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/slab.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/nfs.h>
  45 #include <linux/nfs4.h>
  46 #include <linux/nfs_fs.h>
  47 #include <linux/nfs_page.h>
  48 #include <linux/namei.h>
  49 #include <linux/mount.h>
  50 #include <linux/module.h>
  51 #include <linux/sunrpc/bc_xprt.h>
  52
  53 #include "nfs4_fs.h"
  54 #include "delegation.h"
  55 #include "internal.h"
  56 #include "iostat.h"
  57 #include "callback.h"
  58
  59 #define NFSDBG_FACILITY         NFSDBG_PROC
  60
  61 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  62 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  63
  64 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  65
  66 struct nfs4_opendata;
  67 static int _nfs4_proc_open(struct nfs4_opendata *data);
  68 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  69 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  70 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  71 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  72 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  73
  74 /* Prevent leaks of NFSv4 errors into userland */
  75 static int nfs4_map_errors(int err)
  76 {
  77         if (err >= -1000)
  78                 return err;
  79         switch (err) {
  80         case -NFS4ERR_RESOURCE:
  81                 return -EREMOTEIO;
  82         default:
  83                 dprintk("%s could not handle NFSv4 error %d\n",
  84                                 __func__, -err);
  85                 break;
  86         }
  87         return -EIO;
  88 }
  89
  90 /*
  91  * This is our standard bitmap for GETATTR requests.
  92  */
  93 const u32 nfs4_fattr_bitmap[2] = {
  94         FATTR4_WORD0_TYPE
  95         | FATTR4_WORD0_CHANGE
  96         | FATTR4_WORD0_SIZE
  97         | FATTR4_WORD0_FSID
  98         | FATTR4_WORD0_FILEID,
  99         FATTR4_WORD1_MODE
 100         | FATTR4_WORD1_NUMLINKS
 101         | FATTR4_WORD1_OWNER
 102         | FATTR4_WORD1_OWNER_GROUP
 103         | FATTR4_WORD1_RAWDEV
 104         | FATTR4_WORD1_SPACE_USED
 105         | FATTR4_WORD1_TIME_ACCESS
 106         | FATTR4_WORD1_TIME_METADATA
 107         | FATTR4_WORD1_TIME_MODIFY
 108 };
 109
 110 const u32 nfs4_statfs_bitmap[2] = {
 111         FATTR4_WORD0_FILES_AVAIL
 112         | FATTR4_WORD0_FILES_FREE
 113         | FATTR4_WORD0_FILES_TOTAL,
 114         FATTR4_WORD1_SPACE_AVAIL
 115         | FATTR4_WORD1_SPACE_FREE
 116         | FATTR4_WORD1_SPACE_TOTAL
 117 };
 118
 119 const u32 nfs4_pathconf_bitmap[2] = {
 120         FATTR4_WORD0_MAXLINK
 121         | FATTR4_WORD0_MAXNAME,
 122         0
 123 };
 124
 125 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 126                         | FATTR4_WORD0_MAXREAD
 127                         | FATTR4_WORD0_MAXWRITE
 128                         | FATTR4_WORD0_LEASE_TIME,
 129                         0
 130 };
 131
 132 const u32 nfs4_fs_locations_bitmap[2] = {
 133         FATTR4_WORD0_TYPE
 134         | FATTR4_WORD0_CHANGE
 135         | FATTR4_WORD0_SIZE
 136         | FATTR4_WORD0_FSID
 137         | FATTR4_WORD0_FILEID
 138         | FATTR4_WORD0_FS_LOCATIONS,
 139         FATTR4_WORD1_MODE
 140         | FATTR4_WORD1_NUMLINKS
 141         | FATTR4_WORD1_OWNER
 142         | FATTR4_WORD1_OWNER_GROUP
 143         | FATTR4_WORD1_RAWDEV
 144         | FATTR4_WORD1_SPACE_USED
 145         | FATTR4_WORD1_TIME_ACCESS
 146         | FATTR4_WORD1_TIME_METADATA
 147         | FATTR4_WORD1_TIME_MODIFY
 148         | FATTR4_WORD1_MOUNTED_ON_FILEID
 149 };
 150
 151 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 152                 struct nfs4_readdir_arg *readdir)
 153 {
 154         __be32 *start, *p;
 155
 156         BUG_ON(readdir->count < 80);
 157         if (cookie > 2) {
 158                 readdir->cookie = cookie;
 159                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 160                 return;
 161         }
 162
 163         readdir->cookie = 0;
 164         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 165         if (cookie == 2)
 166                 return;
 167
 168         /*
 169          * NFSv4 servers do not return entries for '.' and '..'
 170          * Therefore, we fake these entries here.  We let '.'
 171          * have cookie 0 and '..' have cookie 1.  Note that
 172          * when talking to the server, we always send cookie 0
 173          * instead of 1 or 2.
 174          */
 175         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 176
 177         if (cookie == 0) {
 178                 *p++ = xdr_one;                                  /* next */
 179                 *p++ = xdr_zero;                   /* cookie, first word */
 180                 *p++ = xdr_one;                   /* cookie, second word */
 181                 *p++ = xdr_one;                             /* entry len */
 182                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 183                 p++;
 184                 *p++ = xdr_one;                         /* bitmap length */
 185                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 186                 *p++ = htonl(8);              /* attribute buffer length */
 187                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 188         }
 189
 190         *p++ = xdr_one;                                  /* next */
 191         *p++ = xdr_zero;                   /* cookie, first word */
 192         *p++ = xdr_two;                   /* cookie, second word */
 193         *p++ = xdr_two;                             /* entry len */
 194         memcpy(p, "..\0\0", 4);                         /* entry */
 195         p++;
 196         *p++ = xdr_one;                         /* bitmap length */
 197         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 198         *p++ = htonl(8);              /* attribute buffer length */
 199         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 200
 201         readdir->pgbase = (char *)p - (char *)start;
 202         readdir->count -= readdir->pgbase;
 203         kunmap_atomic(start, KM_USER0);
 204 }
 205
 206 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 207 {
 208         int res;
 209
 210         might_sleep();
 211
 212         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 213                         nfs_wait_bit_killable, TASK_KILLABLE);
 214         return res;
 215 }
 216
 217 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 218 {
 219         int res = 0;
 220
 221         might_sleep();
 222
 223         if (*timeout <= 0)
 224                 *timeout = NFS4_POLL_RETRY_MIN;
 225         if (*timeout > NFS4_POLL_RETRY_MAX)
 226                 *timeout = NFS4_POLL_RETRY_MAX;
 227         schedule_timeout_killable(*timeout);
 228         if (fatal_signal_pending(current))
 229                 res = -ERESTARTSYS;
 230         *timeout <<= 1;
 231         return res;
 232 }
 233
 234 /* This is the error handling routine for processes that are allowed
 235  * to sleep.
 236  */
 237 static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 238 {
 239         struct nfs_client *clp = server->nfs_client;
 240         struct nfs4_state *state = exception->state;
 241         int ret = errorcode;
 242
 243         exception->retry = 0;
 244         switch(errorcode) {
 245                 case 0:
 246                         return 0;
 247                 case -NFS4ERR_ADMIN_REVOKED:
 248                 case -NFS4ERR_BAD_STATEID:
 249                 case -NFS4ERR_OPENMODE:
 250                         if (state == NULL)
 251                                 break;
 252                         nfs4_state_mark_reclaim_nograce(clp, state);
 253                         goto do_state_recovery;
 254                 case -NFS4ERR_STALE_STATEID:
 255                         if (state == NULL)
 256                                 break;
 257                         nfs4_state_mark_reclaim_reboot(clp, state);
 258                 case -NFS4ERR_STALE_CLIENTID:
 259                 case -NFS4ERR_EXPIRED:
 260                         goto do_state_recovery;
 261 #if defined(CONFIG_NFS_V4_1)
 262                 case -NFS4ERR_BADSESSION:
 263                 case -NFS4ERR_BADSLOT:
 264                 case -NFS4ERR_BAD_HIGH_SLOT:
 265                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 266                 case -NFS4ERR_DEADSESSION:
 267                 case -NFS4ERR_SEQ_FALSE_RETRY:
 268                 case -NFS4ERR_SEQ_MISORDERED:
 269                         dprintk("%s ERROR: %d Reset session\n", __func__,
 270                                 errorcode);
 271                         nfs4_schedule_state_recovery(clp);
 272                         exception->retry = 1;
 273                         break;
 274 #endif /* defined(CONFIG_NFS_V4_1) */
 275                 case -NFS4ERR_FILE_OPEN:
 276                         if (exception->timeout > HZ) {
 277                                 /* We have retried a decent amount, time to
 278                                  * fail
 279                                  */
 280                                 ret = -EBUSY;
 281                                 break;
 282                         }
 283                 case -NFS4ERR_GRACE:
 284                 case -NFS4ERR_DELAY:
 285                 case -EKEYEXPIRED:
 286                         ret = nfs4_delay(server->client, &exception->timeout);
 287                         if (ret != 0)
 288                                 break;
 289                 case -NFS4ERR_OLD_STATEID:
 290                         exception->retry = 1;
 291         }
 292         /* We failed to handle the error */
 293         return nfs4_map_errors(ret);
 294 do_state_recovery:
 295         nfs4_schedule_state_recovery(clp);
 296         ret = nfs4_wait_clnt_recover(clp);
 297         if (ret == 0)
 298                 exception->retry = 1;
 299         return ret;
 300 }
 301
 302
 303 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 304 {
 305         struct nfs_client *clp = server->nfs_client;
 306         spin_lock(&clp->cl_lock);
 307         if (time_before(clp->cl_last_renewal,timestamp))
 308                 clp->cl_last_renewal = timestamp;
 309         spin_unlock(&clp->cl_lock);
 310 }
 311
 312 #if defined(CONFIG_NFS_V4_1)
 313
 314 /*
 315  * nfs4_free_slot - free a slot and efficiently update slot table.
 316  *
 317  * freeing a slot is trivially done by clearing its respective bit
 318  * in the bitmap.
 319  * If the freed slotid equals highest_used_slotid we want to update it
 320  * so that the server would be able to size down the slot table if needed,
 321  * otherwise we know that the highest_used_slotid is still in use.
 322  * When updating highest_used_slotid there may be "holes" in the bitmap
 323  * so we need to scan down from highest_used_slotid to 0 looking for the now
 324  * highest slotid in use.
 325  * If none found, highest_used_slotid is set to -1.
 326  *
 327  * Must be called while holding tbl->slot_tbl_lock
 328  */
 329 static void
 330 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
 331 {
 332         int slotid = free_slotid;
 333
 334         /* clear used bit in bitmap */
 335         __clear_bit(slotid, tbl->used_slots);
 336
 337         /* update highest_used_slotid when it is freed */
 338         if (slotid == tbl->highest_used_slotid) {
 339                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 340                 if (slotid < tbl->max_slots)
 341                         tbl->highest_used_slotid = slotid;
 342                 else
 343                         tbl->highest_used_slotid = -1;
 344         }
 345         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 346                 free_slotid, tbl->highest_used_slotid);
 347 }
 348
 349 /*
 350  * Signal state manager thread if session is drained
 351  */
 352 static void nfs41_check_drain_session_complete(struct nfs4_session *ses)
 353 {
 354         struct rpc_task *task;
 355
 356         if (!test_bit(NFS4CLNT_SESSION_DRAINING, &ses->clp->cl_state)) {
 357                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 358                 if (task)
 359                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 360                 return;
 361         }
 362
 363         if (ses->fc_slot_table.highest_used_slotid != -1)
 364                 return;
 365
 366         dprintk("%s COMPLETE: Session Drained\n", __func__);
 367         complete(&ses->complete);
 368 }
 369
 370 static void nfs41_sequence_free_slot(const struct nfs_client *clp,
 371                               struct nfs4_sequence_res *res)
 372 {
 373         struct nfs4_slot_table *tbl;
 374
 375         tbl = &clp->cl_session->fc_slot_table;
 376         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) {
 377                 /* just wake up the next guy waiting since
 378                  * we may have not consumed a slot after all */
 379                 dprintk("%s: No slot\n", __func__);
 380                 return;
 381         }
 382
 383         spin_lock(&tbl->slot_tbl_lock);
 384         nfs4_free_slot(tbl, res->sr_slotid);
 385         nfs41_check_drain_session_complete(clp->cl_session);
 386         spin_unlock(&tbl->slot_tbl_lock);
 387         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
 388 }
 389
 390 static void nfs41_sequence_done(struct nfs_client *clp,
 391                                 struct nfs4_sequence_res *res,
 392                                 int rpc_status)
 393 {
 394         unsigned long timestamp;
 395         struct nfs4_slot_table *tbl;
 396         struct nfs4_slot *slot;
 397
 398         /*
 399          * sr_status remains 1 if an RPC level error occurred. The server
 400          * may or may not have processed the sequence operation..
 401          * Proceed as if the server received and processed the sequence
 402          * operation.
 403          */
 404         if (res->sr_status == 1)
 405                 res->sr_status = NFS_OK;
 406
 407         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
 408         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE)
 409                 goto out;
 410
 411         /* Check the SEQUENCE operation status */
 412         if (res->sr_status == 0) {
 413                 tbl = &clp->cl_session->fc_slot_table;
 414                 slot = tbl->slots + res->sr_slotid;
 415                 /* Update the slot's sequence and clientid lease timer */
 416                 ++slot->seq_nr;
 417                 timestamp = res->sr_renewal_time;
 418                 spin_lock(&clp->cl_lock);
 419                 if (time_before(clp->cl_last_renewal, timestamp))
 420                         clp->cl_last_renewal = timestamp;
 421                 spin_unlock(&clp->cl_lock);
 422                 /* Check sequence flags */
 423                 if (atomic_read(&clp->cl_count) > 1)
 424                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 425         }
 426 out:
 427         /* The session may be reset by one of the error handlers. */
 428         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 429         nfs41_sequence_free_slot(clp, res);
 430 }
 431
 432 /*
 433  * nfs4_find_slot - efficiently look for a free slot
 434  *
 435  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 436  * If found, we mark the slot as used, update the highest_used_slotid,
 437  * and respectively set up the sequence operation args.
 438  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 439  *
 440  * Note: must be called with under the slot_tbl_lock.
 441  */
 442 static u8
 443 nfs4_find_slot(struct nfs4_slot_table *tbl)
 444 {
 445         int slotid;
 446         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 447         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 448
 449         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 450                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 451                 tbl->max_slots);
 452         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 453         if (slotid >= tbl->max_slots)
 454                 goto out;
 455         __set_bit(slotid, tbl->used_slots);
 456         if (slotid > tbl->highest_used_slotid)
 457                 tbl->highest_used_slotid = slotid;
 458         ret_id = slotid;
 459 out:
 460         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 461                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 462         return ret_id;
 463 }
 464
 465 static int nfs41_setup_sequence(struct nfs4_session *session,
 466                                 struct nfs4_sequence_args *args,
 467                                 struct nfs4_sequence_res *res,
 468                                 int cache_reply,
 469                                 struct rpc_task *task)
 470 {
 471         struct nfs4_slot *slot;
 472         struct nfs4_slot_table *tbl;
 473         u8 slotid;
 474
 475         dprintk("--> %s\n", __func__);
 476         /* slot already allocated? */
 477         if (res->sr_slotid != NFS4_MAX_SLOT_TABLE)
 478                 return 0;
 479
 480         memset(res, 0, sizeof(*res));
 481         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
 482         tbl = &session->fc_slot_table;
 483
 484         spin_lock(&tbl->slot_tbl_lock);
 485         if (test_bit(NFS4CLNT_SESSION_DRAINING, &session->clp->cl_state) &&
 486             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 487                 /*
 488                  * The state manager will wait until the slot table is empty.
 489                  * Schedule the reset thread
 490                  */
 491                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 492                 spin_unlock(&tbl->slot_tbl_lock);
 493                 dprintk("%s Schedule Session Reset\n", __func__);
 494                 return -EAGAIN;
 495         }
 496
 497         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 498             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 499                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 500                 spin_unlock(&tbl->slot_tbl_lock);
 501                 dprintk("%s enforce FIFO order\n", __func__);
 502                 return -EAGAIN;
 503         }
 504
 505         slotid = nfs4_find_slot(tbl);
 506         if (slotid == NFS4_MAX_SLOT_TABLE) {
 507                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 508                 spin_unlock(&tbl->slot_tbl_lock);
 509                 dprintk("<-- %s: no free slots\n", __func__);
 510                 return -EAGAIN;
 511         }
 512         spin_unlock(&tbl->slot_tbl_lock);
 513
 514         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 515         slot = tbl->slots + slotid;
 516         args->sa_session = session;
 517         args->sa_slotid = slotid;
 518         args->sa_cache_this = cache_reply;
 519
 520         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 521
 522         res->sr_session = session;
 523         res->sr_slotid = slotid;
 524         res->sr_renewal_time = jiffies;
 525         /*
 526          * sr_status is only set in decode_sequence, and so will remain
 527          * set to 1 if an rpc level failure occurs.
