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 #include <linux/xattr.h>
  53 #include <linux/utsname.h>
  54
  55 #include "nfs4_fs.h"
  56 #include "delegation.h"
  57 #include "internal.h"
  58 #include "iostat.h"
  59 #include "callback.h"
  60 #include "pnfs.h"
  61
  62 #define NFSDBG_FACILITY         NFSDBG_PROC
  63
  64 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  65 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  66
  67 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  68
  69 struct nfs4_opendata;
  70 static int _nfs4_proc_open(struct nfs4_opendata *data);
  71 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  72 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  73 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  74 static int _nfs4_proc_lookup(struct rpc_clnt *client, struct inode *dir,
  75                              const struct qstr *name, struct nfs_fh *fhandle,
  76                              struct nfs_fattr *fattr);
  77 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  78 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  79                             struct nfs_fattr *fattr, struct iattr *sattr,
  80                             struct nfs4_state *state);
  81
  82 /* Prevent leaks of NFSv4 errors into userland */
  83 static int nfs4_map_errors(int err)
  84 {
  85         if (err >= -1000)
  86                 return err;
  87         switch (err) {
  88         case -NFS4ERR_RESOURCE:
  89                 return -EREMOTEIO;
  90         case -NFS4ERR_WRONGSEC:
  91                 return -EPERM;
  92         case -NFS4ERR_BADOWNER:
  93         case -NFS4ERR_BADNAME:
  94                 return -EINVAL;
  95         default:
  96                 dprintk("%s could not handle NFSv4 error %d\n",
  97                                 __func__, -err);
  98                 break;
  99         }
 100         return -EIO;
 101 }
 102
 103 /*
 104  * This is our standard bitmap for GETATTR requests.
 105  */
 106 const u32 nfs4_fattr_bitmap[2] = {
 107         FATTR4_WORD0_TYPE
 108         | FATTR4_WORD0_CHANGE
 109         | FATTR4_WORD0_SIZE
 110         | FATTR4_WORD0_FSID
 111         | FATTR4_WORD0_FILEID,
 112         FATTR4_WORD1_MODE
 113         | FATTR4_WORD1_NUMLINKS
 114         | FATTR4_WORD1_OWNER
 115         | FATTR4_WORD1_OWNER_GROUP
 116         | FATTR4_WORD1_RAWDEV
 117         | FATTR4_WORD1_SPACE_USED
 118         | FATTR4_WORD1_TIME_ACCESS
 119         | FATTR4_WORD1_TIME_METADATA
 120         | FATTR4_WORD1_TIME_MODIFY
 121 };
 122
 123 const u32 nfs4_statfs_bitmap[2] = {
 124         FATTR4_WORD0_FILES_AVAIL
 125         | FATTR4_WORD0_FILES_FREE
 126         | FATTR4_WORD0_FILES_TOTAL,
 127         FATTR4_WORD1_SPACE_AVAIL
 128         | FATTR4_WORD1_SPACE_FREE
 129         | FATTR4_WORD1_SPACE_TOTAL
 130 };
 131
 132 const u32 nfs4_pathconf_bitmap[2] = {
 133         FATTR4_WORD0_MAXLINK
 134         | FATTR4_WORD0_MAXNAME,
 135         0
 136 };
 137
 138 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 139                         | FATTR4_WORD0_MAXREAD
 140                         | FATTR4_WORD0_MAXWRITE
 141                         | FATTR4_WORD0_LEASE_TIME,
 142                         FATTR4_WORD1_TIME_DELTA
 143                         | FATTR4_WORD1_FS_LAYOUT_TYPES
 144 };
 145
 146 const u32 nfs4_fs_locations_bitmap[2] = {
 147         FATTR4_WORD0_TYPE
 148         | FATTR4_WORD0_CHANGE
 149         | FATTR4_WORD0_SIZE
 150         | FATTR4_WORD0_FSID
 151         | FATTR4_WORD0_FILEID
 152         | FATTR4_WORD0_FS_LOCATIONS,
 153         FATTR4_WORD1_MODE
 154         | FATTR4_WORD1_NUMLINKS
 155         | FATTR4_WORD1_OWNER
 156         | FATTR4_WORD1_OWNER_GROUP
 157         | FATTR4_WORD1_RAWDEV
 158         | FATTR4_WORD1_SPACE_USED
 159         | FATTR4_WORD1_TIME_ACCESS
 160         | FATTR4_WORD1_TIME_METADATA
 161         | FATTR4_WORD1_TIME_MODIFY
 162         | FATTR4_WORD1_MOUNTED_ON_FILEID
 163 };
 164
 165 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 166                 struct nfs4_readdir_arg *readdir)
 167 {
 168         __be32 *start, *p;
 169
 170         BUG_ON(readdir->count < 80);
 171         if (cookie > 2) {
 172                 readdir->cookie = cookie;
 173                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 174                 return;
 175         }
 176
 177         readdir->cookie = 0;
 178         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 179         if (cookie == 2)
 180                 return;
 181
 182         /*
 183          * NFSv4 servers do not return entries for '.' and '..'
 184          * Therefore, we fake these entries here.  We let '.'
 185          * have cookie 0 and '..' have cookie 1.  Note that
 186          * when talking to the server, we always send cookie 0
 187          * instead of 1 or 2.
 188          */
 189         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 190
 191         if (cookie == 0) {
 192                 *p++ = xdr_one;                                  /* next */
 193                 *p++ = xdr_zero;                   /* cookie, first word */
 194                 *p++ = xdr_one;                   /* cookie, second word */
 195                 *p++ = xdr_one;                             /* entry len */
 196                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 197                 p++;
 198                 *p++ = xdr_one;                         /* bitmap length */
 199                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 200                 *p++ = htonl(8);              /* attribute buffer length */
 201                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 202         }
 203
 204         *p++ = xdr_one;                                  /* next */
 205         *p++ = xdr_zero;                   /* cookie, first word */
 206         *p++ = xdr_two;                   /* cookie, second word */
 207         *p++ = xdr_two;                             /* entry len */
 208         memcpy(p, "..\0\0", 4);                         /* entry */
 209         p++;
 210         *p++ = xdr_one;                         /* bitmap length */
 211         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 212         *p++ = htonl(8);              /* attribute buffer length */
 213         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 214
 215         readdir->pgbase = (char *)p - (char *)start;
 216         readdir->count -= readdir->pgbase;
 217         kunmap_atomic(start, KM_USER0);
 218 }
 219
 220 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 221 {
 222         int res;
 223
 224         might_sleep();
 225
 226         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 227                         nfs_wait_bit_killable, TASK_KILLABLE);
 228         return res;
 229 }
 230
 231 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 232 {
 233         int res = 0;
 234
 235         might_sleep();
 236
 237         if (*timeout <= 0)
 238                 *timeout = NFS4_POLL_RETRY_MIN;
 239         if (*timeout > NFS4_POLL_RETRY_MAX)
 240                 *timeout = NFS4_POLL_RETRY_MAX;
 241         schedule_timeout_killable(*timeout);
 242         if (fatal_signal_pending(current))
 243                 res = -ERESTARTSYS;
 244         *timeout <<= 1;
 245         return res;
 246 }
 247
 248 /* This is the error handling routine for processes that are allowed
 249  * to sleep.
 250  */
 251 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 252 {
 253         struct nfs_client *clp = server->nfs_client;
 254         struct nfs4_state *state = exception->state;
 255         int ret = errorcode;
 256
 257         exception->retry = 0;
 258         switch(errorcode) {
 259                 case 0:
 260                         return 0;
 261                 case -NFS4ERR_ADMIN_REVOKED:
 262                 case -NFS4ERR_BAD_STATEID:
 263                 case -NFS4ERR_OPENMODE:
 264                         if (state == NULL)
 265                                 break;
 266                         nfs4_schedule_stateid_recovery(server, state);
 267                         goto wait_on_recovery;
 268                 case -NFS4ERR_STALE_STATEID:
 269                 case -NFS4ERR_STALE_CLIENTID:
 270                 case -NFS4ERR_EXPIRED:
 271                         nfs4_schedule_lease_recovery(clp);
 272                         goto wait_on_recovery;
 273 #if defined(CONFIG_NFS_V4_1)
 274                 case -NFS4ERR_BADSESSION:
 275                 case -NFS4ERR_BADSLOT:
 276                 case -NFS4ERR_BAD_HIGH_SLOT:
 277                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 278                 case -NFS4ERR_DEADSESSION:
 279                 case -NFS4ERR_SEQ_FALSE_RETRY:
 280                 case -NFS4ERR_SEQ_MISORDERED:
 281                         dprintk("%s ERROR: %d Reset session\n", __func__,
 282                                 errorcode);
 283                         nfs4_schedule_session_recovery(clp->cl_session);
 284                         exception->retry = 1;
 285                         break;
 286 #endif /* defined(CONFIG_NFS_V4_1) */
 287                 case -NFS4ERR_FILE_OPEN:
 288                         if (exception->timeout > HZ) {
 289                                 /* We have retried a decent amount, time to
 290                                  * fail
 291                                  */
 292                                 ret = -EBUSY;
 293                                 break;
 294                         }
 295                 case -NFS4ERR_GRACE:
 296                 case -NFS4ERR_DELAY:
 297                 case -EKEYEXPIRED:
 298                         ret = nfs4_delay(server->client, &exception->timeout);
 299                         if (ret != 0)
 300                                 break;
 301                 case -NFS4ERR_OLD_STATEID:
 302                         exception->retry = 1;
 303                         break;
 304                 case -NFS4ERR_BADOWNER:
 305                         /* The following works around a Linux server bug! */
 306                 case -NFS4ERR_BADNAME:
 307                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 308                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 309                                 exception->retry = 1;
 310                                 printk(KERN_WARNING "NFS: v4 server %s "
 311                                                 "does not accept raw "
 312                                                 "uid/gids. "
 313                                                 "Reenabling the idmapper.\n",
 314                                                 server->nfs_client->cl_hostname);
 315                         }
 316         }
 317         /* We failed to handle the error */
 318         return nfs4_map_errors(ret);
 319 wait_on_recovery:
 320         ret = nfs4_wait_clnt_recover(clp);
 321         if (ret == 0)
 322                 exception->retry = 1;
 323         return ret;
 324 }
 325
 326
 327 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 328 {
 329         spin_lock(&clp->cl_lock);
 330         if (time_before(clp->cl_last_renewal,timestamp))
 331                 clp->cl_last_renewal = timestamp;
 332         spin_unlock(&clp->cl_lock);
 333 }
 334
 335 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 336 {
 337         do_renew_lease(server->nfs_client, timestamp);
 338 }
 339
 340 #if defined(CONFIG_NFS_V4_1)
 341
 342 /*
 343  * nfs4_free_slot - free a slot and efficiently update slot table.
 344  *
 345  * freeing a slot is trivially done by clearing its respective bit
 346  * in the bitmap.
 347  * If the freed slotid equals highest_used_slotid we want to update it
 348  * so that the server would be able to size down the slot table if needed,
 349  * otherwise we know that the highest_used_slotid is still in use.
 350  * When updating highest_used_slotid there may be "holes" in the bitmap
 351  * so we need to scan down from highest_used_slotid to 0 looking for the now
 352  * highest slotid in use.
 353  * If none found, highest_used_slotid is set to -1.
 354  *
 355  * Must be called while holding tbl->slot_tbl_lock
 356  */
 357 static void
 358 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 359 {
 360         int free_slotid = free_slot - tbl->slots;
 361         int slotid = free_slotid;
 362
 363         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 364         /* clear used bit in bitmap */
 365         __clear_bit(slotid, tbl->used_slots);
 366
 367         /* update highest_used_slotid when it is freed */
 368         if (slotid == tbl->highest_used_slotid) {
 369                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 370                 if (slotid < tbl->max_slots)
 371                         tbl->highest_used_slotid = slotid;
 372                 else
 373                         tbl->highest_used_slotid = -1;
 374         }
 375         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 376                 free_slotid, tbl->highest_used_slotid);
 377 }
 378
 379 /*
 380  * Signal state manager thread if session fore channel is drained
 381  */
 382 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 383 {
 384         struct rpc_task *task;
 385
 386         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 387                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 388                 if (task)
 389                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 390                 return;
 391         }
 392
 393         if (ses->fc_slot_table.highest_used_slotid != -1)
 394                 return;
 395
 396         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 397         complete(&ses->fc_slot_table.complete);
 398 }
 399
 400 /*
 401  * Signal state manager thread if session back channel is drained
 402  */
 403 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 404 {
 405         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 406             ses->bc_slot_table.highest_used_slotid != -1)
 407                 return;
 408         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 409         complete(&ses->bc_slot_table.complete);
 410 }
 411
 412 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 413 {
 414         struct nfs4_slot_table *tbl;
 415
 416         tbl = &res->sr_session->fc_slot_table;
 417         if (!res->sr_slot) {
 418                 /* just wake up the next guy waiting since
 419                  * we may have not consumed a slot after all */
 420                 dprintk("%s: No slot\n", __func__);
 421                 return;
 422         }
 423
 424         spin_lock(&tbl->slot_tbl_lock);
 425         nfs4_free_slot(tbl, res->sr_slot);
 426         nfs4_check_drain_fc_complete(res->sr_session);
 427         spin_unlock(&tbl->slot_tbl_lock);
 428         res->sr_slot = NULL;
 429 }
 430
 431 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 432 {
 433         unsigned long timestamp;
 434         struct nfs_client *clp;
 435
 436         /*
 437          * sr_status remains 1 if an RPC level error occurred. The server
 438          * may or may not have processed the sequence operation..
 439          * Proceed as if the server received and processed the sequence
 440          * operation.
 441          */
 442         if (res->sr_status == 1)
 443                 res->sr_status = NFS_OK;
 444
 445         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
 446         if (!res->sr_slot)
 447                 goto out;
 448
 449         /* Check the SEQUENCE operation status */
 450         switch (res->sr_status) {
 451         case 0:
 452                 /* Update the slot's sequence and clientid lease timer */
 453                 ++res->sr_slot->seq_nr;
 454                 timestamp = res->sr_renewal_time;
 455                 clp = res->sr_session->clp;
 456                 do_renew_lease(clp, timestamp);
 457                 /* Check sequence flags */
 458                 if (res->sr_status_flags != 0)
 459                         nfs4_schedule_lease_recovery(clp);
 460                 break;
 461         case -NFS4ERR_DELAY:
 462                 /* The server detected a resend of the RPC call and
 463                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 464                  * of RFC5661.
 465                  */
 466                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 467                         __func__,
 468                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 469                         res->sr_slot->seq_nr);
 470                 goto out_retry;
 471         default:
 472                 /* Just update the slot sequence no. */
 473                 ++res->sr_slot->seq_nr;
 474         }
 475 out:
 476         /* The session may be reset by one of the error handlers. */
 477         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 478         nfs41_sequence_free_slot(res);
 479         return 1;
 480 out_retry:
 481         if (!rpc_restart_call(task))
 482                 goto out;
 483         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 484         return 0;
 485 }
 486
 487 static int nfs4_sequence_done(struct rpc_task *task,
 488                                struct nfs4_sequence_res *res)
 489 {
 490         if (res->sr_session == NULL)
 491                 return 1;
 492         return nfs41_sequence_done(task, res);
 493 }
 494
 495 /*
 496  * nfs4_find_slot - efficiently look for a free slot
 497  *
 498  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 499  * If found, we mark the slot as used, update the highest_used_slotid,
 500  * and respectively set up the sequence operation args.
 501  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 502  *
 503  * Note: must be called with under the slot_tbl_lock.
 504  */
 505 static u8
 506 nfs4_find_slot(struct nfs4_slot_table *tbl)
 507 {
 508         int slotid;
 509         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 510         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 511
 512         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 513                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 514                 tbl->max_slots);
 515         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 516         if (slotid >= tbl->max_slots)
 517                 goto out;
 518         __set_bit(slotid, tbl->used_slots);
 519         if (slotid > tbl->highest_used_slotid)
 520                 tbl->highest_used_slotid = slotid;
 521         ret_id = slotid;
 522 out:
 523         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 524                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 525         return ret_id;
 526 }
 527
 528 int nfs41_setup_sequence(struct nfs4_session *session,
 529                                 struct nfs4_sequence_args *args,
 530                                 struct nfs4_sequence_res *res,
 531                                 int cache_reply,
 532                                 struct rpc_task *task)
 533 {
 534         struct nfs4_slot *slot;
 535         struct nfs4_slot_table *tbl;
 536         u8 slotid;
 537
 538         dprintk("--> %s\n", __func__);
 539         /* slot already allocated? */
 540         if (res->sr_slot != NULL)
 541                 return 0;
 542
 543         tbl = &session->fc_slot_table;
 544
 545         spin_lock(&tbl->slot_tbl_lock);
 546         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 547             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 548                 /*
 549                  * The state manager will wait until the slot table is empty.
 550                  * Schedule the reset thread
 551                  */
 552                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 553                 spin_unlock(&tbl->slot_tbl_lock);
 554                 dprintk("%s Schedule Session Reset\n", __func__);
 555                 return -EAGAIN;
 556         }
 557
 558         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 559             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 560                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 561                 spin_unlock(&tbl->slot_tbl_lock);
 562                 dprintk("%s enforce FIFO order\n", __func__);
 563                 return -EAGAIN;
 564         }
 565
 566         slotid = nfs4_find_slot(tbl);
 567         if (slotid == NFS4_MAX_SLOT_TABLE) {
 568                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 569                 spin_unlock(&tbl->slot_tbl_lock);
 570                 dprintk("<-- %s: no free slots\n", __func__);
 571                 return -EAGAIN;
 572         }
 573         spin_unlock(&tbl->slot_tbl_lock);
 574
 575         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 576         slot = tbl->slots + slotid;
 577         args->sa_session = session;
 578         args->sa_slotid = slotid;
 579         args->sa_cache_this = cache_reply;
 580
 581         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 582
 583         res->sr_session = session;
 584         res->sr_slot = slot;
 585         res->sr_renewal_time = jiffies;
 586         res->sr_status_flags = 0;
 587         /*
 588          * sr_status is only set in decode_sequence, and so will remain
 589          * set to 1 if an rpc level failure occurs.
 590          */
 591         res->sr_status = 1;
 592         return 0;
 593 }
 594 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 595
 596 int nfs4_setup_sequence(const struct nfs_server *server,
 597                         struct nfs4_sequence_args *args,
 598                         struct nfs4_sequence_res *res,
 599                         int cache_reply,
 600                         struct rpc_task *task)
 601 {
 602         struct nfs4_session *session = nfs4_get_session(server);
 603         int ret = 0;
 604
 605         if (session == NULL) {
 606                 args->sa_session = NULL;
 607                 res->sr_session = NULL;
 608                 goto out;
 609         }
 610
 611         dprintk("--> %s clp %p session %p sr_slot %td\n",
 612                 __func__, session->clp, session, res->sr_slot ?
