &exception);
6128         } while (exception.retry);
6129         return err;
6130 }
6131
6132 static int _nfs4_free_stateid(struct nfs_server *server, struct nfs4_state *state)
6133 {
6134         int status;
6135         struct nfs41_free_stateid_args args = {
6136                 .stateid = &state->stateid,
6137         };
6138         struct nfs41_free_stateid_res res;
6139         struct rpc_message msg = {
6140                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6141                 .rpc_argp = &args,
6142                 .rpc_resp = &res,
6143         };
6144
6145         args.seq_args.sa_session = res.seq_res.sr_session = NULL;
6146         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 0, 1);
6147         return status;
6148 }
6149
6150 static int nfs41_free_stateid(struct nfs_server *server, struct nfs4_state *state)
6151 {
6152         struct nfs4_exception exception = { };
6153         int err;
6154         do {
6155                 err = nfs4_handle_exception(server,
6156                                 _nfs4_free_stateid(server, state),
6157                                 &exception);
6158         } while (exception.retry);
6159         return err;
6160 }
6161 #endif /* CONFIG_NFS_V4_1 */
6162
6163 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6164         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6165         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6166         .recover_open   = nfs4_open_reclaim,
6167         .recover_lock   = nfs4_lock_reclaim,
6168         .establish_clid = nfs4_init_clientid,
6169         .get_clid_cred  = nfs4_get_setclientid_cred,
6170 };
6171
6172 #if defined(CONFIG_NFS_V4_1)
6173 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6174         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6175         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6176         .recover_open   = nfs4_open_reclaim,
6177         .recover_lock   = nfs4_lock_reclaim,
6178         .establish_clid = nfs41_init_clientid,
6179         .get_clid_cred  = nfs4_get_exchange_id_cred,
6180         .reclaim_complete = nfs41_proc_reclaim_complete,
6181 };
6182 #endif /* CONFIG_NFS_V4_1 */
6183
6184 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6185         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6186         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6187         .recover_open   = nfs4_open_expired,
6188         .recover_lock   = nfs4_lock_expired,
6189         .establish_clid = nfs4_init_clientid,
6190         .get_clid_cred  = nfs4_get_setclientid_cred,
6191 };
6192
6193 #if defined(CONFIG_NFS_V4_1)
6194 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6195         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6196         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6197         .recover_open   = nfs41_open_expired,
6198         .recover_lock   = nfs41_lock_expired,
6199         .establish_clid = nfs41_init_clientid,
6200         .get_clid_cred  = nfs4_get_exchange_id_cred,
6201 };
6202 #endif /* CONFIG_NFS_V4_1 */
6203
6204 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6205         .sched_state_renewal = nfs4_proc_async_renew,
6206         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6207         .renew_lease = nfs4_proc_renew,
6208 };
6209
6210 #if defined(CONFIG_NFS_V4_1)
6211 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6212         .sched_state_renewal = nfs41_proc_async_sequence,
6213         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6214         .renew_lease = nfs4_proc_sequence,
6215 };
6216 #endif
6217
6218 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6219         .minor_version = 0,
6220         .call_sync = _nfs4_call_sync,
6221         .validate_stateid = nfs4_validate_delegation_stateid,
6222         .find_root_sec = nfs4_find_root_sec,
6223         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6224         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6225         .state_renewal_ops = &nfs40_state_renewal_ops,
6226 };
6227
6228 #if defined(CONFIG_NFS_V4_1)
6229 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6230         .minor_version = 1,
6231         .call_sync = _nfs4_call_sync_session,
6232         .validate_stateid = nfs41_validate_delegation_stateid,
6233         .find_root_sec = nfs41_find_root_sec,
6234         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6235         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6236         .state_renewal_ops = &nfs41_state_renewal_ops,
6237 };
6238 #endif
6239
6240 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6241         [0] = &nfs_v4_0_minor_ops,
6242 #if defined(CONFIG_NFS_V4_1)
6243         [1] = &nfs_v4_1_minor_ops,
6244 #endif
6245 };
6246
6247 static const struct inode_operations nfs4_file_inode_operations = {
6248         .permission     = nfs_permission,
6249         .getattr        = nfs_getattr,
6250         .setattr        = nfs_setattr,
6251         .getxattr       = generic_getxattr,
6252         .setxattr       = generic_setxattr,
6253         .listxattr      = generic_listxattr,
6254         .removexattr    = generic_removexattr,
6255 };
6256
6257 const struct nfs_rpc_ops nfs_v4_clientops = {
6258         .version        = 4,                    /* protocol version */
6259         .dentry_ops     = &nfs4_dentry_operations,
6260         .dir_inode_ops  = &nfs4_dir_inode_operations,
6261         .file_inode_ops = &nfs4_file_inode_operations,
6262         .file_ops       = &nfs4_file_operations,
6263         .getroot        = nfs4_proc_get_root,
6264         .getattr        = nfs4_proc_getattr,
6265         .setattr        = nfs4_proc_setattr,
6266         .lookup         = nfs4_proc_lookup,
6267         .access         = nfs4_proc_access,
6268         .readlink       = nfs4_proc_readlink,
6269         .create         = nfs4_proc_create,
6270         .remove         = nfs4_proc_remove,
6271         .unlink_setup   = nfs4_proc_unlink_setup,
6272         .unlink_done    = nfs4_proc_unlink_done,
6273         .rename         = nfs4_proc_rename,
6274         .rename_setup   = nfs4_proc_rename_setup,
6275         .rename_done    = nfs4_proc_rename_done,
6276         .link           = nfs4_proc_link,
6277         .symlink        = nfs4_proc_symlink,
6278         .mkdir          = nfs4_proc_mkdir,
6279         .rmdir          = nfs4_proc_remove,
6280         .readdir        = nfs4_proc_readdir,
6281         .mknod          = nfs4_proc_mknod,
6282         .statfs         = nfs4_proc_statfs,
6283         .fsinfo         = nfs4_proc_fsinfo,
6284         .pathconf       = nfs4_proc_pathconf,
6285         .set_capabilities = nfs4_server_capabilities,
6286         .decode_dirent  = nfs4_decode_dirent,
6287         .read_setup     = nfs4_proc_read_setup,
6288         .read_done      = nfs4_read_done,
6289         .write_setup    = nfs4_proc_write_setup,
6290         .write_done     = nfs4_write_done,
6291         .commit_setup   = nfs4_proc_commit_setup,
6292         .commit_done    = nfs4_commit_done,
6293         .lock           = nfs4_proc_lock,
6294         .clear_acl_cache = nfs4_zap_acl_attr,
6295         .close_context  = nfs4_close_context,
6296         .open_context   = nfs4_atomic_open,
6297         .init_client    = nfs4_init_client,
6298         .secinfo        = nfs4_proc_secinfo,
6299 };
6300
6301 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6302         .prefix = XATTR_NAME_NFSV4_ACL,
6303         .list   = nfs4_xattr_list_nfs4_acl,
6304         .get    = nfs4_xattr_get_nfs4_acl,
6305         .set    = nfs4_xattr_set_nfs4_acl,
6306 };
6307
6308 const struct xattr_handler *nfs4_xattr_handlers[] = {
6309         &nfs4_xattr_nfs4_acl_handler,
6310         NULL
6311 };
6312
6313 /*
6314  * Local variables:
6315  *  c-basic-offset: 8
6316  * End:
6317  */