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/sunrpc/gss_api.h>
  45 #include <linux/nfs.h>
  46 #include <linux/nfs4.h>
  47 #include <linux/nfs_fs.h>
  48 #include <linux/nfs_page.h>
  49 #include <linux/nfs_mount.h>
  50 #include <linux/namei.h>
  51 #include <linux/mount.h>
  52 #include <linux/module.h>
  53 #include <linux/sunrpc/bc_xprt.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56
  57 #include "nfs4_fs.h"
  58 #include "delegation.h"
  59 #include "internal.h"
  60 #include "iostat.h"
  61 #include "callback.h"
  62 #include "pnfs.h"
  63
  64 #define NFSDBG_FACILITY         NFSDBG_PROC
  65
  66 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  67 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  68
  69 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  70
  71 struct nfs4_opendata;
  72 static int _nfs4_proc_open(struct nfs4_opendata *data);
  73 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  74 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  75 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  76 static int _nfs4_proc_lookup(struct rpc_clnt *client, struct inode *dir,
  77                              const struct qstr *name, struct nfs_fh *fhandle,
  78                              struct nfs_fattr *fattr);
  79 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  80 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  81                             struct nfs_fattr *fattr, struct iattr *sattr,
  82                             struct nfs4_state *state);
  83
  84 /* Prevent leaks of NFSv4 errors into userland */
  85 static int nfs4_map_errors(int err)
  86 {
  87         if (err >= -1000)
  88                 return err;
  89         switch (err) {
  90         case -NFS4ERR_RESOURCE:
  91                 return -EREMOTEIO;
  92         case -NFS4ERR_WRONGSEC:
  93                 return -EPERM;
  94         case -NFS4ERR_BADOWNER:
  95         case -NFS4ERR_BADNAME:
  96                 return -EINVAL;
  97         default:
  98                 dprintk("%s could not handle NFSv4 error %d\n",
  99                                 __func__, -err);
 100                 break;
 101         }
 102         return -EIO;
 103 }
 104
 105 /*
 106  * This is our standard bitmap for GETATTR requests.
 107  */
 108 const u32 nfs4_fattr_bitmap[2] = {
 109         FATTR4_WORD0_TYPE
 110         | FATTR4_WORD0_CHANGE
 111         | FATTR4_WORD0_SIZE
 112         | FATTR4_WORD0_FSID
 113         | FATTR4_WORD0_FILEID,
 114         FATTR4_WORD1_MODE
 115         | FATTR4_WORD1_NUMLINKS
 116         | FATTR4_WORD1_OWNER
 117         | FATTR4_WORD1_OWNER_GROUP
 118         | FATTR4_WORD1_RAWDEV
 119         | FATTR4_WORD1_SPACE_USED
 120         | FATTR4_WORD1_TIME_ACCESS
 121         | FATTR4_WORD1_TIME_METADATA
 122         | FATTR4_WORD1_TIME_MODIFY
 123 };
 124
 125 const u32 nfs4_statfs_bitmap[2] = {
 126         FATTR4_WORD0_FILES_AVAIL
 127         | FATTR4_WORD0_FILES_FREE
 128         | FATTR4_WORD0_FILES_TOTAL,
 129         FATTR4_WORD1_SPACE_AVAIL
 130         | FATTR4_WORD1_SPACE_FREE
 131         | FATTR4_WORD1_SPACE_TOTAL
 132 };
 133
 134 const u32 nfs4_pathconf_bitmap[2] = {
 135         FATTR4_WORD0_MAXLINK
 136         | FATTR4_WORD0_MAXNAME,
 137         0
 138 };
 139
 140 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 141                         | FATTR4_WORD0_MAXREAD
 142                         | FATTR4_WORD0_MAXWRITE
 143                         | FATTR4_WORD0_LEASE_TIME,
 144                         FATTR4_WORD1_TIME_DELTA
 145                         | FATTR4_WORD1_FS_LAYOUT_TYPES
 146 };
 147
 148 const u32 nfs4_fs_locations_bitmap[2] = {
 149         FATTR4_WORD0_TYPE
 150         | FATTR4_WORD0_CHANGE
 151         | FATTR4_WORD0_SIZE
 152         | FATTR4_WORD0_FSID
 153         | FATTR4_WORD0_FILEID
 154         | FATTR4_WORD0_FS_LOCATIONS,
 155         FATTR4_WORD1_MODE
 156         | FATTR4_WORD1_NUMLINKS
 157         | FATTR4_WORD1_OWNER
 158         | FATTR4_WORD1_OWNER_GROUP
 159         | FATTR4_WORD1_RAWDEV
 160         | FATTR4_WORD1_SPACE_USED
 161         | FATTR4_WORD1_TIME_ACCESS
 162         | FATTR4_WORD1_TIME_METADATA
 163         | FATTR4_WORD1_TIME_MODIFY
 164         | FATTR4_WORD1_MOUNTED_ON_FILEID
 165 };
 166
 167 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 168                 struct nfs4_readdir_arg *readdir)
 169 {
 170         __be32 *start, *p;
 171
 172         BUG_ON(readdir->count < 80);
 173         if (cookie > 2) {
 174                 readdir->cookie = cookie;
 175                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 176                 return;
 177         }
 178
 179         readdir->cookie = 0;
 180         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 181         if (cookie == 2)
 182                 return;
 183
 184         /*
 185          * NFSv4 servers do not return entries for '.' and '..'
 186          * Therefore, we fake these entries here.  We let '.'
 187          * have cookie 0 and '..' have cookie 1.  Note that
 188          * when talking to the server, we always send cookie 0
 189          * instead of 1 or 2.
 190          */
 191         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 192
 193         if (cookie == 0) {
 194                 *p++ = xdr_one;                                  /* next */
 195                 *p++ = xdr_zero;                   /* cookie, first word */
 196                 *p++ = xdr_one;                   /* cookie, second word */
 197                 *p++ = xdr_one;                             /* entry len */
 198                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 199                 p++;
 200                 *p++ = xdr_one;                         /* bitmap length */
 201                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 202                 *p++ = htonl(8);              /* attribute buffer length */
 203                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 204         }
 205
 206         *p++ = xdr_one;                                  /* next */
 207         *p++ = xdr_zero;                   /* cookie, first word */
 208         *p++ = xdr_two;                   /* cookie, second word */
 209         *p++ = xdr_two;                             /* entry len */
 210         memcpy(p, "..\0\0", 4);                         /* entry */
 211         p++;
 212         *p++ = xdr_one;                         /* bitmap length */
 213         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 214         *p++ = htonl(8);              /* attribute buffer length */
 215         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 216
 217         readdir->pgbase = (char *)p - (char *)start;
 218         readdir->count -= readdir->pgbase;
 219         kunmap_atomic(start, KM_USER0);
 220 }
 221
 222 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 223 {
 224         int res;
 225
 226         might_sleep();
 227
 228         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 229                         nfs_wait_bit_killable, TASK_KILLABLE);
 230         return res;
 231 }
 232
 233 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 234 {
 235         int res = 0;
 236
 237         might_sleep();
 238
 239         if (*timeout <= 0)
 240                 *timeout = NFS4_POLL_RETRY_MIN;
 241         if (*timeout > NFS4_POLL_RETRY_MAX)
 242                 *timeout = NFS4_POLL_RETRY_MAX;
 243         schedule_timeout_killable(*timeout);
 244         if (fatal_signal_pending(current))
 245                 res = -ERESTARTSYS;
 246         *timeout <<= 1;
 247         return res;
 248 }
 249
 250 /* This is the error handling routine for processes that are allowed
 251  * to sleep.
 252  */
 253 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 254 {
 255         struct nfs_client *clp = server->nfs_client;
 256         struct nfs4_state *state = exception->state;
 257         int ret = errorcode;
 258
 259         exception->retry = 0;
 260         switch(errorcode) {
 261                 case 0:
 262                         return 0;
 263                 case -NFS4ERR_ADMIN_REVOKED:
 264                 case -NFS4ERR_BAD_STATEID:
 265                 case -NFS4ERR_OPENMODE:
 266                         if (state == NULL)
 267                                 break;
 268                         nfs4_schedule_stateid_recovery(server, state);
 269                         goto wait_on_recovery;
 270                 case -NFS4ERR_STALE_STATEID:
 271                 case -NFS4ERR_STALE_CLIENTID:
 272                 case -NFS4ERR_EXPIRED:
 273                         nfs4_schedule_lease_recovery(clp);
 274                         goto wait_on_recovery;
 275 #if defined(CONFIG_NFS_V4_1)
 276                 case -NFS4ERR_BADSESSION:
 277                 case -NFS4ERR_BADSLOT:
 278                 case -NFS4ERR_BAD_HIGH_SLOT:
 279                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 280                 case -NFS4ERR_DEADSESSION:
 281                 case -NFS4ERR_SEQ_FALSE_RETRY:
 282                 case -NFS4ERR_SEQ_MISORDERED:
 283                         dprintk("%s ERROR: %d Reset session\n", __func__,
 284                                 errorcode);
 285                         nfs4_schedule_session_recovery(clp->cl_session);
 286                         exception->retry = 1;
 287                         break;
 288 #endif /* defined(CONFIG_NFS_V4_1) */
 289                 case -NFS4ERR_FILE_OPEN:
 290                         if (exception->timeout > HZ) {
 291                                 /* We have retried a decent amount, time to
 292                                  * fail
 293                                  */
 294                                 ret = -EBUSY;
 295                                 break;
 296                         }
 297                 case -NFS4ERR_GRACE:
 298                 case -NFS4ERR_DELAY:
 299                 case -EKEYEXPIRED:
 300                         ret = nfs4_delay(server->client, &exception->timeout);
 301                         if (ret != 0)
 302                                 break;
 303                 case -NFS4ERR_OLD_STATEID:
 304                         exception->retry = 1;
 305                         break;
 306                 case -NFS4ERR_BADOWNER:
 307                         /* The following works around a Linux server bug! */
 308                 case -NFS4ERR_BADNAME:
 309                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 310                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 311                                 exception->retry = 1;
 312                                 printk(KERN_WARNING "NFS: v4 server %s "
 313                                                 "does not accept raw "
 314                                                 "uid/gids. "
 315                                                 "Reenabling the idmapper.\n",
 316                                                 server->nfs_client->cl_hostname);
 317                         }
 318         }
 319         /* We failed to handle the error */
 320         return nfs4_map_errors(ret);
 321 wait_on_recovery:
 322         ret = nfs4_wait_clnt_recover(clp);
 323         if (ret == 0)
 324                 exception->retry = 1;
 325         return ret;
 326 }
 327
 328
 329 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 330 {
 331         spin_lock(&clp->cl_lock);
 332         if (time_before(clp->cl_last_renewal,timestamp))
 333                 clp->cl_last_renewal = timestamp;
 334         spin_unlock(&clp->cl_lock);
 335 }
 336
 337 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 338 {
 339         do_renew_lease(server->nfs_client, timestamp);
 340 }
 341
 342 #if defined(CONFIG_NFS_V4_1)
 343
 344 /*
 345  * nfs4_free_slot - free a slot and efficiently update slot table.