 613                         res->sr_slot - session->fc_slot_table.slots : -1);
 614
 615         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 616                                    task);
 617 out:
 618         dprintk("<-- %s status=%d\n", __func__, ret);
 619         return ret;
 620 }
 621
 622 struct nfs41_call_sync_data {
 623         const struct nfs_server *seq_server;
 624         struct nfs4_sequence_args *seq_args;
 625         struct nfs4_sequence_res *seq_res;
 626         int cache_reply;
 627 };
 628
 629 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 630 {
 631         struct nfs41_call_sync_data *data = calldata;
 632
 633         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 634
 635         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 636                                 data->seq_res, data->cache_reply, task))
 637                 return;
 638         rpc_call_start(task);
 639 }
 640
 641 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 642 {
 643         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 644         nfs41_call_sync_prepare(task, calldata);
 645 }
 646
 647 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 648 {
 649         struct nfs41_call_sync_data *data = calldata;
 650
 651         nfs41_sequence_done(task, data->seq_res);
 652 }
 653
 654 struct rpc_call_ops nfs41_call_sync_ops = {
 655         .rpc_call_prepare = nfs41_call_sync_prepare,
 656         .rpc_call_done = nfs41_call_sync_done,
 657 };
 658
 659 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 660         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 661         .rpc_call_done = nfs41_call_sync_done,
 662 };
 663
 664 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 665                                    struct nfs_server *server,
 666                                    struct rpc_message *msg,
 667                                    struct nfs4_sequence_args *args,
 668                                    struct nfs4_sequence_res *res,
 669                                    int cache_reply,
 670                                    int privileged)
 671 {
 672         int ret;
 673         struct rpc_task *task;
 674         struct nfs41_call_sync_data data = {
 675                 .seq_server = server,
 676                 .seq_args = args,
 677                 .seq_res = res,
 678                 .cache_reply = cache_reply,
 679         };
 680         struct rpc_task_setup task_setup = {
 681                 .rpc_client = clnt,
 682                 .rpc_message = msg,
 683                 .callback_ops = &nfs41_call_sync_ops,
 684                 .callback_data = &data
 685         };
 686
 687         res->sr_slot = NULL;
 688         if (privileged)
 689                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 690         task = rpc_run_task(&task_setup);
 691         if (IS_ERR(task))
 692                 ret = PTR_ERR(task);
 693         else {
 694                 ret = task->tk_status;
 695                 rpc_put_task(task);
 696         }
 697         return ret;
 698 }
 699
 700 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 701                             struct nfs_server *server,
 702                             struct rpc_message *msg,
 703                             struct nfs4_sequence_args *args,
 704                             struct nfs4_sequence_res *res,
 705                             int cache_reply)
 706 {
 707         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 708 }
 709
 710 #else
 711 static int nfs4_sequence_done(struct rpc_task *task,
 712                                struct nfs4_sequence_res *res)
 713 {
 714         return 1;
 715 }
 716 #endif /* CONFIG_NFS_V4_1 */
 717
 718 int _nfs4_call_sync(struct rpc_clnt *clnt,
 719                     struct nfs_server *server,
 720                     struct rpc_message *msg,
 721                     struct nfs4_sequence_args *args,
 722                     struct nfs4_sequence_res *res,
 723                     int cache_reply)
 724 {
 725         args->sa_session = res->sr_session = NULL;
 726         return rpc_call_sync(clnt, msg, 0);
 727 }
 728
 729 static inline
 730 int nfs4_call_sync(struct rpc_clnt *clnt,
 731                    struct nfs_server *server,
 732                    struct rpc_message *msg,
 733                    struct nfs4_sequence_args *args,
 734                    struct nfs4_sequence_res *res,
 735                    int cache_reply)
 736 {
 737         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 738                                                 args, res, cache_reply);
 739 }
 740
 741 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 742 {
 743         struct nfs_inode *nfsi = NFS_I(dir);
 744
 745         spin_lock(&dir->i_lock);
 746         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 747         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 748                 nfs_force_lookup_revalidate(dir);
 749         nfsi->change_attr = cinfo->after;
 750         spin_unlock(&dir->i_lock);
 751 }
 752
 753 struct nfs4_opendata {
 754         struct kref kref;
 755         struct nfs_openargs o_arg;
 756         struct nfs_openres o_res;
 757         struct nfs_open_confirmargs c_arg;
 758         struct nfs_open_confirmres c_res;
 759         struct nfs_fattr f_attr;
 760         struct nfs_fattr dir_attr;
 761         struct path path;
 762         struct dentry *dir;
 763         struct nfs4_state_owner *owner;
 764         struct nfs4_state *state;
 765         struct iattr attrs;
 766         unsigned long timestamp;
 767         unsigned int rpc_done : 1;
 768         int rpc_status;
 769         int cancelled;
 770 };
 771
 772
 773 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 774 {
 775         p->o_res.f_attr = &p->f_attr;
 776         p->o_res.dir_attr = &p->dir_attr;
 777         p->o_res.seqid = p->o_arg.seqid;
 778         p->c_res.seqid = p->c_arg.seqid;
 779         p->o_res.server = p->o_arg.server;
 780         nfs_fattr_init(&p->f_attr);
 781         nfs_fattr_init(&p->dir_attr);
 782 }
 783
 784 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 785                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 786                 const struct iattr *attrs,
 787                 gfp_t gfp_mask)
 788 {
 789         struct dentry *parent = dget_parent(path->dentry);
 790         struct inode *dir = parent->d_inode;
 791         struct nfs_server *server = NFS_SERVER(dir);
 792         struct nfs4_opendata *p;
 793
 794         p = kzalloc(sizeof(*p), gfp_mask);
 795         if (p == NULL)
 796                 goto err;
 797         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 798         if (p->o_arg.seqid == NULL)
 799                 goto err_free;
 800         path_get(path);
 801         p->path = *path;
 802         p->dir = parent;
 803         p->owner = sp;
 804         atomic_inc(&sp->so_count);
 805         p->o_arg.fh = NFS_FH(dir);
 806         p->o_arg.open_flags = flags;
 807         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 808         p->o_arg.clientid = server->nfs_client->cl_clientid;
 809         p->o_arg.id = sp->so_owner_id.id;
 810         p->o_arg.name = &p->path.dentry->d_name;
 811         p->o_arg.server = server;
 812         p->o_arg.bitmask = server->attr_bitmask;
 813         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 814         if (flags & O_CREAT) {
 815                 u32 *s;
 816
 817                 p->o_arg.u.attrs = &p->attrs;
 818                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 819                 s = (u32 *) p->o_arg.u.verifier.data;
 820                 s[0] = jiffies;
 821                 s[1] = current->pid;
 822         }
 823         p->c_arg.fh = &p->o_res.fh;
 824         p->c_arg.stateid = &p->o_res.stateid;
 825         p->c_arg.seqid = p->o_arg.seqid;
 826         nfs4_init_opendata_res(p);
 827         kref_init(&p->kref);
 828         return p;
 829 err_free:
 830         kfree(p);
 831 err:
 832         dput(parent);
 833         return NULL;
 834 }
 835
 836 static void nfs4_opendata_free(struct kref *kref)
 837 {
 838         struct nfs4_opendata *p = container_of(kref,
 839                         struct nfs4_opendata, kref);
 840
 841         nfs_free_seqid(p->o_arg.seqid);
 842         if (p->state != NULL)
 843                 nfs4_put_open_state(p->state);
 844         nfs4_put_state_owner(p->owner);
 845         dput(p->dir);
 846         path_put(&p->path);
 847         kfree(p);
 848 }
 849
 850 static void nfs4_opendata_put(struct nfs4_opendata *p)
 851 {
 852         if (p != NULL)
 853                 kref_put(&p->kref, nfs4_opendata_free);
 854 }
 855
 856 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 857 {
 858         int ret;
 859
 860         ret = rpc_wait_for_completion_task(task);
 861         return ret;
 862 }
 863
 864 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 865 {
 866         int ret = 0;
 867
 868         if (open_mode & O_EXCL)
 869                 goto out;
 870         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 871                 case FMODE_READ:
 872                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 873                                 && state->n_rdonly != 0;
 874                         break;
 875                 case FMODE_WRITE:
 876                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 877                                 && state->n_wronly != 0;
 878                         break;
 879                 case FMODE_READ|FMODE_WRITE:
 880                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 881                                 && state->n_rdwr != 0;
 882         }
 883 out:
 884         return ret;
 885 }
 886
 887 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 888 {
 889         if ((delegation->type & fmode) != fmode)
 890                 return 0;
 891         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 892                 return 0;
 893         nfs_mark_delegation_referenced(delegation);
 894         return 1;
 895 }
 896
 897 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 898 {
 899         switch (fmode) {
 900                 case FMODE_WRITE:
 901                         state->n_wronly++;
 902                         break;
 903                 case FMODE_READ:
 904                         state->n_rdonly++;
 905                         break;
 906                 case FMODE_READ|FMODE_WRITE:
 907                         state->n_rdwr++;
 908         }
 909         nfs4_state_set_mode_locked(state, state->state | fmode);
 910 }
 911
 912 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 913 {
 914         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 915                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 916         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 917         switch (fmode) {
 918                 case FMODE_READ:
 919                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 920                         break;
 921                 case FMODE_WRITE:
 922                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 923                         break;
 924                 case FMODE_READ|FMODE_WRITE:
 925                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 926         }
 927 }
 928
 929 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 930 {
 931         write_seqlock(&state->seqlock);
 932         nfs_set_open_stateid_locked(state, stateid, fmode);
 933         write_sequnlock(&state->seqlock);
 934 }
 935
 936 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 937 {
 938         /*
 939          * Protect the call to nfs4_state_set_mode_locked and
 940          * serialise the stateid update
 941          */
 942         write_seqlock(&state->seqlock);
 943         if (deleg_stateid != NULL) {
 944                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 945                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 946         }
 947         if (open_stateid != NULL)
 948                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 949         write_sequnlock(&state->seqlock);
 950         spin_lock(&state->owner->so_lock);
 951         update_open_stateflags(state, fmode);
 952         spin_unlock(&state->owner->so_lock);
 953 }
 954
 955 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 956 {
 957         struct nfs_inode *nfsi = NFS_I(state->inode);
 958         struct nfs_delegation *deleg_cur;
 959         int ret = 0;
 960
 961         fmode &= (FMODE_READ|FMODE_WRITE);
 962
 963         rcu_read_lock();
 964         deleg_cur = rcu_dereference(nfsi->delegation);
 965         if (deleg_cur == NULL)
 966                 goto no_delegation;
 967
 968         spin_lock(&deleg_cur->lock);
 969         if (nfsi->delegation != deleg_cur ||
 970             (deleg_cur->type & fmode) != fmode)
 971                 goto no_delegation_unlock;
 972
 973         if (delegation == NULL)
 974                 delegation = &deleg_cur->stateid;
 975         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 976                 goto no_delegation_unlock;
 977
 978         nfs_mark_delegation_referenced(deleg_cur);
 979         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 980         ret = 1;
 981 no_delegation_unlock:
 982         spin_unlock(&deleg_cur->lock);
 983 no_delegation:
 984         rcu_read_unlock();
 985
 986         if (!ret && open_stateid != NULL) {
 987                 __update_open_stateid(state, open_stateid, NULL, fmode);
 988                 ret = 1;
 989         }
 990
 991         return ret;
 992 }
 993
 994
 995 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 996 {
 997         struct nfs_delegation *delegation;
 998
 999         rcu_read_lock();
1000         delegation = rcu_dereference(NFS_I(inode)->delegation);
1001         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1002                 rcu_read_unlock();
1003                 return;
1004         }
1005         rcu_read_unlock();
1006         nfs_inode_return_delegation(inode);
1007 }
1008
1009 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1010 {
1011         struct nfs4_state *state = opendata->state;
1012         struct nfs_inode *nfsi = NFS_I(state->inode);
1013         struct nfs_delegation *delegation;
1014         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1015         fmode_t fmode = opendata->o_arg.fmode;
1016         nfs4_stateid stateid;
1017         int ret = -EAGAIN;
1018
1019         for (;;) {
1020                 if (can_open_cached(state, fmode, open_mode)) {
1021                         spin_lock(&state->owner->so_lock);
1022                         if (can_open_cached(state, fmode, open_mode)) {
1023                                 update_open_stateflags(state, fmode);
1024                                 spin_unlock(&state->owner->so_lock);
1025                                 goto out_return_state;
1026                         }
1027                         spin_unlock(&state->owner->so_lock);
1028                 }
1029                 rcu_read_lock();
1030                 delegation = rcu_dereference(nfsi->delegation);
1031                 if (delegation == NULL ||
1032                     !can_open_delegated(delegation, fmode)) {
1033                         rcu_read_unlock();
1034                         break;
1035                 }
1036                 /* Save the delegation */
1037                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1038                 rcu_read_unlock();
1039                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1040                 if (ret != 0)
1041                         goto out;
1042                 ret = -EAGAIN;
1043
1044                 /* Try to update the stateid using the delegation */
1045                 if (update_open_stateid(state, NULL, &stateid, fmode))
1046                         goto out_return_state;
1047         }
1048 out:
1049         return ERR_PTR(ret);
1050 out_return_state:
1051         atomic_inc(&state->count);
1052         return state;
1053 }
1054
1055 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1056 {
1057         struct inode *inode;
1058         struct nfs4_state *state = NULL;
1059         struct nfs_delegation *delegation;
1060         int ret;
1061
1062         if (!data->rpc_done) {
1063                 state = nfs4_try_open_cached(data);
1064                 goto out;
1065         }
1066
1067         ret = -EAGAIN;
1068         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1069                 goto err;
1070         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1071         ret = PTR_ERR(inode);
1072         if (IS_ERR(inode))
1073                 goto err;
1074         ret = -ENOMEM;
1075         state = nfs4_get_open_state(inode, data->owner);
1076         if (state == NULL)
1077                 goto err_put_inode;
1078         if (data->o_res.delegation_type != 0) {
1079                 int delegation_flags = 0;
1080
1081                 rcu_read_lock();
1082                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1083                 if (delegation)
1084                         delegation_flags = delegation->flags;
1085                 rcu_read_unlock();
1086                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1087                         nfs_inode_set_delegation(state->inode,
1088                                         data->owner->so_cred,
1089                                         &data->o_res);
1090                 else
1091                         nfs_inode_reclaim_delegation(state->inode,
1092                                         data->owner->so_cred,
1093                                         &data->o_res);
1094         }
1095
1096         update_open_stateid(state, &data->o_res.stateid, NULL,
1097                         data->o_arg.fmode);
1098         iput(inode);
1099 out:
1100         return state;
1101 err_put_inode:
1102         iput(inode);
1103 err:
1104         return ERR_PTR(ret);
1105 }
1106
1107 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1108 {
1109         struct nfs_inode *nfsi = NFS_I(state->inode);
1110         struct nfs_open_context *ctx;
1111
1112         spin_lock(&state->inode->i_lock);
1113         list_for_each_entry(ctx, &nfsi->open_files, list) {
1114                 if (ctx->state != state)
1115                         continue;
1116                 get_nfs_open_context(ctx);
1117                 spin_unlock(&state->inode->i_lock);
1118                 return ctx;
1119         }
1120         spin_unlock(&state->inode->i_lock);
1121         return ERR_PTR(-ENOENT);
1122 }
1123
1124 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1125 {
1126         struct nfs4_opendata *opendata;
1127
1128         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL, GFP_NOFS);
1129         if (opendata == NULL)
1130                 return ERR_PTR(-ENOMEM);
1131         opendata->state = state;
1132         atomic_inc(&state->count);
1133         return opendata;
1134 }
1135
1136 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1137 {
1138         struct nfs4_state *newstate;
1139         int ret;
1140
1141         opendata->o_arg.open_flags = 0;
1142         opendata->o_arg.fmode = fmode;
1143         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1144         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1145         nfs4_init_opendata_res(opendata);
1146         ret = _nfs4_recover_proc_open(opendata);
1147         if (ret != 0)
1148                 return ret;
1149         newstate = nfs4_opendata_to_nfs4_state(opendata);
1150         if (IS_ERR(newstate))
1151                 return PTR_ERR(newstate);
1152         nfs4_close_state(&opendata->path, newstate, fmode);
1153         *res = newstate;
1154         return 0;
1155 }
1156
1157 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1158 {
1159         struct nfs4_state *newstate;
1160         int ret;
1161
1162         /* memory barrier prior to reading state->n_* */
1163         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1164         smp_rmb();
1165         if (state->n_rdwr != 0) {
1166                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1167                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1168                 if (ret != 0)
1169                         return ret;
1170                 if (newstate != state)
1171                         return -ESTALE;
1172         }
1173         if (state->n_wronly != 0) {
1174                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1175                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1176                 if (ret != 0)
1177                         return ret;
1178                 if (newstate != state)
1179                         return -ESTALE;
1180         }
1181         if (state->n_rdonly != 0) {
1182                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1183                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1184                 if (ret != 0)
1185                         return ret;
1186                 if (newstate != state)
1187                         return -ESTALE;
1188         }
1189         /*
1190          * We may have performed cached opens for all three recoveries.
1191          * Check if we need to update the current stateid.
1192          */
1193         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1194             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1195                 write_seqlock(&state->seqlock);
1196                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1197                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1198                 write_sequnlock(&state->seqlock);
1199         }
1200         return 0;
1201 }
1202
1203 /*
1204  * OPEN_RECLAIM:
1205  *      reclaim state on the server after a reboot.
1206  */
1207 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1208 {
1209         struct nfs_delegation *delegation;
1210         struct nfs4_opendata *opendata;
1211         fmode_t delegation_type = 0;
1212         int status;
1213
1214         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1215         if (IS_ERR(opendata))
1216                 return PTR_ERR(opendata);
1217         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1218         opendata->o_arg.fh = NFS_FH(state->inode);
1219         rcu_read_lock();
1220         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1221         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1222                 delegation_type = delegation->type;
1223         rcu_read_unlock();
1224         opendata->o_arg.u.delegation_type = delegation_type;
1225         status = nfs4_open_recover(opendata, state);
1226         nfs4_opendata_put(opendata);
1227         return status;
1228 }
1229
1230 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1231 {
1232         struct nfs_server *server = NFS_SERVER(state->inode);
1233         struct nfs4_exception exception = { };
1234         int err;
1235         do {
1236                 err = _nfs4_do_open_reclaim(ctx, state);
1237                 if (err != -NFS4ERR_DELAY)
1238                         break;
1239                 nfs4_handle_exception(server, err, &exception);
1240         } while (exception.retry);
1241         return err;
1242 }
1243
1244 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1245 {
1246         struct nfs_open_context *ctx;
1247         int ret;
1248
1249         ctx = nfs4_state_find_open_context(state);
1250         if (IS_ERR(ctx))
1251                 return PTR_ERR(ctx);
1252         ret = nfs4_do_open_reclaim(ctx, state);
1253         put_nfs_open_context(ctx);
1254         return ret;
1255 }
1256
1257 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1258 {
1259         struct nfs4_opendata *opendata;
1260         int ret;
1261
1262         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1263         if (IS_ERR(opendata))
1264                 return PTR_ERR(opendata);
1265         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1266         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1267                         sizeof(opendata->o_arg.u.delegation.data));
1268         ret = nfs4_open_recover(opendata, state);
1269         nfs4_opendata_put(opendata);
1270         return ret;
1271 }
1272
1273 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1274 {
1275         struct nfs4_exception exception = { };
1276         struct nfs_server *server = NFS_SERVER(state->inode);
1277         int err;
1278         do {
1279                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1280                 switch (err) {
1281                         case 0:
1282                         case -ENOENT:
1283                         case -ESTALE:
1284                                 goto out;
1285                         case -NFS4ERR_BADSESSION:
1286                         case -NFS4ERR_BADSLOT:
1287                         case -NFS4ERR_BAD_HIGH_SLOT:
1288                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1289                         case -NFS4ERR_DEADSESSION:
1290                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1291                                 goto out;
1292                         case -NFS4ERR_STALE_CLIENTID:
1293                         case -NFS4ERR_STALE_STATEID:
1294                         case -NFS4ERR_EXPIRED:
1295                                 /* Don't recall a delegation if it was lost */
1296                                 nfs4_schedule_lease_recovery(server->nfs_client);
1297                                 goto out;
1298                         case -ERESTARTSYS:
1299                                 /*
1300                                  * The show must go on: exit, but mark the
1301                                  * stateid as needing recovery.
1302                                  */
1303                         case -NFS4ERR_ADMIN_REVOKED:
1304                         case -NFS4ERR_BAD_STATEID:
1305                                 nfs4_schedule_stateid_recovery(server, state);
1306                         case -EKEYEXPIRED:
1307                                 /*
1308                                  * User RPCSEC_GSS context has expired.
1309                                  * We cannot recover this stateid now, so
1310                                  * skip it and allow recovery thread to
1311                                  * proceed.
1312                                  */
1313                         case -ENOMEM:
1314                                 err = 0;
1315                                 goto out;
1316                 }
1317                 err = nfs4_handle_exception(server, err, &exception);
1318         } while (exception.retry);
1319 out:
1320         return err;
1321 }
1322
1323 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1324 {
1325         struct nfs4_opendata *data = calldata;
1326
1327         data->rpc_status = task->tk_status;
1328         if (data->rpc_status == 0) {
1329                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1330                                 sizeof(data->o_res.stateid.data));
1331                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1332                 renew_lease(data->o_res.server, data->timestamp);
1333                 data->rpc_done = 1;
1334         }
1335 }
1336
1337 static void nfs4_open_confirm_release(void *calldata)
1338 {
1339         struct nfs4_opendata *data = calldata;
1340         struct nfs4_state *state = NULL;
1341
1342         /* If this request hasn't been cancelled, do nothing */
1343         if (data->cancelled == 0)
1344                 goto out_free;
1345         /* In case of error, no cleanup! */
1346         if (!data->rpc_done)
1347                 goto out_free;
1348         state = nfs4_opendata_to_nfs4_state(data);
1349         if (!IS_ERR(state))
1350                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1351 out_free:
1352         nfs4_opendata_put(data);
1353 }
1354
1355 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1356         .rpc_call_done = nfs4_open_confirm_done,
1357         .rpc_release = nfs4_open_confirm_release,
1358 };
1359
1360 /*
1361  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1362  */
1363 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1364 {
1365         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1366         struct rpc_task *task;
1367         struct  rpc_message msg = {
1368                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1369                 .rpc_argp = &data->c_arg,
1370                 .rpc_resp = &data->c_res,
1371                 .rpc_cred = data->owner->so_cred,
1372         };
1373         struct rpc_task_setup task_setup_data = {
1374                 .rpc_client = server->client,
1375                 .rpc_message = &msg,
1376                 .callback_ops = &nfs4_open_confirm_ops,
1377                 .callback_data = data,
1378                 .workqueue = nfsiod_workqueue,
1379                 .flags = RPC_TASK_ASYNC,
1380         };
1381         int status;
1382
1383         kref_get(&data->kref);
1384         data->rpc_done = 0;
1385         data->rpc_status = 0;
1386         data->timestamp = jiffies;
1387         task = rpc_run_task(&task_setup_data);
1388         if (IS_ERR(task))
1389                 return PTR_ERR(task);
1390         status = nfs4_wait_for_completion_rpc_task(task);
1391         if (status != 0) {
1392                 data->cancelled = 1;
1393                 smp_wmb();
1394         } else
1395                 status = data->rpc_status;
1396         rpc_put_task(task);
1397         return status;
1398 }
1399
1400 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1401 {
1402         struct nfs4_opendata *data = calldata;
1403         struct nfs4_state_owner *sp = data->owner;
1404
1405         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1406                 return;
1407         /*
1408          * Check if we still need to send an OPEN call, or if we can use
1409          * a delegation instead.