 528          */
 529         res->sr_status = 1;
 530         return 0;
 531 }
 532
 533 int nfs4_setup_sequence(struct nfs_client *clp,
 534                         struct nfs4_sequence_args *args,
 535                         struct nfs4_sequence_res *res,
 536                         int cache_reply,
 537                         struct rpc_task *task)
 538 {
 539         int ret = 0;
 540
 541         dprintk("--> %s clp %p session %p sr_slotid %d\n",
 542                 __func__, clp, clp->cl_session, res->sr_slotid);
 543
 544         if (!nfs4_has_session(clp))
 545                 goto out;
 546         ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply,
 547                                    task);
 548         if (ret && ret != -EAGAIN) {
 549                 /* terminate rpc task */
 550                 task->tk_status = ret;
 551                 task->tk_action = NULL;
 552         }
 553 out:
 554         dprintk("<-- %s status=%d\n", __func__, ret);
 555         return ret;
 556 }
 557
 558 struct nfs41_call_sync_data {
 559         struct nfs_client *clp;
 560         struct nfs4_sequence_args *seq_args;
 561         struct nfs4_sequence_res *seq_res;
 562         int cache_reply;
 563 };
 564
 565 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 566 {
 567         struct nfs41_call_sync_data *data = calldata;
 568
 569         dprintk("--> %s data->clp->cl_session %p\n", __func__,
 570                 data->clp->cl_session);
 571         if (nfs4_setup_sequence(data->clp, data->seq_args,
 572                                 data->seq_res, data->cache_reply, task))
 573                 return;
 574         rpc_call_start(task);
 575 }
 576
 577 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 578 {
 579         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 580         nfs41_call_sync_prepare(task, calldata);
 581 }
 582
 583 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 584 {
 585         struct nfs41_call_sync_data *data = calldata;
 586
 587         nfs41_sequence_done(data->clp, data->seq_res, task->tk_status);
 588 }
 589
 590 struct rpc_call_ops nfs41_call_sync_ops = {
 591         .rpc_call_prepare = nfs41_call_sync_prepare,
 592         .rpc_call_done = nfs41_call_sync_done,
 593 };
 594
 595 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 596         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 597         .rpc_call_done = nfs41_call_sync_done,
 598 };
 599
 600 static int nfs4_call_sync_sequence(struct nfs_client *clp,
 601                                    struct rpc_clnt *clnt,
 602                                    struct rpc_message *msg,
 603                                    struct nfs4_sequence_args *args,
 604                                    struct nfs4_sequence_res *res,
 605                                    int cache_reply,
 606                                    int privileged)
 607 {
 608         int ret;
 609         struct rpc_task *task;
 610         struct nfs41_call_sync_data data = {
 611                 .clp = clp,
 612                 .seq_args = args,
 613                 .seq_res = res,
 614                 .cache_reply = cache_reply,
 615         };
 616         struct rpc_task_setup task_setup = {
 617                 .rpc_client = clnt,
 618                 .rpc_message = msg,
 619                 .callback_ops = &nfs41_call_sync_ops,
 620                 .callback_data = &data
 621         };
 622
 623         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
 624         if (privileged)
 625                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 626         task = rpc_run_task(&task_setup);
 627         if (IS_ERR(task))
 628                 ret = PTR_ERR(task);
 629         else {
 630                 ret = task->tk_status;
 631                 rpc_put_task(task);
 632         }
 633         return ret;
 634 }
 635
 636 int _nfs4_call_sync_session(struct nfs_server *server,
 637                             struct rpc_message *msg,
 638                             struct nfs4_sequence_args *args,
 639                             struct nfs4_sequence_res *res,
 640                             int cache_reply)
 641 {
 642         return nfs4_call_sync_sequence(server->nfs_client, server->client,
 643                                        msg, args, res, cache_reply, 0);
 644 }
 645
 646 #endif /* CONFIG_NFS_V4_1 */
 647
 648 int _nfs4_call_sync(struct nfs_server *server,
 649                     struct rpc_message *msg,
 650                     struct nfs4_sequence_args *args,
 651                     struct nfs4_sequence_res *res,
 652                     int cache_reply)
 653 {
 654         args->sa_session = res->sr_session = NULL;
 655         return rpc_call_sync(server->client, msg, 0);
 656 }
 657
 658 #define nfs4_call_sync(server, msg, args, res, cache_reply) \
 659         (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \
 660                         &(res)->seq_res, (cache_reply))
 661
 662 static void nfs4_sequence_done(const struct nfs_server *server,
 663                                struct nfs4_sequence_res *res, int rpc_status)
 664 {
 665 #ifdef CONFIG_NFS_V4_1
 666         if (nfs4_has_session(server->nfs_client))
 667                 nfs41_sequence_done(server->nfs_client, res, rpc_status);
 668 #endif /* CONFIG_NFS_V4_1 */
 669 }
 670
 671 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 672 {
 673         struct nfs_inode *nfsi = NFS_I(dir);
 674
 675         spin_lock(&dir->i_lock);
 676         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 677         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 678                 nfs_force_lookup_revalidate(dir);
 679         nfsi->change_attr = cinfo->after;
 680         spin_unlock(&dir->i_lock);
 681 }
 682
 683 struct nfs4_opendata {
 684         struct kref kref;
 685         struct nfs_openargs o_arg;
 686         struct nfs_openres o_res;
 687         struct nfs_open_confirmargs c_arg;
 688         struct nfs_open_confirmres c_res;
 689         struct nfs_fattr f_attr;
 690         struct nfs_fattr dir_attr;
 691         struct path path;
 692         struct dentry *dir;
 693         struct nfs4_state_owner *owner;
 694         struct nfs4_state *state;
 695         struct iattr attrs;
 696         unsigned long timestamp;
 697         unsigned int rpc_done : 1;
 698         int rpc_status;
 699         int cancelled;
 700 };
 701
 702
 703 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 704 {
 705         p->o_res.f_attr = &p->f_attr;
 706         p->o_res.dir_attr = &p->dir_attr;
 707         p->o_res.seqid = p->o_arg.seqid;
 708         p->c_res.seqid = p->c_arg.seqid;
 709         p->o_res.server = p->o_arg.server;
 710         nfs_fattr_init(&p->f_attr);
 711         nfs_fattr_init(&p->dir_attr);
 712         p->o_res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
 713 }
 714
 715 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 716                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 717                 const struct iattr *attrs)
 718 {
 719         struct dentry *parent = dget_parent(path->dentry);
 720         struct inode *dir = parent->d_inode;
 721         struct nfs_server *server = NFS_SERVER(dir);
 722         struct nfs4_opendata *p;
 723
 724         p = kzalloc(sizeof(*p), GFP_KERNEL);
 725         if (p == NULL)
 726                 goto err;
 727         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
 728         if (p->o_arg.seqid == NULL)
 729                 goto err_free;
 730         path_get(path);
 731         p->path = *path;
 732         p->dir = parent;
 733         p->owner = sp;
 734         atomic_inc(&sp->so_count);
 735         p->o_arg.fh = NFS_FH(dir);
 736         p->o_arg.open_flags = flags;
 737         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 738         p->o_arg.clientid = server->nfs_client->cl_clientid;
 739         p->o_arg.id = sp->so_owner_id.id;
 740         p->o_arg.name = &p->path.dentry->d_name;
 741         p->o_arg.server = server;
 742         p->o_arg.bitmask = server->attr_bitmask;
 743         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 744         if (flags & O_EXCL) {
 745                 if (nfs4_has_persistent_session(server->nfs_client)) {
 746                         /* GUARDED */
 747                         p->o_arg.u.attrs = &p->attrs;
 748                         memcpy(&p->attrs, attrs, sizeof(p->attrs));
 749                 } else { /* EXCLUSIVE4_1 */
 750                         u32 *s = (u32 *) p->o_arg.u.verifier.data;
 751                         s[0] = jiffies;
 752                         s[1] = current->pid;
 753                 }
 754         } else if (flags & O_CREAT) {
 755                 p->o_arg.u.attrs = &p->attrs;
 756                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 757         }
 758         p->c_arg.fh = &p->o_res.fh;
 759         p->c_arg.stateid = &p->o_res.stateid;
 760         p->c_arg.seqid = p->o_arg.seqid;
 761         nfs4_init_opendata_res(p);
 762         kref_init(&p->kref);
 763         return p;
 764 err_free:
 765         kfree(p);
 766 err:
 767         dput(parent);
 768         return NULL;
 769 }
 770
 771 static void nfs4_opendata_free(struct kref *kref)
 772 {
 773         struct nfs4_opendata *p = container_of(kref,
 774                         struct nfs4_opendata, kref);
 775
 776         nfs_free_seqid(p->o_arg.seqid);
 777         if (p->state != NULL)
 778                 nfs4_put_open_state(p->state);
 779         nfs4_put_state_owner(p->owner);
 780         dput(p->dir);
 781         path_put(&p->path);
 782         kfree(p);
 783 }
 784
 785 static void nfs4_opendata_put(struct nfs4_opendata *p)
 786 {
 787         if (p != NULL)
 788                 kref_put(&p->kref, nfs4_opendata_free);
 789 }
 790
 791 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 792 {
 793         int ret;
 794
 795         ret = rpc_wait_for_completion_task(task);
 796         return ret;
 797 }
 798
 799 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 800 {
 801         int ret = 0;
 802
 803         if (open_mode & O_EXCL)
 804                 goto out;
 805         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 806                 case FMODE_READ:
 807                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 808                                 && state->n_rdonly != 0;
 809                         break;
 810                 case FMODE_WRITE:
 811                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 812                                 && state->n_wronly != 0;
 813                         break;
 814                 case FMODE_READ|FMODE_WRITE:
 815                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 816                                 && state->n_rdwr != 0;
 817         }
 818 out:
 819         return ret;
 820 }
 821
 822 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 823 {
 824         if ((delegation->type & fmode) != fmode)
 825                 return 0;
 826         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 827                 return 0;
 828         nfs_mark_delegation_referenced(delegation);
 829         return 1;
 830 }
 831
 832 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 833 {
 834         switch (fmode) {
 835                 case FMODE_WRITE:
 836                         state->n_wronly++;
 837                         break;
 838                 case FMODE_READ:
 839                         state->n_rdonly++;
 840                         break;
 841                 case FMODE_READ|FMODE_WRITE:
 842                         state->n_rdwr++;
 843         }
 844         nfs4_state_set_mode_locked(state, state->state | fmode);
 845 }
 846
 847 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 848 {
 849         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 850                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 851         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 852         switch (fmode) {
 853                 case FMODE_READ:
 854                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 855                         break;
 856                 case FMODE_WRITE:
 857                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 858                         break;
 859                 case FMODE_READ|FMODE_WRITE:
 860                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 861         }
 862 }
 863
 864 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 865 {
 866         write_seqlock(&state->seqlock);
 867         nfs_set_open_stateid_locked(state, stateid, fmode);
 868         write_sequnlock(&state->seqlock);
 869 }
 870
 871 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 872 {
 873         /*
 874          * Protect the call to nfs4_state_set_mode_locked and
 875          * serialise the stateid update
 876          */
 877         write_seqlock(&state->seqlock);
 878         if (deleg_stateid != NULL) {
 879                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 880                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 881         }
 882         if (open_stateid != NULL)
 883                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 884         write_sequnlock(&state->seqlock);
 885         spin_lock(&state->owner->so_lock);
 886         update_open_stateflags(state, fmode);
 887         spin_unlock(&state->owner->so_lock);
 888 }
 889
 890 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 891 {
 892         struct nfs_inode *nfsi = NFS_I(state->inode);
 893         struct nfs_delegation *deleg_cur;
 894         int ret = 0;
 895
 896         fmode &= (FMODE_READ|FMODE_WRITE);
 897
 898         rcu_read_lock();
 899         deleg_cur = rcu_dereference(nfsi->delegation);
 900         if (deleg_cur == NULL)
 901                 goto no_delegation;
 902
 903         spin_lock(&deleg_cur->lock);
 904         if (nfsi->delegation != deleg_cur ||
 905             (deleg_cur->type & fmode) != fmode)
 906                 goto no_delegation_unlock;
 907
 908         if (delegation == NULL)
 909                 delegation = &deleg_cur->stateid;
 910         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 911                 goto no_delegation_unlock;
 912
 913         nfs_mark_delegation_referenced(deleg_cur);
 914         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 915         ret = 1;
 916 no_delegation_unlock:
 917         spin_unlock(&deleg_cur->lock);
 918 no_delegation:
 919         rcu_read_unlock();
 920
 921         if (!ret && open_stateid != NULL) {
 922                 __update_open_stateid(state, open_stateid, NULL, fmode);
 923                 ret = 1;
 924         }
 925
 926         return ret;
 927 }
 928
 929
 930 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 931 {
 932         struct nfs_delegation *delegation;
 933
 934         rcu_read_lock();
 935         delegation = rcu_dereference(NFS_I(inode)->delegation);
 936         if (delegation == NULL || (delegation->type & fmode) == fmode) {
 937                 rcu_read_unlock();
 938                 return;
 939         }
 940         rcu_read_unlock();
 941         nfs_inode_return_delegation(inode);
 942 }
 943
 944 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
 945 {
 946         struct nfs4_state *state = opendata->state;
 947         struct nfs_inode *nfsi = NFS_I(state->inode);
 948         struct nfs_delegation *delegation;
 949         int open_mode = opendata->o_arg.open_flags & O_EXCL;
 950         fmode_t fmode = opendata->o_arg.fmode;
 951         nfs4_stateid stateid;
 952         int ret = -EAGAIN;
 953
 954         for (;;) {
 955                 if (can_open_cached(state, fmode, open_mode)) {
 956                         spin_lock(&state->owner->so_lock);
 957                         if (can_open_cached(state, fmode, open_mode)) {
 958                                 update_open_stateflags(state, fmode);
 959                                 spin_unlock(&state->owner->so_lock);
 960                                 goto out_return_state;
 961                         }
 962                         spin_unlock(&state->owner->so_lock);
 963                 }
 964                 rcu_read_lock();
 965                 delegation = rcu_dereference(nfsi->delegation);
 966                 if (delegation == NULL ||
 967                     !can_open_delegated(delegation, fmode)) {
 968                         rcu_read_unlock();
 969                         break;
 970                 }
 971                 /* Save the delegation */
 972                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
 973                 rcu_read_unlock();
 974                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
 975                 if (ret != 0)
 976                         goto out;
 977                 ret = -EAGAIN;
 978
 979                 /* Try to update the stateid using the delegation */
 980                 if (update_open_stateid(state, NULL, &stateid, fmode))
 981                         goto out_return_state;
 982         }
 983 out:
 984         return ERR_PTR(ret);
 985 out_return_state:
 986         atomic_inc(&state->count);
 987         return state;
 988 }
 989
 990 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
 991 {
 992         struct inode *inode;
 993         struct nfs4_state *state = NULL;
 994         struct nfs_delegation *delegation;
 995         int ret;
 996
 997         if (!data->rpc_done) {
 998                 state = nfs4_try_open_cached(data);
 999                 goto out;
1000         }
1001
1002         ret = -EAGAIN;
1003         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1004                 goto err;
1005         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1006         ret = PTR_ERR(inode);
1007         if (IS_ERR(inode))
1008                 goto err;
1009         ret = -ENOMEM;
1010         state = nfs4_get_open_state(inode, data->owner);
1011         if (state == NULL)
1012                 goto err_put_inode;
1013         if (data->o_res.delegation_type != 0) {
1014                 int delegation_flags = 0;
1015
1016                 rcu_read_lock();
1017                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1018                 if (delegation)
1019                         delegation_flags = delegation->flags;
1020                 rcu_read_unlock();
1021                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1022                         nfs_inode_set_delegation(state->inode,
1023                                         data->owner->so_cred,
1024                                         &data->o_res);
1025                 else
1026                         nfs_inode_reclaim_delegation(state->inode,
1027                                         data->owner->so_cred,
1028                                         &data->o_res);
1029         }
1030
1031         update_open_stateid(state, &data->o_res.stateid, NULL,
1032                         data->o_arg.fmode);
1033         iput(inode);
1034 out:
1035         return state;
1036 err_put_inode:
1037         iput(inode);
1038 err:
1039         return ERR_PTR(ret);
1040 }
1041
1042 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1043 {
1044         struct nfs_inode *nfsi = NFS_I(state->inode);
1045         struct nfs_open_context *ctx;
1046
1047         spin_lock(&state->inode->i_lock);
1048         list_for_each_entry(ctx, &nfsi->open_files, list) {
1049                 if (ctx->state != state)
1050                         continue;
1051                 get_nfs_open_context(ctx);
1052                 spin_unlock(&state->inode->i_lock);
1053                 return ctx;
1054         }
1055         spin_unlock(&state->inode->i_lock);
1056         return ERR_PTR(-ENOENT);
1057 }
1058
1059 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1060 {
1061         struct nfs4_opendata *opendata;
1062
1063         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL);
1064         if (opendata == NULL)
1065                 return ERR_PTR(-ENOMEM);
1066         opendata->state = state;
1067         atomic_inc(&state->count);
1068         return opendata;
1069 }
1070
1071 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1072 {
1073         struct nfs4_state *newstate;
1074         int ret;
1075
1076         opendata->o_arg.open_flags = 0;
1077         opendata->o_arg.fmode = fmode;
1078         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1079         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1080         nfs4_init_opendata_res(opendata);
1081         ret = _nfs4_recover_proc_open(opendata);
1082         if (ret != 0)
1083                 return ret;
1084         newstate = nfs4_opendata_to_nfs4_state(opendata);
1085         if (IS_ERR(newstate))
1086                 return PTR_ERR(newstate);
1087         nfs4_close_state(&opendata->path, newstate, fmode);
1088         *res = newstate;
1089         return 0;
1090 }
1091
1092 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1093 {
1094         struct nfs4_state *newstate;
1095         int ret;
1096
1097         /* memory barrier prior to reading state->n_* */
1098         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1099         smp_rmb();
1100         if (state->n_rdwr != 0) {
1101                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1102                 if (ret != 0)
1103                         return ret;
1104                 if (newstate != state)
1105                         return -ESTALE;
1106         }
1107         if (state->n_wronly != 0) {
1108                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1109                 if (ret != 0)
1110                         return ret;
1111                 if (newstate != state)
1112                         return -ESTALE;
1113         }
1114         if (state->n_rdonly != 0) {
1115                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1116                 if (ret != 0)
1117                         return ret;
1118                 if (newstate != state)
1119                         return -ESTALE;
1120         }
1121         /*
1122          * We may have performed cached opens for all three recoveries.
1123          * Check if we need to update the current stateid.
1124          */
1125         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1126             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1127                 write_seqlock(&state->seqlock);
1128                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1129                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1130                 write_sequnlock(&state->seqlock);
1131         }
1132         return 0;
1133 }
1134
1135 /*
1136  * OPEN_RECLAIM:
1137  *      reclaim state on the server after a reboot.
1138  */
1139 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1140 {
1141         struct nfs_delegation *delegation;
1142         struct nfs4_opendata *opendata;
1143         fmode_t delegation_type = 0;
1144         int status;
1145
1146         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1147         if (IS_ERR(opendata))
1148                 return PTR_ERR(opendata);
1149         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1150         opendata->o_arg.fh = NFS_FH(state->inode);
1151         rcu_read_lock();
1152         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1153         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1154                 delegation_type = delegation->type;
1155         rcu_read_unlock();
1156         opendata->o_arg.u.delegation_type = delegation_type;
1157         status = nfs4_open_recover(opendata, state);
1158         nfs4_opendata_put(opendata);
1159         return status;
1160 }
1161
1162 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1163 {
1164         struct nfs_server *server = NFS_SERVER(state->inode);
1165         struct nfs4_exception exception = { };
1166         int err;
1167         do {
1168                 err = _nfs4_do_open_reclaim(ctx, state);
1169                 if (err != -NFS4ERR_DELAY && err != -EKEYEXPIRED)
1170                         break;
1171                 nfs4_handle_exception(server, err, &exception);
1172         } while (exception.retry);
1173         return err;
1174 }
1175
1176 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1177 {
1178         struct nfs_open_context *ctx;
1179         int ret;
1180
1181         ctx = nfs4_state_find_open_context(state);
1182         if (IS_ERR(ctx))
1183                 return PTR_ERR(ctx);
1184         ret = nfs4_do_open_reclaim(ctx, state);
1185         put_nfs_open_context(ctx);
1186         return ret;
1187 }
1188
1189 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1190 {
1191         struct nfs4_opendata *opendata;
1192         int ret;
1193
1194         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1195         if (IS_ERR(opendata))
1196                 return PTR_ERR(opendata);
1197         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1198         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1199                         sizeof(opendata->o_arg.u.delegation.data));
1200         ret = nfs4_open_recover(opendata, state);
1201         nfs4_opendata_put(opendata);
1202         return ret;
1203 }
1204
1205 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1206 {
1207         struct nfs4_exception exception = { };
1208         struct nfs_server *server = NFS_SERVER(state->inode);
1209         int err;
1210         do {
1211                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1212                 switch (err) {
1213                         case 0:
1214                         case -ENOENT:
1215                         case -ESTALE:
1216                                 goto out;
1217                         case -NFS4ERR_BADSESSION:
1218                         case -NFS4ERR_BADSLOT:
1219                         case -NFS4ERR_BAD_HIGH_SLOT:
1220                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1221                         case -NFS4ERR_DEADSESSION:
1222                                 nfs4_schedule_state_recovery(
1223                                         server->nfs_client);
1224                                 goto out;
1225                         case -NFS4ERR_STALE_CLIENTID:
1226                         case -NFS4ERR_STALE_STATEID:
1227                         case -NFS4ERR_EXPIRED:
1228                                 /* Don't recall a delegation if it was lost */
1229                                 nfs4_schedule_state_recovery(server->nfs_client);
1230                                 goto out;
1231                         case -ERESTARTSYS:
1232                                 /*
1233                                  * The show must go on: exit, but mark the
1234                                  * stateid as needing recovery.
1235                                  */
1236                         case -NFS4ERR_ADMIN_REVOKED:
1237                         case -NFS4ERR_BAD_STATEID:
1238                                 nfs4_state_mark_reclaim_nograce(server->nfs_client, state);
1239                         case -ENOMEM:
1240                                 err = 0;
1241                                 goto out;
1242                 }
1243                 err = nfs4_handle_exception(server, err, &exception);
1244         } while (exception.retry);
1245 out:
1246         return err;
1247 }
1248
1249 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1250 {
1251         struct nfs4_opendata *data = calldata;
1252
1253         data->rpc_status = task->tk_status;
1254         if (RPC_ASSASSINATED(task))
1255                 return;
1256         if (data->rpc_status == 0) {
1257                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1258                                 sizeof(data->o_res.stateid.data));
1259                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1260                 renew_lease(data->o_res.server, data->timestamp);
1261                 data->rpc_done = 1;
1262         }
1263 }
1264
1265 static void nfs4_open_confirm_release(void *calldata)
1266 {
1267         struct nfs4_opendata *data = calldata;
1268         struct nfs4_state *state = NULL;
1269
1270         /* If this request hasn't been cancelled, do nothing */
1271         if (data->cancelled == 0)
1272                 goto out_free;
1273         /* In case of error, no cleanup! */
1274         if (!data->rpc_done)
1275                 goto out_free;
1276         state = nfs4_opendata_to_nfs4_state(data);
1277         if (!IS_ERR(state))
1278                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1279 out_free:
1280         nfs4_opendata_put(data);
1281 }
1282
1283 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1284         .rpc_call_done = nfs4_open_confirm_done,
1285         .rpc_release = nfs4_open_confirm_release,
1286 };
1287
1288 /*
1289  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1290  */
1291 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1292 {
1293         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1294         struct rpc_task *task;
1295         struct  rpc_message msg = {
1296                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1297                 .rpc_argp = &data->c_arg,
1298                 .rpc_resp = &data->c_res,
1299                 .rpc_cred = data->owner->so_cred,
1300         };
1301         struct rpc_task_setup task_setup_data = {
1302                 .rpc_client = server->client,
1303                 .rpc_message = &msg,
1304                 .callback_ops = &nfs4_open_confirm_ops,
1305                 .callback_data = data,
1306                 .workqueue = nfsiod_workqueue,
1307                 .flags = RPC_TASK_ASYNC,
1308         };
1309         int status;
1310
1311         kref_get(&data->kref);
1312         data->rpc_done = 0;
1313         data->rpc_status = 0;
1314         data->timestamp = jiffies;
1315         task = rpc_run_task(&task_setup_data);
1316         if (IS_ERR(task))
1317                 return PTR_ERR(task);
1318         status = nfs4_wait_for_completion_rpc_task(task);
1319         if (status != 0) {
1320                 data->cancelled = 1;
1321                 smp_wmb();
1322         } else
1323                 status = data->rpc_status;
1324         rpc_put_task(task);
1325         return status;
1326 }
1327
1328 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1329 {
1330         struct nfs4_opendata *data = calldata;
1331         struct nfs4_state_owner *sp = data->owner;
1332
1333         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1334                 return;
1335         /*
1336          * Check if we still need to send an OPEN call, or if we can use
1337          * a delegation instead.