 346  *
 347  * freeing a slot is trivially done by clearing its respective bit
 348  * in the bitmap.
 349  * If the freed slotid equals highest_used_slotid we want to update it
 350  * so that the server would be able to size down the slot table if needed,
 351  * otherwise we know that the highest_used_slotid is still in use.
 352  * When updating highest_used_slotid there may be "holes" in the bitmap
 353  * so we need to scan down from highest_used_slotid to 0 looking for the now
 354  * highest slotid in use.
 355  * If none found, highest_used_slotid is set to -1.
 356  *
 357  * Must be called while holding tbl->slot_tbl_lock
 358  */
 359 static void
 360 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 361 {
 362         int free_slotid = free_slot - tbl->slots;
 363         int slotid = free_slotid;
 364
 365         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 366         /* clear used bit in bitmap */
 367         __clear_bit(slotid, tbl->used_slots);
 368
 369         /* update highest_used_slotid when it is freed */
 370         if (slotid == tbl->highest_used_slotid) {
 371                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 372                 if (slotid < tbl->max_slots)
 373                         tbl->highest_used_slotid = slotid;
 374                 else
 375                         tbl->highest_used_slotid = -1;
 376         }
 377         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 378                 free_slotid, tbl->highest_used_slotid);
 379 }
 380
 381 /*
 382  * Signal state manager thread if session fore channel is drained
 383  */
 384 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 385 {
 386         struct rpc_task *task;
 387
 388         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 389                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 390                 if (task)
 391                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 392                 return;
 393         }
 394
 395         if (ses->fc_slot_table.highest_used_slotid != -1)
 396                 return;
 397
 398         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 399         complete(&ses->fc_slot_table.complete);
 400 }
 401
 402 /*
 403  * Signal state manager thread if session back channel is drained
 404  */
 405 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 406 {
 407         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 408             ses->bc_slot_table.highest_used_slotid != -1)
 409                 return;
 410         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 411         complete(&ses->bc_slot_table.complete);
 412 }
 413
 414 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 415 {
 416         struct nfs4_slot_table *tbl;
 417
 418         tbl = &res->sr_session->fc_slot_table;
 419         if (!res->sr_slot) {
 420                 /* just wake up the next guy waiting since
 421                  * we may have not consumed a slot after all */
 422                 dprintk("%s: No slot\n", __func__);
 423                 return;
 424         }
 425
 426         spin_lock(&tbl->slot_tbl_lock);
 427         nfs4_free_slot(tbl, res->sr_slot);
 428         nfs4_check_drain_fc_complete(res->sr_session);
 429         spin_unlock(&tbl->slot_tbl_lock);
 430         res->sr_slot = NULL;
 431 }
 432
 433 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 434 {
 435         unsigned long timestamp;
 436         struct nfs_client *clp;
 437
 438         /*
 439          * sr_status remains 1 if an RPC level error occurred. The server
 440          * may or may not have processed the sequence operation..
 441          * Proceed as if the server received and processed the sequence
 442          * operation.
 443          */
 444         if (res->sr_status == 1)
 445                 res->sr_status = NFS_OK;
 446
 447         /* don't increment the sequence number if the task wasn't sent */
 448         if (!RPC_WAS_SENT(task))
 449                 goto out;
 450
 451         /* Check the SEQUENCE operation status */
 452         switch (res->sr_status) {
 453         case 0:
 454                 /* Update the slot's sequence and clientid lease timer */
 455                 ++res->sr_slot->seq_nr;
 456                 timestamp = res->sr_renewal_time;
 457                 clp = res->sr_session->clp;
 458                 do_renew_lease(clp, timestamp);
 459                 /* Check sequence flags */
 460                 if (res->sr_status_flags != 0)
 461                         nfs4_schedule_lease_recovery(clp);
 462                 break;
 463         case -NFS4ERR_DELAY:
 464                 /* The server detected a resend of the RPC call and
 465                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 466                  * of RFC5661.
 467                  */
 468                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 469                         __func__,
 470                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 471                         res->sr_slot->seq_nr);
 472                 goto out_retry;
 473         default:
 474                 /* Just update the slot sequence no. */
 475                 ++res->sr_slot->seq_nr;
 476         }
 477 out:
 478         /* The session may be reset by one of the error handlers. */
 479         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 480         nfs41_sequence_free_slot(res);
 481         return 1;
 482 out_retry:
 483         if (!rpc_restart_call(task))
 484                 goto out;
 485         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 486         return 0;
 487 }
 488
 489 static int nfs4_sequence_done(struct rpc_task *task,
 490                                struct nfs4_sequence_res *res)
 491 {
 492         if (res->sr_session == NULL)
 493                 return 1;
 494         return nfs41_sequence_done(task, res);
 495 }
 496
 497 /*
 498  * nfs4_find_slot - efficiently look for a free slot
 499  *
 500  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 501  * If found, we mark the slot as used, update the highest_used_slotid,
 502  * and respectively set up the sequence operation args.
 503  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 504  *
 505  * Note: must be called with under the slot_tbl_lock.
 506  */
 507 static u8
 508 nfs4_find_slot(struct nfs4_slot_table *tbl)
 509 {
 510         int slotid;
 511         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 512         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 513
 514         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 515                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 516                 tbl->max_slots);
 517         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 518         if (slotid >= tbl->max_slots)
 519                 goto out;
 520         __set_bit(slotid, tbl->used_slots);
 521         if (slotid > tbl->highest_used_slotid)
 522                 tbl->highest_used_slotid = slotid;
 523         ret_id = slotid;
 524 out:
 525         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 526                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 527         return ret_id;
 528 }
 529
 530 int nfs41_setup_sequence(struct nfs4_session *session,
 531                                 struct nfs4_sequence_args *args,
 532                                 struct nfs4_sequence_res *res,
 533                                 int cache_reply,
 534                                 struct rpc_task *task)
 535 {
 536         struct nfs4_slot *slot;
 537         struct nfs4_slot_table *tbl;
 538         u8 slotid;
 539
 540         dprintk("--> %s\n", __func__);
 541         /* slot already allocated? */
 542         if (res->sr_slot != NULL)
 543                 return 0;
 544
 545         tbl = &session->fc_slot_table;
 546
 547         spin_lock(&tbl->slot_tbl_lock);
 548         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 549             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 550                 /*
 551                  * The state manager will wait until the slot table is empty.
 552                  * Schedule the reset thread
 553                  */
 554                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 555                 spin_unlock(&tbl->slot_tbl_lock);
 556                 dprintk("%s Schedule Session Reset\n", __func__);
 557                 return -EAGAIN;
 558         }
 559
 560         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 561             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 562                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 563                 spin_unlock(&tbl->slot_tbl_lock);
 564                 dprintk("%s enforce FIFO order\n", __func__);
 565                 return -EAGAIN;
 566         }
 567
 568         slotid = nfs4_find_slot(tbl);
 569         if (slotid == NFS4_MAX_SLOT_TABLE) {
 570                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 571                 spin_unlock(&tbl->slot_tbl_lock);
 572                 dprintk("<-- %s: no free slots\n", __func__);
 573                 return -EAGAIN;
 574         }
 575         spin_unlock(&tbl->slot_tbl_lock);
 576
 577         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 578         slot = tbl->slots + slotid;
 579         args->sa_session = session;
 580         args->sa_slotid = slotid;
 581         args->sa_cache_this = cache_reply;
 582
 583         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 584
 585         res->sr_session = session;
 586         res->sr_slot = slot;
 587         res->sr_renewal_time = jiffies;
 588         res->sr_status_flags = 0;
 589         /*
 590          * sr_status is only set in decode_sequence, and so will remain
 591          * set to 1 if an rpc level failure occurs.
 592          */
 593         res->sr_status = 1;
 594         return 0;
 595 }
 596 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 597
 598 int nfs4_setup_sequence(const struct nfs_server *server,
 599                         struct nfs4_sequence_args *args,
 600                         struct nfs4_sequence_res *res,
 601                         int cache_reply,
 602                         struct rpc_task *task)
 603 {
 604         struct nfs4_session *session = nfs4_get_session(server);
 605         int ret = 0;
 606
 607         if (session == NULL) {
 608                 args->sa_session = NULL;
 609                 res->sr_session = NULL;
 610                 goto out;
 611         }
 612
 613         dprintk("--> %s clp %p session %p sr_slot %td\n",
 614                 __func__, session->clp, session, res->sr_slot ?