1410          */
1411         if (data->state != NULL) {
1412                 struct nfs_delegation *delegation;
1413
1414                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1415                         goto out_no_action;
1416                 rcu_read_lock();
1417                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1418                 if (delegation != NULL &&
1419                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1420                         rcu_read_unlock();
1421                         goto out_no_action;
1422                 }
1423                 rcu_read_unlock();
1424         }
1425         /* Update sequence id. */
1426         data->o_arg.id = sp->so_owner_id.id;
1427         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1428         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1429                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1430                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1431         }
1432         data->timestamp = jiffies;
1433         if (nfs4_setup_sequence(data->o_arg.server,
1434                                 &data->o_arg.seq_args,
1435                                 &data->o_res.seq_res, 1, task))
1436                 return;
1437         rpc_call_start(task);
1438         return;
1439 out_no_action:
1440         task->tk_action = NULL;
1441
1442 }
1443
1444 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1445 {
1446         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1447         nfs4_open_prepare(task, calldata);
1448 }
1449
1450 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1451 {
1452         struct nfs4_opendata *data = calldata;
1453
1454         data->rpc_status = task->tk_status;
1455
1456         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1457                 return;
1458
1459         if (task->tk_status == 0) {
1460                 switch (data->o_res.f_attr->mode & S_IFMT) {
1461                         case S_IFREG:
1462                                 break;
1463                         case S_IFLNK:
1464                                 data->rpc_status = -ELOOP;
1465                                 break;
1466                         case S_IFDIR:
1467                                 data->rpc_status = -EISDIR;
1468                                 break;
1469                         default:
1470                                 data->rpc_status = -ENOTDIR;
1471                 }
1472                 renew_lease(data->o_res.server, data->timestamp);
1473                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1474                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1475         }
1476         data->rpc_done = 1;
1477 }
1478
1479 static void nfs4_open_release(void *calldata)
1480 {
1481         struct nfs4_opendata *data = calldata;
1482         struct nfs4_state *state = NULL;
1483
1484         /* If this request hasn't been cancelled, do nothing */
1485         if (data->cancelled == 0)
1486                 goto out_free;
1487         /* In case of error, no cleanup! */
1488         if (data->rpc_status != 0 || !data->rpc_done)
1489                 goto out_free;
1490         /* In case we need an open_confirm, no cleanup! */
1491         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1492                 goto out_free;
1493         state = nfs4_opendata_to_nfs4_state(data);
1494         if (!IS_ERR(state))
1495                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1496 out_free:
1497         nfs4_opendata_put(data);
1498 }
1499
1500 static const struct rpc_call_ops nfs4_open_ops = {
1501         .rpc_call_prepare = nfs4_open_prepare,
1502         .rpc_call_done = nfs4_open_done,
1503         .rpc_release = nfs4_open_release,
1504 };
1505
1506 static const struct rpc_call_ops nfs4_recover_open_ops = {
1507         .rpc_call_prepare = nfs4_recover_open_prepare,
1508         .rpc_call_done = nfs4_open_done,
1509         .rpc_release = nfs4_open_release,
1510 };
1511
1512 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1513 {
1514         struct inode *dir = data->dir->d_inode;
1515         struct nfs_server *server = NFS_SERVER(dir);
1516         struct nfs_openargs *o_arg = &data->o_arg;
1517         struct nfs_openres *o_res = &data->o_res;
1518         struct rpc_task *task;
1519         struct rpc_message msg = {
1520                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1521                 .rpc_argp = o_arg,
1522                 .rpc_resp = o_res,
1523                 .rpc_cred = data->owner->so_cred,
1524         };
1525         struct rpc_task_setup task_setup_data = {
1526                 .rpc_client = server->client,
1527                 .rpc_message = &msg,
1528                 .callback_ops = &nfs4_open_ops,
1529                 .callback_data = data,
1530                 .workqueue = nfsiod_workqueue,
1531                 .flags = RPC_TASK_ASYNC,
1532         };
1533         int status;
1534
1535         kref_get(&data->kref);
1536         data->rpc_done = 0;
1537         data->rpc_status = 0;
1538         data->cancelled = 0;
1539         if (isrecover)
1540                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1541         task = rpc_run_task(&task_setup_data);
1542         if (IS_ERR(task))
1543                 return PTR_ERR(task);
1544         status = nfs4_wait_for_completion_rpc_task(task);
1545         if (status != 0) {
1546                 data->cancelled = 1;
1547                 smp_wmb();
1548         } else
1549                 status = data->rpc_status;
1550         rpc_put_task(task);
1551
1552         return status;
1553 }
1554
1555 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1556 {
1557         struct inode *dir = data->dir->d_inode;
1558         struct nfs_openres *o_res = &data->o_res;
1559         int status;
1560
1561         status = nfs4_run_open_task(data, 1);
1562         if (status != 0 || !data->rpc_done)
1563                 return status;
1564
1565         nfs_refresh_inode(dir, o_res->dir_attr);
1566
1567         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1568                 status = _nfs4_proc_open_confirm(data);
1569                 if (status != 0)
1570                         return status;
1571         }
1572
1573         return status;
1574 }
1575
1576 /*
1577  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1578  */
1579 static int _nfs4_proc_open(struct nfs4_opendata *data)
1580 {
1581         struct inode *dir = data->dir->d_inode;
1582         struct nfs_server *server = NFS_SERVER(dir);
1583         struct nfs_openargs *o_arg = &data->o_arg;
1584         struct nfs_openres *o_res = &data->o_res;
1585         int status;
1586
1587         status = nfs4_run_open_task(data, 0);
1588         if (status != 0 || !data->rpc_done)
1589                 return status;
1590
1591         if (o_arg->open_flags & O_CREAT) {
1592                 update_changeattr(dir, &o_res->cinfo);
1593                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1594         } else
1595                 nfs_refresh_inode(dir, o_res->dir_attr);
1596         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1597                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1598         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1599                 status = _nfs4_proc_open_confirm(data);
1600                 if (status != 0)
1601                         return status;
1602         }
1603         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1604                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1605         return 0;
1606 }
1607
1608 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1609 {
1610         unsigned int loop;
1611         int ret;
1612
1613         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1614                 ret = nfs4_wait_clnt_recover(clp);
1615                 if (ret != 0)
1616                         break;
1617                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1618                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1619                         break;
1620                 nfs4_schedule_state_manager(clp);
1621                 ret = -EIO;
1622         }
1623         return ret;
1624 }
1625
1626 static int nfs4_recover_expired_lease(struct nfs_server *server)
1627 {
1628         return nfs4_client_recover_expired_lease(server->nfs_client);
1629 }
1630
1631 /*
1632  * OPEN_EXPIRED:
1633  *      reclaim state on the server after a network partition.
1634  *      Assumes caller holds the appropriate lock
1635  */
1636 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1637 {
1638         struct nfs4_opendata *opendata;
1639         int ret;
1640
1641         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1642         if (IS_ERR(opendata))
1643                 return PTR_ERR(opendata);
1644         ret = nfs4_open_recover(opendata, state);
1645         if (ret == -ESTALE)
1646                 d_drop(ctx->path.dentry);
1647         nfs4_opendata_put(opendata);
1648         return ret;
1649 }
1650
1651 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1652 {
1653         struct nfs_server *server = NFS_SERVER(state->inode);
1654         struct nfs4_exception exception = { };
1655         int err;
1656
1657         do {
1658                 err = _nfs4_open_expired(ctx, state);
1659                 switch (err) {
1660                 default:
1661                         goto out;
1662                 case -NFS4ERR_GRACE:
1663                 case -NFS4ERR_DELAY:
1664                         nfs4_handle_exception(server, err, &exception);
1665                         err = 0;
1666                 }
1667         } while (exception.retry);
1668 out:
1669         return err;
1670 }
1671
1672 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1673 {
1674         struct nfs_open_context *ctx;
1675         int ret;
1676
1677         ctx = nfs4_state_find_open_context(state);
1678         if (IS_ERR(ctx))
1679                 return PTR_ERR(ctx);
1680         ret = nfs4_do_open_expired(ctx, state);
1681         put_nfs_open_context(ctx);
1682         return ret;
1683 }
1684
1685 /*
1686  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1687  * fields corresponding to attributes that were used to store the verifier.
1688  * Make sure we clobber those fields in the later setattr call
1689  */
1690 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1691 {
1692         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1693             !(sattr->ia_valid & ATTR_ATIME_SET))
1694                 sattr->ia_valid |= ATTR_ATIME;
1695
1696         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1697             !(sattr->ia_valid & ATTR_MTIME_SET))
1698                 sattr->ia_valid |= ATTR_MTIME;
1699 }
1700
1701 /*
1702  * Returns a referenced nfs4_state
1703  */
1704 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)
1705 {
1706         struct nfs4_state_owner  *sp;
1707         struct nfs4_state     *state = NULL;
1708         struct nfs_server       *server = NFS_SERVER(dir);
1709         struct nfs4_opendata *opendata;
1710         int status;
1711
1712         /* Protect against reboot recovery conflicts */
1713         status = -ENOMEM;
1714         if (!(sp = nfs4_get_state_owner(server, cred))) {
1715                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1716                 goto out_err;
1717         }
1718         status = nfs4_recover_expired_lease(server);
1719         if (status != 0)
1720                 goto err_put_state_owner;
1721         if (path->dentry->d_inode != NULL)
1722                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1723         status = -ENOMEM;
1724         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr, GFP_KERNEL);
1725         if (opendata == NULL)
1726                 goto err_put_state_owner;
1727
1728         if (path->dentry->d_inode != NULL)
1729                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1730
1731         status = _nfs4_proc_open(opendata);
1732         if (status != 0)
1733                 goto err_opendata_put;
1734
1735         state = nfs4_opendata_to_nfs4_state(opendata);
1736         status = PTR_ERR(state);
1737         if (IS_ERR(state))
1738                 goto err_opendata_put;
1739         if (server->caps & NFS_CAP_POSIX_LOCK)
1740                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1741
1742         if (opendata->o_arg.open_flags & O_EXCL) {
1743                 nfs4_exclusive_attrset(opendata, sattr);
1744
1745                 nfs_fattr_init(opendata->o_res.f_attr);
1746                 status = nfs4_do_setattr(state->inode, cred,
1747                                 opendata->o_res.f_attr, sattr,
1748                                 state);
1749                 if (status == 0)
1750                         nfs_setattr_update_inode(state->inode, sattr);
1751                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1752         }
1753         nfs4_opendata_put(opendata);
1754         nfs4_put_state_owner(sp);
1755         *res = state;
1756         return 0;
1757 err_opendata_put:
1758         nfs4_opendata_put(opendata);
1759 err_put_state_owner:
1760         nfs4_put_state_owner(sp);
1761 out_err:
1762         *res = NULL;
1763         return status;
1764 }
1765
1766
1767 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)
1768 {
1769         struct nfs4_exception exception = { };
1770         struct nfs4_state *res;
1771         int status;
1772
1773         do {
1774                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1775                 if (status == 0)
1776                         break;
1777                 /* NOTE: BAD_SEQID means the server and client disagree about the
1778                  * book-keeping w.r.t. state-changing operations
1779                  * (OPEN/CLOSE/LOCK/LOCKU...)
1780                  * It is actually a sign of a bug on the client or on the server.
1781                  *
1782                  * If we receive a BAD_SEQID error in the particular case of
1783                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1784                  * have unhashed the old state_owner for us, and that we can
1785                  * therefore safely retry using a new one. We should still warn
1786                  * the user though...
1787                  */
1788                 if (status == -NFS4ERR_BAD_SEQID) {
1789                         printk(KERN_WARNING "NFS: v4 server %s "
1790                                         " returned a bad sequence-id error!\n",
1791                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1792                         exception.retry = 1;
1793                         continue;
1794                 }
1795                 /*
1796                  * BAD_STATEID on OPEN means that the server cancelled our
1797                  * state before it received the OPEN_CONFIRM.
1798                  * Recover by retrying the request as per the discussion
1799                  * on Page 181 of RFC3530.
1800                  */
1801                 if (status == -NFS4ERR_BAD_STATEID) {
1802                         exception.retry = 1;
1803                         continue;
1804                 }
1805                 if (status == -EAGAIN) {
1806                         /* We must have found a delegation */
1807                         exception.retry = 1;
1808                         continue;
1809                 }
1810                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1811                                         status, &exception));
1812         } while (exception.retry);
1813         return res;
1814 }
1815
1816 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1817                             struct nfs_fattr *fattr, struct iattr *sattr,
1818                             struct nfs4_state *state)
1819 {
1820         struct nfs_server *server = NFS_SERVER(inode);
1821         struct nfs_setattrargs  arg = {
1822                 .fh             = NFS_FH(inode),
1823                 .iap            = sattr,
1824                 .server         = server,
1825                 .bitmask = server->attr_bitmask,
1826         };
1827         struct nfs_setattrres  res = {
1828                 .fattr          = fattr,
1829                 .server         = server,
1830         };
1831         struct rpc_message msg = {
1832                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1833                 .rpc_argp       = &arg,
1834                 .rpc_resp       = &res,
1835                 .rpc_cred       = cred,
1836         };
1837         unsigned long timestamp = jiffies;
1838         int status;
1839
1840         nfs_fattr_init(fattr);
1841
1842         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1843                 /* Use that stateid */
1844         } else if (state != NULL) {
1845                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1846         } else
1847                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1848
1849         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1850         if (status == 0 && state != NULL)
1851                 renew_lease(server, timestamp);
1852         return status;
1853 }
1854
1855 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1856                            struct nfs_fattr *fattr, struct iattr *sattr,
1857                            struct nfs4_state *state)
1858 {
1859         struct nfs_server *server = NFS_SERVER(inode);
1860         struct nfs4_exception exception = { };
1861         int err;
1862         do {
1863                 err = nfs4_handle_exception(server,
1864                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1865                                 &exception);
1866         } while (exception.retry);
1867         return err;
1868 }
1869
1870 struct nfs4_closedata {
1871         struct path path;
1872         struct inode *inode;
1873         struct nfs4_state *state;
1874         struct nfs_closeargs arg;
1875         struct nfs_closeres res;
1876         struct nfs_fattr fattr;
1877         unsigned long timestamp;
1878         bool roc;
1879         u32 roc_barrier;
1880 };
1881
1882 static void nfs4_free_closedata(void *data)
1883 {
1884         struct nfs4_closedata *calldata = data;
1885         struct nfs4_state_owner *sp = calldata->state->owner;
1886
1887         if (calldata->roc)
1888                 pnfs_roc_release(calldata->state->inode);
1889         nfs4_put_open_state(calldata->state);
1890         nfs_free_seqid(calldata->arg.seqid);
1891         nfs4_put_state_owner(sp);
1892         path_put(&calldata->path);
1893         kfree(calldata);
1894 }
1895
1896 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1897                 fmode_t fmode)
1898 {
1899         spin_lock(&state->owner->so_lock);
1900         if (!(fmode & FMODE_READ))
1901                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1902         if (!(fmode & FMODE_WRITE))
1903                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1904         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1905         spin_unlock(&state->owner->so_lock);
1906 }
1907
1908 static void nfs4_close_done(struct rpc_task *task, void *data)
1909 {
1910         struct nfs4_closedata *calldata = data;
1911         struct nfs4_state *state = calldata->state;
1912         struct nfs_server *server = NFS_SERVER(calldata->inode);
1913
1914         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1915                 return;
1916         /* hmm. we are done with the inode, and in the process of freeing
1917          * the state_owner. we keep this around to process errors
1918          */
1919         switch (task->tk_status) {
1920                 case 0:
1921                         if (calldata->roc)
1922                                 pnfs_roc_set_barrier(state->inode,
1923                                                      calldata->roc_barrier);
1924                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1925                         renew_lease(server, calldata->timestamp);
1926                         nfs4_close_clear_stateid_flags(state,
1927                                         calldata->arg.fmode);
1928                         break;
1929                 case -NFS4ERR_STALE_STATEID:
1930                 case -NFS4ERR_OLD_STATEID:
1931                 case -NFS4ERR_BAD_STATEID:
1932                 case -NFS4ERR_EXPIRED:
1933                         if (calldata->arg.fmode == 0)
1934                                 break;
1935                 default:
1936                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1937                                 rpc_restart_call_prepare(task);
1938         }
1939         nfs_release_seqid(calldata->arg.seqid);
1940         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1941 }
1942
1943 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1944 {
1945         struct nfs4_closedata *calldata = data;
1946         struct nfs4_state *state = calldata->state;
1947         int call_close = 0;
1948
1949         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1950                 return;
1951
1952         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1953         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1954         spin_lock(&state->owner->so_lock);
1955         /* Calculate the change in open mode */
1956         if (state->n_rdwr == 0) {
1957                 if (state->n_rdonly == 0) {
1958                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1959                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1960                         calldata->arg.fmode &= ~FMODE_READ;
1961                 }
1962                 if (state->n_wronly == 0) {
1963                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1964                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1965                         calldata->arg.fmode &= ~FMODE_WRITE;
1966                 }
1967         }
1968         spin_unlock(&state->owner->so_lock);
1969
1970         if (!call_close) {
1971                 /* Note: exit _without_ calling nfs4_close_done */
1972                 task->tk_action = NULL;
1973                 return;
1974         }
1975
1976         if (calldata->arg.fmode == 0) {
1977                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1978                 if (calldata->roc &&
1979                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
1980                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
1981                                      task, NULL);
1982                         return;
1983                 }
1984         }
1985
1986         nfs_fattr_init(calldata->res.fattr);
1987         calldata->timestamp = jiffies;
1988         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
1989                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1990                                 1, task))
1991                 return;
1992         rpc_call_start(task);
1993 }
1994
1995 static const struct rpc_call_ops nfs4_close_ops = {
1996         .rpc_call_prepare = nfs4_close_prepare,
1997         .rpc_call_done = nfs4_close_done,
1998         .rpc_release = nfs4_free_closedata,
1999 };
2000
2001 /*
2002  * It is possible for data to be read/written from a mem-mapped file
2003  * after the sys_close call (which hits the vfs layer as a flush).
2004  * This means that we can't safely call nfsv4 close on a file until
2005  * the inode is cleared. This in turn means that we are not good
2006  * NFSv4 citizens - we do not indicate to the server to update the file's
2007  * share state even when we are done with one of the three share
2008  * stateid's in the inode.
2009  *
2010  * NOTE: Caller must be holding the sp->so_owner semaphore!
2011  */
2012 int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2013 {
2014         struct nfs_server *server = NFS_SERVER(state->inode);
2015         struct nfs4_closedata *calldata;
2016         struct nfs4_state_owner *sp = state->owner;
2017         struct rpc_task *task;
2018         struct rpc_message msg = {
2019                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2020                 .rpc_cred = state->owner->so_cred,
2021         };
2022         struct rpc_task_setup task_setup_data = {
2023                 .rpc_client = server->client,
2024                 .rpc_message = &msg,
2025                 .callback_ops = &nfs4_close_ops,
2026                 .workqueue = nfsiod_workqueue,
2027                 .flags = RPC_TASK_ASYNC,
2028         };
2029         int status = -ENOMEM;
2030
2031         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2032         if (calldata == NULL)
2033                 goto out;
2034         calldata->inode = state->inode;
2035         calldata->state = state;
2036         calldata->arg.fh = NFS_FH(state->inode);
2037         calldata->arg.stateid = &state->open_stateid;
2038         /* Serialization for the sequence id */
2039         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2040         if (calldata->arg.seqid == NULL)
2041                 goto out_free_calldata;
2042         calldata->arg.fmode = 0;
2043         calldata->arg.bitmask = server->cache_consistency_bitmask;
2044         calldata->res.fattr = &calldata->fattr;
2045         calldata->res.seqid = calldata->arg.seqid;
2046         calldata->res.server = server;
2047         calldata->roc = roc;
2048         path_get(path);
2049         calldata->path = *path;
2050
2051         msg.rpc_argp = &calldata->arg;
2052         msg.rpc_resp = &calldata->res;
2053         task_setup_data.callback_data = calldata;
2054         task = rpc_run_task(&task_setup_data);
2055         if (IS_ERR(task))
2056                 return PTR_ERR(task);
2057         status = 0;
2058         if (wait)
2059                 status = rpc_wait_for_completion_task(task);
2060         rpc_put_task(task);
2061         return status;
2062 out_free_calldata:
2063         kfree(calldata);
2064 out:
2065         if (roc)
2066                 pnfs_roc_release(state->inode);
2067         nfs4_put_open_state(state);
2068         nfs4_put_state_owner(sp);
2069         return status;
2070 }
2071
2072 static struct inode *
2073 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2074 {
2075         struct nfs4_state *state;
2076
2077         /* Protect against concurrent sillydeletes */
2078         state = nfs4_do_open(dir, &ctx->path, ctx->mode, open_flags, attr, ctx->cred);
2079         if (IS_ERR(state))
2080                 return ERR_CAST(state);
2081         ctx->state = state;
2082         return igrab(state->inode);
2083 }
2084
2085 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2086 {
2087         if (ctx->state == NULL)
2088                 return;
2089         if (is_sync)
2090                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2091         else
2092                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2093 }
2094
2095 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2096 {
2097         struct nfs4_server_caps_arg args = {
2098                 .fhandle = fhandle,
2099         };
2100         struct nfs4_server_caps_res res = {};
2101         struct rpc_message msg = {
2102                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2103                 .rpc_argp = &args,
2104                 .rpc_resp = &res,
2105         };
2106         int status;
2107
2108         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2109         if (status == 0) {
2110                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2111                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2112                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2113                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2114                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2115                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2116                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2117                         server->caps |= NFS_CAP_ACLS;
2118                 if (res.has_links != 0)
2119                         server->caps |= NFS_CAP_HARDLINKS;
2120                 if (res.has_symlinks != 0)
2121                         server->caps |= NFS_CAP_SYMLINKS;
2122                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2123                         server->caps |= NFS_CAP_FILEID;
2124                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2125                         server->caps |= NFS_CAP_MODE;
2126                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2127                         server->caps |= NFS_CAP_NLINK;
2128                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2129                         server->caps |= NFS_CAP_OWNER;
2130                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2131                         server->caps |= NFS_CAP_OWNER_GROUP;
2132                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2133                         server->caps |= NFS_CAP_ATIME;
2134                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2135                         server->caps |= NFS_CAP_CTIME;
2136                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2137                         server->caps |= NFS_CAP_MTIME;
2138
2139                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2140                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2141                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2142                 server->acl_bitmask = res.acl_bitmask;
2143         }
2144
2145         return status;
2146 }
2147
2148 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2149 {
2150         struct nfs4_exception exception = { };
2151         int err;
2152         do {
2153                 err = nfs4_handle_exception(server,
2154                                 _nfs4_server_capabilities(server, fhandle),
2155                                 &exception);
2156         } while (exception.retry);
2157         return err;
2158 }
2159
2160 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2161                 struct nfs_fsinfo *info)
2162 {
2163         struct nfs4_lookup_root_arg args = {
2164                 .bitmask = nfs4_fattr_bitmap,
2165         };
2166         struct nfs4_lookup_res res = {
2167                 .server = server,
2168                 .fattr = info->fattr,
2169                 .fh = fhandle,
2170         };
2171         struct rpc_message msg = {
2172                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2173                 .rpc_argp = &args,
2174                 .rpc_resp = &res,
2175         };
2176
2177         nfs_fattr_init(info->fattr);
2178         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2179 }
2180
2181 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2182                 struct nfs_fsinfo *info)
2183 {
2184         struct nfs4_exception exception = { };
2185         int err;
2186         do {
2187                 err = nfs4_handle_exception(server,
2188                                 _nfs4_lookup_root(server, fhandle, info),
2189                                 &exception);
2190         } while (exception.retry);
2191         return err;
2192 }
2193
2194 /*
2195  * get the file handle for the "/" directory on the server
2196  */
2197 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2198                               struct nfs_fsinfo *info)
2199 {
2200         int status;
2201
2202         status = nfs4_lookup_root(server, fhandle, info);
2203         if (status == 0)
2204                 status = nfs4_server_capabilities(server, fhandle);
2205         if (status == 0)
2206                 status = nfs4_do_fsinfo(server, fhandle, info);
2207         return nfs4_map_errors(status);
2208 }
2209
2210 /*
2211  * Get locations and (maybe) other attributes of a referral.