1338          */
1339         if (data->state != NULL) {
1340                 struct nfs_delegation *delegation;
1341
1342                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1343                         goto out_no_action;
1344                 rcu_read_lock();
1345                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1346                 if (delegation != NULL &&
1347                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1348                         rcu_read_unlock();
1349                         goto out_no_action;
1350                 }
1351                 rcu_read_unlock();
1352         }
1353         /* Update sequence id. */
1354         data->o_arg.id = sp->so_owner_id.id;
1355         data->o_arg.clientid = sp->so_client->cl_clientid;
1356         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1357                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1358                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1359         }
1360         data->timestamp = jiffies;
1361         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
1362                                 &data->o_arg.seq_args,
1363                                 &data->o_res.seq_res, 1, task))
1364                 return;
1365         rpc_call_start(task);
1366         return;
1367 out_no_action:
1368         task->tk_action = NULL;
1369
1370 }
1371
1372 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1373 {
1374         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1375         nfs4_open_prepare(task, calldata);
1376 }
1377
1378 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1379 {
1380         struct nfs4_opendata *data = calldata;
1381
1382         data->rpc_status = task->tk_status;
1383
1384         nfs4_sequence_done(data->o_arg.server, &data->o_res.seq_res,
1385                         task->tk_status);
1386
1387         if (RPC_ASSASSINATED(task))
1388                 return;
1389         if (task->tk_status == 0) {
1390                 switch (data->o_res.f_attr->mode & S_IFMT) {
1391                         case S_IFREG:
1392                                 break;
1393                         case S_IFLNK:
1394                                 data->rpc_status = -ELOOP;
1395                                 break;
1396                         case S_IFDIR:
1397                                 data->rpc_status = -EISDIR;
1398                                 break;
1399                         default:
1400                                 data->rpc_status = -ENOTDIR;
1401                 }
1402                 renew_lease(data->o_res.server, data->timestamp);
1403                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1404                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1405         }
1406         data->rpc_done = 1;
1407 }
1408
1409 static void nfs4_open_release(void *calldata)
1410 {
1411         struct nfs4_opendata *data = calldata;
1412         struct nfs4_state *state = NULL;
1413
1414         /* If this request hasn't been cancelled, do nothing */
1415         if (data->cancelled == 0)
1416                 goto out_free;
1417         /* In case of error, no cleanup! */
1418         if (data->rpc_status != 0 || !data->rpc_done)
1419                 goto out_free;
1420         /* In case we need an open_confirm, no cleanup! */
1421         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1422                 goto out_free;
1423         state = nfs4_opendata_to_nfs4_state(data);
1424         if (!IS_ERR(state))
1425                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1426 out_free:
1427         nfs4_opendata_put(data);
1428 }
1429
1430 static const struct rpc_call_ops nfs4_open_ops = {
1431         .rpc_call_prepare = nfs4_open_prepare,
1432         .rpc_call_done = nfs4_open_done,
1433         .rpc_release = nfs4_open_release,
1434 };
1435
1436 static const struct rpc_call_ops nfs4_recover_open_ops = {
1437         .rpc_call_prepare = nfs4_recover_open_prepare,
1438         .rpc_call_done = nfs4_open_done,
1439         .rpc_release = nfs4_open_release,
1440 };
1441
1442 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1443 {
1444         struct inode *dir = data->dir->d_inode;
1445         struct nfs_server *server = NFS_SERVER(dir);
1446         struct nfs_openargs *o_arg = &data->o_arg;
1447         struct nfs_openres *o_res = &data->o_res;
1448         struct rpc_task *task;
1449         struct rpc_message msg = {
1450                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1451                 .rpc_argp = o_arg,
1452                 .rpc_resp = o_res,
1453                 .rpc_cred = data->owner->so_cred,
1454         };
1455         struct rpc_task_setup task_setup_data = {
1456                 .rpc_client = server->client,
1457                 .rpc_message = &msg,
1458                 .callback_ops = &nfs4_open_ops,
1459                 .callback_data = data,
1460                 .workqueue = nfsiod_workqueue,
1461                 .flags = RPC_TASK_ASYNC,
1462         };
1463         int status;
1464
1465         kref_get(&data->kref);
1466         data->rpc_done = 0;
1467         data->rpc_status = 0;
1468         data->cancelled = 0;
1469         if (isrecover)
1470                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1471         task = rpc_run_task(&task_setup_data);
1472         if (IS_ERR(task))
1473                 return PTR_ERR(task);
1474         status = nfs4_wait_for_completion_rpc_task(task);
1475         if (status != 0) {
1476                 data->cancelled = 1;
1477                 smp_wmb();
1478         } else
1479                 status = data->rpc_status;
1480         rpc_put_task(task);
1481
1482         return status;
1483 }
1484
1485 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1486 {
1487         struct inode *dir = data->dir->d_inode;
1488         struct nfs_openres *o_res = &data->o_res;
1489         int status;
1490
1491         status = nfs4_run_open_task(data, 1);
1492         if (status != 0 || !data->rpc_done)
1493                 return status;
1494
1495         nfs_refresh_inode(dir, o_res->dir_attr);
1496
1497         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1498                 status = _nfs4_proc_open_confirm(data);
1499                 if (status != 0)
1500                         return status;
1501         }
1502
1503         return status;
1504 }
1505
1506 /*
1507  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1508  */
1509 static int _nfs4_proc_open(struct nfs4_opendata *data)
1510 {
1511         struct inode *dir = data->dir->d_inode;
1512         struct nfs_server *server = NFS_SERVER(dir);
1513         struct nfs_openargs *o_arg = &data->o_arg;
1514         struct nfs_openres *o_res = &data->o_res;
1515         int status;
1516
1517         status = nfs4_run_open_task(data, 0);
1518         if (status != 0 || !data->rpc_done)
1519                 return status;
1520
1521         if (o_arg->open_flags & O_CREAT) {
1522                 update_changeattr(dir, &o_res->cinfo);
1523                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1524         } else
1525                 nfs_refresh_inode(dir, o_res->dir_attr);
1526         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1527                 status = _nfs4_proc_open_confirm(data);
1528                 if (status != 0)
1529                         return status;
1530         }
1531         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1532                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1533         return 0;
1534 }
1535
1536 static int nfs4_recover_expired_lease(struct nfs_server *server)
1537 {
1538         struct nfs_client *clp = server->nfs_client;
1539         unsigned int loop;
1540         int ret;
1541
1542         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1543                 ret = nfs4_wait_clnt_recover(clp);
1544                 if (ret != 0)
1545                         break;
1546                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1547                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1548                         break;
1549                 nfs4_schedule_state_recovery(clp);
1550                 ret = -EIO;
1551         }
1552         return ret;
1553 }
1554
1555 /*
1556  * OPEN_EXPIRED:
1557  *      reclaim state on the server after a network partition.
1558  *      Assumes caller holds the appropriate lock
1559  */
1560 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1561 {
1562         struct nfs4_opendata *opendata;
1563         int ret;
1564
1565         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1566         if (IS_ERR(opendata))
1567                 return PTR_ERR(opendata);
1568         ret = nfs4_open_recover(opendata, state);
1569         if (ret == -ESTALE)
1570                 d_drop(ctx->path.dentry);
1571         nfs4_opendata_put(opendata);
1572         return ret;
1573 }
1574
1575 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1576 {
1577         struct nfs_server *server = NFS_SERVER(state->inode);
1578         struct nfs4_exception exception = { };
1579         int err;
1580
1581         do {
1582                 err = _nfs4_open_expired(ctx, state);
1583                 switch (err) {
1584                 default:
1585                         goto out;
1586                 case -NFS4ERR_GRACE:
1587                 case -NFS4ERR_DELAY:
1588                 case -EKEYEXPIRED:
1589                         nfs4_handle_exception(server, err, &exception);
1590                         err = 0;
1591                 }
1592         } while (exception.retry);
1593 out:
1594         return err;
1595 }
1596
1597 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1598 {
1599         struct nfs_open_context *ctx;
1600         int ret;
1601
1602         ctx = nfs4_state_find_open_context(state);
1603         if (IS_ERR(ctx))
1604                 return PTR_ERR(ctx);
1605         ret = nfs4_do_open_expired(ctx, state);
1606         put_nfs_open_context(ctx);
1607         return ret;
1608 }
1609
1610 /*
1611  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1612  * fields corresponding to attributes that were used to store the verifier.
1613  * Make sure we clobber those fields in the later setattr call
1614  */
1615 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1616 {
1617         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1618             !(sattr->ia_valid & ATTR_ATIME_SET))
1619                 sattr->ia_valid |= ATTR_ATIME;
1620
1621         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1622             !(sattr->ia_valid & ATTR_MTIME_SET))
1623                 sattr->ia_valid |= ATTR_MTIME;
1624 }
1625
1626 /*
1627  * Returns a referenced nfs4_state
1628  */
1629 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1630 {
1631         struct nfs4_state_owner  *sp;
1632         struct nfs4_state     *state = NULL;
1633         struct nfs_server       *server = NFS_SERVER(dir);
1634         struct nfs4_opendata *opendata;
1635         int status;
1636
1637         /* Protect against reboot recovery conflicts */
1638         status = -ENOMEM;
1639         if (!(sp = nfs4_get_state_owner(server, cred))) {
1640                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1641                 goto out_err;
1642         }
1643         status = nfs4_recover_expired_lease(server);
1644         if (status != 0)
1645                 goto err_put_state_owner;
1646         if (path->dentry->d_inode != NULL)
1647                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1648         status = -ENOMEM;
1649         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr);
1650         if (opendata == NULL)
1651                 goto err_put_state_owner;
1652
1653         if (path->dentry->d_inode != NULL)
1654                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1655
1656         status = _nfs4_proc_open(opendata);
1657         if (status != 0)
1658                 goto err_opendata_put;
1659
1660         if (opendata->o_arg.open_flags & O_EXCL)
1661                 nfs4_exclusive_attrset(opendata, sattr);
1662
1663         state = nfs4_opendata_to_nfs4_state(opendata);
1664         status = PTR_ERR(state);
1665         if (IS_ERR(state))
1666                 goto err_opendata_put;
1667         if ((opendata->o_res.rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) != 0)
1668                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1669         nfs4_opendata_put(opendata);
1670         nfs4_put_state_owner(sp);
1671         *res = state;
1672         return 0;
1673 err_opendata_put:
1674         nfs4_opendata_put(opendata);
1675 err_put_state_owner:
1676         nfs4_put_state_owner(sp);
1677 out_err:
1678         *res = NULL;
1679         return status;
1680 }
1681
1682
1683 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1684 {
1685         struct nfs4_exception exception = { };
1686         struct nfs4_state *res;
1687         int status;
1688
1689         do {
1690                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1691                 if (status == 0)
1692                         break;
1693                 /* NOTE: BAD_SEQID means the server and client disagree about the
1694                  * book-keeping w.r.t. state-changing operations
1695                  * (OPEN/CLOSE/LOCK/LOCKU...)
1696                  * It is actually a sign of a bug on the client or on the server.
1697                  *
1698                  * If we receive a BAD_SEQID error in the particular case of
1699                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1700                  * have unhashed the old state_owner for us, and that we can
1701                  * therefore safely retry using a new one. We should still warn
1702                  * the user though...
1703                  */
1704                 if (status == -NFS4ERR_BAD_SEQID) {
1705                         printk(KERN_WARNING "NFS: v4 server %s "
1706                                         " returned a bad sequence-id error!\n",
1707                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1708                         exception.retry = 1;
1709                         continue;
1710                 }
1711                 /*
1712                  * BAD_STATEID on OPEN means that the server cancelled our
1713                  * state before it received the OPEN_CONFIRM.
1714                  * Recover by retrying the request as per the discussion
1715                  * on Page 181 of RFC3530.
1716                  */
1717                 if (status == -NFS4ERR_BAD_STATEID) {
1718                         exception.retry = 1;
1719                         continue;
1720                 }
1721                 if (status == -EAGAIN) {
1722                         /* We must have found a delegation */
1723                         exception.retry = 1;
1724                         continue;
1725                 }
1726                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1727                                         status, &exception));
1728         } while (exception.retry);
1729         return res;
1730 }
1731
1732 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1733                             struct nfs_fattr *fattr, struct iattr *sattr,
1734                             struct nfs4_state *state)
1735 {
1736         struct nfs_server *server = NFS_SERVER(inode);
1737         struct nfs_setattrargs  arg = {
1738                 .fh             = NFS_FH(inode),
1739                 .iap            = sattr,
1740                 .server         = server,
1741                 .bitmask = server->attr_bitmask,
1742         };
1743         struct nfs_setattrres  res = {
1744                 .fattr          = fattr,
1745                 .server         = server,
1746         };
1747         struct rpc_message msg = {
1748                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1749                 .rpc_argp       = &arg,
1750                 .rpc_resp       = &res,
1751                 .rpc_cred       = cred,
1752         };
1753         unsigned long timestamp = jiffies;
1754         int status;
1755
1756         nfs_fattr_init(fattr);
1757
1758         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1759                 /* Use that stateid */
1760         } else if (state != NULL) {
1761                 nfs4_copy_stateid(&arg.stateid, state, current->files);
1762         } else
1763                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1764
1765         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
1766         if (status == 0 && state != NULL)
1767                 renew_lease(server, timestamp);
1768         return status;
1769 }
1770
1771 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1772                            struct nfs_fattr *fattr, struct iattr *sattr,
1773                            struct nfs4_state *state)
1774 {
1775         struct nfs_server *server = NFS_SERVER(inode);
1776         struct nfs4_exception exception = { };
1777         int err;
1778         do {
1779                 err = nfs4_handle_exception(server,
1780                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1781                                 &exception);
1782         } while (exception.retry);
1783         return err;
1784 }
1785
1786 struct nfs4_closedata {
1787         struct path path;
1788         struct inode *inode;
1789         struct nfs4_state *state;
1790         struct nfs_closeargs arg;
1791         struct nfs_closeres res;
1792         struct nfs_fattr fattr;
1793         unsigned long timestamp;
1794 };
1795
1796 static void nfs4_free_closedata(void *data)
1797 {
1798         struct nfs4_closedata *calldata = data;
1799         struct nfs4_state_owner *sp = calldata->state->owner;
1800
1801         nfs4_put_open_state(calldata->state);
1802         nfs_free_seqid(calldata->arg.seqid);
1803         nfs4_put_state_owner(sp);
1804         path_put(&calldata->path);
1805         kfree(calldata);
1806 }
1807
1808 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1809                 fmode_t fmode)
1810 {
1811         spin_lock(&state->owner->so_lock);
1812         if (!(fmode & FMODE_READ))
1813                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1814         if (!(fmode & FMODE_WRITE))
1815                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1816         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1817         spin_unlock(&state->owner->so_lock);
1818 }
1819
1820 static void nfs4_close_done(struct rpc_task *task, void *data)
1821 {
1822         struct nfs4_closedata *calldata = data;
1823         struct nfs4_state *state = calldata->state;
1824         struct nfs_server *server = NFS_SERVER(calldata->inode);
1825
1826         nfs4_sequence_done(server, &calldata->res.seq_res, task->tk_status);
1827         if (RPC_ASSASSINATED(task))
1828                 return;
1829         /* hmm. we are done with the inode, and in the process of freeing
1830          * the state_owner. we keep this around to process errors
1831          */
1832         switch (task->tk_status) {
1833                 case 0:
1834                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1835                         renew_lease(server, calldata->timestamp);
1836                         nfs4_close_clear_stateid_flags(state,
1837                                         calldata->arg.fmode);
1838                         break;
1839                 case -NFS4ERR_STALE_STATEID:
1840                 case -NFS4ERR_OLD_STATEID:
1841                 case -NFS4ERR_BAD_STATEID:
1842                 case -NFS4ERR_EXPIRED:
1843                         if (calldata->arg.fmode == 0)
1844                                 break;
1845                 default:
1846                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1847                                 rpc_restart_call_prepare(task);
1848         }
1849         nfs_release_seqid(calldata->arg.seqid);
1850         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1851 }
1852
1853 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1854 {
1855         struct nfs4_closedata *calldata = data;
1856         struct nfs4_state *state = calldata->state;
1857         int call_close = 0;
1858
1859         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1860                 return;
1861
1862         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1863         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1864         spin_lock(&state->owner->so_lock);
1865         /* Calculate the change in open mode */
1866         if (state->n_rdwr == 0) {
1867                 if (state->n_rdonly == 0) {
1868                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1869                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1870                         calldata->arg.fmode &= ~FMODE_READ;
1871                 }
1872                 if (state->n_wronly == 0) {
1873                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1874                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1875                         calldata->arg.fmode &= ~FMODE_WRITE;
1876                 }
1877         }
1878         spin_unlock(&state->owner->so_lock);
1879
1880         if (!call_close) {
1881                 /* Note: exit _without_ calling nfs4_close_done */
1882                 task->tk_action = NULL;
1883                 return;
1884         }
1885
1886         if (calldata->arg.fmode == 0)
1887                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1888
1889         nfs_fattr_init(calldata->res.fattr);
1890         calldata->timestamp = jiffies;
1891         if (nfs4_setup_sequence((NFS_SERVER(calldata->inode))->nfs_client,
1892                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1893                                 1, task))
1894                 return;
1895         rpc_call_start(task);
1896 }
1897
1898 static const struct rpc_call_ops nfs4_close_ops = {
1899         .rpc_call_prepare = nfs4_close_prepare,
1900         .rpc_call_done = nfs4_close_done,
1901         .rpc_release = nfs4_free_closedata,
1902 };
1903
1904 /*
1905  * It is possible for data to be read/written from a mem-mapped file
1906  * after the sys_close call (which hits the vfs layer as a flush).
1907  * This means that we can't safely call nfsv4 close on a file until
1908  * the inode is cleared. This in turn means that we are not good
1909  * NFSv4 citizens - we do not indicate to the server to update the file's
1910  * share state even when we are done with one of the three share
1911  * stateid's in the inode.
1912  *
1913  * NOTE: Caller must be holding the sp->so_owner semaphore!