 615                         res->sr_slot - session->fc_slot_table.slots : -1);
 616
 617         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 618                                    task);
 619 out:
 620         dprintk("<-- %s status=%d\n", __func__, ret);
 621         return ret;
 622 }
 623
 624 struct nfs41_call_sync_data {
 625         const struct nfs_server *seq_server;
 626         struct nfs4_sequence_args *seq_args;
 627         struct nfs4_sequence_res *seq_res;
 628         int cache_reply;
 629 };
 630
 631 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 632 {
 633         struct nfs41_call_sync_data *data = calldata;
 634
 635         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 636
 637         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 638                                 data->seq_res, data->cache_reply, task))
 639                 return;
 640         rpc_call_start(task);
 641 }
 642
 643 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 644 {
 645         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 646         nfs41_call_sync_prepare(task, calldata);
 647 }
 648
 649 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 650 {
 651         struct nfs41_call_sync_data *data = calldata;
 652
 653         nfs41_sequence_done(task, data->seq_res);
 654 }
 655
 656 struct rpc_call_ops nfs41_call_sync_ops = {
 657         .rpc_call_prepare = nfs41_call_sync_prepare,
 658         .rpc_call_done = nfs41_call_sync_done,
 659 };
 660
 661 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 662         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 663         .rpc_call_done = nfs41_call_sync_done,
 664 };
 665
 666 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 667                                    struct nfs_server *server,
 668                                    struct rpc_message *msg,
 669                                    struct nfs4_sequence_args *args,
 670                                    struct nfs4_sequence_res *res,
 671                                    int cache_reply,
 672                                    int privileged)
 673 {
 674         int ret;
 675         struct rpc_task *task;
 676         struct nfs41_call_sync_data data = {
 677                 .seq_server = server,
 678                 .seq_args = args,
 679                 .seq_res = res,
 680                 .cache_reply = cache_reply,
 681         };
 682         struct rpc_task_setup task_setup = {
 683                 .rpc_client = clnt,
 684                 .rpc_message = msg,
 685                 .callback_ops = &nfs41_call_sync_ops,
 686                 .callback_data = &data
 687         };
 688
 689         res->sr_slot = NULL;
 690         if (privileged)
 691                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 692         task = rpc_run_task(&task_setup);
 693         if (IS_ERR(task))
 694                 ret = PTR_ERR(task);
 695         else {
 696                 ret = task->tk_status;
 697                 rpc_put_task(task);
 698         }
 699         return ret;
 700 }
 701
 702 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 703                             struct nfs_server *server,
 704                             struct rpc_message *msg,
 705                             struct nfs4_sequence_args *args,
 706                             struct nfs4_sequence_res *res,
 707                             int cache_reply)
 708 {
 709         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 710 }
 711
 712 #else
 713 static int nfs4_sequence_done(struct rpc_task *task,
 714                                struct nfs4_sequence_res *res)
 715 {
 716         return 1;
 717 }
 718 #endif /* CONFIG_NFS_V4_1 */
 719
 720 int _nfs4_call_sync(struct rpc_clnt *clnt,
 721                     struct nfs_server *server,
 722                     struct rpc_message *msg,
 723                     struct nfs4_sequence_args *args,
 724                     struct nfs4_sequence_res *res,
 725                     int cache_reply)
 726 {
 727         args->sa_session = res->sr_session = NULL;
 728         return rpc_call_sync(clnt, msg, 0);
 729 }
 730
 731 static inline
 732 int nfs4_call_sync(struct rpc_clnt *clnt,
 733                    struct nfs_server *server,
 734                    struct rpc_message *msg,
 735                    struct nfs4_sequence_args *args,
 736                    struct nfs4_sequence_res *res,
 737                    int cache_reply)
 738 {
 739         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 740                                                 args, res, cache_reply);
 741 }
 742
 743 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 744 {
 745         struct nfs_inode *nfsi = NFS_I(dir);
 746
 747         spin_lock(&dir->i_lock);
 748         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 749         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 750                 nfs_force_lookup_revalidate(dir);
 751         nfsi->change_attr = cinfo->after;
 752         spin_unlock(&dir->i_lock);
 753 }
 754
 755 struct nfs4_opendata {
 756         struct kref kref;
 757         struct nfs_openargs o_arg;
 758         struct nfs_openres o_res;
 759         struct nfs_open_confirmargs c_arg;
 760         struct nfs_open_confirmres c_res;
 761         struct nfs_fattr f_attr;
 762         struct nfs_fattr dir_attr;
 763         struct path path;
 764         struct dentry *dir;
 765         struct nfs4_state_owner *owner;
 766         struct nfs4_state *state;
 767         struct iattr attrs;
 768         unsigned long timestamp;
 769         unsigned int rpc_done : 1;
 770         int rpc_status;
 771         int cancelled;
 772 };
 773
 774
 775 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 776 {
 777         p->o_res.f_attr = &p->f_attr;
 778         p->o_res.dir_attr = &p->dir_attr;
 779         p->o_res.seqid = p->o_arg.seqid;
 780         p->c_res.seqid = p->c_arg.seqid;
 781         p->o_res.server = p->o_arg.server;
 782         nfs_fattr_init(&p->f_attr);
 783         nfs_fattr_init(&p->dir_attr);
 784 }
 785
 786 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 787                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 788                 const struct iattr *attrs,
 789                 gfp_t gfp_mask)
 790 {
 791         struct dentry *parent = dget_parent(path->dentry);
 792         struct inode *dir = parent->d_inode;
 793         struct nfs_server *server = NFS_SERVER(dir);
 794         struct nfs4_opendata *p;
 795
 796         p = kzalloc(sizeof(*p), gfp_mask);
 797         if (p == NULL)
 798                 goto err;
 799         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 800         if (p->o_arg.seqid == NULL)
 801                 goto err_free;
 802         path_get(path);
 803         p->path = *path;
 804         p->dir = parent;
 805         p->owner = sp;
 806         atomic_inc(&sp->so_count);
 807         p->o_arg.fh = NFS_FH(dir);
 808         p->o_arg.open_flags = flags;
 809         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 810         p->o_arg.clientid = server->nfs_client->cl_clientid;
 811         p->o_arg.id = sp->so_owner_id.id;
 812         p->o_arg.name = &p->path.dentry->d_name;
 813         p->o_arg.server = server;
 814         p->o_arg.bitmask = server->attr_bitmask;
 815         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 816         if (flags & O_CREAT) {
 817                 u32 *s;
 818
 819                 p->o_arg.u.attrs = &p->attrs;
 820                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 821                 s = (u32 *) p->o_arg.u.verifier.data;
 822                 s[0] = jiffies;
 823                 s[1] = current->pid;
 824         }
 825         p->c_arg.fh = &p->o_res.fh;
 826         p->c_arg.stateid = &p->o_res.stateid;
 827         p->c_arg.seqid = p->o_arg.seqid;
 828         nfs4_init_opendata_res(p);
 829         kref_init(&p->kref);
 830         return p;
 831 err_free:
 832         kfree(p);
 833 err:
 834         dput(parent);
 835         return NULL;
 836 }
 837
 838 static void nfs4_opendata_free(struct kref *kref)
 839 {
 840         struct nfs4_opendata *p = container_of(kref,
 841                         struct nfs4_opendata, kref);
 842
 843         nfs_free_seqid(p->o_arg.seqid);
 844         if (p->state != NULL)
 845                 nfs4_put_open_state(p->state);
 846         nfs4_put_state_owner(p->owner);
 847         dput(p->dir);
 848         path_put(&p->path);
 849         kfree(p);
 850 }
 851
 852 static void nfs4_opendata_put(struct nfs4_opendata *p)
 853 {
 854         if (p != NULL)
 855                 kref_put(&p->kref, nfs4_opendata_free);
 856 }
 857
 858 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 859 {
 860         int ret;
 861
 862         ret = rpc_wait_for_completion_task(task);
 863         return ret;
 864 }
 865
 866 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 867 {
 868         int ret = 0;
 869
 870         if (open_mode & O_EXCL)
 871                 goto out;
 872         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 873                 case FMODE_READ:
 874                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 875                                 && state->n_rdonly != 0;
 876                         break;
 877                 case FMODE_WRITE:
 878                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 879                                 && state->n_wronly != 0;
 880                         break;
 881                 case FMODE_READ|FMODE_WRITE:
 882                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 883                                 && state->n_rdwr != 0;
 884         }
 885 out:
 886         return ret;
 887 }
 888
 889 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 890 {
 891         if ((delegation->type & fmode) != fmode)
 892                 return 0;
 893         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 894                 return 0;
 895         nfs_mark_delegation_referenced(delegation);
 896         return 1;
 897 }
 898
 899 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 900 {
 901         switch (fmode) {
 902                 case FMODE_WRITE:
 903                         state->n_wronly++;
 904                         break;
 905                 case FMODE_READ:
 906                         state->n_rdonly++;
 907                         break;
 908                 case FMODE_READ|FMODE_WRITE:
 909                         state->n_rdwr++;
 910         }
 911         nfs4_state_set_mode_locked(state, state->state | fmode);
 912 }
 913
 914 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 915 {
 916         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 917                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 918         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 919         switch (fmode) {
 920                 case FMODE_READ:
 921                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 922                         break;
 923                 case FMODE_WRITE:
 924                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 925                         break;
 926                 case FMODE_READ|FMODE_WRITE:
 927                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 928         }
 929 }
 930
 931 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 932 {
 933         write_seqlock(&state->seqlock);
 934         nfs_set_open_stateid_locked(state, stateid, fmode);
 935         write_sequnlock(&state->seqlock);
 936 }
 937
 938 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 939 {
 940         /*
 941          * Protect the call to nfs4_state_set_mode_locked and
 942          * serialise the stateid update
 943          */
 944         write_seqlock(&state->seqlock);
 945         if (deleg_stateid != NULL) {
 946                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 947                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 948         }
 949         if (open_stateid != NULL)
 950                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 951         write_sequnlock(&state->seqlock);
 952         spin_lock(&state->owner->so_lock);
 953         update_open_stateflags(state, fmode);
 954         spin_unlock(&state->owner->so_lock);
 955 }
 956
 957 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 958 {
 959         struct nfs_inode *nfsi = NFS_I(state->inode);
 960         struct nfs_delegation *deleg_cur;
 961         int ret = 0;
 962
 963         fmode &= (FMODE_READ|FMODE_WRITE);
 964
 965         rcu_read_lock();
 966         deleg_cur = rcu_dereference(nfsi->delegation);
 967         if (deleg_cur == NULL)
 968                 goto no_delegation;
 969
 970         spin_lock(&deleg_cur->lock);
 971         if (nfsi->delegation != deleg_cur ||
 972             (deleg_cur->type & fmode) != fmode)
 973                 goto no_delegation_unlock;
 974
 975         if (delegation == NULL)
 976                 delegation = &deleg_cur->stateid;
 977         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 978                 goto no_delegation_unlock;
 979
 980         nfs_mark_delegation_referenced(deleg_cur);
 981         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 982         ret = 1;
 983 no_delegation_unlock:
 984         spin_unlock(&deleg_cur->lock);
 985 no_delegation:
 986         rcu_read_unlock();
 987
 988         if (!ret && open_stateid != NULL) {
 989                 __update_open_stateid(state, open_stateid, NULL, fmode);
 990                 ret = 1;
 991         }
 992
 993         return ret;
 994 }
 995
 996
 997 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 998 {
 999         struct nfs_delegation *delegation;
1000
1001         rcu_read_lock();
1002         delegation = rcu_dereference(NFS_I(inode)->delegation);
1003         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1004                 rcu_read_unlock();
1005                 return;
1006         }
1007         rcu_read_unlock();
1008         nfs_inode_return_delegation(inode);
1009 }
1010
1011 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1012 {
1013         struct nfs4_state *state = opendata->state;
1014         struct nfs_inode *nfsi = NFS_I(state->inode);
1015         struct nfs_delegation *delegation;
1016         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1017         fmode_t fmode = opendata->o_arg.fmode;
1018         nfs4_stateid stateid;
1019         int ret = -EAGAIN;
1020
1021         for (;;) {
1022                 if (can_open_cached(state, fmode, open_mode)) {
1023                         spin_lock(&state->owner->so_lock);
1024                         if (can_open_cached(state, fmode, open_mode)) {
1025                                 update_open_stateflags(state, fmode);
1026                                 spin_unlock(&state->owner->so_lock);
1027                                 goto out_return_state;
1028                         }
1029                         spin_unlock(&state->owner->so_lock);