2212  * Note that we'll actually follow the referral later when
2213  * we detect fsid mismatch in inode revalidation
2214  */
2215 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2216 {
2217         int status = -ENOMEM;
2218         struct page *page = NULL;
2219         struct nfs4_fs_locations *locations = NULL;
2220
2221         page = alloc_page(GFP_KERNEL);
2222         if (page == NULL)
2223                 goto out;
2224         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2225         if (locations == NULL)
2226                 goto out;
2227
2228         status = nfs4_proc_fs_locations(dir, name, locations, page);
2229         if (status != 0)
2230                 goto out;
2231         /* Make sure server returned a different fsid for the referral */
2232         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2233                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2234                 status = -EIO;
2235                 goto out;
2236         }
2237
2238         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2239         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2240         if (!fattr->mode)
2241                 fattr->mode = S_IFDIR;
2242         memset(fhandle, 0, sizeof(struct nfs_fh));
2243 out:
2244         if (page)
2245                 __free_page(page);
2246         kfree(locations);
2247         return status;
2248 }
2249
2250 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2251 {
2252         struct nfs4_getattr_arg args = {
2253                 .fh = fhandle,
2254                 .bitmask = server->attr_bitmask,
2255         };
2256         struct nfs4_getattr_res res = {
2257                 .fattr = fattr,
2258                 .server = server,
2259         };
2260         struct rpc_message msg = {
2261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2262                 .rpc_argp = &args,
2263                 .rpc_resp = &res,
2264         };
2265
2266         nfs_fattr_init(fattr);
2267         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2268 }
2269
2270 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2271 {
2272         struct nfs4_exception exception = { };
2273         int err;
2274         do {
2275                 err = nfs4_handle_exception(server,
2276                                 _nfs4_proc_getattr(server, fhandle, fattr),
2277                                 &exception);
2278         } while (exception.retry);
2279         return err;
2280 }
2281
2282 /*
2283  * The file is not closed if it is opened due to the a request to change
2284  * the size of the file. The open call will not be needed once the
2285  * VFS layer lookup-intents are implemented.
2286  *
2287  * Close is called when the inode is destroyed.
2288  * If we haven't opened the file for O_WRONLY, we
2289  * need to in the size_change case to obtain a stateid.
2290  *
2291  * Got race?
2292  * Because OPEN is always done by name in nfsv4, it is
2293  * possible that we opened a different file by the same
2294  * name.  We can recognize this race condition, but we
2295  * can't do anything about it besides returning an error.
2296  *
2297  * This will be fixed with VFS changes (lookup-intent).
2298  */
2299 static int
2300 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2301                   struct iattr *sattr)
2302 {
2303         struct inode *inode = dentry->d_inode;
2304         struct rpc_cred *cred = NULL;
2305         struct nfs4_state *state = NULL;
2306         int status;
2307
2308         nfs_fattr_init(fattr);
2309
2310         /* Search for an existing open(O_WRITE) file */
2311         if (sattr->ia_valid & ATTR_FILE) {
2312                 struct nfs_open_context *ctx;
2313
2314                 ctx = nfs_file_open_context(sattr->ia_file);
2315                 if (ctx) {
2316                         cred = ctx->cred;
2317                         state = ctx->state;
2318                 }
2319         }
2320
2321         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2322         if (status == 0)
2323                 nfs_setattr_update_inode(inode, sattr);
2324         return status;
2325 }
2326
2327 static int _nfs4_proc_lookupfh(struct rpc_clnt *clnt, struct nfs_server *server,
2328                 const struct nfs_fh *dirfh, const struct qstr *name,
2329                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2330 {
2331         int                    status;
2332         struct nfs4_lookup_arg args = {
2333                 .bitmask = server->attr_bitmask,
2334                 .dir_fh = dirfh,
2335                 .name = name,
2336         };
2337         struct nfs4_lookup_res res = {
2338                 .server = server,
2339                 .fattr = fattr,
2340                 .fh = fhandle,
2341         };
2342         struct rpc_message msg = {
2343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2344                 .rpc_argp = &args,
2345                 .rpc_resp = &res,
2346         };
2347
2348         nfs_fattr_init(fattr);
2349
2350         dprintk("NFS call  lookupfh %s\n", name->name);
2351         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2352         dprintk("NFS reply lookupfh: %d\n", status);
2353         return status;
2354 }
2355
2356 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2357                               struct qstr *name, struct nfs_fh *fhandle,
2358                               struct nfs_fattr *fattr)
2359 {
2360         struct nfs4_exception exception = { };
2361         int err;
2362         do {
2363                 err = _nfs4_proc_lookupfh(server->client, server, dirfh, name, fhandle, fattr);
2364                 /* FIXME: !!!! */
2365                 if (err == -NFS4ERR_MOVED) {
2366                         err = -EREMOTE;
2367                         break;
2368                 }
2369                 err = nfs4_handle_exception(server, err, &exception);
2370         } while (exception.retry);
2371         return err;
2372 }
2373
2374 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2375                 const struct qstr *name, struct nfs_fh *fhandle,
2376                 struct nfs_fattr *fattr)
2377 {
2378         int status;
2379
2380         dprintk("NFS call  lookup %s\n", name->name);
2381         status = _nfs4_proc_lookupfh(clnt, NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2382         if (status == -NFS4ERR_MOVED)
2383                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2384         dprintk("NFS reply lookup: %d\n", status);
2385         return status;
2386 }
2387
2388 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2389 {
2390         memset(fh, 0, sizeof(struct nfs_fh));
2391         fattr->fsid.major = 1;
2392         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2393                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2394         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2395         fattr->nlink = 2;
2396 }
2397
2398 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2399                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2400 {
2401         struct nfs4_exception exception = { };
2402         int err;
2403         do {
2404                 err = nfs4_handle_exception(NFS_SERVER(dir),
2405                                 _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr),
2406                                 &exception);
2407                 if (err == -EPERM)
2408                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2409         } while (exception.retry);
2410         return err;
2411 }
2412
2413 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2414 {
2415         struct nfs_server *server = NFS_SERVER(inode);
2416         struct nfs4_accessargs args = {
2417                 .fh = NFS_FH(inode),
2418                 .bitmask = server->attr_bitmask,
2419         };
2420         struct nfs4_accessres res = {
2421                 .server = server,
2422         };
2423         struct rpc_message msg = {
2424                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2425                 .rpc_argp = &args,
2426                 .rpc_resp = &res,
2427                 .rpc_cred = entry->cred,
2428         };
2429         int mode = entry->mask;
2430         int status;
2431
2432         /*
2433          * Determine which access bits we want to ask for...
2434          */
2435         if (mode & MAY_READ)
2436                 args.access |= NFS4_ACCESS_READ;
2437         if (S_ISDIR(inode->i_mode)) {
2438                 if (mode & MAY_WRITE)
2439                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2440                 if (mode & MAY_EXEC)
2441                         args.access |= NFS4_ACCESS_LOOKUP;
2442         } else {
2443                 if (mode & MAY_WRITE)
2444                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2445                 if (mode & MAY_EXEC)
2446                         args.access |= NFS4_ACCESS_EXECUTE;
2447         }
2448
2449         res.fattr = nfs_alloc_fattr();
2450         if (res.fattr == NULL)
2451                 return -ENOMEM;
2452
2453         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2454         if (!status) {
2455                 entry->mask = 0;
2456                 if (res.access & NFS4_ACCESS_READ)
2457                         entry->mask |= MAY_READ;
2458                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2459                         entry->mask |= MAY_WRITE;
2460                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2461                         entry->mask |= MAY_EXEC;
2462                 nfs_refresh_inode(inode, res.fattr);
2463         }
2464         nfs_free_fattr(res.fattr);
2465         return status;
2466 }
2467
2468 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2469 {
2470         struct nfs4_exception exception = { };
2471         int err;
2472         do {
2473                 err = nfs4_handle_exception(NFS_SERVER(inode),
2474                                 _nfs4_proc_access(inode, entry),
2475                                 &exception);
2476         } while (exception.retry);
2477         return err;
2478 }
2479
2480 /*
2481  * TODO: For the time being, we don't try to get any attributes
2482  * along with any of the zero-copy operations READ, READDIR,
2483  * READLINK, WRITE.
2484  *
2485  * In the case of the first three, we want to put the GETATTR
2486  * after the read-type operation -- this is because it is hard
2487  * to predict the length of a GETATTR response in v4, and thus
2488  * align the READ data correctly.  This means that the GETATTR
2489  * may end up partially falling into the page cache, and we should
2490  * shift it into the 'tail' of the xdr_buf before processing.
2491  * To do this efficiently, we need to know the total length
2492  * of data received, which doesn't seem to be available outside
2493  * of the RPC layer.
2494  *
2495  * In the case of WRITE, we also want to put the GETATTR after
2496  * the operation -- in this case because we want to make sure
2497  * we get the post-operation mtime and size.  This means that
2498  * we can't use xdr_encode_pages() as written: we need a variant
2499  * of it which would leave room in the 'tail' iovec.
2500  *
2501  * Both of these changes to the XDR layer would in fact be quite
2502  * minor, but I decided to leave them for a subsequent patch.
2503  */
2504 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2505                 unsigned int pgbase, unsigned int pglen)
2506 {
2507         struct nfs4_readlink args = {
2508                 .fh       = NFS_FH(inode),
2509                 .pgbase   = pgbase,
2510                 .pglen    = pglen,
2511                 .pages    = &page,
2512         };
2513         struct nfs4_readlink_res res;
2514         struct rpc_message msg = {
2515                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2516                 .rpc_argp = &args,
2517                 .rpc_resp = &res,
2518         };
2519
2520         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2521 }
2522
2523 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2524                 unsigned int pgbase, unsigned int pglen)
2525 {
2526         struct nfs4_exception exception = { };
2527         int err;
2528         do {
2529                 err = nfs4_handle_exception(NFS_SERVER(inode),
2530                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2531                                 &exception);
2532         } while (exception.retry);
2533         return err;
2534 }
2535
2536 /*
2537  * Got race?
2538  * We will need to arrange for the VFS layer to provide an atomic open.
2539  * Until then, this create/open method is prone to inefficiency and race
2540  * conditions due to the lookup, create, and open VFS calls from sys_open()
2541  * placed on the wire.
2542  *
2543  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2544  * The file will be opened again in the subsequent VFS open call
2545  * (nfs4_proc_file_open).
2546  *
2547  * The open for read will just hang around to be used by any process that
2548  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2549  */
2550
2551 static int
2552 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2553                  int flags, struct nfs_open_context *ctx)
2554 {
2555         struct path my_path = {
2556                 .dentry = dentry,
2557         };
2558         struct path *path = &my_path;
2559         struct nfs4_state *state;
2560         struct rpc_cred *cred = NULL;
2561         fmode_t fmode = 0;
2562         int status = 0;
2563
2564         if (ctx != NULL) {
2565                 cred = ctx->cred;
2566                 path = &ctx->path;
2567                 fmode = ctx->mode;
2568         }
2569         sattr->ia_mode &= ~current_umask();
2570         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2571         d_drop(dentry);
2572         if (IS_ERR(state)) {
2573                 status = PTR_ERR(state);
2574                 goto out;
2575         }
2576         d_add(dentry, igrab(state->inode));
2577         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2578         if (ctx != NULL)
2579                 ctx->state = state;
2580         else
2581                 nfs4_close_sync(path, state, fmode);
2582 out:
2583         return status;
2584 }
2585
2586 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2587 {
2588         struct nfs_server *server = NFS_SERVER(dir);
2589         struct nfs_removeargs args = {
2590                 .fh = NFS_FH(dir),
2591                 .name.len = name->len,
2592                 .name.name = name->name,
2593                 .bitmask = server->attr_bitmask,
2594         };
2595         struct nfs_removeres res = {
2596                 .server = server,
2597         };
2598         struct rpc_message msg = {
2599                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2600                 .rpc_argp = &args,
2601                 .rpc_resp = &res,
2602         };
2603         int status = -ENOMEM;
2604
2605         res.dir_attr = nfs_alloc_fattr();
2606         if (res.dir_attr == NULL)
2607                 goto out;
2608
2609         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2610         if (status == 0) {
2611                 update_changeattr(dir, &res.cinfo);
2612                 nfs_post_op_update_inode(dir, res.dir_attr);
2613         }
2614         nfs_free_fattr(res.dir_attr);
2615 out:
2616         return status;
2617 }
2618
2619 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2620 {
2621         struct nfs4_exception exception = { };
2622         int err;
2623         do {
2624                 err = nfs4_handle_exception(NFS_SERVER(dir),
2625                                 _nfs4_proc_remove(dir, name),
2626                                 &exception);
2627         } while (exception.retry);
2628         return err;
2629 }
2630
2631 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2632 {
2633         struct nfs_server *server = NFS_SERVER(dir);
2634         struct nfs_removeargs *args = msg->rpc_argp;
2635         struct nfs_removeres *res = msg->rpc_resp;
2636
2637         args->bitmask = server->cache_consistency_bitmask;
2638         res->server = server;
2639         res->seq_res.sr_slot = NULL;
2640         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2641 }
2642
2643 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2644 {
2645         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2646
2647         if (!nfs4_sequence_done(task, &res->seq_res))
2648                 return 0;
2649         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2650                 return 0;
2651         update_changeattr(dir, &res->cinfo);
2652         nfs_post_op_update_inode(dir, res->dir_attr);
2653         return 1;
2654 }
2655
2656 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2657 {
2658         struct nfs_server *server = NFS_SERVER(dir);
2659         struct nfs_renameargs *arg = msg->rpc_argp;
2660         struct nfs_renameres *res = msg->rpc_resp;
2661
2662         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2663         arg->bitmask = server->attr_bitmask;
2664         res->server = server;
2665 }
2666
2667 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2668                                  struct inode *new_dir)
2669 {
2670         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2671
2672         if (!nfs4_sequence_done(task, &res->seq_res))
2673                 return 0;
2674         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2675                 return 0;
2676
2677         update_changeattr(old_dir, &res->old_cinfo);
2678         nfs_post_op_update_inode(old_dir, res->old_fattr);
2679         update_changeattr(new_dir, &res->new_cinfo);
2680         nfs_post_op_update_inode(new_dir, res->new_fattr);
2681         return 1;
2682 }
2683
2684 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2685                 struct inode *new_dir, struct qstr *new_name)
2686 {
2687         struct nfs_server *server = NFS_SERVER(old_dir);
2688         struct nfs_renameargs arg = {
2689                 .old_dir = NFS_FH(old_dir),
2690                 .new_dir = NFS_FH(new_dir),
2691                 .old_name = old_name,
2692                 .new_name = new_name,
2693                 .bitmask = server->attr_bitmask,
2694         };
2695         struct nfs_renameres res = {
2696                 .server = server,
2697         };
2698         struct rpc_message msg = {
2699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2700                 .rpc_argp = &arg,
2701                 .rpc_resp = &res,
2702         };
2703         int status = -ENOMEM;
2704
2705         res.old_fattr = nfs_alloc_fattr();
2706         res.new_fattr = nfs_alloc_fattr();
2707         if (res.old_fattr == NULL || res.new_fattr == NULL)
2708                 goto out;
2709
2710         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2711         if (!status) {
2712                 update_changeattr(old_dir, &res.old_cinfo);
2713                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2714                 update_changeattr(new_dir, &res.new_cinfo);
2715                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2716         }
2717 out:
2718         nfs_free_fattr(res.new_fattr);
2719         nfs_free_fattr(res.old_fattr);
2720         return status;
2721 }
2722
2723 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2724                 struct inode *new_dir, struct qstr *new_name)
2725 {
2726         struct nfs4_exception exception = { };
2727         int err;
2728         do {
2729                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2730                                 _nfs4_proc_rename(old_dir, old_name,
2731                                         new_dir, new_name),
2732                                 &exception);
2733         } while (exception.retry);
2734         return err;
2735 }
2736
2737 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2738 {
2739         struct nfs_server *server = NFS_SERVER(inode);
2740         struct nfs4_link_arg arg = {
2741                 .fh     = NFS_FH(inode),
2742                 .dir_fh = NFS_FH(dir),
2743                 .name   = name,
2744                 .bitmask = server->attr_bitmask,
2745         };
2746         struct nfs4_link_res res = {
2747                 .server = server,
2748         };
2749         struct rpc_message msg = {
2750                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2751                 .rpc_argp = &arg,
2752                 .rpc_resp = &res,
2753         };
2754         int status = -ENOMEM;
2755
2756         res.fattr = nfs_alloc_fattr();
2757         res.dir_attr = nfs_alloc_fattr();
2758         if (res.fattr == NULL || res.dir_attr == NULL)
2759                 goto out;
2760
2761         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2762         if (!status) {
2763                 update_changeattr(dir, &res.cinfo);
2764                 nfs_post_op_update_inode(dir, res.dir_attr);
2765                 nfs_post_op_update_inode(inode, res.fattr);
2766         }
2767 out:
2768         nfs_free_fattr(res.dir_attr);
2769         nfs_free_fattr(res.fattr);
2770         return status;
2771 }
2772
2773 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2774 {
2775         struct nfs4_exception exception = { };
2776         int err;
2777         do {
2778                 err = nfs4_handle_exception(NFS_SERVER(inode),
2779                                 _nfs4_proc_link(inode, dir, name),
2780                                 &exception);
2781         } while (exception.retry);
2782         return err;
2783 }
2784
2785 struct nfs4_createdata {
2786         struct rpc_message msg;
2787         struct nfs4_create_arg arg;
2788         struct nfs4_create_res res;
2789         struct nfs_fh fh;
2790         struct nfs_fattr fattr;
2791         struct nfs_fattr dir_fattr;
2792 };
2793
2794 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2795                 struct qstr *name, struct iattr *sattr, u32 ftype)
2796 {
2797         struct nfs4_createdata *data;
2798
2799         data = kzalloc(sizeof(*data), GFP_KERNEL);
2800         if (data != NULL) {
2801                 struct nfs_server *server = NFS_SERVER(dir);
2802
2803                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2804                 data->msg.rpc_argp = &data->arg;
2805                 data->msg.rpc_resp = &data->res;
2806                 data->arg.dir_fh = NFS_FH(dir);
2807                 data->arg.server = server;
2808                 data->arg.name = name;
2809                 data->arg.attrs = sattr;
2810                 data->arg.ftype = ftype;
2811                 data->arg.bitmask = server->attr_bitmask;
2812                 data->res.server = server;
2813                 data->res.fh = &data->fh;
2814                 data->res.fattr = &data->fattr;
2815                 data->res.dir_fattr = &data->dir_fattr;
2816                 nfs_fattr_init(data->res.fattr);
2817                 nfs_fattr_init(data->res.dir_fattr);
2818         }
2819         return data;
2820 }
2821
2822 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2823 {
2824         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2825                                     &data->arg.seq_args, &data->res.seq_res, 1);
2826         if (status == 0) {
2827                 update_changeattr(dir, &data->res.dir_cinfo);
2828                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2829                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2830         }
2831         return status;
2832 }
2833
2834 static void nfs4_free_createdata(struct nfs4_createdata *data)
2835 {
2836         kfree(data);
2837 }
2838
2839 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2840                 struct page *page, unsigned int len, struct iattr *sattr)
2841 {
2842         struct nfs4_createdata *data;
2843         int status = -ENAMETOOLONG;
2844
2845         if (len > NFS4_MAXPATHLEN)
2846                 goto out;
2847
2848         status = -ENOMEM;
2849         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2850         if (data == NULL)
2851                 goto out;
2852
2853         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2854         data->arg.u.symlink.pages = &page;
2855         data->arg.u.symlink.len = len;
2856
2857         status = nfs4_do_create(dir, dentry, data);
2858
2859         nfs4_free_createdata(data);
2860 out:
2861         return status;
2862 }
2863
2864 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2865                 struct page *page, unsigned int len, struct iattr *sattr)
2866 {
2867         struct nfs4_exception exception = { };
2868         int err;