1914  */
1915 int nfs4_do_close(struct path *path, struct nfs4_state *state, int wait)
1916 {
1917         struct nfs_server *server = NFS_SERVER(state->inode);
1918         struct nfs4_closedata *calldata;
1919         struct nfs4_state_owner *sp = state->owner;
1920         struct rpc_task *task;
1921         struct rpc_message msg = {
1922                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
1923                 .rpc_cred = state->owner->so_cred,
1924         };
1925         struct rpc_task_setup task_setup_data = {
1926                 .rpc_client = server->client,
1927                 .rpc_message = &msg,
1928                 .callback_ops = &nfs4_close_ops,
1929                 .workqueue = nfsiod_workqueue,
1930                 .flags = RPC_TASK_ASYNC,
1931         };
1932         int status = -ENOMEM;
1933
1934         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
1935         if (calldata == NULL)
1936                 goto out;
1937         calldata->inode = state->inode;
1938         calldata->state = state;
1939         calldata->arg.fh = NFS_FH(state->inode);
1940         calldata->arg.stateid = &state->open_stateid;
1941         /* Serialization for the sequence id */
1942         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
1943         if (calldata->arg.seqid == NULL)
1944                 goto out_free_calldata;
1945         calldata->arg.fmode = 0;
1946         calldata->arg.bitmask = server->cache_consistency_bitmask;
1947         calldata->res.fattr = &calldata->fattr;
1948         calldata->res.seqid = calldata->arg.seqid;
1949         calldata->res.server = server;
1950         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
1951         path_get(path);
1952         calldata->path = *path;
1953
1954         msg.rpc_argp = &calldata->arg,
1955         msg.rpc_resp = &calldata->res,
1956         task_setup_data.callback_data = calldata;
1957         task = rpc_run_task(&task_setup_data);
1958         if (IS_ERR(task))
1959                 return PTR_ERR(task);
1960         status = 0;
1961         if (wait)
1962                 status = rpc_wait_for_completion_task(task);
1963         rpc_put_task(task);
1964         return status;
1965 out_free_calldata:
1966         kfree(calldata);
1967 out:
1968         nfs4_put_open_state(state);
1969         nfs4_put_state_owner(sp);
1970         return status;
1971 }
1972
1973 static int nfs4_intent_set_file(struct nameidata *nd, struct path *path, struct nfs4_state *state, fmode_t fmode)
1974 {
1975         struct file *filp;
1976         int ret;
1977
1978         /* If the open_intent is for execute, we have an extra check to make */
1979         if (fmode & FMODE_EXEC) {
1980                 ret = nfs_may_open(state->inode,
1981                                 state->owner->so_cred,
1982                                 nd->intent.open.flags);
1983                 if (ret < 0)
1984                         goto out_close;
1985         }
1986         filp = lookup_instantiate_filp(nd, path->dentry, NULL);
1987         if (!IS_ERR(filp)) {
1988                 struct nfs_open_context *ctx;
1989                 ctx = nfs_file_open_context(filp);
1990                 ctx->state = state;
1991                 return 0;
1992         }
1993         ret = PTR_ERR(filp);
1994 out_close:
1995         nfs4_close_sync(path, state, fmode & (FMODE_READ|FMODE_WRITE));
1996         return ret;
1997 }
1998
1999 struct dentry *
2000 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
2001 {
2002         struct path path = {
2003                 .mnt = nd->path.mnt,
2004                 .dentry = dentry,
2005         };
2006         struct dentry *parent;
2007         struct iattr attr;
2008         struct rpc_cred *cred;
2009         struct nfs4_state *state;
2010         struct dentry *res;
2011         fmode_t fmode = nd->intent.open.flags & (FMODE_READ | FMODE_WRITE | FMODE_EXEC);
2012
2013         if (nd->flags & LOOKUP_CREATE) {
2014                 attr.ia_mode = nd->intent.open.create_mode;
2015                 attr.ia_valid = ATTR_MODE;
2016                 if (!IS_POSIXACL(dir))
2017                         attr.ia_mode &= ~current_umask();
2018         } else {
2019                 attr.ia_valid = 0;
2020                 BUG_ON(nd->intent.open.flags & O_CREAT);
2021         }
2022
2023         cred = rpc_lookup_cred();
2024         if (IS_ERR(cred))
2025                 return (struct dentry *)cred;
2026         parent = dentry->d_parent;
2027         /* Protect against concurrent sillydeletes */
2028         nfs_block_sillyrename(parent);
2029         state = nfs4_do_open(dir, &path, fmode, nd->intent.open.flags, &attr, cred);
2030         put_rpccred(cred);
2031         if (IS_ERR(state)) {
2032                 if (PTR_ERR(state) == -ENOENT) {
2033                         d_add(dentry, NULL);
2034                         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2035                 }
2036                 nfs_unblock_sillyrename(parent);
2037                 return (struct dentry *)state;
2038         }
2039         res = d_add_unique(dentry, igrab(state->inode));
2040         if (res != NULL)
2041                 path.dentry = res;
2042         nfs_set_verifier(path.dentry, nfs_save_change_attribute(dir));
2043         nfs_unblock_sillyrename(parent);
2044         nfs4_intent_set_file(nd, &path, state, fmode);
2045         return res;
2046 }
2047
2048 int
2049 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
2050 {
2051         struct path path = {
2052                 .mnt = nd->path.mnt,
2053                 .dentry = dentry,
2054         };
2055         struct rpc_cred *cred;
2056         struct nfs4_state *state;
2057         fmode_t fmode = openflags & (FMODE_READ | FMODE_WRITE);
2058
2059         cred = rpc_lookup_cred();
2060         if (IS_ERR(cred))
2061                 return PTR_ERR(cred);
2062         state = nfs4_do_open(dir, &path, fmode, openflags, NULL, cred);
2063         put_rpccred(cred);
2064         if (IS_ERR(state)) {
2065                 switch (PTR_ERR(state)) {
2066                         case -EPERM:
2067                         case -EACCES:
2068                         case -EDQUOT:
2069                         case -ENOSPC:
2070                         case -EROFS:
2071                                 lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
2072                                 return 1;
2073                         default:
2074                                 goto out_drop;
2075                 }
2076         }
2077         if (state->inode == dentry->d_inode) {
2078                 nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2079                 nfs4_intent_set_file(nd, &path, state, fmode);
2080                 return 1;
2081         }
2082         nfs4_close_sync(&path, state, fmode);
2083 out_drop:
2084         d_drop(dentry);
2085         return 0;
2086 }
2087
2088 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2089 {
2090         if (ctx->state == NULL)
2091                 return;
2092         if (is_sync)
2093                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2094         else
2095                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2096 }
2097
2098 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2099 {
2100         struct nfs4_server_caps_arg args = {
2101                 .fhandle = fhandle,
2102         };
2103         struct nfs4_server_caps_res res = {};
2104         struct rpc_message msg = {
2105                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2106                 .rpc_argp = &args,
2107                 .rpc_resp = &res,
2108         };
2109         int status;
2110
2111         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2112         if (status == 0) {
2113                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2114                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2115                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2116                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2117                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2118                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2119                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2120                         server->caps |= NFS_CAP_ACLS;
2121                 if (res.has_links != 0)
2122                         server->caps |= NFS_CAP_HARDLINKS;
2123                 if (res.has_symlinks != 0)
2124                         server->caps |= NFS_CAP_SYMLINKS;
2125                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2126                         server->caps |= NFS_CAP_FILEID;
2127                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2128                         server->caps |= NFS_CAP_MODE;
2129                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2130                         server->caps |= NFS_CAP_NLINK;
2131                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2132                         server->caps |= NFS_CAP_OWNER;
2133                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2134                         server->caps |= NFS_CAP_OWNER_GROUP;
2135                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2136                         server->caps |= NFS_CAP_ATIME;
2137                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2138                         server->caps |= NFS_CAP_CTIME;
2139                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2140                         server->caps |= NFS_CAP_MTIME;
2141
2142                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2143                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2144                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2145                 server->acl_bitmask = res.acl_bitmask;
2146         }
2147
2148         return status;
2149 }
2150
2151 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2152 {
2153         struct nfs4_exception exception = { };
2154         int err;
2155         do {
2156                 err = nfs4_handle_exception(server,
2157                                 _nfs4_server_capabilities(server, fhandle),
2158                                 &exception);
2159         } while (exception.retry);
2160         return err;
2161 }
2162
2163 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2164                 struct nfs_fsinfo *info)
2165 {
2166         struct nfs4_lookup_root_arg args = {
2167                 .bitmask = nfs4_fattr_bitmap,
2168         };
2169         struct nfs4_lookup_res res = {
2170                 .server = server,
2171                 .fattr = info->fattr,
2172                 .fh = fhandle,
2173         };
2174         struct rpc_message msg = {
2175                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2176                 .rpc_argp = &args,
2177                 .rpc_resp = &res,
2178         };
2179
2180         nfs_fattr_init(info->fattr);
2181         return nfs4_call_sync(server, &msg, &args, &res, 0);
2182 }
2183
2184 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2185                 struct nfs_fsinfo *info)
2186 {
2187         struct nfs4_exception exception = { };
2188         int err;
2189         do {
2190                 err = nfs4_handle_exception(server,
2191                                 _nfs4_lookup_root(server, fhandle, info),
2192                                 &exception);
2193         } while (exception.retry);
2194         return err;
2195 }
2196
2197 /*
2198  * get the file handle for the "/" directory on the server
2199  */
2200 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2201                               struct nfs_fsinfo *info)
2202 {
2203         int status;
2204
2205         status = nfs4_lookup_root(server, fhandle, info);
2206         if (status == 0)
2207                 status = nfs4_server_capabilities(server, fhandle);
2208         if (status == 0)
2209                 status = nfs4_do_fsinfo(server, fhandle, info);
2210         return nfs4_map_errors(status);
2211 }
2212
2213 /*
2214  * Get locations and (maybe) other attributes of a referral.
2215  * Note that we'll actually follow the referral later when
2216  * we detect fsid mismatch in inode revalidation
2217  */
2218 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2219 {
2220         int status = -ENOMEM;
2221         struct page *page = NULL;
2222         struct nfs4_fs_locations *locations = NULL;
2223
2224         page = alloc_page(GFP_KERNEL);
2225         if (page == NULL)
2226                 goto out;
2227         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2228         if (locations == NULL)
2229                 goto out;
2230
2231         status = nfs4_proc_fs_locations(dir, name, locations, page);
2232         if (status != 0)
2233                 goto out;
2234         /* Make sure server returned a different fsid for the referral */
2235         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2236                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2237                 status = -EIO;
2238                 goto out;
2239         }
2240
2241         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2242         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2243         if (!fattr->mode)
2244                 fattr->mode = S_IFDIR;
2245         memset(fhandle, 0, sizeof(struct nfs_fh));
2246 out:
2247         if (page)
2248                 __free_page(page);
2249         if (locations)
2250                 kfree(locations);
2251         return status;
2252 }
2253
2254 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2255 {
2256         struct nfs4_getattr_arg args = {
2257                 .fh = fhandle,
2258                 .bitmask = server->attr_bitmask,
2259         };
2260         struct nfs4_getattr_res res = {
2261                 .fattr = fattr,
2262                 .server = server,
2263         };
2264         struct rpc_message msg = {
2265                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2266                 .rpc_argp = &args,
2267                 .rpc_resp = &res,
2268         };
2269
2270         nfs_fattr_init(fattr);
2271         return nfs4_call_sync(server, &msg, &args, &res, 0);
2272 }
2273
2274 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2275 {
2276         struct nfs4_exception exception = { };
2277         int err;
2278         do {
2279                 err = nfs4_handle_exception(server,
2280                                 _nfs4_proc_getattr(server, fhandle, fattr),
2281                                 &exception);
2282         } while (exception.retry);
2283         return err;
2284 }
2285
2286 /*
2287  * The file is not closed if it is opened due to the a request to change
2288  * the size of the file. The open call will not be needed once the
2289  * VFS layer lookup-intents are implemented.
2290  *
2291  * Close is called when the inode is destroyed.
2292  * If we haven't opened the file for O_WRONLY, we
2293  * need to in the size_change case to obtain a stateid.
2294  *
2295  * Got race?
2296  * Because OPEN is always done by name in nfsv4, it is
2297  * possible that we opened a different file by the same
2298  * name.  We can recognize this race condition, but we
2299  * can't do anything about it besides returning an error.
2300  *
2301  * This will be fixed with VFS changes (lookup-intent).
2302  */
2303 static int
2304 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2305                   struct iattr *sattr)
2306 {
2307         struct inode *inode = dentry->d_inode;
2308         struct rpc_cred *cred = NULL;
2309         struct nfs4_state *state = NULL;
2310         int status;
2311
2312         nfs_fattr_init(fattr);
2313
2314         /* Search for an existing open(O_WRITE) file */
2315         if (sattr->ia_valid & ATTR_FILE) {
2316                 struct nfs_open_context *ctx;
2317
2318                 ctx = nfs_file_open_context(sattr->ia_file);
2319                 if (ctx) {
2320                         cred = ctx->cred;
2321                         state = ctx->state;
2322                 }
2323         }
2324
2325         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2326         if (status == 0)
2327                 nfs_setattr_update_inode(inode, sattr);
2328         return status;
2329 }
2330
2331 static int _nfs4_proc_lookupfh(struct nfs_server *server, const struct nfs_fh *dirfh,
2332                 const struct qstr *name, struct nfs_fh *fhandle,
2333                 struct nfs_fattr *fattr)
2334 {
2335         int                    status;
2336         struct nfs4_lookup_arg args = {
2337                 .bitmask = server->attr_bitmask,
2338                 .dir_fh = dirfh,
2339                 .name = name,
2340         };
2341         struct nfs4_lookup_res res = {
2342                 .server = server,
2343                 .fattr = fattr,
2344                 .fh = fhandle,
2345         };
2346         struct rpc_message msg = {
2347                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2348                 .rpc_argp = &args,
2349                 .rpc_resp = &res,
2350         };
2351
2352         nfs_fattr_init(fattr);
2353
2354         dprintk("NFS call  lookupfh %s\n", name->name);
2355         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2356         dprintk("NFS reply lookupfh: %d\n", status);
2357         return status;
2358 }
2359
2360 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2361                               struct qstr *name, struct nfs_fh *fhandle,
2362                               struct nfs_fattr *fattr)
2363 {
2364         struct nfs4_exception exception = { };
2365         int err;
2366         do {
2367                 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
2368                 /* FIXME: !!!! */
2369                 if (err == -NFS4ERR_MOVED) {
2370                         err = -EREMOTE;
2371                         break;
2372                 }
2373                 err = nfs4_handle_exception(server, err, &exception);
2374         } while (exception.retry);
2375         return err;
2376 }
2377
2378 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
2379                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2380 {
2381         int status;
2382
2383         dprintk("NFS call  lookup %s\n", name->name);
2384         status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2385         if (status == -NFS4ERR_MOVED)
2386                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2387         dprintk("NFS reply lookup: %d\n", status);
2388         return status;
2389 }
2390
2391 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2392 {
2393         struct nfs4_exception exception = { };
2394         int err;
2395         do {
2396                 err = nfs4_handle_exception(NFS_SERVER(dir),
2397                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
2398                                 &exception);
2399         } while (exception.retry);
2400         return err;
2401 }
2402
2403 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2404 {
2405         struct nfs_server *server = NFS_SERVER(inode);
2406         struct nfs_fattr fattr;
2407         struct nfs4_accessargs args = {
2408                 .fh = NFS_FH(inode),
2409                 .bitmask = server->attr_bitmask,
2410         };
2411         struct nfs4_accessres res = {
2412                 .server = server,
2413                 .fattr = &fattr,
2414         };
2415         struct rpc_message msg = {
2416                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2417                 .rpc_argp = &args,
2418                 .rpc_resp = &res,
2419                 .rpc_cred = entry->cred,
2420         };
2421         int mode = entry->mask;
2422         int status;
2423
2424         /*
2425          * Determine which access bits we want to ask for...
2426          */
2427         if (mode & MAY_READ)
2428                 args.access |= NFS4_ACCESS_READ;
2429         if (S_ISDIR(inode->i_mode)) {
2430                 if (mode & MAY_WRITE)
2431                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2432                 if (mode & MAY_EXEC)
2433                         args.access |= NFS4_ACCESS_LOOKUP;
2434         } else {
2435                 if (mode & MAY_WRITE)
2436                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2437                 if (mode & MAY_EXEC)
2438                         args.access |= NFS4_ACCESS_EXECUTE;
2439         }
2440         nfs_fattr_init(&fattr);
2441         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2442         if (!status) {
2443                 entry->mask = 0;
2444                 if (res.access & NFS4_ACCESS_READ)
2445                         entry->mask |= MAY_READ;
2446                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2447                         entry->mask |= MAY_WRITE;
2448                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2449                         entry->mask |= MAY_EXEC;
2450                 nfs_refresh_inode(inode, &fattr);
2451         }
2452         return status;
2453 }
2454
2455 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2456 {
2457         struct nfs4_exception exception = { };
2458         int err;
2459         do {
2460                 err = nfs4_handle_exception(NFS_SERVER(inode),
2461                                 _nfs4_proc_access(inode, entry),
2462                                 &exception);
2463         } while (exception.retry);
2464         return err;
2465 }
2466
2467 /*
2468  * TODO: For the time being, we don't try to get any attributes
2469  * along with any of the zero-copy operations READ, READDIR,
2470  * READLINK, WRITE.
2471  *
2472  * In the case of the first three, we want to put the GETATTR
2473  * after the read-type operation -- this is because it is hard
2474  * to predict the length of a GETATTR response in v4, and thus
2475  * align the READ data correctly.  This means that the GETATTR
2476  * may end up partially falling into the page cache, and we should
2477  * shift it into the 'tail' of the xdr_buf before processing.
2478  * To do this efficiently, we need to know the total length
2479  * of data received, which doesn't seem to be available outside
2480  * of the RPC layer.
2481  *
2482  * In the case of WRITE, we also want to put the GETATTR after
2483  * the operation -- in this case because we want to make sure
2484  * we get the post-operation mtime and size.  This means that
2485  * we can't use xdr_encode_pages() as written: we need a variant
2486  * of it which would leave room in the 'tail' iovec.
2487  *
2488  * Both of these changes to the XDR layer would in fact be quite
2489  * minor, but I decided to leave them for a subsequent patch.
2490  */
2491 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2492                 unsigned int pgbase, unsigned int pglen)
2493 {
2494         struct nfs4_readlink args = {
2495                 .fh       = NFS_FH(inode),
2496                 .pgbase   = pgbase,
2497                 .pglen    = pglen,
2498                 .pages    = &page,
2499         };
2500         struct nfs4_readlink_res res;
2501         struct rpc_message msg = {
2502                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2503                 .rpc_argp = &args,
2504                 .rpc_resp = &res,
2505         };
2506
2507         return nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
2508 }
2509
2510 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2511                 unsigned int pgbase, unsigned int pglen)
2512 {
2513         struct nfs4_exception exception = { };
2514         int err;
2515         do {
2516                 err = nfs4_handle_exception(NFS_SERVER(inode),
2517                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2518                                 &exception);
2519         } while (exception.retry);
2520         return err;
2521 }
2522
2523 /*
2524  * Got race?
2525  * We will need to arrange for the VFS layer to provide an atomic open.
2526  * Until then, this create/open method is prone to inefficiency and race
2527  * conditions due to the lookup, create, and open VFS calls from sys_open()
2528  * placed on the wire.
2529  *
2530  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2531  * The file will be opened again in the subsequent VFS open call
2532  * (nfs4_proc_file_open).