1030                 }
1031                 rcu_read_lock();
1032                 delegation = rcu_dereference(nfsi->delegation);
1033                 if (delegation == NULL ||
1034                     !can_open_delegated(delegation, fmode)) {
1035                         rcu_read_unlock();
1036                         break;
1037                 }
1038                 /* Save the delegation */
1039                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1040                 rcu_read_unlock();
1041                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1042                 if (ret != 0)
1043                         goto out;
1044                 ret = -EAGAIN;
1045
1046                 /* Try to update the stateid using the delegation */
1047                 if (update_open_stateid(state, NULL, &stateid, fmode))
1048                         goto out_return_state;
1049         }
1050 out:
1051         return ERR_PTR(ret);
1052 out_return_state:
1053         atomic_inc(&state->count);
1054         return state;
1055 }
1056
1057 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1058 {
1059         struct inode *inode;
1060         struct nfs4_state *state = NULL;
1061         struct nfs_delegation *delegation;
1062         int ret;
1063
1064         if (!data->rpc_done) {
1065                 state = nfs4_try_open_cached(data);
1066                 goto out;
1067         }
1068
1069         ret = -EAGAIN;
1070         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1071                 goto err;
1072         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1073         ret = PTR_ERR(inode);
1074         if (IS_ERR(inode))
1075                 goto err;
1076         ret = -ENOMEM;
1077         state = nfs4_get_open_state(inode, data->owner);
1078         if (state == NULL)
1079                 goto err_put_inode;
1080         if (data->o_res.delegation_type != 0) {
1081                 int delegation_flags = 0;
1082
1083                 rcu_read_lock();
1084                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1085                 if (delegation)
1086                         delegation_flags = delegation->flags;
1087                 rcu_read_unlock();
1088                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1089                         nfs_inode_set_delegation(state->inode,
1090                                         data->owner->so_cred,
1091                                         &data->o_res);
1092                 else
1093                         nfs_inode_reclaim_delegation(state->inode,
1094                                         data->owner->so_cred,
1095                                         &data->o_res);
1096         }
1097
1098         update_open_stateid(state, &data->o_res.stateid, NULL,
1099                         data->o_arg.fmode);
1100         iput(inode);
1101 out:
1102         return state;
1103 err_put_inode:
1104         iput(inode);
1105 err:
1106         return ERR_PTR(ret);
1107 }
1108
1109 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1110 {
1111         struct nfs_inode *nfsi = NFS_I(state->inode);
1112         struct nfs_open_context *ctx;
1113
1114         spin_lock(&state->inode->i_lock);
1115         list_for_each_entry(ctx, &nfsi->open_files, list) {
1116                 if (ctx->state != state)
1117                         continue;
1118                 get_nfs_open_context(ctx);
1119                 spin_unlock(&state->inode->i_lock);
1120                 return ctx;
1121         }
1122         spin_unlock(&state->inode->i_lock);
1123         return ERR_PTR(-ENOENT);
1124 }
1125
1126 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1127 {
1128         struct nfs4_opendata *opendata;
1129
1130         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL, GFP_NOFS);
1131         if (opendata == NULL)
1132                 return ERR_PTR(-ENOMEM);
1133         opendata->state = state;
1134         atomic_inc(&state->count);
1135         return opendata;
1136 }
1137
1138 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1139 {
1140         struct nfs4_state *newstate;
1141         int ret;
1142
1143         opendata->o_arg.open_flags = 0;
1144         opendata->o_arg.fmode = fmode;
1145         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1146         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1147         nfs4_init_opendata_res(opendata);
1148         ret = _nfs4_recover_proc_open(opendata);
1149         if (ret != 0)
1150                 return ret;
1151         newstate = nfs4_opendata_to_nfs4_state(opendata);
1152         if (IS_ERR(newstate))
1153                 return PTR_ERR(newstate);
1154         nfs4_close_state(&opendata->path, newstate, fmode);
1155         *res = newstate;
1156         return 0;
1157 }
1158
1159 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1160 {
1161         struct nfs4_state *newstate;
1162         int ret;
1163
1164         /* memory barrier prior to reading state->n_* */
1165         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1166         smp_rmb();
1167         if (state->n_rdwr != 0) {
1168                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1169                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1170                 if (ret != 0)
1171                         return ret;
1172                 if (newstate != state)
1173                         return -ESTALE;
1174         }
1175         if (state->n_wronly != 0) {
1176                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1177                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1178                 if (ret != 0)
1179                         return ret;
1180                 if (newstate != state)
1181                         return -ESTALE;
1182         }
1183         if (state->n_rdonly != 0) {
1184                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1185                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1186                 if (ret != 0)
1187                         return ret;
1188                 if (newstate != state)
1189                         return -ESTALE;
1190         }
1191         /*
1192          * We may have performed cached opens for all three recoveries.
1193          * Check if we need to update the current stateid.
1194          */
1195         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1196             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1197                 write_seqlock(&state->seqlock);
1198                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1199                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1200                 write_sequnlock(&state->seqlock);
1201         }
1202         return 0;
1203 }
1204
1205 /*
1206  * OPEN_RECLAIM:
1207  *      reclaim state on the server after a reboot.
1208  */
1209 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1210 {
1211         struct nfs_delegation *delegation;
1212         struct nfs4_opendata *opendata;
1213         fmode_t delegation_type = 0;
1214         int status;
1215
1216         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1217         if (IS_ERR(opendata))
1218                 return PTR_ERR(opendata);
1219         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1220         opendata->o_arg.fh = NFS_FH(state->inode);
1221         rcu_read_lock();
1222         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1223         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1224                 delegation_type = delegation->type;
1225         rcu_read_unlock();
1226         opendata->o_arg.u.delegation_type = delegation_type;
1227         status = nfs4_open_recover(opendata, state);
1228         nfs4_opendata_put(opendata);
1229         return status;
1230 }
1231
1232 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1233 {
1234         struct nfs_server *server = NFS_SERVER(state->inode);
1235         struct nfs4_exception exception = { };
1236         int err;
1237         do {
1238                 err = _nfs4_do_open_reclaim(ctx, state);
1239                 if (err != -NFS4ERR_DELAY)
1240                         break;
1241                 nfs4_handle_exception(server, err, &exception);
1242         } while (exception.retry);
1243         return err;
1244 }
1245
1246 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1247 {
1248         struct nfs_open_context *ctx;
1249         int ret;
1250
1251         ctx = nfs4_state_find_open_context(state);
1252         if (IS_ERR(ctx))
1253                 return PTR_ERR(ctx);
1254         ret = nfs4_do_open_reclaim(ctx, state);
1255         put_nfs_open_context(ctx);
1256         return ret;
1257 }
1258
1259 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1260 {
1261         struct nfs4_opendata *opendata;
1262         int ret;
1263
1264         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1265         if (IS_ERR(opendata))
1266                 return PTR_ERR(opendata);
1267         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1268         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1269                         sizeof(opendata->o_arg.u.delegation.data));
1270         ret = nfs4_open_recover(opendata, state);
1271         nfs4_opendata_put(opendata);
1272         return ret;
1273 }
1274
1275 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1276 {
1277         struct nfs4_exception exception = { };
1278         struct nfs_server *server = NFS_SERVER(state->inode);
1279         int err;
1280         do {
1281                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1282                 switch (err) {
1283                         case 0:
1284                         case -ENOENT:
1285                         case -ESTALE:
1286                                 goto out;
1287                         case -NFS4ERR_BADSESSION:
1288                         case -NFS4ERR_BADSLOT:
1289                         case -NFS4ERR_BAD_HIGH_SLOT:
1290                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1291                         case -NFS4ERR_DEADSESSION:
1292                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1293                                 goto out;
1294                         case -NFS4ERR_STALE_CLIENTID:
1295                         case -NFS4ERR_STALE_STATEID:
1296                         case -NFS4ERR_EXPIRED:
1297                                 /* Don't recall a delegation if it was lost */
1298                                 nfs4_schedule_lease_recovery(server->nfs_client);
1299                                 goto out;
1300                         case -ERESTARTSYS:
1301                                 /*
1302                                  * The show must go on: exit, but mark the
1303                                  * stateid as needing recovery.
1304                                  */
1305                         case -NFS4ERR_ADMIN_REVOKED:
1306                         case -NFS4ERR_BAD_STATEID:
1307                                 nfs4_schedule_stateid_recovery(server, state);
1308                         case -EKEYEXPIRED:
1309                                 /*
1310                                  * User RPCSEC_GSS context has expired.
1311                                  * We cannot recover this stateid now, so
1312                                  * skip it and allow recovery thread to
1313                                  * proceed.
1314                                  */
1315                         case -ENOMEM:
1316                                 err = 0;
1317                                 goto out;
1318                 }
1319                 err = nfs4_handle_exception(server, err, &exception);
1320         } while (exception.retry);
1321 out:
1322         return err;
1323 }
1324
1325 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1326 {
1327         struct nfs4_opendata *data = calldata;
1328
1329         data->rpc_status = task->tk_status;
1330         if (data->rpc_status == 0) {
1331                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1332                                 sizeof(data->o_res.stateid.data));
1333                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1334                 renew_lease(data->o_res.server, data->timestamp);
1335                 data->rpc_done = 1;
1336         }
1337 }
1338
1339 static void nfs4_open_confirm_release(void *calldata)
1340 {
1341         struct nfs4_opendata *data = calldata;
1342         struct nfs4_state *state = NULL;
1343
1344         /* If this request hasn't been cancelled, do nothing */
1345         if (data->cancelled == 0)
1346                 goto out_free;
1347         /* In case of error, no cleanup! */
1348         if (!data->rpc_done)
1349                 goto out_free;
1350         state = nfs4_opendata_to_nfs4_state(data);
1351         if (!IS_ERR(state))
1352                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1353 out_free:
1354         nfs4_opendata_put(data);
1355 }
1356
1357 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1358         .rpc_call_done = nfs4_open_confirm_done,
1359         .rpc_release = nfs4_open_confirm_release,
1360 };
1361
1362 /*
1363  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1364  */
1365 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1366 {
1367         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1368         struct rpc_task *task;
1369         struct  rpc_message msg = {
1370                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1371                 .rpc_argp = &data->c_arg,
1372                 .rpc_resp = &data->c_res,
1373                 .rpc_cred = data->owner->so_cred,
1374         };
1375         struct rpc_task_setup task_setup_data = {
1376                 .rpc_client = server->client,
1377                 .rpc_message = &msg,
1378                 .callback_ops = &nfs4_open_confirm_ops,
1379                 .callback_data = data,
1380                 .workqueue = nfsiod_workqueue,
1381                 .flags = RPC_TASK_ASYNC,
1382         };
1383         int status;
1384
1385         kref_get(&data->kref);
1386         data->rpc_done = 0;
1387         data->rpc_status = 0;
1388         data->timestamp = jiffies;
1389         task = rpc_run_task(&task_setup_data);
1390         if (IS_ERR(task))
1391                 return PTR_ERR(task);
1392         status = nfs4_wait_for_completion_rpc_task(task);
1393         if (status != 0) {
1394                 data->cancelled = 1;
1395                 smp_wmb();
1396         } else
1397                 status = data->rpc_status;
1398         rpc_put_task(task);
1399         return status;
1400 }
1401
1402 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1403 {
1404         struct nfs4_opendata *data = calldata;
1405         struct nfs4_state_owner *sp = data->owner;
1406
1407         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1408                 return;
1409         /*
1410          * Check if we still need to send an OPEN call, or if we can use
1411          * a delegation instead.