2869         do {
2870                 err = nfs4_handle_exception(NFS_SERVER(dir),
2871                                 _nfs4_proc_symlink(dir, dentry, page,
2872                                                         len, sattr),
2873                                 &exception);
2874         } while (exception.retry);
2875         return err;
2876 }
2877
2878 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2879                 struct iattr *sattr)
2880 {
2881         struct nfs4_createdata *data;
2882         int status = -ENOMEM;
2883
2884         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2885         if (data == NULL)
2886                 goto out;
2887
2888         status = nfs4_do_create(dir, dentry, data);
2889
2890         nfs4_free_createdata(data);
2891 out:
2892         return status;
2893 }
2894
2895 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2896                 struct iattr *sattr)
2897 {
2898         struct nfs4_exception exception = { };
2899         int err;
2900
2901         sattr->ia_mode &= ~current_umask();
2902         do {
2903                 err = nfs4_handle_exception(NFS_SERVER(dir),
2904                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2905                                 &exception);
2906         } while (exception.retry);
2907         return err;
2908 }
2909
2910 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2911                 u64 cookie, struct page **pages, unsigned int count, int plus)
2912 {
2913         struct inode            *dir = dentry->d_inode;
2914         struct nfs4_readdir_arg args = {
2915                 .fh = NFS_FH(dir),
2916                 .pages = pages,
2917                 .pgbase = 0,
2918                 .count = count,
2919                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2920                 .plus = plus,
2921         };
2922         struct nfs4_readdir_res res;
2923         struct rpc_message msg = {
2924                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2925                 .rpc_argp = &args,
2926                 .rpc_resp = &res,
2927                 .rpc_cred = cred,
2928         };
2929         int                     status;
2930
2931         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2932                         dentry->d_parent->d_name.name,
2933                         dentry->d_name.name,
2934                         (unsigned long long)cookie);
2935         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2936         res.pgbase = args.pgbase;
2937         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
2938         if (status >= 0) {
2939                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2940                 status += args.pgbase;
2941         }
2942
2943         nfs_invalidate_atime(dir);
2944
2945         dprintk("%s: returns %d\n", __func__, status);
2946         return status;
2947 }
2948
2949 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2950                 u64 cookie, struct page **pages, unsigned int count, int plus)
2951 {
2952         struct nfs4_exception exception = { };
2953         int err;
2954         do {
2955                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2956                                 _nfs4_proc_readdir(dentry, cred, cookie,
2957                                         pages, count, plus),
2958                                 &exception);
2959         } while (exception.retry);
2960         return err;
2961 }
2962
2963 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2964                 struct iattr *sattr, dev_t rdev)
2965 {
2966         struct nfs4_createdata *data;
2967         int mode = sattr->ia_mode;
2968         int status = -ENOMEM;
2969
2970         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2971         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2972
2973         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2974         if (data == NULL)
2975                 goto out;
2976
2977         if (S_ISFIFO(mode))
2978                 data->arg.ftype = NF4FIFO;
2979         else if (S_ISBLK(mode)) {
2980                 data->arg.ftype = NF4BLK;
2981                 data->arg.u.device.specdata1 = MAJOR(rdev);
2982                 data->arg.u.device.specdata2 = MINOR(rdev);
2983         }
2984         else if (S_ISCHR(mode)) {
2985                 data->arg.ftype = NF4CHR;
2986                 data->arg.u.device.specdata1 = MAJOR(rdev);
2987                 data->arg.u.device.specdata2 = MINOR(rdev);
2988         }
2989
2990         status = nfs4_do_create(dir, dentry, data);
2991
2992         nfs4_free_createdata(data);
2993 out:
2994         return status;
2995 }
2996
2997 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2998                 struct iattr *sattr, dev_t rdev)
2999 {
3000         struct nfs4_exception exception = { };
3001         int err;
3002
3003         sattr->ia_mode &= ~current_umask();
3004         do {
3005                 err = nfs4_handle_exception(NFS_SERVER(dir),
3006                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3007                                 &exception);
3008         } while (exception.retry);
3009         return err;
3010 }
3011
3012 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3013                  struct nfs_fsstat *fsstat)
3014 {
3015         struct nfs4_statfs_arg args = {
3016                 .fh = fhandle,
3017                 .bitmask = server->attr_bitmask,
3018         };
3019         struct nfs4_statfs_res res = {
3020                 .fsstat = fsstat,
3021         };
3022         struct rpc_message msg = {
3023                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3024                 .rpc_argp = &args,
3025                 .rpc_resp = &res,
3026         };
3027
3028         nfs_fattr_init(fsstat->fattr);
3029         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3030 }
3031
3032 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3033 {
3034         struct nfs4_exception exception = { };
3035         int err;
3036         do {
3037                 err = nfs4_handle_exception(server,
3038                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3039                                 &exception);
3040         } while (exception.retry);
3041         return err;
3042 }
3043
3044 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3045                 struct nfs_fsinfo *fsinfo)
3046 {
3047         struct nfs4_fsinfo_arg args = {
3048                 .fh = fhandle,
3049                 .bitmask = server->attr_bitmask,
3050         };
3051         struct nfs4_fsinfo_res res = {
3052                 .fsinfo = fsinfo,
3053         };
3054         struct rpc_message msg = {
3055                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3056                 .rpc_argp = &args,
3057                 .rpc_resp = &res,
3058         };
3059
3060         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3061 }
3062
3063 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3064 {
3065         struct nfs4_exception exception = { };
3066         int err;
3067
3068         do {
3069                 err = nfs4_handle_exception(server,
3070                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3071                                 &exception);
3072         } while (exception.retry);
3073         return err;
3074 }
3075
3076 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3077 {
3078         nfs_fattr_init(fsinfo->fattr);
3079         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3080 }
3081
3082 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3083                 struct nfs_pathconf *pathconf)
3084 {
3085         struct nfs4_pathconf_arg args = {
3086                 .fh = fhandle,
3087                 .bitmask = server->attr_bitmask,
3088         };
3089         struct nfs4_pathconf_res res = {
3090                 .pathconf = pathconf,
3091         };
3092         struct rpc_message msg = {
3093                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3094                 .rpc_argp = &args,
3095                 .rpc_resp = &res,
3096         };
3097
3098         /* None of the pathconf attributes are mandatory to implement */
3099         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3100                 memset(pathconf, 0, sizeof(*pathconf));
3101                 return 0;
3102         }
3103
3104         nfs_fattr_init(pathconf->fattr);
3105         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3106 }
3107
3108 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3109                 struct nfs_pathconf *pathconf)
3110 {
3111         struct nfs4_exception exception = { };
3112         int err;
3113
3114         do {
3115                 err = nfs4_handle_exception(server,
3116                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3117                                 &exception);
3118         } while (exception.retry);
3119         return err;
3120 }
3121
3122 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3123 {
3124         struct nfs_server *server = NFS_SERVER(data->inode);
3125
3126         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3127                 nfs_restart_rpc(task, server->nfs_client);
3128                 return -EAGAIN;
3129         }
3130
3131         nfs_invalidate_atime(data->inode);
3132         if (task->tk_status > 0)
3133                 renew_lease(server, data->timestamp);
3134         return 0;
3135 }
3136
3137 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3138 {
3139
3140         dprintk("--> %s\n", __func__);
3141
3142         if (!nfs4_sequence_done(task, &data->res.seq_res))
3143                 return -EAGAIN;
3144
3145         return data->read_done_cb(task, data);
3146 }
3147
3148 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3149 {
3150         data->timestamp   = jiffies;
3151         data->read_done_cb = nfs4_read_done_cb;
3152         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3153 }
3154
3155 /* Reset the the nfs_read_data to send the read to the MDS. */
3156 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3157 {
3158         dprintk("%s Reset task for i/o through\n", __func__);
3159         put_lseg(data->lseg);
3160         data->lseg = NULL;
3161         /* offsets will differ in the dense stripe case */
3162         data->args.offset = data->mds_offset;
3163         data->ds_clp = NULL;
3164         data->args.fh     = NFS_FH(data->inode);
3165         data->read_done_cb = nfs4_read_done_cb;
3166         task->tk_ops = data->mds_ops;
3167         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3168 }
3169 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3170
3171 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3172 {
3173         struct inode *inode = data->inode;
3174
3175         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3176                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3177                 return -EAGAIN;
3178         }
3179         if (task->tk_status >= 0) {
3180                 renew_lease(NFS_SERVER(inode), data->timestamp);
3181                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3182         }
3183         return 0;
3184 }
3185
3186 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3187 {
3188         if (!nfs4_sequence_done(task, &data->res.seq_res))
3189                 return -EAGAIN;
3190         return data->write_done_cb(task, data);
3191 }
3192
3193 /* Reset the the nfs_write_data to send the write to the MDS. */
3194 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3195 {
3196         dprintk("%s Reset task for i/o through\n", __func__);
3197         put_lseg(data->lseg);
3198         data->lseg          = NULL;
3199         data->ds_clp        = NULL;
3200         data->write_done_cb = nfs4_write_done_cb;
3201         data->args.fh       = NFS_FH(data->inode);
3202         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3203         data->args.offset   = data->mds_offset;
3204         data->res.fattr     = &data->fattr;
3205         task->tk_ops        = data->mds_ops;
3206         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3207 }
3208 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3209
3210 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3211 {
3212         struct nfs_server *server = NFS_SERVER(data->inode);
3213
3214         if (data->lseg) {
3215                 data->args.bitmask = NULL;
3216                 data->res.fattr = NULL;
3217         } else
3218                 data->args.bitmask = server->cache_consistency_bitmask;
3219         if (!data->write_done_cb)
3220                 data->write_done_cb = nfs4_write_done_cb;
3221         data->res.server = server;
3222         data->timestamp   = jiffies;
3223
3224         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3225 }
3226
3227 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3228 {
3229         struct inode *inode = data->inode;
3230
3231         if (!nfs4_sequence_done(task, &data->res.seq_res))
3232                 return -EAGAIN;
3233
3234         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3235                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3236                 return -EAGAIN;
3237         }
3238         nfs_refresh_inode(inode, data->res.fattr);
3239         return 0;
3240 }
3241
3242 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3243 {
3244         struct nfs_server *server = NFS_SERVER(data->inode);
3245
3246         data->args.bitmask = server->cache_consistency_bitmask;
3247         data->res.server = server;
3248         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3249 }
3250
3251 struct nfs4_renewdata {
3252         struct nfs_client       *client;
3253         unsigned long           timestamp;
3254 };
3255
3256 /*
3257  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3258  * standalone procedure for queueing an asynchronous RENEW.
3259  */
3260 static void nfs4_renew_release(void *calldata)
3261 {
3262         struct nfs4_renewdata *data = calldata;
3263         struct nfs_client *clp = data->client;
3264
3265         if (atomic_read(&clp->cl_count) > 1)
3266                 nfs4_schedule_state_renewal(clp);
3267         nfs_put_client(clp);
3268         kfree(data);
3269 }
3270
3271 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3272 {
3273         struct nfs4_renewdata *data = calldata;
3274         struct nfs_client *clp = data->client;
3275         unsigned long timestamp = data->timestamp;
3276
3277         if (task->tk_status < 0) {
3278                 /* Unless we're shutting down, schedule state recovery! */
3279                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3280                         nfs4_schedule_lease_recovery(clp);
3281                 return;
3282         }
3283         do_renew_lease(clp, timestamp);
3284 }
3285
3286 static const struct rpc_call_ops nfs4_renew_ops = {
3287         .rpc_call_done = nfs4_renew_done,
3288         .rpc_release = nfs4_renew_release,
3289 };
3290
3291 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3292 {
3293         struct rpc_message msg = {
3294                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3295                 .rpc_argp       = clp,
3296                 .rpc_cred       = cred,
3297         };
3298         struct nfs4_renewdata *data;
3299
3300         if (!atomic_inc_not_zero(&clp->cl_count))
3301                 return -EIO;
3302         data = kmalloc(sizeof(*data), GFP_KERNEL);
3303         if (data == NULL)
3304                 return -ENOMEM;
3305         data->client = clp;
3306         data->timestamp = jiffies;
3307         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3308                         &nfs4_renew_ops, data);
3309 }
3310
3311 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3312 {
3313         struct rpc_message msg = {
3314                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3315                 .rpc_argp       = clp,
3316                 .rpc_cred       = cred,
3317         };
3318         unsigned long now = jiffies;
3319         int status;
3320
3321         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3322         if (status < 0)
3323                 return status;
3324         do_renew_lease(clp, now);
3325         return 0;
3326 }
3327
3328 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3329 {
3330         return (server->caps & NFS_CAP_ACLS)
3331                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3332                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3333 }
3334
3335 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3336  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3337  * the stack.
3338  */
3339 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3340
3341 static void buf_to_pages(const void *buf, size_t buflen,
3342                 struct page **pages, unsigned int *pgbase)
3343 {
3344         const void *p = buf;
3345
3346         *pgbase = offset_in_page(buf);
3347         p -= *pgbase;
3348         while (p < buf + buflen) {
3349                 *(pages++) = virt_to_page(p);
3350                 p += PAGE_CACHE_SIZE;
3351         }
3352 }
3353
3354 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3355                 struct page **pages, unsigned int *pgbase)
3356 {
3357         struct page *newpage, **spages;
3358         int rc = 0;
3359         size_t len;
3360         spages = pages;
3361
3362         do {
3363                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3364                 newpage = alloc_page(GFP_KERNEL);
3365
3366                 if (newpage == NULL)
3367                         goto unwind;
3368                 memcpy(page_address(newpage), buf, len);
3369                 buf += len;
3370                 buflen -= len;
3371                 *pages++ = newpage;
3372                 rc++;
3373         } while (buflen != 0);
3374
3375         return rc;
3376
3377 unwind:
3378         for(; rc > 0; rc--)
3379                 __free_page(spages[rc-1]);
3380         return -ENOMEM;
3381 }
3382
3383 struct nfs4_cached_acl {
3384         int cached;
3385         size_t len;
3386         char data[0];
3387 };
3388
3389 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3390 {
3391         struct nfs_inode *nfsi = NFS_I(inode);
3392
3393         spin_lock(&inode->i_lock);
3394         kfree(nfsi->nfs4_acl);
3395         nfsi->nfs4_acl = acl;
3396         spin_unlock(&inode->i_lock);
3397 }
3398
3399 static void nfs4_zap_acl_attr(struct inode *inode)
3400 {
3401         nfs4_set_cached_acl(inode, NULL);
3402 }
3403
3404 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3405 {
3406         struct nfs_inode *nfsi = NFS_I(inode);
3407         struct nfs4_cached_acl *acl;
3408         int ret = -ENOENT;
3409
3410         spin_lock(&inode->i_lock);
3411         acl = nfsi->nfs4_acl;
3412         if (acl == NULL)
3413                 goto out;
3414         if (buf == NULL) /* user is just asking for length */
3415                 goto out_len;
3416         if (acl->cached == 0)
3417                 goto out;
3418         ret = -ERANGE; /* see getxattr(2) man page */
3419         if (acl->len > buflen)
3420                 goto out;
3421         memcpy(buf, acl->data, acl->len);
3422 out_len:
3423         ret = acl->len;
3424 out:
3425         spin_unlock(&inode->i_lock);
3426         return ret;
3427 }
3428
3429 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3430 {
3431         struct nfs4_cached_acl *acl;
3432
3433         if (buf && acl_len <= PAGE_SIZE) {
3434                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3435                 if (acl == NULL)
3436                         goto out;
3437                 acl->cached = 1;
3438                 memcpy(acl->data, buf, acl_len);
3439         } else {
3440                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3441                 if (acl == NULL)
3442                         goto out;
3443                 acl->cached = 0;
3444         }
3445         acl->len = acl_len;
3446 out:
3447         nfs4_set_cached_acl(inode, acl);
3448 }
3449
3450 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3451 {
3452         struct page *pages[NFS4ACL_MAXPAGES];
3453         struct nfs_getaclargs args = {
3454                 .fh = NFS_FH(inode),
3455                 .acl_pages = pages,
3456                 .acl_len = buflen,
3457         };
3458         struct nfs_getaclres res = {
3459                 .acl_len = buflen,
3460         };
3461         void *resp_buf;
3462         struct rpc_message msg = {
3463                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3464                 .rpc_argp = &args,
3465                 .rpc_resp = &res,
3466         };
3467         struct page *localpage = NULL;
3468         int ret;
3469
3470         if (buflen < PAGE_SIZE) {
3471                 /* As long as we're doing a round trip to the server anyway,
3472                  * let's be prepared for a page of acl data. */
3473                 localpage = alloc_page(GFP_KERNEL);
3474                 resp_buf = page_address(localpage);
3475                 if (localpage == NULL)
3476                         return -ENOMEM;
3477                 args.acl_pages[0] = localpage;
3478                 args.acl_pgbase = 0;
3479                 args.acl_len = PAGE_SIZE;
3480         } else {
3481                 resp_buf = buf;
3482                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3483         }
3484         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3485         if (ret)
3486                 goto out_free;
3487         if (res.acl_len > args.acl_len)
3488                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3489         else
3490                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3491         if (buf) {
3492                 ret = -ERANGE;
3493                 if (res.acl_len > buflen)
3494                         goto out_free;
3495                 if (localpage)
3496                         memcpy(buf, resp_buf, res.acl_len);
3497         }
3498         ret = res.acl_len;
3499 out_free:
3500         if (localpage)
3501                 __free_page(localpage);
3502         return ret;
3503 }
3504
3505 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3506 {
3507         struct nfs4_exception exception = { };
3508         ssize_t ret;
3509         do {
3510                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3511                 if (ret >= 0)
3512                         break;
3513                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3514         } while (exception.retry);
3515         return ret;
3516 }
3517
3518 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3519 {
3520         struct nfs_server *server = NFS_SERVER(inode);
3521         int ret;
3522
3523         if (!nfs4_server_supports_acls(server))
3524                 return -EOPNOTSUPP;
3525         ret = nfs_revalidate_inode(server, inode);
3526         if (ret < 0)
3527                 return ret;
3528         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3529                 nfs_zap_acl_cache(inode);
3530         ret = nfs4_read_cached_acl(inode, buf, buflen);
3531         if (ret != -ENOENT)
3532                 return ret;
3533         return nfs4_get_acl_uncached(inode, buf, buflen);
3534 }
3535
3536 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3537 {
3538         struct nfs_server *server = NFS_SERVER(inode);
3539         struct page *pages[NFS4ACL_MAXPAGES];
3540         struct nfs_setaclargs arg = {
3541                 .fh             = NFS_FH(inode),
3542                 .acl_pages      = pages,
3543                 .acl_len        = buflen,
3544         };
3545         struct nfs_setaclres res;
3546         struct rpc_message msg = {
3547                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3548                 .rpc_argp       = &arg,
3549                 .rpc_resp       = &res,
3550         };
3551         int ret, i;
3552
3553         if (!nfs4_server_supports_acls(server))
3554                 return -EOPNOTSUPP;
3555         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3556         if (i < 0)
3557                 return i;
3558         nfs_inode_return_delegation(inode);
3559         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3560
3561         /*
3562          * Free each page after tx, so the only ref left is
3563          * held by the network stack
3564          */
3565         for (; i > 0; i--)
3566                 put_page(pages[i-1]);
3567
3568         /*
3569          * Acl update can result in inode attribute update.
3570          * so mark the attribute cache invalid.