2533  *
2534  * The open for read will just hang around to be used by any process that
2535  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2536  */
2537
2538 static int
2539 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2540                  int flags, struct nameidata *nd)
2541 {
2542         struct path path = {
2543                 .mnt = nd->path.mnt,
2544                 .dentry = dentry,
2545         };
2546         struct nfs4_state *state;
2547         struct rpc_cred *cred;
2548         fmode_t fmode = flags & (FMODE_READ | FMODE_WRITE);
2549         int status = 0;
2550
2551         cred = rpc_lookup_cred();
2552         if (IS_ERR(cred)) {
2553                 status = PTR_ERR(cred);
2554                 goto out;
2555         }
2556         state = nfs4_do_open(dir, &path, fmode, flags, sattr, cred);
2557         d_drop(dentry);
2558         if (IS_ERR(state)) {
2559                 status = PTR_ERR(state);
2560                 goto out_putcred;
2561         }
2562         d_add(dentry, igrab(state->inode));
2563         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2564         if (flags & O_EXCL) {
2565                 struct nfs_fattr fattr;
2566                 status = nfs4_do_setattr(state->inode, cred, &fattr, sattr, state);
2567                 if (status == 0)
2568                         nfs_setattr_update_inode(state->inode, sattr);
2569                 nfs_post_op_update_inode(state->inode, &fattr);
2570         }
2571         if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0)
2572                 status = nfs4_intent_set_file(nd, &path, state, fmode);
2573         else
2574                 nfs4_close_sync(&path, state, fmode);
2575 out_putcred:
2576         put_rpccred(cred);
2577 out:
2578         return status;
2579 }
2580
2581 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2582 {
2583         struct nfs_server *server = NFS_SERVER(dir);
2584         struct nfs_removeargs args = {
2585                 .fh = NFS_FH(dir),
2586                 .name.len = name->len,
2587                 .name.name = name->name,
2588                 .bitmask = server->attr_bitmask,
2589         };
2590         struct nfs_removeres res = {
2591                 .server = server,
2592         };
2593         struct rpc_message msg = {
2594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2595                 .rpc_argp = &args,
2596                 .rpc_resp = &res,
2597         };
2598         int                     status;
2599
2600         nfs_fattr_init(&res.dir_attr);
2601         status = nfs4_call_sync(server, &msg, &args, &res, 1);
2602         if (status == 0) {
2603                 update_changeattr(dir, &res.cinfo);
2604                 nfs_post_op_update_inode(dir, &res.dir_attr);
2605         }
2606         return status;
2607 }
2608
2609 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2610 {
2611         struct nfs4_exception exception = { };
2612         int err;
2613         do {
2614                 err = nfs4_handle_exception(NFS_SERVER(dir),
2615                                 _nfs4_proc_remove(dir, name),
2616                                 &exception);
2617         } while (exception.retry);
2618         return err;
2619 }
2620
2621 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2622 {
2623         struct nfs_server *server = NFS_SERVER(dir);
2624         struct nfs_removeargs *args = msg->rpc_argp;
2625         struct nfs_removeres *res = msg->rpc_resp;
2626
2627         args->bitmask = server->cache_consistency_bitmask;
2628         res->server = server;
2629         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2630 }
2631
2632 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2633 {
2634         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2635
2636         nfs4_sequence_done(res->server, &res->seq_res, task->tk_status);
2637         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2638                 return 0;
2639         update_changeattr(dir, &res->cinfo);
2640         nfs_post_op_update_inode(dir, &res->dir_attr);
2641         return 1;
2642 }
2643
2644 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2645                 struct inode *new_dir, struct qstr *new_name)
2646 {
2647         struct nfs_server *server = NFS_SERVER(old_dir);
2648         struct nfs4_rename_arg arg = {
2649                 .old_dir = NFS_FH(old_dir),
2650                 .new_dir = NFS_FH(new_dir),
2651                 .old_name = old_name,
2652                 .new_name = new_name,
2653                 .bitmask = server->attr_bitmask,
2654         };
2655         struct nfs_fattr old_fattr, new_fattr;
2656         struct nfs4_rename_res res = {
2657                 .server = server,
2658                 .old_fattr = &old_fattr,
2659                 .new_fattr = &new_fattr,
2660         };
2661         struct rpc_message msg = {
2662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2663                 .rpc_argp = &arg,
2664                 .rpc_resp = &res,
2665         };
2666         int                     status;
2667
2668         nfs_fattr_init(res.old_fattr);
2669         nfs_fattr_init(res.new_fattr);
2670         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2671
2672         if (!status) {
2673                 update_changeattr(old_dir, &res.old_cinfo);
2674                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2675                 update_changeattr(new_dir, &res.new_cinfo);
2676                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2677         }
2678         return status;
2679 }
2680
2681 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2682                 struct inode *new_dir, struct qstr *new_name)
2683 {
2684         struct nfs4_exception exception = { };
2685         int err;
2686         do {
2687                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2688                                 _nfs4_proc_rename(old_dir, old_name,
2689                                         new_dir, new_name),
2690                                 &exception);
2691         } while (exception.retry);
2692         return err;
2693 }
2694
2695 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2696 {
2697         struct nfs_server *server = NFS_SERVER(inode);
2698         struct nfs4_link_arg arg = {
2699                 .fh     = NFS_FH(inode),
2700                 .dir_fh = NFS_FH(dir),
2701                 .name   = name,
2702                 .bitmask = server->attr_bitmask,
2703         };
2704         struct nfs_fattr fattr, dir_attr;
2705         struct nfs4_link_res res = {
2706                 .server = server,
2707                 .fattr = &fattr,
2708                 .dir_attr = &dir_attr,
2709         };
2710         struct rpc_message msg = {
2711                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2712                 .rpc_argp = &arg,
2713                 .rpc_resp = &res,
2714         };
2715         int                     status;
2716
2717         nfs_fattr_init(res.fattr);
2718         nfs_fattr_init(res.dir_attr);
2719         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2720         if (!status) {
2721                 update_changeattr(dir, &res.cinfo);
2722                 nfs_post_op_update_inode(dir, res.dir_attr);
2723                 nfs_post_op_update_inode(inode, res.fattr);
2724         }
2725
2726         return status;
2727 }
2728
2729 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2730 {
2731         struct nfs4_exception exception = { };
2732         int err;
2733         do {
2734                 err = nfs4_handle_exception(NFS_SERVER(inode),
2735                                 _nfs4_proc_link(inode, dir, name),
2736                                 &exception);
2737         } while (exception.retry);
2738         return err;
2739 }
2740
2741 struct nfs4_createdata {
2742         struct rpc_message msg;
2743         struct nfs4_create_arg arg;
2744         struct nfs4_create_res res;
2745         struct nfs_fh fh;
2746         struct nfs_fattr fattr;
2747         struct nfs_fattr dir_fattr;
2748 };
2749
2750 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2751                 struct qstr *name, struct iattr *sattr, u32 ftype)
2752 {
2753         struct nfs4_createdata *data;
2754
2755         data = kzalloc(sizeof(*data), GFP_KERNEL);
2756         if (data != NULL) {
2757                 struct nfs_server *server = NFS_SERVER(dir);
2758
2759                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2760                 data->msg.rpc_argp = &data->arg;
2761                 data->msg.rpc_resp = &data->res;
2762                 data->arg.dir_fh = NFS_FH(dir);
2763                 data->arg.server = server;
2764                 data->arg.name = name;
2765                 data->arg.attrs = sattr;
2766                 data->arg.ftype = ftype;
2767                 data->arg.bitmask = server->attr_bitmask;
2768                 data->res.server = server;
2769                 data->res.fh = &data->fh;
2770                 data->res.fattr = &data->fattr;
2771                 data->res.dir_fattr = &data->dir_fattr;
2772                 nfs_fattr_init(data->res.fattr);
2773                 nfs_fattr_init(data->res.dir_fattr);
2774         }
2775         return data;
2776 }
2777
2778 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2779 {
2780         int status = nfs4_call_sync(NFS_SERVER(dir), &data->msg,
2781                                     &data->arg, &data->res, 1);
2782         if (status == 0) {
2783                 update_changeattr(dir, &data->res.dir_cinfo);
2784                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2785                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2786         }
2787         return status;
2788 }
2789
2790 static void nfs4_free_createdata(struct nfs4_createdata *data)
2791 {
2792         kfree(data);
2793 }
2794
2795 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2796                 struct page *page, unsigned int len, struct iattr *sattr)
2797 {
2798         struct nfs4_createdata *data;
2799         int status = -ENAMETOOLONG;
2800
2801         if (len > NFS4_MAXPATHLEN)
2802                 goto out;
2803
2804         status = -ENOMEM;
2805         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2806         if (data == NULL)
2807                 goto out;
2808
2809         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2810         data->arg.u.symlink.pages = &page;
2811         data->arg.u.symlink.len = len;
2812
2813         status = nfs4_do_create(dir, dentry, data);
2814
2815         nfs4_free_createdata(data);
2816 out:
2817         return status;
2818 }
2819
2820 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2821                 struct page *page, unsigned int len, struct iattr *sattr)
2822 {
2823         struct nfs4_exception exception = { };
2824         int err;
2825         do {
2826                 err = nfs4_handle_exception(NFS_SERVER(dir),
2827                                 _nfs4_proc_symlink(dir, dentry, page,
2828                                                         len, sattr),
2829                                 &exception);
2830         } while (exception.retry);
2831         return err;
2832 }
2833
2834 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2835                 struct iattr *sattr)
2836 {
2837         struct nfs4_createdata *data;
2838         int status = -ENOMEM;
2839
2840         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2841         if (data == NULL)
2842                 goto out;
2843
2844         status = nfs4_do_create(dir, dentry, data);
2845
2846         nfs4_free_createdata(data);
2847 out:
2848         return status;
2849 }
2850
2851 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2852                 struct iattr *sattr)
2853 {
2854         struct nfs4_exception exception = { };
2855         int err;
2856         do {
2857                 err = nfs4_handle_exception(NFS_SERVER(dir),
2858                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2859                                 &exception);
2860         } while (exception.retry);
2861         return err;
2862 }
2863
2864 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2865                   u64 cookie, struct page *page, unsigned int count, int plus)
2866 {
2867         struct inode            *dir = dentry->d_inode;
2868         struct nfs4_readdir_arg args = {
2869                 .fh = NFS_FH(dir),
2870                 .pages = &page,
2871                 .pgbase = 0,
2872                 .count = count,
2873                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2874         };
2875         struct nfs4_readdir_res res;
2876         struct rpc_message msg = {
2877                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2878                 .rpc_argp = &args,
2879                 .rpc_resp = &res,
2880                 .rpc_cred = cred,
2881         };
2882         int                     status;
2883
2884         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2885                         dentry->d_parent->d_name.name,
2886                         dentry->d_name.name,
2887                         (unsigned long long)cookie);
2888         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2889         res.pgbase = args.pgbase;
2890         status = nfs4_call_sync(NFS_SERVER(dir), &msg, &args, &res, 0);
2891         if (status == 0)
2892                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2893
2894         nfs_invalidate_atime(dir);
2895
2896         dprintk("%s: returns %d\n", __func__, status);
2897         return status;
2898 }
2899
2900 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2901                   u64 cookie, struct page *page, unsigned int count, int plus)
2902 {
2903         struct nfs4_exception exception = { };
2904         int err;
2905         do {
2906                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2907                                 _nfs4_proc_readdir(dentry, cred, cookie,
2908                                         page, count, plus),
2909                                 &exception);
2910         } while (exception.retry);
2911         return err;
2912 }
2913
2914 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2915                 struct iattr *sattr, dev_t rdev)
2916 {
2917         struct nfs4_createdata *data;
2918         int mode = sattr->ia_mode;
2919         int status = -ENOMEM;
2920
2921         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2922         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2923
2924         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2925         if (data == NULL)
2926                 goto out;
2927
2928         if (S_ISFIFO(mode))
2929                 data->arg.ftype = NF4FIFO;
2930         else if (S_ISBLK(mode)) {
2931                 data->arg.ftype = NF4BLK;
2932                 data->arg.u.device.specdata1 = MAJOR(rdev);
2933                 data->arg.u.device.specdata2 = MINOR(rdev);
2934         }
2935         else if (S_ISCHR(mode)) {
2936                 data->arg.ftype = NF4CHR;
2937                 data->arg.u.device.specdata1 = MAJOR(rdev);
2938                 data->arg.u.device.specdata2 = MINOR(rdev);
2939         }
2940
2941         status = nfs4_do_create(dir, dentry, data);
2942
2943         nfs4_free_createdata(data);
2944 out:
2945         return status;
2946 }
2947
2948 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2949                 struct iattr *sattr, dev_t rdev)
2950 {
2951         struct nfs4_exception exception = { };
2952         int err;
2953         do {
2954                 err = nfs4_handle_exception(NFS_SERVER(dir),
2955                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
2956                                 &exception);
2957         } while (exception.retry);
2958         return err;
2959 }
2960
2961 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
2962                  struct nfs_fsstat *fsstat)
2963 {
2964         struct nfs4_statfs_arg args = {
2965                 .fh = fhandle,
2966                 .bitmask = server->attr_bitmask,
2967         };
2968         struct nfs4_statfs_res res = {
2969                 .fsstat = fsstat,
2970         };
2971         struct rpc_message msg = {
2972                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
2973                 .rpc_argp = &args,
2974                 .rpc_resp = &res,
2975         };
2976
2977         nfs_fattr_init(fsstat->fattr);
2978         return  nfs4_call_sync(server, &msg, &args, &res, 0);
2979 }
2980
2981 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
2982 {
2983         struct nfs4_exception exception = { };
2984         int err;
2985         do {
2986                 err = nfs4_handle_exception(server,
2987                                 _nfs4_proc_statfs(server, fhandle, fsstat),
2988                                 &exception);
2989         } while (exception.retry);
2990         return err;
2991 }
2992
2993 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
2994                 struct nfs_fsinfo *fsinfo)
2995 {
2996         struct nfs4_fsinfo_arg args = {
2997                 .fh = fhandle,
2998                 .bitmask = server->attr_bitmask,
2999         };
3000         struct nfs4_fsinfo_res res = {
3001                 .fsinfo = fsinfo,
3002         };
3003         struct rpc_message msg = {
3004                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3005                 .rpc_argp = &args,
3006                 .rpc_resp = &res,
3007         };
3008
3009         return nfs4_call_sync(server, &msg, &args, &res, 0);
3010 }
3011
3012 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3013 {
3014         struct nfs4_exception exception = { };
3015         int err;
3016
3017         do {
3018                 err = nfs4_handle_exception(server,
3019                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3020                                 &exception);
3021         } while (exception.retry);
3022         return err;
3023 }
3024
3025 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3026 {
3027         nfs_fattr_init(fsinfo->fattr);
3028         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3029 }
3030
3031 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3032                 struct nfs_pathconf *pathconf)
3033 {
3034         struct nfs4_pathconf_arg args = {
3035                 .fh = fhandle,
3036                 .bitmask = server->attr_bitmask,
3037         };
3038         struct nfs4_pathconf_res res = {
3039                 .pathconf = pathconf,
3040         };
3041         struct rpc_message msg = {
3042                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3043                 .rpc_argp = &args,
3044                 .rpc_resp = &res,
3045         };
3046
3047         /* None of the pathconf attributes are mandatory to implement */
3048         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3049                 memset(pathconf, 0, sizeof(*pathconf));
3050                 return 0;
3051         }
3052
3053         nfs_fattr_init(pathconf->fattr);
3054         return nfs4_call_sync(server, &msg, &args, &res, 0);
3055 }
3056
3057 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3058                 struct nfs_pathconf *pathconf)
3059 {
3060         struct nfs4_exception exception = { };
3061         int err;
3062
3063         do {
3064                 err = nfs4_handle_exception(server,
3065                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3066                                 &exception);
3067         } while (exception.retry);
3068         return err;
3069 }
3070
3071 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3072 {
3073         struct nfs_server *server = NFS_SERVER(data->inode);
3074
3075         dprintk("--> %s\n", __func__);
3076
3077         nfs4_sequence_done(server, &data->res.seq_res, task->tk_status);
3078
3079         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3080                 nfs_restart_rpc(task, server->nfs_client);
3081                 return -EAGAIN;
3082         }
3083
3084         nfs_invalidate_atime(data->inode);
3085         if (task->tk_status > 0)
3086                 renew_lease(server, data->timestamp);
3087         return 0;
3088 }
3089
3090 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3091 {
3092         data->timestamp   = jiffies;
3093         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3094 }
3095
3096 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3097 {
3098         struct inode *inode = data->inode;
3099
3100         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
3101                            task->tk_status);
3102
3103         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3104                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3105                 return -EAGAIN;
3106         }
3107         if (task->tk_status >= 0) {
3108                 renew_lease(NFS_SERVER(inode), data->timestamp);
3109                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3110         }
3111         return 0;
3112 }
3113
3114 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3115 {
3116         struct nfs_server *server = NFS_SERVER(data->inode);
3117
3118         data->args.bitmask = server->cache_consistency_bitmask;
3119         data->res.server = server;
3120         data->timestamp   = jiffies;
3121
3122         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3123 }
3124
3125 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3126 {
3127         struct inode *inode = data->inode;
3128
3129         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
3130                            task->tk_status);
3131         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3132                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3133                 return -EAGAIN;
3134         }
3135         nfs_refresh_inode(inode, data->res.fattr);
3136         return 0;
3137 }
3138
3139 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3140 {
3141         struct nfs_server *server = NFS_SERVER(data->inode);
3142
3143         data->args.bitmask = server->cache_consistency_bitmask;
3144         data->res.server = server;
3145         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3146 }
3147
3148 /*
3149  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3150  * standalone procedure for queueing an asynchronous RENEW.
3151  */
3152 static void nfs4_renew_release(void *data)
3153 {
3154         struct nfs_client *clp = data;
3155
3156         if (atomic_read(&clp->cl_count) > 1)
3157                 nfs4_schedule_state_renewal(clp);
3158         nfs_put_client(clp);
3159 }
3160
3161 static void nfs4_renew_done(struct rpc_task *task, void *data)
3162 {
3163         struct nfs_client *clp = data;
3164         unsigned long timestamp = task->tk_start;
3165
3166         if (task->tk_status < 0) {
3167                 /* Unless we're shutting down, schedule state recovery! */
3168                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3169                         nfs4_schedule_state_recovery(clp);
3170                 return;
3171         }
3172         spin_lock(&clp->cl_lock);
3173         if (time_before(clp->cl_last_renewal,timestamp))
3174                 clp->cl_last_renewal = timestamp;
3175         spin_unlock(&clp->cl_lock);
3176 }
3177
3178 static const struct rpc_call_ops nfs4_renew_ops = {
3179         .rpc_call_done = nfs4_renew_done,
3180         .rpc_release = nfs4_renew_release,
3181 };
3182
3183 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3184 {
3185         struct rpc_message msg = {
3186                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3187                 .rpc_argp       = clp,
3188                 .rpc_cred       = cred,
3189         };
3190
3191         if (!atomic_inc_not_zero(&clp->cl_count))
3192                 return -EIO;
3193         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3194                         &nfs4_renew_ops, clp);
3195 }
3196
3197 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3198 {
3199         struct rpc_message msg = {
3200                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3201                 .rpc_argp       = clp,
3202                 .rpc_cred       = cred,
3203         };
3204         unsigned long now = jiffies;
3205         int status;
3206
3207         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3208         if (status < 0)
3209                 return status;
3210         spin_lock(&clp->cl_lock);
3211         if (time_before(clp->cl_last_renewal,now))
3212                 clp->cl_last_renewal = now;
3213         spin_unlock(&clp->cl_lock);
3214         return 0;
3215 }
3216
3217 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3218 {
3219         return (server->caps & NFS_CAP_ACLS)
3220                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3221                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3222 }
3223
3224 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3225  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3226  * the stack.