1412          */
1413         if (data->state != NULL) {
1414                 struct nfs_delegation *delegation;
1415
1416                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1417                         goto out_no_action;
1418                 rcu_read_lock();
1419                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1420                 if (delegation != NULL &&
1421                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1422                         rcu_read_unlock();
1423                         goto out_no_action;
1424                 }
1425                 rcu_read_unlock();
1426         }
1427         /* Update sequence id. */
1428         data->o_arg.id = sp->so_owner_id.id;
1429         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1430         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1431                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1432                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1433         }
1434         data->timestamp = jiffies;
1435         if (nfs4_setup_sequence(data->o_arg.server,
1436                                 &data->o_arg.seq_args,
1437                                 &data->o_res.seq_res, 1, task))
1438                 return;
1439         rpc_call_start(task);
1440         return;
1441 out_no_action:
1442         task->tk_action = NULL;
1443
1444 }
1445
1446 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1447 {
1448         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1449         nfs4_open_prepare(task, calldata);
1450 }
1451
1452 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1453 {
1454         struct nfs4_opendata *data = calldata;
1455
1456         data->rpc_status = task->tk_status;
1457
1458         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1459                 return;
1460
1461         if (task->tk_status == 0) {
1462                 switch (data->o_res.f_attr->mode & S_IFMT) {
1463                         case S_IFREG:
1464                                 break;
1465                         case S_IFLNK:
1466                                 data->rpc_status = -ELOOP;
1467                                 break;
1468                         case S_IFDIR:
1469                                 data->rpc_status = -EISDIR;
1470                                 break;
1471                         default:
1472                                 data->rpc_status = -ENOTDIR;
1473                 }
1474                 renew_lease(data->o_res.server, data->timestamp);
1475                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1476                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1477         }
1478         data->rpc_done = 1;
1479 }
1480
1481 static void nfs4_open_release(void *calldata)
1482 {
1483         struct nfs4_opendata *data = calldata;
1484         struct nfs4_state *state = NULL;
1485
1486         /* If this request hasn't been cancelled, do nothing */
1487         if (data->cancelled == 0)
1488                 goto out_free;
1489         /* In case of error, no cleanup! */
1490         if (data->rpc_status != 0 || !data->rpc_done)
1491                 goto out_free;
1492         /* In case we need an open_confirm, no cleanup! */
1493         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1494                 goto out_free;
1495         state = nfs4_opendata_to_nfs4_state(data);
1496         if (!IS_ERR(state))
1497                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1498 out_free:
1499         nfs4_opendata_put(data);
1500 }
1501
1502 static const struct rpc_call_ops nfs4_open_ops = {
1503         .rpc_call_prepare = nfs4_open_prepare,
1504         .rpc_call_done = nfs4_open_done,
1505         .rpc_release = nfs4_open_release,
1506 };
1507
1508 static const struct rpc_call_ops nfs4_recover_open_ops = {
1509         .rpc_call_prepare = nfs4_recover_open_prepare,
1510         .rpc_call_done = nfs4_open_done,
1511         .rpc_release = nfs4_open_release,
1512 };
1513
1514 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1515 {
1516         struct inode *dir = data->dir->d_inode;
1517         struct nfs_server *server = NFS_SERVER(dir);
1518         struct nfs_openargs *o_arg = &data->o_arg;
1519         struct nfs_openres *o_res = &data->o_res;
1520         struct rpc_task *task;
1521         struct rpc_message msg = {
1522                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1523                 .rpc_argp = o_arg,
1524                 .rpc_resp = o_res,
1525                 .rpc_cred = data->owner->so_cred,
1526         };
1527         struct rpc_task_setup task_setup_data = {
1528                 .rpc_client = server->client,
1529                 .rpc_message = &msg,
1530                 .callback_ops = &nfs4_open_ops,
1531                 .callback_data = data,
1532                 .workqueue = nfsiod_workqueue,
1533                 .flags = RPC_TASK_ASYNC,
1534         };
1535         int status;
1536
1537         kref_get(&data->kref);
1538         data->rpc_done = 0;
1539         data->rpc_status = 0;
1540         data->cancelled = 0;
1541         if (isrecover)
1542                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1543         task = rpc_run_task(&task_setup_data);
1544         if (IS_ERR(task))
1545                 return PTR_ERR(task);
1546         status = nfs4_wait_for_completion_rpc_task(task);
1547         if (status != 0) {
1548                 data->cancelled = 1;
1549                 smp_wmb();
1550         } else
1551                 status = data->rpc_status;
1552         rpc_put_task(task);
1553
1554         return status;
1555 }
1556
1557 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1558 {
1559         struct inode *dir = data->dir->d_inode;
1560         struct nfs_openres *o_res = &data->o_res;
1561         int status;
1562
1563         status = nfs4_run_open_task(data, 1);
1564         if (status != 0 || !data->rpc_done)
1565                 return status;
1566
1567         nfs_refresh_inode(dir, o_res->dir_attr);
1568
1569         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1570                 status = _nfs4_proc_open_confirm(data);
1571                 if (status != 0)
1572                         return status;
1573         }
1574
1575         return status;
1576 }
1577
1578 /*
1579  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1580  */
1581 static int _nfs4_proc_open(struct nfs4_opendata *data)
1582 {
1583         struct inode *dir = data->dir->d_inode;
1584         struct nfs_server *server = NFS_SERVER(dir);
1585         struct nfs_openargs *o_arg = &data->o_arg;
1586         struct nfs_openres *o_res = &data->o_res;
1587         int status;
1588
1589         status = nfs4_run_open_task(data, 0);
1590         if (status != 0 || !data->rpc_done)
1591                 return status;
1592
1593         if (o_arg->open_flags & O_CREAT) {
1594                 update_changeattr(dir, &o_res->cinfo);
1595                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1596         } else
1597                 nfs_refresh_inode(dir, o_res->dir_attr);
1598         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1599                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1600         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1601                 status = _nfs4_proc_open_confirm(data);
1602                 if (status != 0)
1603                         return status;
1604         }
1605         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1606                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1607         return 0;
1608 }
1609
1610 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1611 {
1612         unsigned int loop;
1613         int ret;
1614
1615         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1616                 ret = nfs4_wait_clnt_recover(clp);
1617                 if (ret != 0)
1618                         break;
1619                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1620                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1621                         break;
1622                 nfs4_schedule_state_manager(clp);
1623                 ret = -EIO;
1624         }
1625         return ret;
1626 }
1627
1628 static int nfs4_recover_expired_lease(struct nfs_server *server)
1629 {
1630         return nfs4_client_recover_expired_lease(server->nfs_client);
1631 }
1632
1633 /*
1634  * OPEN_EXPIRED:
1635  *      reclaim state on the server after a network partition.
1636  *      Assumes caller holds the appropriate lock
1637  */
1638 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1639 {
1640         struct nfs4_opendata *opendata;
1641         int ret;
1642
1643         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1644         if (IS_ERR(opendata))
1645                 return PTR_ERR(opendata);
1646         ret = nfs4_open_recover(opendata, state);
1647         if (ret == -ESTALE)
1648                 d_drop(ctx->path.dentry);
1649         nfs4_opendata_put(opendata);
1650         return ret;
1651 }
1652
1653 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1654 {
1655         struct nfs_server *server = NFS_SERVER(state->inode);
1656         struct nfs4_exception exception = { };
1657         int err;
1658
1659         do {
1660                 err = _nfs4_open_expired(ctx, state);
1661                 switch (err) {
1662                 default:
1663                         goto out;
1664                 case -NFS4ERR_GRACE:
1665                 case -NFS4ERR_DELAY:
1666                         nfs4_handle_exception(server, err, &exception);
1667                         err = 0;
1668                 }
1669         } while (exception.retry);
1670 out:
1671         return err;
1672 }
1673
1674 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1675 {
1676         struct nfs_open_context *ctx;
1677         int ret;
1678
1679         ctx = nfs4_state_find_open_context(state);
1680         if (IS_ERR(ctx))
1681                 return PTR_ERR(ctx);
1682         ret = nfs4_do_open_expired(ctx, state);
1683         put_nfs_open_context(ctx);
1684         return ret;
1685 }
1686
1687 /*
1688  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1689  * fields corresponding to attributes that were used to store the verifier.
1690  * Make sure we clobber those fields in the later setattr call
1691  */
1692 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1693 {
1694         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1695             !(sattr->ia_valid & ATTR_ATIME_SET))
1696                 sattr->ia_valid |= ATTR_ATIME;
1697
1698         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1699             !(sattr->ia_valid & ATTR_MTIME_SET))
1700                 sattr->ia_valid |= ATTR_MTIME;
1701 }
1702
1703 /*
1704  * Returns a referenced nfs4_state
1705  */
1706 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)
1707 {
1708         struct nfs4_state_owner  *sp;
1709         struct nfs4_state     *state = NULL;
1710         struct nfs_server       *server = NFS_SERVER(dir);
1711         struct nfs4_opendata *opendata;
1712         int status;
1713
1714         /* Protect against reboot recovery conflicts */
1715         status = -ENOMEM;
1716         if (!(sp = nfs4_get_state_owner(server, cred))) {
1717                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1718                 goto out_err;
1719         }
1720         status = nfs4_recover_expired_lease(server);
1721         if (status != 0)
1722                 goto err_put_state_owner;
1723         if (path->dentry->d_inode != NULL)
1724                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1725         status = -ENOMEM;
1726         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr, GFP_KERNEL);
1727         if (opendata == NULL)
1728                 goto err_put_state_owner;
1729
1730         if (path->dentry->d_inode != NULL)
1731                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1732
1733         status = _nfs4_proc_open(opendata);
1734         if (status != 0)
1735                 goto err_opendata_put;
1736
1737         state = nfs4_opendata_to_nfs4_state(opendata);
1738         status = PTR_ERR(state);
1739         if (IS_ERR(state))
1740                 goto err_opendata_put;
1741         if (server->caps & NFS_CAP_POSIX_LOCK)
1742                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1743
1744         if (opendata->o_arg.open_flags & O_EXCL) {
1745                 nfs4_exclusive_attrset(opendata, sattr);
1746
1747                 nfs_fattr_init(opendata->o_res.f_attr);
1748                 status = nfs4_do_setattr(state->inode, cred,
1749                                 opendata->o_res.f_attr, sattr,
1750                                 state);
1751                 if (status == 0)
1752                         nfs_setattr_update_inode(state->inode, sattr);
1753                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1754         }
1755         nfs4_opendata_put(opendata);
1756         nfs4_put_state_owner(sp);
1757         *res = state;
1758         return 0;
1759 err_opendata_put:
1760         nfs4_opendata_put(opendata);
1761 err_put_state_owner:
1762         nfs4_put_state_owner(sp);
1763 out_err:
1764         *res = NULL;
1765         return status;
1766 }
1767
1768
1769 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)
1770 {
1771         struct nfs4_exception exception = { };
1772         struct nfs4_state *res;
1773         int status;
1774
1775         do {
1776                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1777                 if (status == 0)
1778                         break;
1779                 /* NOTE: BAD_SEQID means the server and client disagree about the
1780                  * book-keeping w.r.t. state-changing operations
1781                  * (OPEN/CLOSE/LOCK/LOCKU...)