3571          */
3572         spin_lock(&inode->i_lock);
3573         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3574         spin_unlock(&inode->i_lock);
3575         nfs_access_zap_cache(inode);
3576         nfs_zap_acl_cache(inode);
3577         return ret;
3578 }
3579
3580 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3581 {
3582         struct nfs4_exception exception = { };
3583         int err;
3584         do {
3585                 err = nfs4_handle_exception(NFS_SERVER(inode),
3586                                 __nfs4_proc_set_acl(inode, buf, buflen),
3587                                 &exception);
3588         } while (exception.retry);
3589         return err;
3590 }
3591
3592 static int
3593 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3594 {
3595         struct nfs_client *clp = server->nfs_client;
3596
3597         if (task->tk_status >= 0)
3598                 return 0;
3599         switch(task->tk_status) {
3600                 case -NFS4ERR_ADMIN_REVOKED:
3601                 case -NFS4ERR_BAD_STATEID:
3602                 case -NFS4ERR_OPENMODE:
3603                         if (state == NULL)
3604                                 break;
3605                         nfs4_schedule_stateid_recovery(server, state);
3606                         goto wait_on_recovery;
3607                 case -NFS4ERR_STALE_STATEID:
3608                 case -NFS4ERR_STALE_CLIENTID:
3609                 case -NFS4ERR_EXPIRED:
3610                         nfs4_schedule_lease_recovery(clp);
3611                         goto wait_on_recovery;
3612 #if defined(CONFIG_NFS_V4_1)
3613                 case -NFS4ERR_BADSESSION:
3614                 case -NFS4ERR_BADSLOT:
3615                 case -NFS4ERR_BAD_HIGH_SLOT:
3616                 case -NFS4ERR_DEADSESSION:
3617                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3618                 case -NFS4ERR_SEQ_FALSE_RETRY:
3619                 case -NFS4ERR_SEQ_MISORDERED:
3620                         dprintk("%s ERROR %d, Reset session\n", __func__,
3621                                 task->tk_status);
3622                         nfs4_schedule_session_recovery(clp->cl_session);
3623                         task->tk_status = 0;
3624                         return -EAGAIN;
3625 #endif /* CONFIG_NFS_V4_1 */
3626                 case -NFS4ERR_DELAY:
3627                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3628                 case -NFS4ERR_GRACE:
3629                 case -EKEYEXPIRED:
3630                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3631                         task->tk_status = 0;
3632                         return -EAGAIN;
3633                 case -NFS4ERR_OLD_STATEID:
3634                         task->tk_status = 0;
3635                         return -EAGAIN;
3636         }
3637         task->tk_status = nfs4_map_errors(task->tk_status);
3638         return 0;
3639 wait_on_recovery:
3640         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3641         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3642                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3643         task->tk_status = 0;
3644         return -EAGAIN;
3645 }
3646
3647 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3648                 unsigned short port, struct rpc_cred *cred,
3649                 struct nfs4_setclientid_res *res)
3650 {
3651         nfs4_verifier sc_verifier;
3652         struct nfs4_setclientid setclientid = {
3653                 .sc_verifier = &sc_verifier,
3654                 .sc_prog = program,
3655                 .sc_cb_ident = clp->cl_cb_ident,
3656         };
3657         struct rpc_message msg = {
3658                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3659                 .rpc_argp = &setclientid,
3660                 .rpc_resp = res,
3661                 .rpc_cred = cred,
3662         };
3663         __be32 *p;
3664         int loop = 0;
3665         int status;
3666
3667         p = (__be32*)sc_verifier.data;
3668         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3669         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3670
3671         for(;;) {
3672                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3673                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3674                                 clp->cl_ipaddr,
3675                                 rpc_peeraddr2str(clp->cl_rpcclient,
3676                                                         RPC_DISPLAY_ADDR),
3677                                 rpc_peeraddr2str(clp->cl_rpcclient,
3678                                                         RPC_DISPLAY_PROTO),
3679                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3680                                 clp->cl_id_uniquifier);
3681                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3682                                 sizeof(setclientid.sc_netid),
3683                                 rpc_peeraddr2str(clp->cl_rpcclient,
3684                                                         RPC_DISPLAY_NETID));
3685                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3686                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3687                                 clp->cl_ipaddr, port >> 8, port & 255);
3688
3689                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3690                 if (status != -NFS4ERR_CLID_INUSE)
3691                         break;
3692                 if (signalled())
3693                         break;
3694                 if (loop++ & 1)
3695                         ssleep(clp->cl_lease_time / HZ + 1);
3696                 else
3697                         if (++clp->cl_id_uniquifier == 0)
3698                                 break;
3699         }
3700         return status;
3701 }
3702
3703 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3704                 struct nfs4_setclientid_res *arg,
3705                 struct rpc_cred *cred)
3706 {
3707         struct nfs_fsinfo fsinfo;
3708         struct rpc_message msg = {
3709                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3710                 .rpc_argp = arg,
3711                 .rpc_resp = &fsinfo,
3712                 .rpc_cred = cred,
3713         };
3714         unsigned long now;
3715         int status;
3716
3717         now = jiffies;
3718         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3719         if (status == 0) {
3720                 spin_lock(&clp->cl_lock);
3721                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3722                 clp->cl_last_renewal = now;
3723                 spin_unlock(&clp->cl_lock);
3724         }
3725         return status;
3726 }
3727
3728 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3729                 struct nfs4_setclientid_res *arg,
3730                 struct rpc_cred *cred)
3731 {
3732         long timeout = 0;
3733         int err;
3734         do {
3735                 err = _nfs4_proc_setclientid_confirm(clp, arg, cred);
3736                 switch (err) {
3737                         case 0:
3738                                 return err;
3739                         case -NFS4ERR_RESOURCE:
3740                                 /* The IBM lawyers misread another document! */
3741                         case -NFS4ERR_DELAY:
3742                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3743                 }
3744         } while (err == 0);
3745         return err;
3746 }
3747
3748 struct nfs4_delegreturndata {
3749         struct nfs4_delegreturnargs args;
3750         struct nfs4_delegreturnres res;
3751         struct nfs_fh fh;
3752         nfs4_stateid stateid;
3753         unsigned long timestamp;
3754         struct nfs_fattr fattr;
3755         int rpc_status;
3756 };
3757
3758 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3759 {
3760         struct nfs4_delegreturndata *data = calldata;
3761
3762         if (!nfs4_sequence_done(task, &data->res.seq_res))
3763                 return;
3764
3765         switch (task->tk_status) {
3766         case -NFS4ERR_STALE_STATEID:
3767         case -NFS4ERR_EXPIRED:
3768         case 0:
3769                 renew_lease(data->res.server, data->timestamp);
3770                 break;
3771         default:
3772                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3773                                 -EAGAIN) {
3774                         nfs_restart_rpc(task, data->res.server->nfs_client);
3775                         return;
3776                 }
3777         }
3778         data->rpc_status = task->tk_status;
3779 }
3780
3781 static void nfs4_delegreturn_release(void *calldata)
3782 {
3783         kfree(calldata);
3784 }
3785
3786 #if defined(CONFIG_NFS_V4_1)
3787 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3788 {
3789         struct nfs4_delegreturndata *d_data;
3790
3791         d_data = (struct nfs4_delegreturndata *)data;
3792
3793         if (nfs4_setup_sequence(d_data->res.server,
3794                                 &d_data->args.seq_args,
3795                                 &d_data->res.seq_res, 1, task))
3796                 return;
3797         rpc_call_start(task);
3798 }
3799 #endif /* CONFIG_NFS_V4_1 */
3800
3801 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3802 #if defined(CONFIG_NFS_V4_1)
3803         .rpc_call_prepare = nfs4_delegreturn_prepare,
3804 #endif /* CONFIG_NFS_V4_1 */
3805         .rpc_call_done = nfs4_delegreturn_done,
3806         .rpc_release = nfs4_delegreturn_release,
3807 };
3808
3809 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3810 {
3811         struct nfs4_delegreturndata *data;
3812         struct nfs_server *server = NFS_SERVER(inode);
3813         struct rpc_task *task;
3814         struct rpc_message msg = {
3815                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3816                 .rpc_cred = cred,
3817         };
3818         struct rpc_task_setup task_setup_data = {
3819                 .rpc_client = server->client,
3820                 .rpc_message = &msg,
3821                 .callback_ops = &nfs4_delegreturn_ops,
3822                 .flags = RPC_TASK_ASYNC,
3823         };
3824         int status = 0;
3825
3826         data = kzalloc(sizeof(*data), GFP_NOFS);
3827         if (data == NULL)
3828                 return -ENOMEM;
3829         data->args.fhandle = &data->fh;
3830         data->args.stateid = &data->stateid;
3831         data->args.bitmask = server->attr_bitmask;
3832         nfs_copy_fh(&data->fh, NFS_FH(inode));
3833         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3834         data->res.fattr = &data->fattr;
3835         data->res.server = server;
3836         nfs_fattr_init(data->res.fattr);
3837         data->timestamp = jiffies;
3838         data->rpc_status = 0;
3839
3840         task_setup_data.callback_data = data;
3841         msg.rpc_argp = &data->args;
3842         msg.rpc_resp = &data->res;
3843         task = rpc_run_task(&task_setup_data);
3844         if (IS_ERR(task))
3845                 return PTR_ERR(task);
3846         if (!issync)
3847                 goto out;
3848         status = nfs4_wait_for_completion_rpc_task(task);
3849         if (status != 0)
3850                 goto out;
3851         status = data->rpc_status;
3852         if (status != 0)
3853                 goto out;
3854         nfs_refresh_inode(inode, &data->fattr);
3855 out:
3856         rpc_put_task(task);
3857         return status;
3858 }
3859
3860 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3861 {
3862         struct nfs_server *server = NFS_SERVER(inode);
3863         struct nfs4_exception exception = { };
3864         int err;
3865         do {
3866                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3867                 switch (err) {
3868                         case -NFS4ERR_STALE_STATEID:
3869                         case -NFS4ERR_EXPIRED:
3870                         case 0:
3871                                 return 0;
3872                 }
3873                 err = nfs4_handle_exception(server, err, &exception);
3874         } while (exception.retry);
3875         return err;
3876 }
3877
3878 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3879 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3880
3881 /*
3882  * sleep, with exponential backoff, and retry the LOCK operation.
3883  */
3884 static unsigned long
3885 nfs4_set_lock_task_retry(unsigned long timeout)
3886 {
3887         schedule_timeout_killable(timeout);
3888         timeout <<= 1;
3889         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3890                 return NFS4_LOCK_MAXTIMEOUT;
3891         return timeout;
3892 }
3893
3894 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3895 {
3896         struct inode *inode = state->inode;
3897         struct nfs_server *server = NFS_SERVER(inode);
3898         struct nfs_client *clp = server->nfs_client;
3899         struct nfs_lockt_args arg = {
3900                 .fh = NFS_FH(inode),
3901                 .fl = request,
3902         };
3903         struct nfs_lockt_res res = {
3904                 .denied = request,
3905         };
3906         struct rpc_message msg = {
3907                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3908                 .rpc_argp       = &arg,
3909                 .rpc_resp       = &res,
3910                 .rpc_cred       = state->owner->so_cred,
3911         };
3912         struct nfs4_lock_state *lsp;
3913         int status;
3914
3915         arg.lock_owner.clientid = clp->cl_clientid;
3916         status = nfs4_set_lock_state(state, request);
3917         if (status != 0)
3918                 goto out;
3919         lsp = request->fl_u.nfs4_fl.owner;
3920         arg.lock_owner.id = lsp->ls_id.id;
3921         arg.lock_owner.s_dev = server->s_dev;
3922         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3923         switch (status) {
3924                 case 0:
3925                         request->fl_type = F_UNLCK;
3926                         break;
3927                 case -NFS4ERR_DENIED:
3928                         status = 0;
3929         }
3930         request->fl_ops->fl_release_private(request);
3931 out:
3932         return status;
3933 }
3934
3935 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3936 {
3937         struct nfs4_exception exception = { };
3938         int err;
3939
3940         do {
3941                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3942                                 _nfs4_proc_getlk(state, cmd, request),
3943                                 &exception);
3944         } while (exception.retry);
3945         return err;
3946 }
3947
3948 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3949 {
3950         int res = 0;
3951         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3952                 case FL_POSIX:
3953                         res = posix_lock_file_wait(file, fl);
3954                         break;
3955                 case FL_FLOCK:
3956                         res = flock_lock_file_wait(file, fl);
3957                         break;
3958                 default:
3959                         BUG();
3960         }
3961         return res;
3962 }
3963
3964 struct nfs4_unlockdata {
3965         struct nfs_locku_args arg;
3966         struct nfs_locku_res res;
3967         struct nfs4_lock_state *lsp;
3968         struct nfs_open_context *ctx;
3969         struct file_lock fl;
3970         const struct nfs_server *server;
3971         unsigned long timestamp;
3972 };
3973
3974 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3975                 struct nfs_open_context *ctx,
3976                 struct nfs4_lock_state *lsp,
3977                 struct nfs_seqid *seqid)
3978 {
3979         struct nfs4_unlockdata *p;
3980         struct inode *inode = lsp->ls_state->inode;
3981
3982         p = kzalloc(sizeof(*p), GFP_NOFS);
3983         if (p == NULL)
3984                 return NULL;
3985         p->arg.fh = NFS_FH(inode);
3986         p->arg.fl = &p->fl;
3987         p->arg.seqid = seqid;
3988         p->res.seqid = seqid;
3989         p->arg.stateid = &lsp->ls_stateid;
3990         p->lsp = lsp;
3991         atomic_inc(&lsp->ls_count);
3992         /* Ensure we don't close file until we're done freeing locks! */
3993         p->ctx = get_nfs_open_context(ctx);
3994         memcpy(&p->fl, fl, sizeof(p->fl));
3995         p->server = NFS_SERVER(inode);
3996         return p;
3997 }
3998
3999 static void nfs4_locku_release_calldata(void *data)
4000 {
4001         struct nfs4_unlockdata *calldata = data;
4002         nfs_free_seqid(calldata->arg.seqid);
4003         nfs4_put_lock_state(calldata->lsp);
4004         put_nfs_open_context(calldata->ctx);
4005         kfree(calldata);
4006 }
4007
4008 static void nfs4_locku_done(struct rpc_task *task, void *data)
4009 {
4010         struct nfs4_unlockdata *calldata = data;
4011
4012         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4013                 return;
4014         switch (task->tk_status) {
4015                 case 0:
4016                         memcpy(calldata->lsp->ls_stateid.data,
4017                                         calldata->res.stateid.data,
4018                                         sizeof(calldata->lsp->ls_stateid.data));
4019                         renew_lease(calldata->server, calldata->timestamp);
4020                         break;
4021                 case -NFS4ERR_BAD_STATEID:
4022                 case -NFS4ERR_OLD_STATEID:
4023                 case -NFS4ERR_STALE_STATEID:
4024                 case -NFS4ERR_EXPIRED:
4025                         break;
4026                 default:
4027                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4028                                 nfs_restart_rpc(task,
4029                                                  calldata->server->nfs_client);
4030         }
4031 }
4032
4033 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4034 {
4035         struct nfs4_unlockdata *calldata = data;
4036
4037         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4038                 return;
4039         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4040                 /* Note: exit _without_ running nfs4_locku_done */
4041                 task->tk_action = NULL;
4042                 return;
4043         }
4044         calldata->timestamp = jiffies;
4045         if (nfs4_setup_sequence(calldata->server,
4046                                 &calldata->arg.seq_args,
4047                                 &calldata->res.seq_res, 1, task))
4048                 return;
4049         rpc_call_start(task);
4050 }
4051
4052 static const struct rpc_call_ops nfs4_locku_ops = {
4053         .rpc_call_prepare = nfs4_locku_prepare,
4054         .rpc_call_done = nfs4_locku_done,
4055         .rpc_release = nfs4_locku_release_calldata,
4056 };
4057
4058 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4059                 struct nfs_open_context *ctx,
4060                 struct nfs4_lock_state *lsp,
4061                 struct nfs_seqid *seqid)
4062 {
4063         struct nfs4_unlockdata *data;
4064         struct rpc_message msg = {
4065                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4066                 .rpc_cred = ctx->cred,
4067         };
4068         struct rpc_task_setup task_setup_data = {
4069                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4070                 .rpc_message = &msg,
4071                 .callback_ops = &nfs4_locku_ops,
4072                 .workqueue = nfsiod_workqueue,
4073                 .flags = RPC_TASK_ASYNC,
4074         };
4075
4076         /* Ensure this is an unlock - when canceling a lock, the
4077          * canceled lock is passed in, and it won't be an unlock.
4078          */
4079         fl->fl_type = F_UNLCK;
4080
4081         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4082         if (data == NULL) {
4083                 nfs_free_seqid(seqid);
4084                 return ERR_PTR(-ENOMEM);
4085         }
4086
4087         msg.rpc_argp = &data->arg;
4088         msg.rpc_resp = &data->res;
4089         task_setup_data.callback_data = data;
4090         return rpc_run_task(&task_setup_data);
4091 }
4092
4093 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4094 {
4095         struct nfs_inode *nfsi = NFS_I(state->inode);
4096         struct nfs_seqid *seqid;
4097         struct nfs4_lock_state *lsp;
4098         struct rpc_task *task;
4099         int status = 0;
4100         unsigned char fl_flags = request->fl_flags;
4101
4102         status = nfs4_set_lock_state(state, request);
4103         /* Unlock _before_ we do the RPC call */
4104         request->fl_flags |= FL_EXISTS;
4105         down_read(&nfsi->rwsem);
4106         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4107                 up_read(&nfsi->rwsem);
4108                 goto out;
4109         }
4110         up_read(&nfsi->rwsem);
4111         if (status != 0)
4112                 goto out;
4113         /* Is this a delegated lock? */
4114         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4115                 goto out;
4116         lsp = request->fl_u.nfs4_fl.owner;
4117         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4118         status = -ENOMEM;
4119         if (seqid == NULL)
4120                 goto out;
4121         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4122         status = PTR_ERR(task);
4123         if (IS_ERR(task))
4124                 goto out;
4125         status = nfs4_wait_for_completion_rpc_task(task);
4126         rpc_put_task(task);
4127 out:
4128         request->fl_flags = fl_flags;
4129         return status;
4130 }
4131
4132 struct nfs4_lockdata {
4133         struct nfs_lock_args arg;
4134         struct nfs_lock_res res;
4135         struct nfs4_lock_state *lsp;
4136         struct nfs_open_context *ctx;
4137         struct file_lock fl;
4138         unsigned long timestamp;
4139         int rpc_status;
4140         int cancelled;
4141         struct nfs_server *server;
4142 };
4143
4144 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4145                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4146                 gfp_t gfp_mask)
4147 {
4148         struct nfs4_lockdata *p;
4149         struct inode *inode = lsp->ls_state->inode;
4150         struct nfs_server *server = NFS_SERVER(inode);
4151
4152         p = kzalloc(sizeof(*p), gfp_mask);
4153         if (p == NULL)
4154                 return NULL;
4155
4156         p->arg.fh = NFS_FH(inode);
4157         p->arg.fl = &p->fl;
4158         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4159         if (p->arg.open_seqid == NULL)
4160                 goto out_free;
4161         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4162         if (p->arg.lock_seqid == NULL)
4163                 goto out_free_seqid;
4164         p->arg.lock_stateid = &lsp->ls_stateid;
4165         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4166         p->arg.lock_owner.id = lsp->ls_id.id;
4167         p->arg.lock_owner.s_dev = server->s_dev;
4168         p->res.lock_seqid = p->arg.lock_seqid;
4169         p->lsp = lsp;
4170         p->server = server;
4171         atomic_inc(&lsp->ls_count);
4172         p->ctx = get_nfs_open_context(ctx);
4173         memcpy(&p->fl, fl, sizeof(p->fl));
4174         return p;
4175 out_free_seqid:
4176         nfs_free_seqid(p->arg.open_seqid);
4177 out_free:
4178         kfree(p);
4179         return NULL;
4180 }
4181
4182 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4183 {
4184         struct nfs4_lockdata *data = calldata;
4185         struct nfs4_state *state = data->lsp->ls_state;
4186
4187         dprintk("%s: begin!\n", __func__);
4188         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4189                 return;
4190         /* Do we need to do an open_to_lock_owner? */
4191         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4192                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4193                         return;
4194                 data->arg.open_stateid = &state->stateid;
4195                 data->arg.new_lock_owner = 1;
4196                 data->res.open_seqid = data->arg.open_seqid;
4197         } else
4198                 data->arg.new_lock_owner = 0;
4199         data->timestamp = jiffies;
4200         if (nfs4_setup_sequence(data->server,
4201                                 &data->arg.seq_args,
4202                                 &data->res.seq_res, 1, task))
4203                 return;
4204         rpc_call_start(task);
4205         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4206 }
4207
4208 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4209 {
4210         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4211         nfs4_lock_prepare(task, calldata);
4212 }
4213
4214 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4215 {
4216         struct nfs4_lockdata *data = calldata;
4217
4218         dprintk("%s: begin!\n", __func__);
4219
4220         if (!nfs4_sequence_done(task, &data->res.seq_res))
4221                 return;
4222
4223         data->rpc_status = task->tk_status;
4224         if (data->arg.new_lock_owner != 0) {
4225                 if (data->rpc_status == 0)
4226                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4227                 else
4228                         goto out;
4229         }
4230         if (data->rpc_status == 0) {
4231                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4232                                         sizeof(data->lsp->ls_stateid.data));
4233                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4234                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4235         }
4236 out:
4237         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4238 }
4239
4240 static void nfs4_lock_release(void *calldata)
4241 {
4242         struct nfs4_lockdata *data = calldata;
4243
4244         dprintk("%s: begin!\n", __func__);
4245         nfs_free_seqid(data->arg.open_seqid);
4246         if (data->cancelled != 0) {
4247                 struct rpc_task *task;
4248                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4249                                 data->arg.lock_seqid);
4250                 if (!IS_ERR(task))
4251                         rpc_put_task_async(task);
4252                 dprintk("%s: cancelling lock!\n", __func__);
4253         } else
4254                 nfs_free_seqid(data->arg.lock_seqid);
4255         nfs4_put_lock_state(data->lsp);
4256         put_nfs_open_context(data->ctx);
4257         kfree(data);
4258         dprintk("%s: done!\n", __func__);
4259 }
4260
4261 static const struct rpc_call_ops nfs4_lock_ops = {
4262         .rpc_call_prepare = nfs4_lock_prepare,
4263         .rpc_call_done = nfs4_lock_done,
4264         .rpc_release = nfs4_lock_release,
4265 };
4266
4267 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4268         .rpc_call_prepare = nfs4_recover_lock_prepare,
4269         .rpc_call_done = nfs4_lock_done,
4270         .rpc_release = nfs4_lock_release,
4271 };
4272
4273 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4274 {
4275         switch (error) {
4276         case -NFS4ERR_ADMIN_REVOKED:
4277         case -NFS4ERR_BAD_STATEID:
4278                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4279                 if (new_lock_owner != 0 ||
4280                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4281                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4282                 break;
4283         case -NFS4ERR_STALE_STATEID:
4284                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4285         case -NFS4ERR_EXPIRED:
4286                 nfs4_schedule_lease_recovery(server->nfs_client);
4287         };
4288 }
4289
4290 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4291 {
4292         struct nfs4_lockdata *data;
4293         struct rpc_task *task;
4294         struct rpc_message msg = {
4295                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4296                 .rpc_cred = state->owner->so_cred,
4297         };
4298         struct rpc_task_setup task_setup_data = {
4299                 .rpc_client = NFS_CLIENT(state->inode),
4300                 .rpc_message = &msg,
4301                 .callback_ops = &nfs4_lock_ops,
4302                 .workqueue = nfsiod_workqueue,
4303                 .flags = RPC_TASK_ASYNC,
4304         };
4305         int ret;
4306
4307         dprintk("%s: begin!\n", __func__);
4308         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4309                         fl->fl_u.nfs4_fl.owner,
4310                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4311         if (data == NULL)
4312                 return -ENOMEM;
4313         if (IS_SETLKW(cmd))
4314                 data->arg.block = 1;
4315         if (recovery_type > NFS_LOCK_NEW) {
4316                 if (recovery_type == NFS_LOCK_RECLAIM)
4317                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4318                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4319         }
4320         msg.rpc_argp = &data->arg;
4321         msg.rpc_resp = &data->res;
4322         task_setup_data.callback_data = data;
4323         task = rpc_run_task(&task_setup_data);
4324         if (IS_ERR(task))
4325                 return PTR_ERR(task);
4326         ret = nfs4_wait_for_completion_rpc_task(task);
4327         if (ret == 0) {
4328                 ret = data->rpc_status;
4329                 if (ret)
4330                         nfs4_handle_setlk_error(data->server, data->lsp,
4331                                         data->arg.new_lock_owner, ret);
4332         } else
4333                 data->cancelled = 1;
4334         rpc_put_task(task);
4335         dprintk("%s: done, ret = %d!\n", __func__, ret);
4336         return ret;
4337 }
4338
4339 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4340 {
4341         struct nfs_server *server = NFS_SERVER(state->inode);
4342         struct nfs4_exception exception = { };
4343         int err;
4344
4345         do {
4346                 /* Cache the lock if possible... */
4347                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4348                         return 0;
4349                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4350                 if (err != -NFS4ERR_DELAY)
4351                         break;
4352                 nfs4_handle_exception(server, err, &exception);
4353         } while (exception.retry);
4354         return err;
4355 }
4356
4357 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4358 {
4359         struct nfs_server *server = NFS_SERVER(state->inode);
4360         struct nfs4_exception exception = { };
4361         int err;
4362
4363         err = nfs4_set_lock_state(state, request);
4364         if (err != 0)
4365                 return err;
4366         do {
4367                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4368                         return 0;
4369                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4370                 switch (err) {
4371                 default:
4372                         goto out;
4373                 case -NFS4ERR_GRACE:
4374                 case -NFS4ERR_DELAY:
4375                         nfs4_handle_exception(server, err, &exception);
4376                         err = 0;
4377                 }
4378         } while (exception.retry);
4379 out:
4380         return err;
4381 }
4382
4383 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4384 {
4385         struct nfs_inode *nfsi = NFS_I(state->inode);
4386         unsigned char fl_flags = request->fl_flags;
4387         int status = -ENOLCK;
4388
4389         if ((fl_flags & FL_POSIX) &&
4390                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4391                 goto out;
4392         /* Is this a delegated open? */
4393         status = nfs4_set_lock_state(state, request);
4394         if (status != 0)
4395                 goto out;
4396         request->fl_flags |= FL_ACCESS;
4397         status = do_vfs_lock(request->fl_file, request);
4398         if (status < 0)
4399                 goto out;
4400         down_read(&nfsi->rwsem);
4401         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4402                 /* Yes: cache locks! */
4403                 /* ...but avoid races with delegation recall... */
4404                 request->fl_flags = fl_flags & ~FL_SLEEP;
4405                 status = do_vfs_lock(request->fl_file, request);
4406                 goto out_unlock;
4407         }
4408         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4409         if (status != 0)
4410                 goto out_unlock;
4411         /* Note: we always want to sleep here! */
4412         request->fl_flags = fl_flags | FL_SLEEP;
4413         if (do_vfs_lock(request->fl_file, request) < 0)
4414                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4415 out_unlock:
4416         up_read(&nfsi->rwsem);
4417 out:
4418         request->fl_flags = fl_flags;
4419         return status;
4420 }
4421
4422 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4423 {
4424         struct nfs4_exception exception = { };
4425         int err;
4426
4427         do {
4428                 err = _nfs4_proc_setlk(state, cmd, request);
4429                 if (err == -NFS4ERR_DENIED)
4430                         err = -EAGAIN;
4431                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4432                                 err, &exception);
4433         } while (exception.retry);
4434         return err;
4435 }
4436
4437 static int
4438 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4439 {
4440         struct nfs_open_context *ctx;
4441         struct nfs4_state *state;
4442         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4443         int status;
4444
4445         /* verify open state */
4446         ctx = nfs_file_open_context(filp);
4447         state = ctx->state;
4448
4449         if (request->fl_start < 0 || request->fl_end < 0)
4450                 return -EINVAL;
4451
4452         if (IS_GETLK(cmd)) {
4453                 if (state != NULL)
4454                         return nfs4_proc_getlk(state, F_GETLK, request);
4455                 return 0;
4456         }
4457
4458         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4459                 return -EINVAL;
4460
4461         if (request->fl_type == F_UNLCK) {
4462                 if (state != NULL)
4463                         return nfs4_proc_unlck(state, cmd, request);
4464                 return 0;
4465         }
4466
4467         if (state == NULL)
4468                 return -ENOLCK;
4469         do {
4470                 status = nfs4_proc_setlk(state, cmd, request);
4471                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4472                         break;
4473                 timeout = nfs4_set_lock_task_retry(timeout);
4474                 status = -ERESTARTSYS;
4475                 if (signalled())
4476                         break;
4477         } while(status < 0);
4478         return status;
4479 }
4480
4481 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4482 {
4483         struct nfs_server *server = NFS_SERVER(state->inode);
4484         struct nfs4_exception exception = { };
4485         int err;
4486
4487         err = nfs4_set_lock_state(state, fl);
4488         if (err != 0)
4489                 goto out;
4490         do {
4491                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4492                 switch (err) {
4493                         default:
4494                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4495                                                 __func__, err);
4496                         case 0:
4497                         case -ESTALE:
4498                                 goto out;
4499                         case -NFS4ERR_EXPIRED:
4500                         case -NFS4ERR_STALE_CLIENTID:
4501                         case -NFS4ERR_STALE_STATEID:
4502                                 nfs4_schedule_lease_recovery(server->nfs_client);
4503                                 goto out;
4504                         case -NFS4ERR_BADSESSION:
4505                         case -NFS4ERR_BADSLOT:
4506                         case -NFS4ERR_BAD_HIGH_SLOT:
4507                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4508                         case -NFS4ERR_DEADSESSION:
4509                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4510                                 goto out;
4511                         case -ERESTARTSYS:
4512                                 /*
4513                                  * The show must go on: exit, but mark the
4514                                  * stateid as needing recovery.