3227  */
3228 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3229
3230 static void buf_to_pages(const void *buf, size_t buflen,
3231                 struct page **pages, unsigned int *pgbase)
3232 {
3233         const void *p = buf;
3234
3235         *pgbase = offset_in_page(buf);
3236         p -= *pgbase;
3237         while (p < buf + buflen) {
3238                 *(pages++) = virt_to_page(p);
3239                 p += PAGE_CACHE_SIZE;
3240         }
3241 }
3242
3243 struct nfs4_cached_acl {
3244         int cached;
3245         size_t len;
3246         char data[0];
3247 };
3248
3249 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3250 {
3251         struct nfs_inode *nfsi = NFS_I(inode);
3252
3253         spin_lock(&inode->i_lock);
3254         kfree(nfsi->nfs4_acl);
3255         nfsi->nfs4_acl = acl;
3256         spin_unlock(&inode->i_lock);
3257 }
3258
3259 static void nfs4_zap_acl_attr(struct inode *inode)
3260 {
3261         nfs4_set_cached_acl(inode, NULL);
3262 }
3263
3264 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3265 {
3266         struct nfs_inode *nfsi = NFS_I(inode);
3267         struct nfs4_cached_acl *acl;
3268         int ret = -ENOENT;
3269
3270         spin_lock(&inode->i_lock);
3271         acl = nfsi->nfs4_acl;
3272         if (acl == NULL)
3273                 goto out;
3274         if (buf == NULL) /* user is just asking for length */
3275                 goto out_len;
3276         if (acl->cached == 0)
3277                 goto out;
3278         ret = -ERANGE; /* see getxattr(2) man page */
3279         if (acl->len > buflen)
3280                 goto out;
3281         memcpy(buf, acl->data, acl->len);
3282 out_len:
3283         ret = acl->len;
3284 out:
3285         spin_unlock(&inode->i_lock);
3286         return ret;
3287 }
3288
3289 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3290 {
3291         struct nfs4_cached_acl *acl;
3292
3293         if (buf && acl_len <= PAGE_SIZE) {
3294                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3295                 if (acl == NULL)
3296                         goto out;
3297                 acl->cached = 1;
3298                 memcpy(acl->data, buf, acl_len);
3299         } else {
3300                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3301                 if (acl == NULL)
3302                         goto out;
3303                 acl->cached = 0;
3304         }
3305         acl->len = acl_len;
3306 out:
3307         nfs4_set_cached_acl(inode, acl);
3308 }
3309
3310 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3311 {
3312         struct page *pages[NFS4ACL_MAXPAGES];
3313         struct nfs_getaclargs args = {
3314                 .fh = NFS_FH(inode),
3315                 .acl_pages = pages,
3316                 .acl_len = buflen,
3317         };
3318         struct nfs_getaclres res = {
3319                 .acl_len = buflen,
3320         };
3321         void *resp_buf;
3322         struct rpc_message msg = {
3323                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3324                 .rpc_argp = &args,
3325                 .rpc_resp = &res,
3326         };
3327         struct page *localpage = NULL;
3328         int ret;
3329
3330         if (buflen < PAGE_SIZE) {
3331                 /* As long as we're doing a round trip to the server anyway,
3332                  * let's be prepared for a page of acl data. */
3333                 localpage = alloc_page(GFP_KERNEL);
3334                 resp_buf = page_address(localpage);
3335                 if (localpage == NULL)
3336                         return -ENOMEM;
3337                 args.acl_pages[0] = localpage;
3338                 args.acl_pgbase = 0;
3339                 args.acl_len = PAGE_SIZE;
3340         } else {
3341                 resp_buf = buf;
3342                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3343         }
3344         ret = nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
3345         if (ret)
3346                 goto out_free;
3347         if (res.acl_len > args.acl_len)
3348                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3349         else
3350                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3351         if (buf) {
3352                 ret = -ERANGE;
3353                 if (res.acl_len > buflen)
3354                         goto out_free;
3355                 if (localpage)
3356                         memcpy(buf, resp_buf, res.acl_len);
3357         }
3358         ret = res.acl_len;
3359 out_free:
3360         if (localpage)
3361                 __free_page(localpage);
3362         return ret;
3363 }
3364
3365 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3366 {
3367         struct nfs4_exception exception = { };
3368         ssize_t ret;
3369         do {
3370                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3371                 if (ret >= 0)
3372                         break;
3373                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3374         } while (exception.retry);
3375         return ret;
3376 }
3377
3378 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3379 {
3380         struct nfs_server *server = NFS_SERVER(inode);
3381         int ret;
3382
3383         if (!nfs4_server_supports_acls(server))
3384                 return -EOPNOTSUPP;
3385         ret = nfs_revalidate_inode(server, inode);
3386         if (ret < 0)
3387                 return ret;
3388         ret = nfs4_read_cached_acl(inode, buf, buflen);
3389         if (ret != -ENOENT)
3390                 return ret;
3391         return nfs4_get_acl_uncached(inode, buf, buflen);
3392 }
3393
3394 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3395 {
3396         struct nfs_server *server = NFS_SERVER(inode);
3397         struct page *pages[NFS4ACL_MAXPAGES];
3398         struct nfs_setaclargs arg = {
3399                 .fh             = NFS_FH(inode),
3400                 .acl_pages      = pages,
3401                 .acl_len        = buflen,
3402         };
3403         struct nfs_setaclres res;
3404         struct rpc_message msg = {
3405                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3406                 .rpc_argp       = &arg,
3407                 .rpc_resp       = &res,
3408         };
3409         int ret;
3410
3411         if (!nfs4_server_supports_acls(server))
3412                 return -EOPNOTSUPP;
3413         nfs_inode_return_delegation(inode);
3414         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3415         ret = nfs4_call_sync(server, &msg, &arg, &res, 1);
3416         nfs_access_zap_cache(inode);
3417         nfs_zap_acl_cache(inode);
3418         return ret;
3419 }
3420
3421 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3422 {
3423         struct nfs4_exception exception = { };
3424         int err;
3425         do {
3426                 err = nfs4_handle_exception(NFS_SERVER(inode),
3427                                 __nfs4_proc_set_acl(inode, buf, buflen),
3428                                 &exception);
3429         } while (exception.retry);
3430         return err;
3431 }
3432
3433 static int
3434 _nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs_client *clp, struct nfs4_state *state)
3435 {
3436         if (!clp || task->tk_status >= 0)
3437                 return 0;
3438         switch(task->tk_status) {
3439                 case -NFS4ERR_ADMIN_REVOKED:
3440                 case -NFS4ERR_BAD_STATEID:
3441                 case -NFS4ERR_OPENMODE:
3442                         if (state == NULL)
3443                                 break;
3444                         nfs4_state_mark_reclaim_nograce(clp, state);
3445                         goto do_state_recovery;
3446                 case -NFS4ERR_STALE_STATEID:
3447                         if (state == NULL)
3448                                 break;
3449                         nfs4_state_mark_reclaim_reboot(clp, state);
3450                 case -NFS4ERR_STALE_CLIENTID:
3451                 case -NFS4ERR_EXPIRED:
3452                         goto do_state_recovery;
3453 #if defined(CONFIG_NFS_V4_1)
3454                 case -NFS4ERR_BADSESSION:
3455                 case -NFS4ERR_BADSLOT:
3456                 case -NFS4ERR_BAD_HIGH_SLOT:
3457                 case -NFS4ERR_DEADSESSION:
3458                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3459                 case -NFS4ERR_SEQ_FALSE_RETRY:
3460                 case -NFS4ERR_SEQ_MISORDERED:
3461                         dprintk("%s ERROR %d, Reset session\n", __func__,
3462                                 task->tk_status);
3463                         nfs4_schedule_state_recovery(clp);
3464                         task->tk_status = 0;
3465                         return -EAGAIN;
3466 #endif /* CONFIG_NFS_V4_1 */
3467                 case -NFS4ERR_DELAY:
3468                         if (server)
3469                                 nfs_inc_server_stats(server, NFSIOS_DELAY);
3470                 case -NFS4ERR_GRACE:
3471                 case -EKEYEXPIRED:
3472                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3473                         task->tk_status = 0;
3474                         return -EAGAIN;
3475                 case -NFS4ERR_OLD_STATEID:
3476                         task->tk_status = 0;
3477                         return -EAGAIN;
3478         }
3479         task->tk_status = nfs4_map_errors(task->tk_status);
3480         return 0;
3481 do_state_recovery:
3482         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3483         nfs4_schedule_state_recovery(clp);
3484         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3485                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3486         task->tk_status = 0;
3487         return -EAGAIN;
3488 }
3489
3490 static int
3491 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3492 {
3493         return _nfs4_async_handle_error(task, server, server->nfs_client, state);
3494 }
3495
3496 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program, unsigned short port, struct rpc_cred *cred)
3497 {
3498         nfs4_verifier sc_verifier;
3499         struct nfs4_setclientid setclientid = {
3500                 .sc_verifier = &sc_verifier,
3501                 .sc_prog = program,
3502         };
3503         struct rpc_message msg = {
3504                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3505                 .rpc_argp = &setclientid,
3506                 .rpc_resp = clp,
3507                 .rpc_cred = cred,
3508         };
3509         __be32 *p;
3510         int loop = 0;
3511         int status;
3512
3513         p = (__be32*)sc_verifier.data;
3514         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3515         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3516
3517         for(;;) {
3518                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3519                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3520                                 clp->cl_ipaddr,
3521                                 rpc_peeraddr2str(clp->cl_rpcclient,
3522                                                         RPC_DISPLAY_ADDR),
3523                                 rpc_peeraddr2str(clp->cl_rpcclient,
3524                                                         RPC_DISPLAY_PROTO),
3525                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3526                                 clp->cl_id_uniquifier);
3527                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3528                                 sizeof(setclientid.sc_netid),
3529                                 rpc_peeraddr2str(clp->cl_rpcclient,
3530                                                         RPC_DISPLAY_NETID));
3531                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3532                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3533                                 clp->cl_ipaddr, port >> 8, port & 255);
3534
3535                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3536                 if (status != -NFS4ERR_CLID_INUSE)
3537                         break;
3538                 if (signalled())
3539                         break;
3540                 if (loop++ & 1)
3541                         ssleep(clp->cl_lease_time + 1);
3542                 else
3543                         if (++clp->cl_id_uniquifier == 0)
3544                                 break;
3545         }
3546         return status;
3547 }
3548
3549 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3550 {
3551         struct nfs_fsinfo fsinfo;
3552         struct rpc_message msg = {
3553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3554                 .rpc_argp = clp,
3555                 .rpc_resp = &fsinfo,
3556                 .rpc_cred = cred,
3557         };
3558         unsigned long now;
3559         int status;
3560
3561         now = jiffies;
3562         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3563         if (status == 0) {
3564                 spin_lock(&clp->cl_lock);
3565                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3566                 clp->cl_last_renewal = now;
3567                 spin_unlock(&clp->cl_lock);
3568         }
3569         return status;
3570 }
3571
3572 int nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3573 {
3574         long timeout = 0;
3575         int err;
3576         do {
3577                 err = _nfs4_proc_setclientid_confirm(clp, cred);
3578                 switch (err) {
3579                         case 0:
3580                                 return err;
3581                         case -NFS4ERR_RESOURCE:
3582                                 /* The IBM lawyers misread another document! */
3583                         case -NFS4ERR_DELAY:
3584                         case -EKEYEXPIRED:
3585                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3586                 }
3587         } while (err == 0);
3588         return err;
3589 }
3590
3591 struct nfs4_delegreturndata {
3592         struct nfs4_delegreturnargs args;
3593         struct nfs4_delegreturnres res;
3594         struct nfs_fh fh;
3595         nfs4_stateid stateid;
3596         unsigned long timestamp;
3597         struct nfs_fattr fattr;
3598         int rpc_status;
3599 };
3600
3601 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3602 {
3603         struct nfs4_delegreturndata *data = calldata;
3604
3605         nfs4_sequence_done(data->res.server, &data->res.seq_res,
3606                         task->tk_status);
3607
3608         switch (task->tk_status) {
3609         case -NFS4ERR_STALE_STATEID:
3610         case -NFS4ERR_EXPIRED:
3611         case 0:
3612                 renew_lease(data->res.server, data->timestamp);
3613                 break;
3614         default:
3615                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3616                                 -EAGAIN) {
3617                         nfs_restart_rpc(task, data->res.server->nfs_client);
3618                         return;
3619                 }
3620         }
3621         data->rpc_status = task->tk_status;
3622 }
3623
3624 static void nfs4_delegreturn_release(void *calldata)
3625 {
3626         kfree(calldata);
3627 }
3628
3629 #if defined(CONFIG_NFS_V4_1)
3630 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3631 {
3632         struct nfs4_delegreturndata *d_data;
3633
3634         d_data = (struct nfs4_delegreturndata *)data;
3635
3636         if (nfs4_setup_sequence(d_data->res.server->nfs_client,
3637                                 &d_data->args.seq_args,
3638                                 &d_data->res.seq_res, 1, task))
3639                 return;
3640         rpc_call_start(task);
3641 }
3642 #endif /* CONFIG_NFS_V4_1 */
3643
3644 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3645 #if defined(CONFIG_NFS_V4_1)
3646         .rpc_call_prepare = nfs4_delegreturn_prepare,
3647 #endif /* CONFIG_NFS_V4_1 */
3648         .rpc_call_done = nfs4_delegreturn_done,
3649         .rpc_release = nfs4_delegreturn_release,
3650 };
3651
3652 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3653 {
3654         struct nfs4_delegreturndata *data;
3655         struct nfs_server *server = NFS_SERVER(inode);
3656         struct rpc_task *task;
3657         struct rpc_message msg = {
3658                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3659                 .rpc_cred = cred,
3660         };
3661         struct rpc_task_setup task_setup_data = {
3662                 .rpc_client = server->client,
3663                 .rpc_message = &msg,
3664                 .callback_ops = &nfs4_delegreturn_ops,
3665                 .flags = RPC_TASK_ASYNC,
3666         };
3667         int status = 0;
3668
3669         data = kzalloc(sizeof(*data), GFP_KERNEL);
3670         if (data == NULL)
3671                 return -ENOMEM;
3672         data->args.fhandle = &data->fh;
3673         data->args.stateid = &data->stateid;
3674         data->args.bitmask = server->attr_bitmask;
3675         nfs_copy_fh(&data->fh, NFS_FH(inode));
3676         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3677         data->res.fattr = &data->fattr;
3678         data->res.server = server;
3679         data->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3680         nfs_fattr_init(data->res.fattr);
3681         data->timestamp = jiffies;
3682         data->rpc_status = 0;
3683
3684         task_setup_data.callback_data = data;
3685         msg.rpc_argp = &data->args,
3686         msg.rpc_resp = &data->res,
3687         task = rpc_run_task(&task_setup_data);
3688         if (IS_ERR(task))
3689                 return PTR_ERR(task);
3690         if (!issync)
3691                 goto out;
3692         status = nfs4_wait_for_completion_rpc_task(task);
3693         if (status != 0)
3694                 goto out;
3695         status = data->rpc_status;
3696         if (status != 0)
3697                 goto out;
3698         nfs_refresh_inode(inode, &data->fattr);
3699 out:
3700         rpc_put_task(task);
3701         return status;
3702 }
3703
3704 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3705 {
3706         struct nfs_server *server = NFS_SERVER(inode);
3707         struct nfs4_exception exception = { };
3708         int err;
3709         do {
3710                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3711                 switch (err) {
3712                         case -NFS4ERR_STALE_STATEID:
3713                         case -NFS4ERR_EXPIRED:
3714                         case 0:
3715                                 return 0;
3716                 }
3717                 err = nfs4_handle_exception(server, err, &exception);
3718         } while (exception.retry);
3719         return err;
3720 }
3721
3722 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3723 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3724
3725 /*
3726  * sleep, with exponential backoff, and retry the LOCK operation.
3727  */
3728 static unsigned long
3729 nfs4_set_lock_task_retry(unsigned long timeout)
3730 {
3731         schedule_timeout_killable(timeout);
3732         timeout <<= 1;
3733         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3734                 return NFS4_LOCK_MAXTIMEOUT;
3735         return timeout;
3736 }
3737
3738 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3739 {
3740         struct inode *inode = state->inode;
3741         struct nfs_server *server = NFS_SERVER(inode);
3742         struct nfs_client *clp = server->nfs_client;
3743         struct nfs_lockt_args arg = {
3744                 .fh = NFS_FH(inode),
3745                 .fl = request,
3746         };
3747         struct nfs_lockt_res res = {
3748                 .denied = request,
3749         };
3750         struct rpc_message msg = {
3751                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3752                 .rpc_argp       = &arg,
3753                 .rpc_resp       = &res,
3754                 .rpc_cred       = state->owner->so_cred,
3755         };
3756         struct nfs4_lock_state *lsp;
3757         int status;
3758
3759         arg.lock_owner.clientid = clp->cl_clientid;
3760         status = nfs4_set_lock_state(state, request);
3761         if (status != 0)
3762                 goto out;
3763         lsp = request->fl_u.nfs4_fl.owner;
3764         arg.lock_owner.id = lsp->ls_id.id;
3765         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
3766         switch (status) {
3767                 case 0:
3768                         request->fl_type = F_UNLCK;
3769                         break;
3770                 case -NFS4ERR_DENIED:
3771                         status = 0;
3772         }
3773         request->fl_ops->fl_release_private(request);
3774 out:
3775         return status;
3776 }
3777
3778 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3779 {
3780         struct nfs4_exception exception = { };
3781         int err;
3782
3783         do {
3784                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3785                                 _nfs4_proc_getlk(state, cmd, request),
3786                                 &exception);
3787         } while (exception.retry);
3788         return err;
3789 }
3790
3791 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3792 {
3793         int res = 0;
3794         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3795                 case FL_POSIX:
3796                         res = posix_lock_file_wait(file, fl);
3797                         break;
3798                 case FL_FLOCK:
3799                         res = flock_lock_file_wait(file, fl);
3800                         break;
3801                 default:
3802                         BUG();
3803         }
3804         return res;
3805 }
3806
3807 struct nfs4_unlockdata {
3808         struct nfs_locku_args arg;
3809         struct nfs_locku_res res;
3810         struct nfs4_lock_state *lsp;
3811         struct nfs_open_context *ctx;
3812         struct file_lock fl;
3813         const struct nfs_server *server;
3814         unsigned long timestamp;
3815 };
3816
3817 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3818                 struct nfs_open_context *ctx,
3819                 struct nfs4_lock_state *lsp,
3820                 struct nfs_seqid *seqid)
3821 {
3822         struct nfs4_unlockdata *p;
3823         struct inode *inode = lsp->ls_state->inode;
3824
3825         p = kzalloc(sizeof(*p), GFP_KERNEL);
3826         if (p == NULL)
3827                 return NULL;
3828         p->arg.fh = NFS_FH(inode);
3829         p->arg.fl = &p->fl;
3830         p->arg.seqid = seqid;
3831         p->res.seqid = seqid;
3832         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3833         p->arg.stateid = &lsp->ls_stateid;
3834         p->lsp = lsp;
3835         atomic_inc(&lsp->ls_count);
3836         /* Ensure we don't close file until we're done freeing locks! */
3837         p->ctx = get_nfs_open_context(ctx);
3838         memcpy(&p->fl, fl, sizeof(p->fl));
3839         p->server = NFS_SERVER(inode);
3840         return p;
3841 }
3842
3843 static void nfs4_locku_release_calldata(void *data)
3844 {
3845         struct nfs4_unlockdata *calldata = data;
3846         nfs_free_seqid(calldata->arg.seqid);
3847         nfs4_put_lock_state(calldata->lsp);
3848         put_nfs_open_context(calldata->ctx);
3849         kfree(calldata);
3850 }
3851
3852 static void nfs4_locku_done(struct rpc_task *task, void *data)
3853 {
3854         struct nfs4_unlockdata *calldata = data;
3855
3856         nfs4_sequence_done(calldata->server, &calldata->res.seq_res,
3857                            task->tk_status);
3858         if (RPC_ASSASSINATED(task))
3859                 return;
3860         switch (task->tk_status) {
3861                 case 0:
3862                         memcpy(calldata->lsp->ls_stateid.data,
3863                                         calldata->res.stateid.data,
3864                                         sizeof(calldata->lsp->ls_stateid.data));
3865                         renew_lease(calldata->server, calldata->timestamp);
3866                         break;
3867                 case -NFS4ERR_BAD_STATEID:
3868                 case -NFS4ERR_OLD_STATEID:
3869                 case -NFS4ERR_STALE_STATEID:
3870                 case -NFS4ERR_EXPIRED:
3871                         break;
3872                 default:
3873                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
3874                                 nfs_restart_rpc(task,
3875                                                  calldata->server->nfs_client);
3876         }
3877 }
3878
3879 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
3880 {
3881         struct nfs4_unlockdata *calldata = data;
3882
3883         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3884                 return;
3885         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
3886                 /* Note: exit _without_ running nfs4_locku_done */
3887                 task->tk_action = NULL;
3888                 return;
3889         }
3890         calldata->timestamp = jiffies;
3891         if (nfs4_setup_sequence(calldata->server->nfs_client,
3892                                 &calldata->arg.seq_args,
3893                                 &calldata->res.seq_res, 1, task))
3894                 return;
3895         rpc_call_start(task);
3896 }
3897
3898 static const struct rpc_call_ops nfs4_locku_ops = {
3899         .rpc_call_prepare = nfs4_locku_prepare,
3900         .rpc_call_done = nfs4_locku_done,
3901         .rpc_release = nfs4_locku_release_calldata,
3902 };
3903
3904 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
3905                 struct nfs_open_context *ctx,
3906                 struct nfs4_lock_state *lsp,
3907                 struct nfs_seqid *seqid)
3908 {
3909         struct nfs4_unlockdata *data;
3910         struct rpc_message msg = {
3911                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
3912                 .rpc_cred = ctx->cred,
3913         };
3914         struct rpc_task_setup task_setup_data = {
3915                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
3916                 .rpc_message = &msg,
3917                 .callback_ops = &nfs4_locku_ops,
3918                 .workqueue = nfsiod_workqueue,
3919                 .flags = RPC_TASK_ASYNC,
3920         };
3921
3922         /* Ensure this is an unlock - when canceling a lock, the
3923          * canceled lock is passed in, and it won't be an unlock.