1782                  * It is actually a sign of a bug on the client or on the server.
1783                  *
1784                  * If we receive a BAD_SEQID error in the particular case of
1785                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1786                  * have unhashed the old state_owner for us, and that we can
1787                  * therefore safely retry using a new one. We should still warn
1788                  * the user though...
1789                  */
1790                 if (status == -NFS4ERR_BAD_SEQID) {
1791                         printk(KERN_WARNING "NFS: v4 server %s "
1792                                         " returned a bad sequence-id error!\n",
1793                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1794                         exception.retry = 1;
1795                         continue;
1796                 }
1797                 /*
1798                  * BAD_STATEID on OPEN means that the server cancelled our
1799                  * state before it received the OPEN_CONFIRM.
1800                  * Recover by retrying the request as per the discussion
1801                  * on Page 181 of RFC3530.
1802                  */
1803                 if (status == -NFS4ERR_BAD_STATEID) {
1804                         exception.retry = 1;
1805                         continue;
1806                 }
1807                 if (status == -EAGAIN) {
1808                         /* We must have found a delegation */
1809                         exception.retry = 1;
1810                         continue;
1811                 }
1812                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1813                                         status, &exception));
1814         } while (exception.retry);
1815         return res;
1816 }
1817
1818 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1819                             struct nfs_fattr *fattr, struct iattr *sattr,
1820                             struct nfs4_state *state)
1821 {
1822         struct nfs_server *server = NFS_SERVER(inode);
1823         struct nfs_setattrargs  arg = {
1824                 .fh             = NFS_FH(inode),
1825                 .iap            = sattr,
1826                 .server         = server,
1827                 .bitmask = server->attr_bitmask,
1828         };
1829         struct nfs_setattrres  res = {
1830                 .fattr          = fattr,
1831                 .server         = server,
1832         };
1833         struct rpc_message msg = {
1834                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1835                 .rpc_argp       = &arg,
1836                 .rpc_resp       = &res,
1837                 .rpc_cred       = cred,
1838         };
1839         unsigned long timestamp = jiffies;
1840         int status;
1841
1842         nfs_fattr_init(fattr);
1843
1844         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1845                 /* Use that stateid */
1846         } else if (state != NULL) {
1847                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1848         } else
1849                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1850
1851         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1852         if (status == 0 && state != NULL)
1853                 renew_lease(server, timestamp);
1854         return status;
1855 }
1856
1857 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1858                            struct nfs_fattr *fattr, struct iattr *sattr,
1859                            struct nfs4_state *state)
1860 {
1861         struct nfs_server *server = NFS_SERVER(inode);
1862         struct nfs4_exception exception = { };
1863         int err;
1864         do {
1865                 err = nfs4_handle_exception(server,
1866                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1867                                 &exception);
1868         } while (exception.retry);
1869         return err;
1870 }
1871
1872 struct nfs4_closedata {
1873         struct path path;
1874         struct inode *inode;
1875         struct nfs4_state *state;
1876         struct nfs_closeargs arg;
1877         struct nfs_closeres res;
1878         struct nfs_fattr fattr;
1879         unsigned long timestamp;
1880         bool roc;
1881         u32 roc_barrier;
1882 };
1883
1884 static void nfs4_free_closedata(void *data)
1885 {
1886         struct nfs4_closedata *calldata = data;
1887         struct nfs4_state_owner *sp = calldata->state->owner;
1888
1889         if (calldata->roc)
1890                 pnfs_roc_release(calldata->state->inode);
1891         nfs4_put_open_state(calldata->state);
1892         nfs_free_seqid(calldata->arg.seqid);
1893         nfs4_put_state_owner(sp);
1894         path_put(&calldata->path);
1895         kfree(calldata);
1896 }
1897
1898 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1899                 fmode_t fmode)
1900 {
1901         spin_lock(&state->owner->so_lock);
1902         if (!(fmode & FMODE_READ))
1903                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1904         if (!(fmode & FMODE_WRITE))
1905                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1906         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1907         spin_unlock(&state->owner->so_lock);
1908 }
1909
1910 static void nfs4_close_done(struct rpc_task *task, void *data)
1911 {
1912         struct nfs4_closedata *calldata = data;
1913         struct nfs4_state *state = calldata->state;
1914         struct nfs_server *server = NFS_SERVER(calldata->inode);
1915
1916         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1917                 return;
1918         /* hmm. we are done with the inode, and in the process of freeing
1919          * the state_owner. we keep this around to process errors
1920          */
1921         switch (task->tk_status) {
1922                 case 0:
1923                         if (calldata->roc)
1924                                 pnfs_roc_set_barrier(state->inode,
1925                                                      calldata->roc_barrier);
1926                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1927                         renew_lease(server, calldata->timestamp);
1928                         nfs4_close_clear_stateid_flags(state,
1929                                         calldata->arg.fmode);
1930                         break;
1931                 case -NFS4ERR_STALE_STATEID:
1932                 case -NFS4ERR_OLD_STATEID:
1933                 case -NFS4ERR_BAD_STATEID:
1934                 case -NFS4ERR_EXPIRED:
1935                         if (calldata->arg.fmode == 0)
1936                                 break;
1937                 default:
1938                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1939                                 rpc_restart_call_prepare(task);
1940         }
1941         nfs_release_seqid(calldata->arg.seqid);
1942         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1943 }
1944
1945 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1946 {
1947         struct nfs4_closedata *calldata = data;
1948         struct nfs4_state *state = calldata->state;
1949         int call_close = 0;
1950
1951         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1952                 return;
1953
1954         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1955         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1956         spin_lock(&state->owner->so_lock);
1957         /* Calculate the change in open mode */
1958         if (state->n_rdwr == 0) {
1959                 if (state->n_rdonly == 0) {
1960                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1961                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1962                         calldata->arg.fmode &= ~FMODE_READ;
1963                 }
1964                 if (state->n_wronly == 0) {
1965                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1966                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1967                         calldata->arg.fmode &= ~FMODE_WRITE;
1968                 }
1969         }
1970         spin_unlock(&state->owner->so_lock);
1971
1972         if (!call_close) {
1973                 /* Note: exit _without_ calling nfs4_close_done */
1974                 task->tk_action = NULL;
1975                 return;
1976         }
1977
1978         if (calldata->arg.fmode == 0) {
1979                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1980                 if (calldata->roc &&
1981                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
1982                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
1983                                      task, NULL);
1984                         return;
1985                 }
1986         }
1987
1988         nfs_fattr_init(calldata->res.fattr);
1989         calldata->timestamp = jiffies;
1990         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
1991                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1992                                 1, task))
1993                 return;
1994         rpc_call_start(task);
1995 }
1996
1997 static const struct rpc_call_ops nfs4_close_ops = {
1998         .rpc_call_prepare = nfs4_close_prepare,
1999         .rpc_call_done = nfs4_close_done,
2000         .rpc_release = nfs4_free_closedata,
2001 };
2002
2003 /*
2004  * It is possible for data to be read/written from a mem-mapped file
2005  * after the sys_close call (which hits the vfs layer as a flush).
2006  * This means that we can't safely call nfsv4 close on a file until
2007  * the inode is cleared. This in turn means that we are not good
2008  * NFSv4 citizens - we do not indicate to the server to update the file's
2009  * share state even when we are done with one of the three share
2010  * stateid's in the inode.
2011  *
2012  * NOTE: Caller must be holding the sp->so_owner semaphore!