4515                                  */
4516                         case -NFS4ERR_ADMIN_REVOKED:
4517                         case -NFS4ERR_BAD_STATEID:
4518                         case -NFS4ERR_OPENMODE:
4519                                 nfs4_schedule_stateid_recovery(server, state);
4520                                 err = 0;
4521                                 goto out;
4522                         case -EKEYEXPIRED:
4523                                 /*
4524                                  * User RPCSEC_GSS context has expired.
4525                                  * We cannot recover this stateid now, so
4526                                  * skip it and allow recovery thread to
4527                                  * proceed.
4528                                  */
4529                                 err = 0;
4530                                 goto out;
4531                         case -ENOMEM:
4532                         case -NFS4ERR_DENIED:
4533                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4534                                 err = 0;
4535                                 goto out;
4536                         case -NFS4ERR_DELAY:
4537                                 break;
4538                 }
4539                 err = nfs4_handle_exception(server, err, &exception);
4540         } while (exception.retry);
4541 out:
4542         return err;
4543 }
4544
4545 static void nfs4_release_lockowner_release(void *calldata)
4546 {
4547         kfree(calldata);
4548 }
4549
4550 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4551         .rpc_release = nfs4_release_lockowner_release,
4552 };
4553
4554 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4555 {
4556         struct nfs_server *server = lsp->ls_state->owner->so_server;
4557         struct nfs_release_lockowner_args *args;
4558         struct rpc_message msg = {
4559                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4560         };
4561
4562         if (server->nfs_client->cl_mvops->minor_version != 0)
4563                 return;
4564         args = kmalloc(sizeof(*args), GFP_NOFS);
4565         if (!args)
4566                 return;
4567         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4568         args->lock_owner.id = lsp->ls_id.id;
4569         args->lock_owner.s_dev = server->s_dev;
4570         msg.rpc_argp = args;
4571         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4572 }
4573
4574 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4575
4576 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4577                                    const void *buf, size_t buflen,
4578                                    int flags, int type)
4579 {
4580         if (strcmp(key, "") != 0)
4581                 return -EINVAL;
4582
4583         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4584 }
4585
4586 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4587                                    void *buf, size_t buflen, int type)
4588 {
4589         if (strcmp(key, "") != 0)
4590                 return -EINVAL;
4591
4592         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4593 }
4594
4595 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4596                                        size_t list_len, const char *name,
4597                                        size_t name_len, int type)
4598 {
4599         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4600
4601         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4602                 return 0;
4603
4604         if (list && len <= list_len)
4605                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4606         return len;
4607 }
4608
4609 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4610 {
4611         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4612                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4613                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4614                 return;
4615
4616         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4617                 NFS_ATTR_FATTR_NLINK;
4618         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4619         fattr->nlink = 2;
4620 }
4621
4622 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4623                 struct nfs4_fs_locations *fs_locations, struct page *page)
4624 {
4625         struct nfs_server *server = NFS_SERVER(dir);
4626         u32 bitmask[2] = {
4627                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4628                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4629         };
4630         struct nfs4_fs_locations_arg args = {
4631                 .dir_fh = NFS_FH(dir),
4632                 .name = name,
4633                 .page = page,
4634                 .bitmask = bitmask,
4635         };
4636         struct nfs4_fs_locations_res res = {
4637                 .fs_locations = fs_locations,
4638         };
4639         struct rpc_message msg = {
4640                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4641                 .rpc_argp = &args,
4642                 .rpc_resp = &res,
4643         };
4644         int status;
4645
4646         dprintk("%s: start\n", __func__);
4647         nfs_fattr_init(&fs_locations->fattr);
4648         fs_locations->server = server;
4649         fs_locations->nlocations = 0;
4650         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4651         nfs_fixup_referral_attributes(&fs_locations->fattr);
4652         dprintk("%s: returned status = %d\n", __func__, status);
4653         return status;
4654 }
4655
4656 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4657 {
4658         int status;
4659         struct nfs4_secinfo_arg args = {
4660                 .dir_fh = NFS_FH(dir),
4661                 .name   = name,
4662         };
4663         struct nfs4_secinfo_res res = {
4664                 .flavors     = flavors,
4665         };
4666         struct rpc_message msg = {
4667                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4668                 .rpc_argp = &args,
4669                 .rpc_resp = &res,
4670         };
4671
4672         dprintk("NFS call  secinfo %s\n", name->name);
4673         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4674         dprintk("NFS reply  secinfo: %d\n", status);
4675         return status;
4676 }
4677
4678 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4679 {
4680         struct nfs4_exception exception = { };
4681         int err;
4682         do {
4683                 err = nfs4_handle_exception(NFS_SERVER(dir),
4684                                 _nfs4_proc_secinfo(dir, name, flavors),
4685                                 &exception);
4686         } while (exception.retry);
4687         return err;
4688 }
4689
4690 #ifdef CONFIG_NFS_V4_1
4691 /*
4692  * Check the exchange flags returned by the server for invalid flags, having
4693  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4694  * DS flags set.
4695  */
4696 static int nfs4_check_cl_exchange_flags(u32 flags)
4697 {
4698         if (flags & ~EXCHGID4_FLAG_MASK_R)
4699                 goto out_inval;
4700         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4701             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4702                 goto out_inval;
4703         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4704                 goto out_inval;
4705         return NFS_OK;
4706 out_inval:
4707         return -NFS4ERR_INVAL;
4708 }
4709
4710 /*
4711  * nfs4_proc_exchange_id()
4712  *
4713  * Since the clientid has expired, all compounds using sessions
4714  * associated with the stale clientid will be returning
4715  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4716  * be in some phase of session reset.
4717  */
4718 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4719 {
4720         nfs4_verifier verifier;
4721         struct nfs41_exchange_id_args args = {
4722                 .client = clp,
4723                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4724         };
4725         struct nfs41_exchange_id_res res = {
4726                 .client = clp,
4727         };
4728         int status;
4729         struct rpc_message msg = {
4730                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4731                 .rpc_argp = &args,
4732                 .rpc_resp = &res,
4733                 .rpc_cred = cred,
4734         };
4735         __be32 *p;
4736
4737         dprintk("--> %s\n", __func__);
4738         BUG_ON(clp == NULL);
4739
4740         p = (u32 *)verifier.data;
4741         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4742         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4743         args.verifier = &verifier;
4744
4745         args.id_len = scnprintf(args.id, sizeof(args.id),
4746                                 "%s/%s.%s/%u",
4747                                 clp->cl_ipaddr,
4748                                 init_utsname()->nodename,
4749                                 init_utsname()->domainname,
4750                                 clp->cl_rpcclient->cl_auth->au_flavor);
4751
4752         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4753         if (!status)
4754                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4755         dprintk("<-- %s status= %d\n", __func__, status);
4756         return status;
4757 }
4758
4759 struct nfs4_get_lease_time_data {
4760         struct nfs4_get_lease_time_args *args;
4761         struct nfs4_get_lease_time_res *res;
4762         struct nfs_client *clp;
4763 };
4764
4765 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4766                                         void *calldata)
4767 {
4768         int ret;
4769         struct nfs4_get_lease_time_data *data =
4770                         (struct nfs4_get_lease_time_data *)calldata;
4771
4772         dprintk("--> %s\n", __func__);
4773         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4774         /* just setup sequence, do not trigger session recovery
4775            since we're invoked within one */
4776         ret = nfs41_setup_sequence(data->clp->cl_session,
4777                                    &data->args->la_seq_args,
4778                                    &data->res->lr_seq_res, 0, task);
4779
4780         BUG_ON(ret == -EAGAIN);
4781         rpc_call_start(task);
4782         dprintk("<-- %s\n", __func__);
4783 }
4784
4785 /*
4786  * Called from nfs4_state_manager thread for session setup, so don't recover
4787  * from sequence operation or clientid errors.
4788  */
4789 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4790 {
4791         struct nfs4_get_lease_time_data *data =
4792                         (struct nfs4_get_lease_time_data *)calldata;
4793
4794         dprintk("--> %s\n", __func__);
4795         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4796                 return;
4797         switch (task->tk_status) {
4798         case -NFS4ERR_DELAY:
4799         case -NFS4ERR_GRACE:
4800                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4801                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4802                 task->tk_status = 0;
4803                 nfs_restart_rpc(task, data->clp);
4804                 return;
4805         }
4806         dprintk("<-- %s\n", __func__);
4807 }
4808
4809 struct rpc_call_ops nfs4_get_lease_time_ops = {
4810         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4811         .rpc_call_done = nfs4_get_lease_time_done,
4812 };
4813
4814 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4815 {
4816         struct rpc_task *task;
4817         struct nfs4_get_lease_time_args args;
4818         struct nfs4_get_lease_time_res res = {
4819                 .lr_fsinfo = fsinfo,
4820         };
4821         struct nfs4_get_lease_time_data data = {
4822                 .args = &args,
4823                 .res = &res,
4824                 .clp = clp,
4825         };
4826         struct rpc_message msg = {
4827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4828                 .rpc_argp = &args,
4829                 .rpc_resp = &res,
4830         };
4831         struct rpc_task_setup task_setup = {
4832                 .rpc_client = clp->cl_rpcclient,
4833                 .rpc_message = &msg,
4834                 .callback_ops = &nfs4_get_lease_time_ops,
4835                 .callback_data = &data
4836         };
4837         int status;
4838
4839         dprintk("--> %s\n", __func__);
4840         task = rpc_run_task(&task_setup);
4841
4842         if (IS_ERR(task))
4843                 status = PTR_ERR(task);
4844         else {
4845                 status = task->tk_status;
4846                 rpc_put_task(task);
4847         }
4848         dprintk("<-- %s return %d\n", __func__, status);
4849
4850         return status;
4851 }
4852
4853 /*
4854  * Reset a slot table
4855  */
4856 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4857                                  int ivalue)
4858 {
4859         struct nfs4_slot *new = NULL;
4860         int i;
4861         int ret = 0;
4862
4863         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4864                 max_reqs, tbl->max_slots);
4865
4866         /* Does the newly negotiated max_reqs match the existing slot table? */
4867         if (max_reqs != tbl->max_slots) {
4868                 ret = -ENOMEM;
4869                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4870                               GFP_NOFS);
4871                 if (!new)
4872                         goto out;
4873                 ret = 0;
4874                 kfree(tbl->slots);
4875         }
4876         spin_lock(&tbl->slot_tbl_lock);
4877         if (new) {
4878                 tbl->slots = new;
4879                 tbl->max_slots = max_reqs;
4880         }
4881         for (i = 0; i < tbl->max_slots; ++i)
4882                 tbl->slots[i].seq_nr = ivalue;
4883         spin_unlock(&tbl->slot_tbl_lock);
4884         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4885                 tbl, tbl->slots, tbl->max_slots);
4886 out:
4887         dprintk("<-- %s: return %d\n", __func__, ret);
4888         return ret;
4889 }
4890
4891 /*
4892  * Reset the forechannel and backchannel slot tables
4893  */
4894 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4895 {
4896         int status;
4897
4898         status = nfs4_reset_slot_table(&session->fc_slot_table,
4899                         session->fc_attrs.max_reqs, 1);
4900         if (status)
4901                 return status;
4902
4903         status = nfs4_reset_slot_table(&session->bc_slot_table,
4904                         session->bc_attrs.max_reqs, 0);
4905         return status;
4906 }
4907
4908 /* Destroy the slot table */
4909 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4910 {
4911         if (session->fc_slot_table.slots != NULL) {
4912                 kfree(session->fc_slot_table.slots);
4913                 session->fc_slot_table.slots = NULL;
4914         }
4915         if (session->bc_slot_table.slots != NULL) {
4916                 kfree(session->bc_slot_table.slots);
4917                 session->bc_slot_table.slots = NULL;
4918         }
4919         return;
4920 }
4921
4922 /*
4923  * Initialize slot table
4924  */
4925 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4926                 int max_slots, int ivalue)
4927 {
4928         struct nfs4_slot *slot;
4929         int ret = -ENOMEM;
4930
4931         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4932
4933         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4934
4935         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4936         if (!slot)
4937                 goto out;
4938         ret = 0;
4939
4940         spin_lock(&tbl->slot_tbl_lock);
4941         tbl->max_slots = max_slots;
4942         tbl->slots = slot;
4943         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4944         spin_unlock(&tbl->slot_tbl_lock);
4945         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4946                 tbl, tbl->slots, tbl->max_slots);
4947 out:
4948         dprintk("<-- %s: return %d\n", __func__, ret);
4949         return ret;
4950 }
4951
4952 /*
4953  * Initialize the forechannel and backchannel tables
4954  */
4955 static int nfs4_init_slot_tables(struct nfs4_session *session)
4956 {
4957         struct nfs4_slot_table *tbl;
4958         int status = 0;
4959
4960         tbl = &session->fc_slot_table;
4961         if (tbl->slots == NULL) {
4962                 status = nfs4_init_slot_table(tbl,
4963                                 session->fc_attrs.max_reqs, 1);
4964                 if (status)
4965                         return status;
4966         }
4967
4968         tbl = &session->bc_slot_table;
4969         if (tbl->slots == NULL) {
4970                 status = nfs4_init_slot_table(tbl,
4971                                 session->bc_attrs.max_reqs, 0);
4972                 if (status)
4973                         nfs4_destroy_slot_tables(session);
4974         }
4975
4976         return status;
4977 }
4978
4979 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4980 {
4981         struct nfs4_session *session;
4982         struct nfs4_slot_table *tbl;
4983
4984         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
4985         if (!session)
4986                 return NULL;
4987
4988         tbl = &session->fc_slot_table;
4989         tbl->highest_used_slotid = -1;
4990         spin_lock_init(&tbl->slot_tbl_lock);
4991         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
4992         init_completion(&tbl->complete);
4993
4994         tbl = &session->bc_slot_table;
4995         tbl->highest_used_slotid = -1;
4996         spin_lock_init(&tbl->slot_tbl_lock);
4997         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
4998         init_completion(&tbl->complete);
4999
5000         session->session_state = 1<<NFS4_SESSION_INITING;
5001
5002         session->clp = clp;
5003         return session;
5004 }
5005
5006 void nfs4_destroy_session(struct nfs4_session *session)
5007 {
5008         nfs4_proc_destroy_session(session);
5009         dprintk("%s Destroy backchannel for xprt %p\n",
5010                 __func__, session->clp->cl_rpcclient->cl_xprt);
5011         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5012                                 NFS41_BC_MIN_CALLBACKS);
5013         nfs4_destroy_slot_tables(session);
5014         kfree(session);
5015 }
5016
5017 /*
5018  * Initialize the values to be used by the client in CREATE_SESSION
5019  * If nfs4_init_session set the fore channel request and response sizes,
5020  * use them.
5021  *
5022  * Set the back channel max_resp_sz_cached to zero to force the client to
5023  * always set csa_cachethis to FALSE because the current implementation
5024  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5025  */
5026 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5027 {
5028         struct nfs4_session *session = args->client->cl_session;
5029         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5030                      mxresp_sz = session->fc_attrs.max_resp_sz;
5031
5032         if (mxrqst_sz == 0)
5033                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5034         if (mxresp_sz == 0)
5035                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5036         /* Fore channel attributes */
5037         args->fc_attrs.headerpadsz = 0;
5038         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5039         args->fc_attrs.max_resp_sz = mxresp_sz;
5040         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5041         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5042
5043         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5044                 "max_ops=%u max_reqs=%u\n",
5045                 __func__,
5046                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5047                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5048
5049         /* Back channel attributes */
5050         args->bc_attrs.headerpadsz = 0;
5051         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5052         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5053         args->bc_attrs.max_resp_sz_cached = 0;
5054         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5055         args->bc_attrs.max_reqs = 1;
5056
5057         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5058                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5059                 __func__,
5060                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5061                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5062                 args->bc_attrs.max_reqs);
5063 }
5064
5065 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5066 {
5067         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5068         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5069
5070         if (rcvd->headerpadsz > sent->headerpadsz)
5071                 return -EINVAL;
5072         if (rcvd->max_resp_sz > sent->max_resp_sz)
5073                 return -EINVAL;
5074         /*
5075          * Our requested max_ops is the minimum we need; we're not
5076          * prepared to break up compounds into smaller pieces than that.
5077          * So, no point even trying to continue if the server won't
5078          * cooperate:
5079          */
5080         if (rcvd->max_ops < sent->max_ops)
5081                 return -EINVAL;
5082         if (rcvd->max_reqs == 0)
5083                 return -EINVAL;
5084         return 0;
5085 }
5086
5087 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5088 {
5089         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5090         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5091
5092         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5093                 return -EINVAL;
5094         if (rcvd->max_resp_sz < sent->max_resp_sz)
5095                 return -EINVAL;
5096         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5097                 return -EINVAL;
5098         /* These would render the backchannel useless: */
5099         if (rcvd->max_ops  == 0)
5100                 return -EINVAL;
5101         if (rcvd->max_reqs == 0)
5102                 return -EINVAL;
5103         return 0;
5104 }
5105
5106 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5107                                      struct nfs4_session *session)
5108 {
5109         int ret;
5110
5111         ret = nfs4_verify_fore_channel_attrs(args, session);
5112         if (ret)
5113                 return ret;
5114         return nfs4_verify_back_channel_attrs(args, session);
5115 }
5116
5117 static int _nfs4_proc_create_session(struct nfs_client *clp)
5118 {
5119         struct nfs4_session *session = clp->cl_session;
5120         struct nfs41_create_session_args args = {
5121                 .client = clp,
5122                 .cb_program = NFS4_CALLBACK,
5123         };
5124         struct nfs41_create_session_res res = {
5125                 .client = clp,
5126         };
5127         struct rpc_message msg = {
5128                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5129                 .rpc_argp = &args,
5130                 .rpc_resp = &res,
5131         };
5132         int status;
5133
5134         nfs4_init_channel_attrs(&args);
5135         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5136
5137         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5138
5139         if (!status)
5140                 /* Verify the session's negotiated channel_attrs values */
5141                 status = nfs4_verify_channel_attrs(&args, session);
5142         if (!status) {
5143                 /* Increment the clientid slot sequence id */
5144                 clp->cl_seqid++;
5145         }
5146
5147         return status;
5148 }
5149
5150 /*
5151  * Issues a CREATE_SESSION operation to the server.