3924          */
3925         fl->fl_type = F_UNLCK;
3926
3927         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
3928         if (data == NULL) {
3929                 nfs_free_seqid(seqid);
3930                 return ERR_PTR(-ENOMEM);
3931         }
3932
3933         msg.rpc_argp = &data->arg,
3934         msg.rpc_resp = &data->res,
3935         task_setup_data.callback_data = data;
3936         return rpc_run_task(&task_setup_data);
3937 }
3938
3939 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
3940 {
3941         struct nfs_inode *nfsi = NFS_I(state->inode);
3942         struct nfs_seqid *seqid;
3943         struct nfs4_lock_state *lsp;
3944         struct rpc_task *task;
3945         int status = 0;
3946         unsigned char fl_flags = request->fl_flags;
3947
3948         status = nfs4_set_lock_state(state, request);
3949         /* Unlock _before_ we do the RPC call */
3950         request->fl_flags |= FL_EXISTS;
3951         down_read(&nfsi->rwsem);
3952         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
3953                 up_read(&nfsi->rwsem);
3954                 goto out;
3955         }
3956         up_read(&nfsi->rwsem);
3957         if (status != 0)
3958                 goto out;
3959         /* Is this a delegated lock? */
3960         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
3961                 goto out;
3962         lsp = request->fl_u.nfs4_fl.owner;
3963         seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3964         status = -ENOMEM;
3965         if (seqid == NULL)
3966                 goto out;
3967         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
3968         status = PTR_ERR(task);
3969         if (IS_ERR(task))
3970                 goto out;
3971         status = nfs4_wait_for_completion_rpc_task(task);
3972         rpc_put_task(task);
3973 out:
3974         request->fl_flags = fl_flags;
3975         return status;
3976 }
3977
3978 struct nfs4_lockdata {
3979         struct nfs_lock_args arg;
3980         struct nfs_lock_res res;
3981         struct nfs4_lock_state *lsp;
3982         struct nfs_open_context *ctx;
3983         struct file_lock fl;
3984         unsigned long timestamp;
3985         int rpc_status;
3986         int cancelled;
3987         struct nfs_server *server;
3988 };
3989
3990 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
3991                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp)
3992 {
3993         struct nfs4_lockdata *p;
3994         struct inode *inode = lsp->ls_state->inode;
3995         struct nfs_server *server = NFS_SERVER(inode);
3996
3997         p = kzalloc(sizeof(*p), GFP_KERNEL);
3998         if (p == NULL)
3999                 return NULL;
4000
4001         p->arg.fh = NFS_FH(inode);
4002         p->arg.fl = &p->fl;
4003         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid);
4004         if (p->arg.open_seqid == NULL)
4005                 goto out_free;
4006         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
4007         if (p->arg.lock_seqid == NULL)
4008                 goto out_free_seqid;
4009         p->arg.lock_stateid = &lsp->ls_stateid;
4010         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4011         p->arg.lock_owner.id = lsp->ls_id.id;
4012         p->res.lock_seqid = p->arg.lock_seqid;
4013         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
4014         p->lsp = lsp;
4015         p->server = server;
4016         atomic_inc(&lsp->ls_count);
4017         p->ctx = get_nfs_open_context(ctx);
4018         memcpy(&p->fl, fl, sizeof(p->fl));
4019         return p;
4020 out_free_seqid:
4021         nfs_free_seqid(p->arg.open_seqid);
4022 out_free:
4023         kfree(p);
4024         return NULL;
4025 }
4026
4027 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4028 {
4029         struct nfs4_lockdata *data = calldata;
4030         struct nfs4_state *state = data->lsp->ls_state;
4031
4032         dprintk("%s: begin!\n", __func__);
4033         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4034                 return;
4035         /* Do we need to do an open_to_lock_owner? */
4036         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4037                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4038                         return;
4039                 data->arg.open_stateid = &state->stateid;
4040                 data->arg.new_lock_owner = 1;
4041                 data->res.open_seqid = data->arg.open_seqid;
4042         } else
4043                 data->arg.new_lock_owner = 0;
4044         data->timestamp = jiffies;
4045         if (nfs4_setup_sequence(data->server->nfs_client, &data->arg.seq_args,
4046                                 &data->res.seq_res, 1, task))
4047                 return;
4048         rpc_call_start(task);
4049         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4050 }
4051
4052 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4053 {
4054         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4055         nfs4_lock_prepare(task, calldata);
4056 }
4057
4058 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4059 {
4060         struct nfs4_lockdata *data = calldata;
4061
4062         dprintk("%s: begin!\n", __func__);
4063
4064         nfs4_sequence_done(data->server, &data->res.seq_res,
4065                         task->tk_status);
4066
4067         data->rpc_status = task->tk_status;
4068         if (RPC_ASSASSINATED(task))
4069                 goto out;
4070         if (data->arg.new_lock_owner != 0) {
4071                 if (data->rpc_status == 0)
4072                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4073                 else
4074                         goto out;
4075         }
4076         if (data->rpc_status == 0) {
4077                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4078                                         sizeof(data->lsp->ls_stateid.data));
4079                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4080                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4081         }
4082 out:
4083         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4084 }
4085
4086 static void nfs4_lock_release(void *calldata)
4087 {
4088         struct nfs4_lockdata *data = calldata;
4089
4090         dprintk("%s: begin!\n", __func__);
4091         nfs_free_seqid(data->arg.open_seqid);
4092         if (data->cancelled != 0) {
4093                 struct rpc_task *task;
4094                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4095                                 data->arg.lock_seqid);
4096                 if (!IS_ERR(task))
4097                         rpc_put_task(task);
4098                 dprintk("%s: cancelling lock!\n", __func__);
4099         } else
4100                 nfs_free_seqid(data->arg.lock_seqid);
4101         nfs4_put_lock_state(data->lsp);
4102         put_nfs_open_context(data->ctx);
4103         kfree(data);
4104         dprintk("%s: done!\n", __func__);
4105 }
4106
4107 static const struct rpc_call_ops nfs4_lock_ops = {
4108         .rpc_call_prepare = nfs4_lock_prepare,
4109         .rpc_call_done = nfs4_lock_done,
4110         .rpc_release = nfs4_lock_release,
4111 };
4112
4113 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4114         .rpc_call_prepare = nfs4_recover_lock_prepare,
4115         .rpc_call_done = nfs4_lock_done,
4116         .rpc_release = nfs4_lock_release,
4117 };
4118
4119 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4120 {
4121         struct nfs_client *clp = server->nfs_client;
4122         struct nfs4_state *state = lsp->ls_state;
4123
4124         switch (error) {
4125         case -NFS4ERR_ADMIN_REVOKED:
4126         case -NFS4ERR_BAD_STATEID:
4127         case -NFS4ERR_EXPIRED:
4128                 if (new_lock_owner != 0 ||
4129                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4130                         nfs4_state_mark_reclaim_nograce(clp, state);
4131                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4132                 break;
4133         case -NFS4ERR_STALE_STATEID:
4134                 if (new_lock_owner != 0 ||
4135                     (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4136                         nfs4_state_mark_reclaim_reboot(clp, state);
4137                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4138         };
4139 }
4140
4141 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4142 {
4143         struct nfs4_lockdata *data;
4144         struct rpc_task *task;
4145         struct rpc_message msg = {
4146                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4147                 .rpc_cred = state->owner->so_cred,
4148         };
4149         struct rpc_task_setup task_setup_data = {
4150                 .rpc_client = NFS_CLIENT(state->inode),
4151                 .rpc_message = &msg,
4152                 .callback_ops = &nfs4_lock_ops,
4153                 .workqueue = nfsiod_workqueue,
4154                 .flags = RPC_TASK_ASYNC,
4155         };
4156         int ret;
4157
4158         dprintk("%s: begin!\n", __func__);
4159         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4160                         fl->fl_u.nfs4_fl.owner);
4161         if (data == NULL)
4162                 return -ENOMEM;
4163         if (IS_SETLKW(cmd))
4164                 data->arg.block = 1;
4165         if (recovery_type > NFS_LOCK_NEW) {
4166                 if (recovery_type == NFS_LOCK_RECLAIM)
4167                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4168                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4169         }
4170         msg.rpc_argp = &data->arg,
4171         msg.rpc_resp = &data->res,
4172         task_setup_data.callback_data = data;
4173         task = rpc_run_task(&task_setup_data);
4174         if (IS_ERR(task))
4175                 return PTR_ERR(task);
4176         ret = nfs4_wait_for_completion_rpc_task(task);
4177         if (ret == 0) {
4178                 ret = data->rpc_status;
4179                 if (ret)
4180                         nfs4_handle_setlk_error(data->server, data->lsp,
4181                                         data->arg.new_lock_owner, ret);
4182         } else
4183                 data->cancelled = 1;
4184         rpc_put_task(task);
4185         dprintk("%s: done, ret = %d!\n", __func__, ret);
4186         return ret;
4187 }
4188
4189 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4190 {
4191         struct nfs_server *server = NFS_SERVER(state->inode);
4192         struct nfs4_exception exception = { };
4193         int err;
4194
4195         do {
4196                 /* Cache the lock if possible... */
4197                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4198                         return 0;
4199                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4200                 if (err != -NFS4ERR_DELAY && err != -EKEYEXPIRED)
4201                         break;
4202                 nfs4_handle_exception(server, err, &exception);
4203         } while (exception.retry);
4204         return err;
4205 }
4206
4207 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4208 {
4209         struct nfs_server *server = NFS_SERVER(state->inode);
4210         struct nfs4_exception exception = { };
4211         int err;
4212
4213         err = nfs4_set_lock_state(state, request);
4214         if (err != 0)
4215                 return err;
4216         do {
4217                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4218                         return 0;
4219                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4220                 switch (err) {
4221                 default:
4222                         goto out;
4223                 case -NFS4ERR_GRACE:
4224                 case -NFS4ERR_DELAY:
4225                 case -EKEYEXPIRED:
4226                         nfs4_handle_exception(server, err, &exception);
4227                         err = 0;
4228                 }
4229         } while (exception.retry);
4230 out:
4231         return err;
4232 }
4233
4234 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4235 {
4236         struct nfs_inode *nfsi = NFS_I(state->inode);
4237         unsigned char fl_flags = request->fl_flags;
4238         int status = -ENOLCK;
4239
4240         if ((fl_flags & FL_POSIX) &&
4241                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4242                 goto out;
4243         /* Is this a delegated open? */
4244         status = nfs4_set_lock_state(state, request);
4245         if (status != 0)
4246                 goto out;
4247         request->fl_flags |= FL_ACCESS;
4248         status = do_vfs_lock(request->fl_file, request);
4249         if (status < 0)
4250                 goto out;
4251         down_read(&nfsi->rwsem);
4252         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4253                 /* Yes: cache locks! */
4254                 /* ...but avoid races with delegation recall... */
4255                 request->fl_flags = fl_flags & ~FL_SLEEP;
4256                 status = do_vfs_lock(request->fl_file, request);
4257                 goto out_unlock;
4258         }
4259         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4260         if (status != 0)
4261                 goto out_unlock;
4262         /* Note: we always want to sleep here! */
4263         request->fl_flags = fl_flags | FL_SLEEP;
4264         if (do_vfs_lock(request->fl_file, request) < 0)
4265                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4266 out_unlock:
4267         up_read(&nfsi->rwsem);
4268 out:
4269         request->fl_flags = fl_flags;
4270         return status;
4271 }
4272
4273 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4274 {
4275         struct nfs4_exception exception = { };
4276         int err;
4277
4278         do {
4279                 err = _nfs4_proc_setlk(state, cmd, request);
4280                 if (err == -NFS4ERR_DENIED)
4281                         err = -EAGAIN;
4282                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4283                                 err, &exception);
4284         } while (exception.retry);
4285         return err;
4286 }
4287
4288 static int
4289 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4290 {
4291         struct nfs_open_context *ctx;
4292         struct nfs4_state *state;
4293         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4294         int status;
4295
4296         /* verify open state */
4297         ctx = nfs_file_open_context(filp);
4298         state = ctx->state;
4299
4300         if (request->fl_start < 0 || request->fl_end < 0)
4301                 return -EINVAL;
4302
4303         if (IS_GETLK(cmd)) {
4304                 if (state != NULL)
4305                         return nfs4_proc_getlk(state, F_GETLK, request);
4306                 return 0;
4307         }
4308
4309         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4310                 return -EINVAL;
4311
4312         if (request->fl_type == F_UNLCK) {
4313                 if (state != NULL)
4314                         return nfs4_proc_unlck(state, cmd, request);
4315                 return 0;
4316         }
4317
4318         if (state == NULL)
4319                 return -ENOLCK;
4320         do {
4321                 status = nfs4_proc_setlk(state, cmd, request);
4322                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4323                         break;
4324                 timeout = nfs4_set_lock_task_retry(timeout);
4325                 status = -ERESTARTSYS;
4326                 if (signalled())
4327                         break;
4328         } while(status < 0);
4329         return status;
4330 }
4331
4332 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4333 {
4334         struct nfs_server *server = NFS_SERVER(state->inode);
4335         struct nfs4_exception exception = { };
4336         int err;
4337
4338         err = nfs4_set_lock_state(state, fl);
4339         if (err != 0)
4340                 goto out;
4341         do {
4342                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4343                 switch (err) {
4344                         default:
4345                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4346                                                 __func__, err);
4347                         case 0:
4348                         case -ESTALE:
4349                                 goto out;
4350                         case -NFS4ERR_EXPIRED:
4351                         case -NFS4ERR_STALE_CLIENTID:
4352                         case -NFS4ERR_STALE_STATEID:
4353                         case -NFS4ERR_BADSESSION:
4354                         case -NFS4ERR_BADSLOT:
4355                         case -NFS4ERR_BAD_HIGH_SLOT:
4356                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4357                         case -NFS4ERR_DEADSESSION:
4358                                 nfs4_schedule_state_recovery(server->nfs_client);
4359                                 goto out;
4360                         case -ERESTARTSYS:
4361                                 /*
4362                                  * The show must go on: exit, but mark the
4363                                  * stateid as needing recovery.
4364                                  */
4365                         case -NFS4ERR_ADMIN_REVOKED:
4366                         case -NFS4ERR_BAD_STATEID:
4367                         case -NFS4ERR_OPENMODE:
4368                                 nfs4_state_mark_reclaim_nograce(server->nfs_client, state);
4369                                 err = 0;
4370                                 goto out;
4371                         case -ENOMEM:
4372                         case -NFS4ERR_DENIED:
4373                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4374                                 err = 0;
4375                                 goto out;
4376                         case -NFS4ERR_DELAY:
4377                         case -EKEYEXPIRED:
4378                                 break;
4379                 }
4380                 err = nfs4_handle_exception(server, err, &exception);
4381         } while (exception.retry);
4382 out:
4383         return err;
4384 }
4385
4386 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4387
4388 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
4389                 size_t buflen, int flags)
4390 {
4391         struct inode *inode = dentry->d_inode;
4392
4393         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4394                 return -EOPNOTSUPP;
4395
4396         return nfs4_proc_set_acl(inode, buf, buflen);
4397 }
4398
4399 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
4400  * and that's what we'll do for e.g. user attributes that haven't been set.
4401  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
4402  * attributes in kernel-managed attribute namespaces. */
4403 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
4404                 size_t buflen)
4405 {
4406         struct inode *inode = dentry->d_inode;
4407
4408         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4409                 return -EOPNOTSUPP;
4410
4411         return nfs4_proc_get_acl(inode, buf, buflen);
4412 }
4413
4414 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
4415 {
4416         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
4417
4418         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4419                 return 0;
4420         if (buf && buflen < len)
4421                 return -ERANGE;
4422         if (buf)
4423                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
4424         return len;
4425 }
4426
4427 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4428 {
4429         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4430                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4431                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4432                 return;
4433
4434         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4435                 NFS_ATTR_FATTR_NLINK;
4436         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4437         fattr->nlink = 2;
4438 }
4439
4440 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4441                 struct nfs4_fs_locations *fs_locations, struct page *page)
4442 {
4443         struct nfs_server *server = NFS_SERVER(dir);
4444         u32 bitmask[2] = {
4445                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4446                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4447         };
4448         struct nfs4_fs_locations_arg args = {
4449                 .dir_fh = NFS_FH(dir),
4450                 .name = name,
4451                 .page = page,
4452                 .bitmask = bitmask,
4453         };
4454         struct nfs4_fs_locations_res res = {
4455                 .fs_locations = fs_locations,
4456         };
4457         struct rpc_message msg = {
4458                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4459                 .rpc_argp = &args,
4460                 .rpc_resp = &res,
4461         };
4462         int status;
4463
4464         dprintk("%s: start\n", __func__);
4465         nfs_fattr_init(&fs_locations->fattr);
4466         fs_locations->server = server;
4467         fs_locations->nlocations = 0;
4468         status = nfs4_call_sync(server, &msg, &args, &res, 0);
4469         nfs_fixup_referral_attributes(&fs_locations->fattr);
4470         dprintk("%s: returned status = %d\n", __func__, status);
4471         return status;
4472 }
4473
4474 #ifdef CONFIG_NFS_V4_1
4475 /*
4476  * nfs4_proc_exchange_id()
4477  *
4478  * Since the clientid has expired, all compounds using sessions
4479  * associated with the stale clientid will be returning
4480  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4481  * be in some phase of session reset.
4482  */
4483 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4484 {
4485         nfs4_verifier verifier;
4486         struct nfs41_exchange_id_args args = {
4487                 .client = clp,
4488                 .flags = clp->cl_exchange_flags,
4489         };
4490         struct nfs41_exchange_id_res res = {
4491                 .client = clp,
4492         };
4493         int status;
4494         struct rpc_message msg = {
4495                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4496                 .rpc_argp = &args,
4497                 .rpc_resp = &res,
4498                 .rpc_cred = cred,
4499         };
4500         __be32 *p;
4501
4502         dprintk("--> %s\n", __func__);
4503         BUG_ON(clp == NULL);
4504
4505         /* Remove server-only flags */
4506         args.flags &= ~EXCHGID4_FLAG_CONFIRMED_R;
4507
4508         p = (u32 *)verifier.data;
4509         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4510         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4511         args.verifier = &verifier;
4512
4513         while (1) {
4514                 args.id_len = scnprintf(args.id, sizeof(args.id),
4515                                         "%s/%s %u",
4516                                         clp->cl_ipaddr,
4517                                         rpc_peeraddr2str(clp->cl_rpcclient,
4518                                                          RPC_DISPLAY_ADDR),
4519                                         clp->cl_id_uniquifier);
4520
4521                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4522
4523                 if (status != -NFS4ERR_CLID_INUSE)
4524                         break;
4525
4526                 if (signalled())
4527                         break;
4528
4529                 if (++clp->cl_id_uniquifier == 0)
4530                         break;
4531         }
4532
4533         dprintk("<-- %s status= %d\n", __func__, status);
4534         return status;
4535 }
4536
4537 struct nfs4_get_lease_time_data {
4538         struct nfs4_get_lease_time_args *args;
4539         struct nfs4_get_lease_time_res *res;
4540         struct nfs_client *clp;
4541 };
4542
4543 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4544                                         void *calldata)
4545 {
4546         int ret;
4547         struct nfs4_get_lease_time_data *data =
4548                         (struct nfs4_get_lease_time_data *)calldata;
4549
4550         dprintk("--> %s\n", __func__);
4551         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4552         /* just setup sequence, do not trigger session recovery
4553            since we're invoked within one */
4554         ret = nfs41_setup_sequence(data->clp->cl_session,
4555                                    &data->args->la_seq_args,
4556                                    &data->res->lr_seq_res, 0, task);
4557
4558         BUG_ON(ret == -EAGAIN);
4559         rpc_call_start(task);
4560         dprintk("<-- %s\n", __func__);
4561 }
4562
4563 /*
4564  * Called from nfs4_state_manager thread for session setup, so don't recover
4565  * from sequence operation or clientid errors.
4566  */
4567 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4568 {
4569         struct nfs4_get_lease_time_data *data =
4570                         (struct nfs4_get_lease_time_data *)calldata;
4571
4572         dprintk("--> %s\n", __func__);
4573         nfs41_sequence_done(data->clp, &data->res->lr_seq_res, task->tk_status);
4574         switch (task->tk_status) {
4575         case -NFS4ERR_DELAY:
4576         case -NFS4ERR_GRACE:
4577         case -EKEYEXPIRED:
4578                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4579                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4580                 task->tk_status = 0;
4581                 nfs_restart_rpc(task, data->clp);
4582                 return;
4583         }
4584         dprintk("<-- %s\n", __func__);
4585 }
4586
4587 struct rpc_call_ops nfs4_get_lease_time_ops = {
4588         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4589         .rpc_call_done = nfs4_get_lease_time_done,
4590 };
4591
4592 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4593 {
4594         struct rpc_task *task;
4595         struct nfs4_get_lease_time_args args;
4596         struct nfs4_get_lease_time_res res = {
4597                 .lr_fsinfo = fsinfo,
4598         };
4599         struct nfs4_get_lease_time_data data = {
4600                 .args = &args,
4601                 .res = &res,
4602                 .clp = clp,
4603         };
4604         struct rpc_message msg = {
4605                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4606                 .rpc_argp = &args,
4607                 .rpc_resp = &res,
4608         };
4609         struct rpc_task_setup task_setup = {
4610                 .rpc_client = clp->cl_rpcclient,
4611                 .rpc_message = &msg,
4612                 .callback_ops = &nfs4_get_lease_time_ops,
4613                 .callback_data = &data
4614         };
4615         int status;
4616
4617         res.lr_seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
4618         dprintk("--> %s\n", __func__);
4619         task = rpc_run_task(&task_setup);
4620
4621         if (IS_ERR(task))
4622                 status = PTR_ERR(task);
4623         else {
4624                 status = task->tk_status;
4625                 rpc_put_task(task);
4626         }
4627         dprintk("<-- %s return %d\n", __func__, status);
4628
4629         return status;
4630 }
4631
4632 /*
4633  * Reset a slot table
4634  */
4635 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4636                                  int ivalue)
4637 {
4638         struct nfs4_slot *new = NULL;
4639         int i;
4640         int ret = 0;
4641
4642         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4643                 max_reqs, tbl->max_slots);
4644
4645         /* Does the newly negotiated max_reqs match the existing slot table? */
4646         if (max_reqs != tbl->max_slots) {
4647                 ret = -ENOMEM;
4648                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4649                               GFP_KERNEL);
4650                 if (!new)
4651                         goto out;
4652                 ret = 0;
4653                 kfree(tbl->slots);
4654         }
4655         spin_lock(&tbl->slot_tbl_lock);
4656         if (new) {
4657                 tbl->slots = new;
4658                 tbl->max_slots = max_reqs;
4659         }
4660         for (i = 0; i < tbl->max_slots; ++i)
4661                 tbl->slots[i].seq_nr = ivalue;
4662         spin_unlock(&tbl->slot_tbl_lock);
4663         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4664                 tbl, tbl->slots, tbl->max_slots);
4665 out:
4666         dprintk("<-- %s: return %d\n", __func__, ret);
4667         return ret;
4668 }
4669
4670 /*
4671  * Reset the forechannel and backchannel slot tables
4672  */
4673 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4674 {
4675         int status;
4676
4677         status = nfs4_reset_slot_table(&session->fc_slot_table,
4678                         session->fc_attrs.max_reqs, 1);
4679         if (status)
4680                 return status;
4681
4682         status = nfs4_reset_slot_table(&session->bc_slot_table,
4683                         session->bc_attrs.max_reqs, 0);
4684         return status;
4685 }
4686
4687 /* Destroy the slot table */
4688 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4689 {
4690         if (session->fc_slot_table.slots != NULL) {
4691                 kfree(session->fc_slot_table.slots);
4692                 session->fc_slot_table.slots = NULL;
4693         }
4694         if (session->bc_slot_table.slots != NULL) {
4695                 kfree(session->bc_slot_table.slots);
4696                 session->bc_slot_table.slots = NULL;
4697         }
4698         return;
4699 }
4700
4701 /*
4702  * Initialize slot table
4703  */
4704 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4705                 int max_slots, int ivalue)
4706 {
4707         struct nfs4_slot *slot;
4708         int ret = -ENOMEM;
4709
4710         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4711
4712         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4713
4714         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_KERNEL);
4715         if (!slot)
4716                 goto out;
4717         ret = 0;
4718
4719         spin_lock(&tbl->slot_tbl_lock);
4720         tbl->max_slots = max_slots;
4721         tbl->slots = slot;
4722         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4723         spin_unlock(&tbl->slot_tbl_lock);
4724         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4725                 tbl, tbl->slots, tbl->max_slots);
4726 out:
4727         dprintk("<-- %s: return %d\n", __func__, ret);
4728         return ret;
4729 }
4730
4731 /*
4732  * Initialize the forechannel and backchannel tables
4733  */
4734 static int nfs4_init_slot_tables(struct nfs4_session *session)
4735 {
4736         struct nfs4_slot_table *tbl;
4737         int status = 0;
4738
4739         tbl = &session->fc_slot_table;
4740         if (tbl->slots == NULL) {
4741                 status = nfs4_init_slot_table(tbl,
4742                                 session->fc_attrs.max_reqs, 1);
4743                 if (status)
4744                         return status;
4745         }
4746
4747         tbl = &session->bc_slot_table;
4748         if (tbl->slots == NULL) {
4749                 status = nfs4_init_slot_table(tbl,
4750                                 session->bc_attrs.max_reqs, 0);
4751                 if (status)
4752                         nfs4_destroy_slot_tables(session);
4753         }
4754
4755         return status;
4756 }
4757
4758 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4759 {
4760         struct nfs4_session *session;
4761         struct nfs4_slot_table *tbl;
4762
4763         session = kzalloc(sizeof(struct nfs4_session), GFP_KERNEL);
4764         if (!session)
4765                 return NULL;
4766
4767         /*
4768          * The create session reply races with the server back
4769          * channel probe. Mark the client NFS_CS_SESSION_INITING
4770          * so that the client back channel can find the
4771          * nfs_client struct
4772          */
4773         clp->cl_cons_state = NFS_CS_SESSION_INITING;
4774         init_completion(&session->complete);
4775
4776         tbl = &session->fc_slot_table;
4777         tbl->highest_used_slotid = -1;
4778         spin_lock_init(&tbl->slot_tbl_lock);
4779         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
4780
4781         tbl = &session->bc_slot_table;
4782         tbl->highest_used_slotid = -1;
4783         spin_lock_init(&tbl->slot_tbl_lock);
4784         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
4785
4786         session->clp = clp;
4787         return session;
4788 }
4789
4790 void nfs4_destroy_session(struct nfs4_session *session)
4791 {
4792         nfs4_proc_destroy_session(session);
4793         dprintk("%s Destroy backchannel for xprt %p\n",
4794                 __func__, session->clp->cl_rpcclient->cl_xprt);
4795         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
4796                                 NFS41_BC_MIN_CALLBACKS);
4797         nfs4_destroy_slot_tables(session);
4798         kfree(session);
4799 }
4800
4801 /*
4802  * Initialize the values to be used by the client in CREATE_SESSION
4803  * If nfs4_init_session set the fore channel request and response sizes,
4804  * use them.