2013  */
2014 int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2015 {
2016         struct nfs_server *server = NFS_SERVER(state->inode);
2017         struct nfs4_closedata *calldata;
2018         struct nfs4_state_owner *sp = state->owner;
2019         struct rpc_task *task;
2020         struct rpc_message msg = {
2021                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2022                 .rpc_cred = state->owner->so_cred,
2023         };
2024         struct rpc_task_setup task_setup_data = {
2025                 .rpc_client = server->client,
2026                 .rpc_message = &msg,
2027                 .callback_ops = &nfs4_close_ops,
2028                 .workqueue = nfsiod_workqueue,
2029                 .flags = RPC_TASK_ASYNC,
2030         };
2031         int status = -ENOMEM;
2032
2033         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2034         if (calldata == NULL)
2035                 goto out;
2036         calldata->inode = state->inode;
2037         calldata->state = state;
2038         calldata->arg.fh = NFS_FH(state->inode);
2039         calldata->arg.stateid = &state->open_stateid;
2040         /* Serialization for the sequence id */
2041         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2042         if (calldata->arg.seqid == NULL)
2043                 goto out_free_calldata;
2044         calldata->arg.fmode = 0;
2045         calldata->arg.bitmask = server->cache_consistency_bitmask;
2046         calldata->res.fattr = &calldata->fattr;
2047         calldata->res.seqid = calldata->arg.seqid;
2048         calldata->res.server = server;
2049         calldata->roc = roc;
2050         path_get(path);
2051         calldata->path = *path;
2052
2053         msg.rpc_argp = &calldata->arg;
2054         msg.rpc_resp = &calldata->res;
2055         task_setup_data.callback_data = calldata;
2056         task = rpc_run_task(&task_setup_data);
2057         if (IS_ERR(task))
2058                 return PTR_ERR(task);
2059         status = 0;
2060         if (wait)
2061                 status = rpc_wait_for_completion_task(task);
2062         rpc_put_task(task);
2063         return status;
2064 out_free_calldata:
2065         kfree(calldata);
2066 out:
2067         if (roc)
2068                 pnfs_roc_release(state->inode);
2069         nfs4_put_open_state(state);
2070         nfs4_put_state_owner(sp);
2071         return status;
2072 }
2073
2074 static struct inode *
2075 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2076 {
2077         struct nfs4_state *state;
2078
2079         /* Protect against concurrent sillydeletes */
2080         state = nfs4_do_open(dir, &ctx->path, ctx->mode, open_flags, attr, ctx->cred);
2081         if (IS_ERR(state))
2082                 return ERR_CAST(state);
2083         ctx->state = state;
2084         return igrab(state->inode);
2085 }
2086
2087 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2088 {
2089         if (ctx->state == NULL)
2090                 return;
2091         if (is_sync)
2092                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2093         else
2094                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2095 }
2096
2097 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2098 {
2099         struct nfs4_server_caps_arg args = {
2100                 .fhandle = fhandle,
2101         };
2102         struct nfs4_server_caps_res res = {};
2103         struct rpc_message msg = {
2104                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2105                 .rpc_argp = &args,
2106                 .rpc_resp = &res,
2107         };
2108         int status;
2109
2110         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2111         if (status == 0) {
2112                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2113                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2114                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2115                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2116                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2117                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2118                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2119                         server->caps |= NFS_CAP_ACLS;
2120                 if (res.has_links != 0)
2121                         server->caps |= NFS_CAP_HARDLINKS;
2122                 if (res.has_symlinks != 0)
2123                         server->caps |= NFS_CAP_SYMLINKS;
2124                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2125                         server->caps |= NFS_CAP_FILEID;
2126                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2127                         server->caps |= NFS_CAP_MODE;
2128                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2129                         server->caps |= NFS_CAP_NLINK;
2130                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2131                         server->caps |= NFS_CAP_OWNER;
2132                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2133                         server->caps |= NFS_CAP_OWNER_GROUP;
2134                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2135                         server->caps |= NFS_CAP_ATIME;
2136                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2137                         server->caps |= NFS_CAP_CTIME;
2138                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2139                         server->caps |= NFS_CAP_MTIME;
2140
2141                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2142                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2143                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2144                 server->acl_bitmask = res.acl_bitmask;
2145         }
2146
2147         return status;
2148 }
2149
2150 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2151 {
2152         struct nfs4_exception exception = { };
2153         int err;
2154         do {
2155                 err = nfs4_handle_exception(server,
2156                                 _nfs4_server_capabilities(server, fhandle),
2157                                 &exception);
2158         } while (exception.retry);
2159         return err;
2160 }
2161
2162 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2163                 struct nfs_fsinfo *info)
2164 {
2165         struct nfs4_lookup_root_arg args = {
2166                 .bitmask = nfs4_fattr_bitmap,
2167         };
2168         struct nfs4_lookup_res res = {
2169                 .server = server,
2170                 .fattr = info->fattr,
2171                 .fh = fhandle,
2172         };
2173         struct rpc_message msg = {
2174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2175                 .rpc_argp = &args,
2176                 .rpc_resp = &res,
2177         };
2178
2179         nfs_fattr_init(info->fattr);
2180         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2181 }
2182
2183 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2184                 struct nfs_fsinfo *info)
2185 {
2186         struct nfs4_exception exception = { };
2187         int err;
2188         do {
2189                 err = nfs4_handle_exception(server,
2190                                 _nfs4_lookup_root(server, fhandle, info),
2191                                 &exception);
2192         } while (exception.retry);
2193         return err;
2194 }
2195
2196 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2197                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2198 {
2199         struct rpc_auth *auth;
2200         int ret;
2201
2202         auth = rpcauth_create(flavor, server->client);
2203         if (!auth) {
2204                 ret = -EIO;
2205                 goto out;
2206         }
2207         ret = nfs4_lookup_root(server, fhandle, info);
2208 out:
2209         return ret;
2210 }
2211
2212 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2213                               struct nfs_fsinfo *info)
2214 {
2215         int i, len, status = 0;
2216         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2217
2218         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2219         flav_array[len] = RPC_AUTH_NULL;
2220         len += 1;
2221
2222         for (i = 0; i < len; i++) {
2223                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2224                 if (status == -EPERM || status == -EACCES)
2225                         continue;
2226                 break;
2227         }
2228         return status;
2229 }
2230
2231 /*
2232  * get the file handle for the "/" directory on the server
2233  */
2234 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2235                               struct nfs_fsinfo *info)
2236 {
2237         int status = nfs4_lookup_root(server, fhandle, info);
2238         if ((status == -EPERM) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2239                 status = nfs4_find_root_sec(server, fhandle, info);
2240         if (status == 0)
2241                 status = nfs4_server_capabilities(server, fhandle);
2242         if (status == 0)
2243                 status = nfs4_do_fsinfo(server, fhandle, info);
2244         return nfs4_map_errors(status);
2245 }
2246
2247 /*
2248  * Get locations and (maybe) other attributes of a referral.
2249  * Note that we'll actually follow the referral later when
2250  * we detect fsid mismatch in inode revalidation
2251  */
2252 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2253 {
2254         int status = -ENOMEM;
2255         struct page *page = NULL;
2256         struct nfs4_fs_locations *locations = NULL;
2257
2258         page = alloc_page(GFP_KERNEL);
2259         if (page == NULL)
2260                 goto out;
2261         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2262         if (locations == NULL)
2263                 goto out;
2264
2265         status = nfs4_proc_fs_locations(dir, name, locations, page);
2266         if (status != 0)
2267                 goto out;
2268         /* Make sure server returned a different fsid for the referral */
2269         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2270                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2271                 status = -EIO;
2272                 goto out;
2273         }
2274
2275         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2276         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2277         if (!fattr->mode)
2278                 fattr->mode = S_IFDIR;
2279         memset(fhandle, 0, sizeof(struct nfs_fh));
2280 out:
2281         if (page)
2282                 __free_page(page);
2283         kfree(locations);
2284         return status;
2285 }
2286
2287 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2288 {
2289         struct nfs4_getattr_arg args = {
2290                 .fh = fhandle,
2291                 .bitmask = server->attr_bitmask,
2292         };
2293         struct nfs4_getattr_res res = {
2294                 .fattr = fattr,
2295                 .server = server,
2296         };
2297         struct rpc_message msg = {
2298                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2299                 .rpc_argp = &args,
2300                 .rpc_resp = &res,
2301         };
2302
2303         nfs_fattr_init(fattr);
2304         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2305 }
2306
2307 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2308 {
2309         struct nfs4_exception exception = { };
2310         int err;
2311         do {
2312                 err = nfs4_handle_exception(server,
2313                                 _nfs4_proc_getattr(server, fhandle, fattr),
2314                                 &exception);
2315         } while (exception.retry);
2316         return err;
2317 }
2318
2319 /*
2320  * The file is not closed if it is opened due to the a request to change
2321  * the size of the file. The open call will not be needed once the
2322  * VFS layer lookup-intents are implemented.
2323  *
2324  * Close is called when the inode is destroyed.
2325  * If we haven't opened the file for O_WRONLY, we
2326  * need to in the size_change case to obtain a stateid.
2327  *
2328  * Got race?
2329  * Because OPEN is always done by name in nfsv4, it is
2330  * possible that we opened a different file by the same
2331  * name.  We can recognize this race condition, but we
2332  * can't do anything about it besides returning an error.
2333  *
2334  * This will be fixed with VFS changes (lookup-intent).
2335  */
2336 static int
2337 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2338                   struct iattr *sattr)
2339 {
2340         struct inode *inode = dentry->d_inode;
2341         struct rpc_cred *cred = NULL;
2342         struct nfs4_state *state = NULL;
2343         int status;
2344
2345         nfs_fattr_init(fattr);
2346
2347         /* Search for an existing open(O_WRITE) file */
2348         if (sattr->ia_valid & ATTR_FILE) {
2349                 struct nfs_open_context *ctx;
2350
2351                 ctx = nfs_file_open_context(sattr->ia_file);
2352                 if (ctx) {
2353                         cred = ctx->cred;
2354                         state = ctx->state;
2355                 }
2356         }
2357
2358         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2359         if (status == 0)
2360                 nfs_setattr_update_inode(inode, sattr);
2361         return status;
2362 }
2363
2364 static int _nfs4_proc_lookupfh(struct rpc_clnt *clnt, struct nfs_server *server,
2365                 const struct nfs_fh *dirfh, const struct qstr *name,
2366                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2367 {
2368         int                    status;
2369         struct nfs4_lookup_arg args = {
2370                 .bitmask = server->attr_bitmask,
2371                 .dir_fh = dirfh,
2372                 .name = name,
2373         };
2374         struct nfs4_lookup_res res = {
2375                 .server = server,
2376                 .fattr = fattr,
2377                 .fh = fhandle,
2378         };
2379         struct rpc_message msg = {
2380                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2381                 .rpc_argp = &args,
2382                 .rpc_resp = &res,
2383         };
2384
2385         nfs_fattr_init(fattr);
2386
2387         dprintk("NFS call  lookupfh %s\n", name->name);
2388         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2389         dprintk("NFS reply lookupfh: %d\n", status);
2390         return status;
2391 }
2392
2393 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2394                               struct qstr *name, struct nfs_fh *fhandle,
2395                               struct nfs_fattr *fattr)
2396 {
2397         struct nfs4_exception exception = { };
2398         int err;
2399         do {
2400                 err = _nfs4_proc_lookupfh(server->client, server, dirfh, name, fhandle, fattr);
2401                 /* FIXME: !!!! */
2402                 if (err == -NFS4ERR_MOVED) {
2403                         err = -EREMOTE;
2404                         break;
2405                 }
2406                 err = nfs4_handle_exception(server, err, &exception);
2407         } while (exception.retry);
2408         return err;
2409 }
2410
2411 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2412                 const struct qstr *name, struct nfs_fh *fhandle,
2413                 struct nfs_fattr *fattr)
2414 {
2415         int status;
2416
2417         dprintk("NFS call  lookup %s\n", name->name);
2418         status = _nfs4_proc_lookupfh(clnt, NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2419         if (status == -NFS4ERR_MOVED)
2420                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2421         dprintk("NFS reply lookup: %d\n", status);
2422         return status;
2423 }
2424
2425 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2426 {
2427         memset(fh, 0, sizeof(struct nfs_fh));
2428         fattr->fsid.major = 1;
2429         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2430                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2431         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2432         fattr->nlink = 2;
2433 }
2434
2435 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2436                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2437 {
2438         struct nfs4_exception exception = { };
2439         int err;
2440         do {
2441                 err = nfs4_handle_exception(NFS_SERVER(dir),
2442                                 _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr),
2443                                 &exception);
2444                 if (err == -EPERM)
2445                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2446         } while (exception.retry);
2447         return err;
2448 }
2449
2450 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2451 {
2452         struct nfs_server *server = NFS_SERVER(inode);
2453         struct nfs4_accessargs args = {
2454                 .fh = NFS_FH(inode),
2455                 .bitmask = server->attr_bitmask,
2456         };
2457         struct nfs4_accessres res = {
2458                 .server = server,
2459         };
2460         struct rpc_message msg = {
2461                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2462                 .rpc_argp = &args,
2463                 .rpc_resp = &res,
2464                 .rpc_cred = entry->cred,
2465         };
2466         int mode = entry->mask;
2467         int status;
2468
2469         /*
2470          * Determine which access bits we want to ask for...