5152  * It is the responsibility of the caller to verify the session is
5153  * expired before calling this routine.
5154  */
5155 int nfs4_proc_create_session(struct nfs_client *clp)
5156 {
5157         int status;
5158         unsigned *ptr;
5159         struct nfs4_session *session = clp->cl_session;
5160         long timeout = 0;
5161         int err;
5162
5163         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5164
5165         do {
5166                 status = _nfs4_proc_create_session(clp);
5167                 if (status == -NFS4ERR_DELAY) {
5168                         err = nfs4_delay(clp->cl_rpcclient, &timeout);
5169                         if (err)
5170                                 status = err;
5171                 }
5172         } while (status == -NFS4ERR_DELAY);
5173
5174         if (status)
5175                 goto out;
5176
5177         /* Init and reset the fore channel */
5178         status = nfs4_init_slot_tables(session);
5179         dprintk("slot table initialization returned %d\n", status);
5180         if (status)
5181                 goto out;
5182         status = nfs4_reset_slot_tables(session);
5183         dprintk("slot table reset returned %d\n", status);
5184         if (status)
5185                 goto out;
5186
5187         ptr = (unsigned *)&session->sess_id.data[0];
5188         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5189                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5190 out:
5191         dprintk("<-- %s\n", __func__);
5192         return status;
5193 }
5194
5195 /*
5196  * Issue the over-the-wire RPC DESTROY_SESSION.
5197  * The caller must serialize access to this routine.
5198  */
5199 int nfs4_proc_destroy_session(struct nfs4_session *session)
5200 {
5201         int status = 0;
5202         struct rpc_message msg;
5203
5204         dprintk("--> nfs4_proc_destroy_session\n");
5205
5206         /* session is still being setup */
5207         if (session->clp->cl_cons_state != NFS_CS_READY)
5208                 return status;
5209
5210         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5211         msg.rpc_argp = session;
5212         msg.rpc_resp = NULL;
5213         msg.rpc_cred = NULL;
5214         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5215
5216         if (status)
5217                 printk(KERN_WARNING
5218                         "Got error %d from the server on DESTROY_SESSION. "
5219                         "Session has been destroyed regardless...\n", status);
5220
5221         dprintk("<-- nfs4_proc_destroy_session\n");
5222         return status;
5223 }
5224
5225 int nfs4_init_session(struct nfs_server *server)
5226 {
5227         struct nfs_client *clp = server->nfs_client;
5228         struct nfs4_session *session;
5229         unsigned int rsize, wsize;
5230         int ret;
5231
5232         if (!nfs4_has_session(clp))
5233                 return 0;
5234
5235         session = clp->cl_session;
5236         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5237                 return 0;
5238
5239         rsize = server->rsize;
5240         if (rsize == 0)
5241                 rsize = NFS_MAX_FILE_IO_SIZE;
5242         wsize = server->wsize;
5243         if (wsize == 0)
5244                 wsize = NFS_MAX_FILE_IO_SIZE;
5245
5246         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5247         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5248
5249         ret = nfs4_recover_expired_lease(server);
5250         if (!ret)
5251                 ret = nfs4_check_client_ready(clp);
5252         return ret;
5253 }
5254
5255 int nfs4_init_ds_session(struct nfs_client *clp)
5256 {
5257         struct nfs4_session *session = clp->cl_session;
5258         int ret;
5259
5260         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5261                 return 0;
5262
5263         ret = nfs4_client_recover_expired_lease(clp);
5264         if (!ret)
5265                 /* Test for the DS role */
5266                 if (!is_ds_client(clp))
5267                         ret = -ENODEV;
5268         if (!ret)
5269                 ret = nfs4_check_client_ready(clp);
5270         return ret;
5271
5272 }
5273 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5274
5275
5276 /*
5277  * Renew the cl_session lease.
5278  */
5279 struct nfs4_sequence_data {
5280         struct nfs_client *clp;
5281         struct nfs4_sequence_args args;
5282         struct nfs4_sequence_res res;
5283 };
5284
5285 static void nfs41_sequence_release(void *data)
5286 {
5287         struct nfs4_sequence_data *calldata = data;
5288         struct nfs_client *clp = calldata->clp;
5289
5290         if (atomic_read(&clp->cl_count) > 1)
5291                 nfs4_schedule_state_renewal(clp);
5292         nfs_put_client(clp);
5293         kfree(calldata);
5294 }
5295
5296 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5297 {
5298         switch(task->tk_status) {
5299         case -NFS4ERR_DELAY:
5300                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5301                 return -EAGAIN;
5302         default:
5303                 nfs4_schedule_lease_recovery(clp);
5304         }
5305         return 0;
5306 }
5307
5308 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5309 {
5310         struct nfs4_sequence_data *calldata = data;
5311         struct nfs_client *clp = calldata->clp;
5312
5313         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5314                 return;
5315
5316         if (task->tk_status < 0) {
5317                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5318                 if (atomic_read(&clp->cl_count) == 1)
5319                         goto out;
5320
5321                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5322                         rpc_restart_call_prepare(task);
5323                         return;
5324                 }
5325         }
5326         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5327 out:
5328         dprintk("<-- %s\n", __func__);
5329 }
5330
5331 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5332 {
5333         struct nfs4_sequence_data *calldata = data;
5334         struct nfs_client *clp = calldata->clp;
5335         struct nfs4_sequence_args *args;
5336         struct nfs4_sequence_res *res;
5337
5338         args = task->tk_msg.rpc_argp;
5339         res = task->tk_msg.rpc_resp;
5340
5341         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5342                 return;
5343         rpc_call_start(task);
5344 }
5345
5346 static const struct rpc_call_ops nfs41_sequence_ops = {
5347         .rpc_call_done = nfs41_sequence_call_done,
5348         .rpc_call_prepare = nfs41_sequence_prepare,
5349         .rpc_release = nfs41_sequence_release,
5350 };
5351
5352 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5353 {
5354         struct nfs4_sequence_data *calldata;
5355         struct rpc_message msg = {
5356                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5357                 .rpc_cred = cred,
5358         };
5359         struct rpc_task_setup task_setup_data = {
5360                 .rpc_client = clp->cl_rpcclient,
5361                 .rpc_message = &msg,
5362                 .callback_ops = &nfs41_sequence_ops,
5363                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5364         };
5365
5366         if (!atomic_inc_not_zero(&clp->cl_count))
5367                 return ERR_PTR(-EIO);
5368         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5369         if (calldata == NULL) {
5370                 nfs_put_client(clp);
5371                 return ERR_PTR(-ENOMEM);
5372         }
5373         msg.rpc_argp = &calldata->args;
5374         msg.rpc_resp = &calldata->res;
5375         calldata->clp = clp;
5376         task_setup_data.callback_data = calldata;
5377
5378         return rpc_run_task(&task_setup_data);
5379 }
5380
5381 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5382 {
5383         struct rpc_task *task;
5384         int ret = 0;
5385
5386         task = _nfs41_proc_sequence(clp, cred);
5387         if (IS_ERR(task))
5388                 ret = PTR_ERR(task);
5389         else
5390                 rpc_put_task_async(task);
5391         dprintk("<-- %s status=%d\n", __func__, ret);
5392         return ret;
5393 }
5394
5395 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5396 {
5397         struct rpc_task *task;
5398         int ret;
5399
5400         task = _nfs41_proc_sequence(clp, cred);
5401         if (IS_ERR(task)) {
5402                 ret = PTR_ERR(task);
5403                 goto out;
5404         }
5405         ret = rpc_wait_for_completion_task(task);
5406         if (!ret) {
5407                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5408
5409                 if (task->tk_status == 0)
5410                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5411                 ret = task->tk_status;
5412         }
5413         rpc_put_task(task);
5414 out:
5415         dprintk("<-- %s status=%d\n", __func__, ret);
5416         return ret;
5417 }
5418
5419 struct nfs4_reclaim_complete_data {
5420         struct nfs_client *clp;
5421         struct nfs41_reclaim_complete_args arg;
5422         struct nfs41_reclaim_complete_res res;
5423 };
5424
5425 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5426 {
5427         struct nfs4_reclaim_complete_data *calldata = data;
5428
5429         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5430         if (nfs41_setup_sequence(calldata->clp->cl_session,
5431                                 &calldata->arg.seq_args,
5432                                 &calldata->res.seq_res, 0, task))
5433                 return;
5434
5435         rpc_call_start(task);
5436 }
5437
5438 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5439 {
5440         switch(task->tk_status) {
5441         case 0:
5442         case -NFS4ERR_COMPLETE_ALREADY:
5443         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5444                 break;
5445         case -NFS4ERR_DELAY:
5446                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5447                 return -EAGAIN;
5448         default:
5449                 nfs4_schedule_lease_recovery(clp);
5450         }
5451         return 0;
5452 }
5453
5454 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5455 {
5456         struct nfs4_reclaim_complete_data *calldata = data;
5457         struct nfs_client *clp = calldata->clp;
5458         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5459
5460         dprintk("--> %s\n", __func__);
5461         if (!nfs41_sequence_done(task, res))
5462                 return;
5463
5464         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5465                 rpc_restart_call_prepare(task);
5466                 return;
5467         }
5468         dprintk("<-- %s\n", __func__);
5469 }
5470
5471 static void nfs4_free_reclaim_complete_data(void *data)
5472 {
5473         struct nfs4_reclaim_complete_data *calldata = data;
5474
5475         kfree(calldata);
5476 }
5477
5478 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5479         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5480         .rpc_call_done = nfs4_reclaim_complete_done,
5481         .rpc_release = nfs4_free_reclaim_complete_data,
5482 };
5483
5484 /*
5485  * Issue a global reclaim complete.
5486  */
5487 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5488 {
5489         struct nfs4_reclaim_complete_data *calldata;
5490         struct rpc_task *task;
5491         struct rpc_message msg = {
5492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5493         };
5494         struct rpc_task_setup task_setup_data = {
5495                 .rpc_client = clp->cl_rpcclient,
5496                 .rpc_message = &msg,
5497                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5498                 .flags = RPC_TASK_ASYNC,
5499         };
5500         int status = -ENOMEM;
5501
5502         dprintk("--> %s\n", __func__);
5503         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5504         if (calldata == NULL)
5505                 goto out;
5506         calldata->clp = clp;
5507         calldata->arg.one_fs = 0;
5508
5509         msg.rpc_argp = &calldata->arg;
5510         msg.rpc_resp = &calldata->res;
5511         task_setup_data.callback_data = calldata;
5512         task = rpc_run_task(&task_setup_data);
5513         if (IS_ERR(task)) {
5514                 status = PTR_ERR(task);
5515                 goto out;
5516         }
5517         status = nfs4_wait_for_completion_rpc_task(task);
5518         if (status == 0)
5519                 status = task->tk_status;
5520         rpc_put_task(task);
5521         return 0;
5522 out:
5523         dprintk("<-- %s status=%d\n", __func__, status);
5524         return status;
5525 }
5526
5527 static void
5528 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5529 {
5530         struct nfs4_layoutget *lgp = calldata;
5531         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5532
5533         dprintk("--> %s\n", __func__);
5534         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5535          * right now covering the LAYOUTGET we are about to send.
5536          * However, that is not so catastrophic, and there seems
5537          * to be no way to prevent it completely.
5538          */
5539         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5540                                 &lgp->res.seq_res, 0, task))
5541                 return;
5542         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5543                                           NFS_I(lgp->args.inode)->layout,
5544                                           lgp->args.ctx->state)) {
5545                 rpc_exit(task, NFS4_OK);
5546                 return;
5547         }
5548         rpc_call_start(task);
5549 }
5550
5551 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5552 {
5553         struct nfs4_layoutget *lgp = calldata;
5554         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5555
5556         dprintk("--> %s\n", __func__);
5557
5558         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5559                 return;
5560
5561         switch (task->tk_status) {
5562         case 0:
5563                 break;
5564         case -NFS4ERR_LAYOUTTRYLATER:
5565         case -NFS4ERR_RECALLCONFLICT:
5566                 task->tk_status = -NFS4ERR_DELAY;
5567                 /* Fall through */
5568         default:
5569                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5570                         rpc_restart_call_prepare(task);
5571                         return;
5572                 }
5573         }
5574         dprintk("<-- %s\n", __func__);
5575 }
5576
5577 static void nfs4_layoutget_release(void *calldata)
5578 {
5579         struct nfs4_layoutget *lgp = calldata;
5580
5581         dprintk("--> %s\n", __func__);
5582         if (lgp->res.layout.buf != NULL)
5583                 free_page((unsigned long) lgp->res.layout.buf);
5584         put_nfs_open_context(lgp->args.ctx);
5585         kfree(calldata);
5586         dprintk("<-- %s\n", __func__);
5587 }
5588
5589 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5590         .rpc_call_prepare = nfs4_layoutget_prepare,
5591         .rpc_call_done = nfs4_layoutget_done,
5592         .rpc_release = nfs4_layoutget_release,
5593 };
5594
5595 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5596 {
5597         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5598         struct rpc_task *task;
5599         struct rpc_message msg = {
5600                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5601                 .rpc_argp = &lgp->args,
5602                 .rpc_resp = &lgp->res,
5603         };
5604         struct rpc_task_setup task_setup_data = {
5605                 .rpc_client = server->client,
5606                 .rpc_message = &msg,
5607                 .callback_ops = &nfs4_layoutget_call_ops,
5608                 .callback_data = lgp,
5609                 .flags = RPC_TASK_ASYNC,
5610         };
5611         int status = 0;
5612
5613         dprintk("--> %s\n", __func__);
5614
5615         lgp->res.layout.buf = (void *)__get_free_page(GFP_NOFS);
5616         if (lgp->res.layout.buf == NULL) {
5617                 nfs4_layoutget_release(lgp);
5618                 return -ENOMEM;
5619         }
5620
5621         lgp->res.seq_res.sr_slot = NULL;
5622         task = rpc_run_task(&task_setup_data);
5623         if (IS_ERR(task))
5624                 return PTR_ERR(task);
5625         status = nfs4_wait_for_completion_rpc_task(task);
5626         if (status == 0)
5627                 status = task->tk_status;
5628         if (status == 0)
5629                 status = pnfs_layout_process(lgp);
5630         rpc_put_task(task);
5631         dprintk("<-- %s status=%d\n", __func__, status);
5632         return status;
5633 }
5634
5635 static int
5636 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5637 {
5638         struct nfs4_getdeviceinfo_args args = {
5639                 .pdev = pdev,
5640         };
5641         struct nfs4_getdeviceinfo_res res = {
5642                 .pdev = pdev,
5643         };
5644         struct rpc_message msg = {
5645                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5646                 .rpc_argp = &args,
5647                 .rpc_resp = &res,
5648         };
5649         int status;
5650
5651         dprintk("--> %s\n", __func__);
5652         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5653         dprintk("<-- %s status=%d\n", __func__, status);
5654
5655         return status;
5656 }
5657
5658 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5659 {
5660         struct nfs4_exception exception = { };
5661         int err;
5662
5663         do {
5664                 err = nfs4_handle_exception(server,
5665                                         _nfs4_proc_getdeviceinfo(server, pdev),
5666                                         &exception);
5667         } while (exception.retry);
5668         return err;
5669 }
5670 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5671
5672 #endif /* CONFIG_NFS_V4_1 */
5673
5674 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5675         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5676         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5677         .recover_open   = nfs4_open_reclaim,
5678         .recover_lock   = nfs4_lock_reclaim,
5679         .establish_clid = nfs4_init_clientid,
5680         .get_clid_cred  = nfs4_get_setclientid_cred,
5681 };
5682
5683 #if defined(CONFIG_NFS_V4_1)
5684 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5685         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5686         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5687         .recover_open   = nfs4_open_reclaim,
5688         .recover_lock   = nfs4_lock_reclaim,
5689         .establish_clid = nfs41_init_clientid,
5690         .get_clid_cred  = nfs4_get_exchange_id_cred,
5691         .reclaim_complete = nfs41_proc_reclaim_complete,
5692 };
5693 #endif /* CONFIG_NFS_V4_1 */
5694
5695 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5696         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5697         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5698         .recover_open   = nfs4_open_expired,
5699         .recover_lock   = nfs4_lock_expired,
5700         .establish_clid = nfs4_init_clientid,
5701         .get_clid_cred  = nfs4_get_setclientid_cred,
5702 };
5703
5704 #if defined(CONFIG_NFS_V4_1)
5705 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5706         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5707         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5708         .recover_open   = nfs4_open_expired,
5709         .recover_lock   = nfs4_lock_expired,
5710         .establish_clid = nfs41_init_clientid,
5711         .get_clid_cred  = nfs4_get_exchange_id_cred,
5712 };
5713 #endif /* CONFIG_NFS_V4_1 */
5714
5715 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5716         .sched_state_renewal = nfs4_proc_async_renew,
5717         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5718         .renew_lease = nfs4_proc_renew,
5719 };
5720
5721 #if defined(CONFIG_NFS_V4_1)
5722 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5723         .sched_state_renewal = nfs41_proc_async_sequence,
5724         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5725         .renew_lease = nfs4_proc_sequence,
5726 };
5727 #endif
5728
5729 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5730         .minor_version = 0,
5731         .call_sync = _nfs4_call_sync,
5732         .validate_stateid = nfs4_validate_delegation_stateid,
5733         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5734         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5735         .state_renewal_ops = &nfs40_state_renewal_ops,
5736 };
5737
5738 #if defined(CONFIG_NFS_V4_1)
5739 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5740         .minor_version = 1,
5741         .call_sync = _nfs4_call_sync_session,
5742         .validate_stateid = nfs41_validate_delegation_stateid,
5743         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5744         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5745         .state_renewal_ops = &nfs41_state_renewal_ops,
5746 };
5747 #endif
5748
5749 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5750         [0] = &nfs_v4_0_minor_ops,
5751 #if defined(CONFIG_NFS_V4_1)
5752         [1] = &nfs_v4_1_minor_ops,
5753 #endif
5754 };
5755
5756 static const struct inode_operations nfs4_file_inode_operations = {
5757         .permission     = nfs_permission,
5758         .getattr        = nfs_getattr,
5759         .setattr        = nfs_setattr,
5760         .getxattr       = generic_getxattr,
5761         .setxattr       = generic_setxattr,
5762         .listxattr      = generic_listxattr,
5763         .removexattr    = generic_removexattr,
5764 };
5765
5766 const struct nfs_rpc_ops nfs_v4_clientops = {
5767         .version        = 4,                    /* protocol version */
5768         .dentry_ops     = &nfs4_dentry_operations,
5769         .dir_inode_ops  = &nfs4_dir_inode_operations,
5770         .file_inode_ops = &nfs4_file_inode_operations,
5771         .getroot        = nfs4_proc_get_root,
5772         .getattr        = nfs4_proc_getattr,
5773         .setattr        = nfs4_proc_setattr,
5774         .lookupfh       = nfs4_proc_lookupfh,
5775         .lookup         = nfs4_proc_lookup,
5776         .access         = nfs4_proc_access,
5777         .readlink       = nfs4_proc_readlink,
5778         .create         = nfs4_proc_create,
5779         .remove         = nfs4_proc_remove,
5780         .unlink_setup   = nfs4_proc_unlink_setup,
5781         .unlink_done    = nfs4_proc_unlink_done,
5782         .rename         = nfs4_proc_rename,
5783         .rename_setup   = nfs4_proc_rename_setup,
5784         .rename_done    = nfs4_proc_rename_done,
5785         .link           = nfs4_proc_link,
5786         .symlink        = nfs4_proc_symlink,
5787         .mkdir          = nfs4_proc_mkdir,
5788         .rmdir          = nfs4_proc_remove,
5789         .readdir        = nfs4_proc_readdir,
5790         .mknod          = nfs4_proc_mknod,
5791         .statfs         = nfs4_proc_statfs,
5792         .fsinfo         = nfs4_proc_fsinfo,
5793         .pathconf       = nfs4_proc_pathconf,
5794         .set_capabilities = nfs4_server_capabilities,
5795         .decode_dirent  = nfs4_decode_dirent,
5796         .read_setup     = nfs4_proc_read_setup,
5797         .read_done      = nfs4_read_done,
5798         .write_setup    = nfs4_proc_write_setup,
5799         .write_done     = nfs4_write_done,
5800         .commit_setup   = nfs4_proc_commit_setup,
5801         .commit_done    = nfs4_commit_done,
5802         .lock           = nfs4_proc_lock,
5803         .clear_acl_cache = nfs4_zap_acl_attr,
5804         .close_context  = nfs4_close_context,
5805         .open_context   = nfs4_atomic_open,
5806         .init_client    = nfs4_init_client,
5807         .secinfo        = nfs4_proc_secinfo,
5808 };
5809
5810 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5811         .prefix = XATTR_NAME_NFSV4_ACL,
5812         .list   = nfs4_xattr_list_nfs4_acl,
5813         .get    = nfs4_xattr_get_nfs4_acl,
5814         .set    = nfs4_xattr_set_nfs4_acl,
5815 };
5816
5817 const struct xattr_handler *nfs4_xattr_handlers[] = {
5818         &nfs4_xattr_nfs4_acl_handler,
5819         NULL
5820 };
5821
5822 /*
5823  * Local variables:
5824  *  c-basic-offset: 8
5825  * End:
5826  */