4805  *
4806  * Set the back channel max_resp_sz_cached to zero to force the client to
4807  * always set csa_cachethis to FALSE because the current implementation
4808  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
4809  */
4810 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
4811 {
4812         struct nfs4_session *session = args->client->cl_session;
4813         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
4814                      mxresp_sz = session->fc_attrs.max_resp_sz;
4815
4816         if (mxrqst_sz == 0)
4817                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
4818         if (mxresp_sz == 0)
4819                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
4820         /* Fore channel attributes */
4821         args->fc_attrs.headerpadsz = 0;
4822         args->fc_attrs.max_rqst_sz = mxrqst_sz;
4823         args->fc_attrs.max_resp_sz = mxresp_sz;
4824         args->fc_attrs.max_ops = NFS4_MAX_OPS;
4825         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
4826
4827         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
4828                 "max_ops=%u max_reqs=%u\n",
4829                 __func__,
4830                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
4831                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
4832
4833         /* Back channel attributes */
4834         args->bc_attrs.headerpadsz = 0;
4835         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
4836         args->bc_attrs.max_resp_sz = PAGE_SIZE;
4837         args->bc_attrs.max_resp_sz_cached = 0;
4838         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
4839         args->bc_attrs.max_reqs = 1;
4840
4841         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
4842                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
4843                 __func__,
4844                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
4845                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
4846                 args->bc_attrs.max_reqs);
4847 }
4848
4849 static int _verify_channel_attr(char *chan, char *attr_name, u32 sent, u32 rcvd)
4850 {
4851         if (rcvd <= sent)
4852                 return 0;
4853         printk(KERN_WARNING "%s: Session INVALID: %s channel %s increased. "
4854                 "sent=%u rcvd=%u\n", __func__, chan, attr_name, sent, rcvd);
4855         return -EINVAL;
4856 }
4857
4858 #define _verify_fore_channel_attr(_name_) \
4859         _verify_channel_attr("fore", #_name_, \
4860                              args->fc_attrs._name_, \
4861                              session->fc_attrs._name_)
4862
4863 #define _verify_back_channel_attr(_name_) \
4864         _verify_channel_attr("back", #_name_, \
4865                              args->bc_attrs._name_, \
4866                              session->bc_attrs._name_)
4867
4868 /*
4869  * The server is not allowed to increase the fore channel header pad size,
4870  * maximum response size, or maximum number of operations.
4871  *
4872  * The back channel attributes are only negotiatied down: We send what the
4873  * (back channel) server insists upon.
4874  */
4875 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
4876                                      struct nfs4_session *session)
4877 {
4878         int ret = 0;
4879
4880         ret |= _verify_fore_channel_attr(headerpadsz);
4881         ret |= _verify_fore_channel_attr(max_resp_sz);
4882         ret |= _verify_fore_channel_attr(max_ops);
4883
4884         ret |= _verify_back_channel_attr(headerpadsz);
4885         ret |= _verify_back_channel_attr(max_rqst_sz);
4886         ret |= _verify_back_channel_attr(max_resp_sz);
4887         ret |= _verify_back_channel_attr(max_resp_sz_cached);
4888         ret |= _verify_back_channel_attr(max_ops);
4889         ret |= _verify_back_channel_attr(max_reqs);
4890
4891         return ret;
4892 }
4893
4894 static int _nfs4_proc_create_session(struct nfs_client *clp)
4895 {
4896         struct nfs4_session *session = clp->cl_session;
4897         struct nfs41_create_session_args args = {
4898                 .client = clp,
4899                 .cb_program = NFS4_CALLBACK,
4900         };
4901         struct nfs41_create_session_res res = {
4902                 .client = clp,
4903         };
4904         struct rpc_message msg = {
4905                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
4906                 .rpc_argp = &args,
4907                 .rpc_resp = &res,
4908         };
4909         int status;
4910
4911         nfs4_init_channel_attrs(&args);
4912         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
4913
4914         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4915
4916         if (!status)
4917                 /* Verify the session's negotiated channel_attrs values */
4918                 status = nfs4_verify_channel_attrs(&args, session);
4919         if (!status) {
4920                 /* Increment the clientid slot sequence id */
4921                 clp->cl_seqid++;
4922         }
4923
4924         return status;
4925 }
4926
4927 /*
4928  * Issues a CREATE_SESSION operation to the server.
4929  * It is the responsibility of the caller to verify the session is
4930  * expired before calling this routine.
4931  */
4932 int nfs4_proc_create_session(struct nfs_client *clp)
4933 {
4934         int status;
4935         unsigned *ptr;
4936         struct nfs4_session *session = clp->cl_session;
4937
4938         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
4939
4940         status = _nfs4_proc_create_session(clp);
4941         if (status)
4942                 goto out;
4943
4944         /* Init and reset the fore channel */
4945         status = nfs4_init_slot_tables(session);
4946         dprintk("slot table initialization returned %d\n", status);
4947         if (status)
4948                 goto out;
4949         status = nfs4_reset_slot_tables(session);
4950         dprintk("slot table reset returned %d\n", status);
4951         if (status)
4952                 goto out;
4953
4954         ptr = (unsigned *)&session->sess_id.data[0];
4955         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
4956                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
4957 out:
4958         dprintk("<-- %s\n", __func__);
4959         return status;
4960 }
4961
4962 /*
4963  * Issue the over-the-wire RPC DESTROY_SESSION.
4964  * The caller must serialize access to this routine.
4965  */
4966 int nfs4_proc_destroy_session(struct nfs4_session *session)
4967 {
4968         int status = 0;
4969         struct rpc_message msg;
4970
4971         dprintk("--> nfs4_proc_destroy_session\n");
4972
4973         /* session is still being setup */
4974         if (session->clp->cl_cons_state != NFS_CS_READY)
4975                 return status;
4976
4977         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
4978         msg.rpc_argp = session;
4979         msg.rpc_resp = NULL;
4980         msg.rpc_cred = NULL;
4981         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4982
4983         if (status)
4984                 printk(KERN_WARNING
4985                         "Got error %d from the server on DESTROY_SESSION. "
4986                         "Session has been destroyed regardless...\n", status);
4987
4988         dprintk("<-- nfs4_proc_destroy_session\n");
4989         return status;
4990 }
4991
4992 int nfs4_init_session(struct nfs_server *server)
4993 {
4994         struct nfs_client *clp = server->nfs_client;
4995         struct nfs4_session *session;
4996         unsigned int rsize, wsize;
4997         int ret;
4998
4999         if (!nfs4_has_session(clp))
5000                 return 0;
5001
5002         rsize = server->rsize;
5003         if (rsize == 0)
5004                 rsize = NFS_MAX_FILE_IO_SIZE;
5005         wsize = server->wsize;
5006         if (wsize == 0)
5007                 wsize = NFS_MAX_FILE_IO_SIZE;
5008
5009         session = clp->cl_session;
5010         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5011         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5012
5013         ret = nfs4_recover_expired_lease(server);
5014         if (!ret)
5015                 ret = nfs4_check_client_ready(clp);
5016         return ret;
5017 }
5018
5019 /*
5020  * Renew the cl_session lease.
5021  */
5022 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5023 {
5024         struct nfs4_sequence_args args;
5025         struct nfs4_sequence_res res;
5026
5027         struct rpc_message msg = {
5028                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5029                 .rpc_argp = &args,
5030                 .rpc_resp = &res,
5031                 .rpc_cred = cred,
5032         };
5033
5034         args.sa_cache_this = 0;
5035
5036         return nfs4_call_sync_sequence(clp, clp->cl_rpcclient, &msg, &args,
5037                                        &res, args.sa_cache_this, 1);
5038 }
5039
5040 static void nfs41_sequence_release(void *data)
5041 {
5042         struct nfs_client *clp = (struct nfs_client *)data;
5043
5044         if (atomic_read(&clp->cl_count) > 1)
5045                 nfs4_schedule_state_renewal(clp);
5046         nfs_put_client(clp);
5047 }
5048
5049 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5050 {
5051         struct nfs_client *clp = (struct nfs_client *)data;
5052
5053         nfs41_sequence_done(clp, task->tk_msg.rpc_resp, task->tk_status);
5054
5055         if (task->tk_status < 0) {
5056                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5057                 if (atomic_read(&clp->cl_count) == 1)
5058                         goto out;
5059
5060                 if (_nfs4_async_handle_error(task, NULL, clp, NULL)
5061                                                                 == -EAGAIN) {
5062                         nfs_restart_rpc(task, clp);
5063                         return;
5064                 }
5065         }
5066         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5067 out:
5068         kfree(task->tk_msg.rpc_argp);
5069         kfree(task->tk_msg.rpc_resp);
5070
5071         dprintk("<-- %s\n", __func__);
5072 }
5073
5074 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5075 {
5076         struct nfs_client *clp;
5077         struct nfs4_sequence_args *args;
5078         struct nfs4_sequence_res *res;
5079
5080         clp = (struct nfs_client *)data;
5081         args = task->tk_msg.rpc_argp;
5082         res = task->tk_msg.rpc_resp;
5083
5084         if (nfs4_setup_sequence(clp, args, res, 0, task))
5085                 return;
5086         rpc_call_start(task);
5087 }
5088
5089 static const struct rpc_call_ops nfs41_sequence_ops = {
5090         .rpc_call_done = nfs41_sequence_call_done,
5091         .rpc_call_prepare = nfs41_sequence_prepare,
5092         .rpc_release = nfs41_sequence_release,
5093 };
5094
5095 static int nfs41_proc_async_sequence(struct nfs_client *clp,
5096                                      struct rpc_cred *cred)
5097 {
5098         struct nfs4_sequence_args *args;
5099         struct nfs4_sequence_res *res;
5100         struct rpc_message msg = {
5101                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5102                 .rpc_cred = cred,
5103         };
5104
5105         if (!atomic_inc_not_zero(&clp->cl_count))
5106                 return -EIO;
5107         args = kzalloc(sizeof(*args), GFP_KERNEL);
5108         res = kzalloc(sizeof(*res), GFP_KERNEL);
5109         if (!args || !res) {
5110                 kfree(args);
5111                 kfree(res);
5112                 nfs_put_client(clp);
5113                 return -ENOMEM;
5114         }
5115         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
5116         msg.rpc_argp = args;
5117         msg.rpc_resp = res;
5118
5119         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
5120                               &nfs41_sequence_ops, (void *)clp);
5121 }
5122
5123 struct nfs4_reclaim_complete_data {
5124         struct nfs_client *clp;
5125         struct nfs41_reclaim_complete_args arg;
5126         struct nfs41_reclaim_complete_res res;
5127 };
5128
5129 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5130 {
5131         struct nfs4_reclaim_complete_data *calldata = data;
5132
5133         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5134         if (nfs4_setup_sequence(calldata->clp, &calldata->arg.seq_args,
5135                                 &calldata->res.seq_res, 0, task))
5136                 return;
5137
5138         rpc_call_start(task);
5139 }
5140
5141 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5142 {
5143         struct nfs4_reclaim_complete_data *calldata = data;
5144         struct nfs_client *clp = calldata->clp;
5145         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5146
5147         dprintk("--> %s\n", __func__);
5148         nfs41_sequence_done(clp, res, task->tk_status);
5149         switch (task->tk_status) {
5150         case 0:
5151         case -NFS4ERR_COMPLETE_ALREADY:
5152                 break;
5153         case -NFS4ERR_BADSESSION:
5154         case -NFS4ERR_DEADSESSION:
5155                 /*
5156                  * Handle the session error, but do not retry the operation, as
5157                  * we have no way of telling whether the clientid had to be
5158                  * reset before we got our reply.  If reset, a new wave of
5159                  * reclaim operations will follow, containing their own reclaim
5160                  * complete.  We don't want our retry to get on the way of
5161                  * recovery by incorrectly indicating to the server that we're
5162                  * done reclaiming state since the process had to be restarted.
5163                  */
5164                 _nfs4_async_handle_error(task, NULL, clp, NULL);
5165                 break;
5166         default:
5167                 if (_nfs4_async_handle_error(
5168                                 task, NULL, clp, NULL) == -EAGAIN) {
5169                         rpc_restart_call_prepare(task);
5170                         return;
5171                 }
5172         }
5173
5174         dprintk("<-- %s\n", __func__);
5175 }
5176
5177 static void nfs4_free_reclaim_complete_data(void *data)
5178 {
5179         struct nfs4_reclaim_complete_data *calldata = data;
5180
5181         kfree(calldata);
5182 }
5183
5184 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5185         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5186         .rpc_call_done = nfs4_reclaim_complete_done,
5187         .rpc_release = nfs4_free_reclaim_complete_data,
5188 };
5189
5190 /*
5191  * Issue a global reclaim complete.
5192  */
5193 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5194 {
5195         struct nfs4_reclaim_complete_data *calldata;
5196         struct rpc_task *task;
5197         struct rpc_message msg = {
5198                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5199         };
5200         struct rpc_task_setup task_setup_data = {
5201                 .rpc_client = clp->cl_rpcclient,
5202                 .rpc_message = &msg,
5203                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5204                 .flags = RPC_TASK_ASYNC,
5205         };
5206         int status = -ENOMEM;
5207
5208         dprintk("--> %s\n", __func__);
5209         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
5210         if (calldata == NULL)
5211                 goto out;
5212         calldata->clp = clp;
5213         calldata->arg.one_fs = 0;
5214         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
5215
5216         msg.rpc_argp = &calldata->arg;
5217         msg.rpc_resp = &calldata->res;
5218         task_setup_data.callback_data = calldata;
5219         task = rpc_run_task(&task_setup_data);
5220         if (IS_ERR(task))
5221                 status = PTR_ERR(task);
5222         rpc_put_task(task);
5223 out:
5224         dprintk("<-- %s status=%d\n", __func__, status);
5225         return status;
5226 }
5227 #endif /* CONFIG_NFS_V4_1 */
5228
5229 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5230         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5231         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5232         .recover_open   = nfs4_open_reclaim,
5233         .recover_lock   = nfs4_lock_reclaim,
5234         .establish_clid = nfs4_init_clientid,
5235         .get_clid_cred  = nfs4_get_setclientid_cred,
5236 };
5237
5238 #if defined(CONFIG_NFS_V4_1)
5239 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5240         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5241         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5242         .recover_open   = nfs4_open_reclaim,
5243         .recover_lock   = nfs4_lock_reclaim,
5244         .establish_clid = nfs41_init_clientid,
5245         .get_clid_cred  = nfs4_get_exchange_id_cred,
5246         .reclaim_complete = nfs41_proc_reclaim_complete,
5247 };
5248 #endif /* CONFIG_NFS_V4_1 */
5249
5250 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5251         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5252         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5253         .recover_open   = nfs4_open_expired,
5254         .recover_lock   = nfs4_lock_expired,
5255         .establish_clid = nfs4_init_clientid,
5256         .get_clid_cred  = nfs4_get_setclientid_cred,
5257 };
5258
5259 #if defined(CONFIG_NFS_V4_1)
5260 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5261         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5262         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5263         .recover_open   = nfs4_open_expired,
5264         .recover_lock   = nfs4_lock_expired,
5265         .establish_clid = nfs41_init_clientid,
5266         .get_clid_cred  = nfs4_get_exchange_id_cred,
5267 };
5268 #endif /* CONFIG_NFS_V4_1 */
5269
5270 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5271         .sched_state_renewal = nfs4_proc_async_renew,
5272         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5273         .renew_lease = nfs4_proc_renew,
5274 };
5275
5276 #if defined(CONFIG_NFS_V4_1)
5277 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5278         .sched_state_renewal = nfs41_proc_async_sequence,
5279         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5280         .renew_lease = nfs4_proc_sequence,
5281 };
5282 #endif
5283
5284 /*
5285  * Per minor version reboot and network partition recovery ops
5286  */
5287
5288 struct nfs4_state_recovery_ops *nfs4_reboot_recovery_ops[] = {
5289         &nfs40_reboot_recovery_ops,
5290 #if defined(CONFIG_NFS_V4_1)
5291         &nfs41_reboot_recovery_ops,
5292 #endif
5293 };
5294
5295 struct nfs4_state_recovery_ops *nfs4_nograce_recovery_ops[] = {
5296         &nfs40_nograce_recovery_ops,
5297 #if defined(CONFIG_NFS_V4_1)
5298         &nfs41_nograce_recovery_ops,
5299 #endif
5300 };
5301
5302 struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[] = {
5303         &nfs40_state_renewal_ops,
5304 #if defined(CONFIG_NFS_V4_1)
5305         &nfs41_state_renewal_ops,
5306 #endif
5307 };
5308
5309 static const struct inode_operations nfs4_file_inode_operations = {
5310         .permission     = nfs_permission,
5311         .getattr        = nfs_getattr,
5312         .setattr        = nfs_setattr,
5313         .getxattr       = nfs4_getxattr,
5314         .setxattr       = nfs4_setxattr,
5315         .listxattr      = nfs4_listxattr,
5316 };
5317
5318 const struct nfs_rpc_ops nfs_v4_clientops = {
5319         .version        = 4,                    /* protocol version */
5320         .dentry_ops     = &nfs4_dentry_operations,
5321         .dir_inode_ops  = &nfs4_dir_inode_operations,
5322         .file_inode_ops = &nfs4_file_inode_operations,
5323         .getroot        = nfs4_proc_get_root,
5324         .getattr        = nfs4_proc_getattr,
5325         .setattr        = nfs4_proc_setattr,
5326         .lookupfh       = nfs4_proc_lookupfh,
5327         .lookup         = nfs4_proc_lookup,
5328         .access         = nfs4_proc_access,
5329         .readlink       = nfs4_proc_readlink,
5330         .create         = nfs4_proc_create,
5331         .remove         = nfs4_proc_remove,
5332         .unlink_setup   = nfs4_proc_unlink_setup,
5333         .unlink_done    = nfs4_proc_unlink_done,
5334         .rename         = nfs4_proc_rename,
5335         .link           = nfs4_proc_link,
5336         .symlink        = nfs4_proc_symlink,
5337         .mkdir          = nfs4_proc_mkdir,
5338         .rmdir          = nfs4_proc_remove,
5339         .readdir        = nfs4_proc_readdir,
5340         .mknod          = nfs4_proc_mknod,
5341         .statfs         = nfs4_proc_statfs,
5342         .fsinfo         = nfs4_proc_fsinfo,
5343         .pathconf       = nfs4_proc_pathconf,
5344         .set_capabilities = nfs4_server_capabilities,
5345         .decode_dirent  = nfs4_decode_dirent,
5346         .read_setup     = nfs4_proc_read_setup,
5347         .read_done      = nfs4_read_done,
5348         .write_setup    = nfs4_proc_write_setup,
5349         .write_done     = nfs4_write_done,
5350         .commit_setup   = nfs4_proc_commit_setup,
5351         .commit_done    = nfs4_commit_done,
5352         .lock           = nfs4_proc_lock,
5353         .clear_acl_cache = nfs4_zap_acl_attr,
5354         .close_context  = nfs4_close_context,
5355 };
5356
5357 /*
5358  * Local variables:
5359  *  c-basic-offset: 8
5360  * End:
5361  */