2471          */
2472         if (mode & MAY_READ)
2473                 args.access |= NFS4_ACCESS_READ;
2474         if (S_ISDIR(inode->i_mode)) {
2475                 if (mode & MAY_WRITE)
2476                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2477                 if (mode & MAY_EXEC)
2478                         args.access |= NFS4_ACCESS_LOOKUP;
2479         } else {
2480                 if (mode & MAY_WRITE)
2481                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2482                 if (mode & MAY_EXEC)
2483                         args.access |= NFS4_ACCESS_EXECUTE;
2484         }
2485
2486         res.fattr = nfs_alloc_fattr();
2487         if (res.fattr == NULL)
2488                 return -ENOMEM;
2489
2490         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2491         if (!status) {
2492                 entry->mask = 0;
2493                 if (res.access & NFS4_ACCESS_READ)
2494                         entry->mask |= MAY_READ;
2495                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2496                         entry->mask |= MAY_WRITE;
2497                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2498                         entry->mask |= MAY_EXEC;
2499                 nfs_refresh_inode(inode, res.fattr);
2500         }
2501         nfs_free_fattr(res.fattr);
2502         return status;
2503 }
2504
2505 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2506 {
2507         struct nfs4_exception exception = { };
2508         int err;
2509         do {
2510                 err = nfs4_handle_exception(NFS_SERVER(inode),
2511                                 _nfs4_proc_access(inode, entry),
2512                                 &exception);
2513         } while (exception.retry);
2514         return err;
2515 }
2516
2517 /*
2518  * TODO: For the time being, we don't try to get any attributes
2519  * along with any of the zero-copy operations READ, READDIR,
2520  * READLINK, WRITE.
2521  *
2522  * In the case of the first three, we want to put the GETATTR
2523  * after the read-type operation -- this is because it is hard
2524  * to predict the length of a GETATTR response in v4, and thus
2525  * align the READ data correctly.  This means that the GETATTR
2526  * may end up partially falling into the page cache, and we should
2527  * shift it into the 'tail' of the xdr_buf before processing.
2528  * To do this efficiently, we need to know the total length
2529  * of data received, which doesn't seem to be available outside
2530  * of the RPC layer.
2531  *
2532  * In the case of WRITE, we also want to put the GETATTR after
2533  * the operation -- in this case because we want to make sure
2534  * we get the post-operation mtime and size.  This means that
2535  * we can't use xdr_encode_pages() as written: we need a variant
2536  * of it which would leave room in the 'tail' iovec.
2537  *
2538  * Both of these changes to the XDR layer would in fact be quite
2539  * minor, but I decided to leave them for a subsequent patch.
2540  */
2541 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2542                 unsigned int pgbase, unsigned int pglen)
2543 {
2544         struct nfs4_readlink args = {
2545                 .fh       = NFS_FH(inode),
2546                 .pgbase   = pgbase,
2547                 .pglen    = pglen,
2548                 .pages    = &page,
2549         };
2550         struct nfs4_readlink_res res;
2551         struct rpc_message msg = {
2552                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2553                 .rpc_argp = &args,
2554                 .rpc_resp = &res,
2555         };
2556
2557         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2558 }
2559
2560 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2561                 unsigned int pgbase, unsigned int pglen)
2562 {
2563         struct nfs4_exception exception = { };
2564         int err;
2565         do {
2566                 err = nfs4_handle_exception(NFS_SERVER(inode),
2567                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2568                                 &exception);
2569         } while (exception.retry);
2570         return err;
2571 }
2572
2573 /*
2574  * Got race?
2575  * We will need to arrange for the VFS layer to provide an atomic open.
2576  * Until then, this create/open method is prone to inefficiency and race
2577  * conditions due to the lookup, create, and open VFS calls from sys_open()
2578  * placed on the wire.
2579  *
2580  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2581  * The file will be opened again in the subsequent VFS open call
2582  * (nfs4_proc_file_open).
2583  *
2584  * The open for read will just hang around to be used by any process that
2585  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2586  */
2587
2588 static int
2589 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2590                  int flags, struct nfs_open_context *ctx)
2591 {
2592         struct path my_path = {
2593                 .dentry = dentry,
2594         };
2595         struct path *path = &my_path;
2596         struct nfs4_state *state;
2597         struct rpc_cred *cred = NULL;
2598         fmode_t fmode = 0;
2599         int status = 0;
2600
2601         if (ctx != NULL) {
2602                 cred = ctx->cred;
2603                 path = &ctx->path;
2604                 fmode = ctx->mode;
2605         }
2606         sattr->ia_mode &= ~current_umask();
2607         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2608         d_drop(dentry);
2609         if (IS_ERR(state)) {
2610                 status = PTR_ERR(state);
2611                 goto out;
2612         }
2613         d_add(dentry, igrab(state->inode));
2614         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2615         if (ctx != NULL)
2616                 ctx->state = state;
2617         else
2618                 nfs4_close_sync(path, state, fmode);
2619 out:
2620         return status;
2621 }
2622
2623 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2624 {
2625         struct nfs_server *server = NFS_SERVER(dir);
2626         struct nfs_removeargs args = {
2627                 .fh = NFS_FH(dir),
2628                 .name.len = name->len,
2629                 .name.name = name->name,
2630                 .bitmask = server->attr_bitmask,
2631         };
2632         struct nfs_removeres res = {
2633                 .server = server,
2634         };
2635         struct rpc_message msg = {
2636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2637                 .rpc_argp = &args,
2638                 .rpc_resp = &res,
2639         };
2640         int status = -ENOMEM;
2641
2642         res.dir_attr = nfs_alloc_fattr();
2643         if (res.dir_attr == NULL)
2644                 goto out;
2645
2646         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2647         if (status == 0) {
2648                 update_changeattr(dir, &res.cinfo);
2649                 nfs_post_op_update_inode(dir, res.dir_attr);
2650         }
2651         nfs_free_fattr(res.dir_attr);
2652 out:
2653         return status;
2654 }
2655
2656 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2657 {
2658         struct nfs4_exception exception = { };
2659         int err;
2660         do {
2661                 err = nfs4_handle_exception(NFS_SERVER(dir),
2662                                 _nfs4_proc_remove(dir, name),
2663                                 &exception);
2664         } while (exception.retry);
2665         return err;
2666 }
2667
2668 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2669 {
2670         struct nfs_server *server = NFS_SERVER(dir);
2671         struct nfs_removeargs *args = msg->rpc_argp;
2672         struct nfs_removeres *res = msg->rpc_resp;
2673
2674         args->bitmask = server->cache_consistency_bitmask;
2675         res->server = server;
2676         res->seq_res.sr_slot = NULL;
2677         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2678 }
2679
2680 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2681 {
2682         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2683
2684         if (!nfs4_sequence_done(task, &res->seq_res))
2685                 return 0;
2686         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2687                 return 0;
2688         update_changeattr(dir, &res->cinfo);
2689         nfs_post_op_update_inode(dir, res->dir_attr);
2690         return 1;
2691 }
2692
2693 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2694 {
2695         struct nfs_server *server = NFS_SERVER(dir);
2696         struct nfs_renameargs *arg = msg->rpc_argp;
2697         struct nfs_renameres *res = msg->rpc_resp;
2698
2699         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2700         arg->bitmask = server->attr_bitmask;
2701         res->server = server;
2702 }
2703
2704 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2705                                  struct inode *new_dir)
2706 {
2707         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2708
2709         if (!nfs4_sequence_done(task, &res->seq_res))
2710                 return 0;
2711         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2712                 return 0;
2713
2714         update_changeattr(old_dir, &res->old_cinfo);
2715         nfs_post_op_update_inode(old_dir, res->old_fattr);
2716         update_changeattr(new_dir, &res->new_cinfo);
2717         nfs_post_op_update_inode(new_dir, res->new_fattr);
2718         return 1;
2719 }
2720
2721 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2722                 struct inode *new_dir, struct qstr *new_name)
2723 {
2724         struct nfs_server *server = NFS_SERVER(old_dir);
2725         struct nfs_renameargs arg = {
2726                 .old_dir = NFS_FH(old_dir),
2727                 .new_dir = NFS_FH(new_dir),
2728                 .old_name = old_name,
2729                 .new_name = new_name,
2730                 .bitmask = server->attr_bitmask,
2731         };
2732         struct nfs_renameres res = {
2733                 .server = server,
2734         };
2735         struct rpc_message msg = {
2736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2737                 .rpc_argp = &arg,
2738                 .rpc_resp = &res,
2739         };
2740         int status = -ENOMEM;
2741
2742         res.old_fattr = nfs_alloc_fattr();
2743         res.new_fattr = nfs_alloc_fattr();
2744         if (res.old_fattr == NULL || res.new_fattr == NULL)
2745                 goto out;
2746
2747         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2748         if (!status) {
2749                 update_changeattr(old_dir, &res.old_cinfo);
2750                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2751                 update_changeattr(new_dir, &res.new_cinfo);
2752                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2753         }
2754 out:
2755         nfs_free_fattr(res.new_fattr);
2756         nfs_free_fattr(res.old_fattr);
2757         return status;
2758 }
2759
2760 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2761                 struct inode *new_dir, struct qstr *new_name)
2762 {
2763         struct nfs4_exception exception = { };
2764         int err;
2765         do {
2766                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2767                                 _nfs4_proc_rename(old_dir, old_name,
2768                                         new_dir, new_name),
2769                                 &exception);
2770         } while (exception.retry);
2771         return err;
2772 }
2773
2774 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2775 {
2776         struct nfs_server *server = NFS_SERVER(inode);
2777         struct nfs4_link_arg arg = {
2778                 .fh     = NFS_FH(inode),
2779                 .dir_fh = NFS_FH(dir),
2780                 .name   = name,
2781                 .bitmask = server->attr_bitmask,
2782         };
2783         struct nfs4_link_res res = {
2784                 .server = server,
2785         };
2786         struct rpc_message msg = {
2787                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2788                 .rpc_argp = &arg,
2789                 .rpc_resp = &res,
2790         };
2791         int status = -ENOMEM;
2792
2793         res.fattr = nfs_alloc_fattr();
2794         res.dir_attr = nfs_alloc_fattr();
2795         if (res.fattr == NULL || res.dir_attr == NULL)
2796                 goto out;
2797
2798         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2799         if (!status) {
2800                 update_changeattr(dir, &res.cinfo);
2801                 nfs_post_op_update_inode(dir, res.dir_attr);
2802                 nfs_post_op_update_inode(inode, res.fattr);
2803         }
2804 out:
2805         nfs_free_fattr(res.dir_attr);
2806         nfs_free_fattr(res.fattr);
2807         return status;
2808 }
2809
2810 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2811 {
2812         struct nfs4_exception exception = { };
2813         int err;
2814         do {
2815                 err = nfs4_handle_exception(NFS_SERVER(inode),
2816                                 _nfs4_proc_link(inode, dir, name),
2817                                 &exception);
2818         } while (exception.retry);
2819         return err;
2820 }
2821
2822 struct nfs4_createdata {
2823         struct rpc_message msg;
2824         struct nfs4_create_arg arg;
2825         struct nfs4_create_res res;
2826         struct nfs_fh fh;
2827         struct nfs_fattr fattr;
2828         struct nfs_fattr dir_fattr;
2829 };
2830
2831 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2832                 struct qstr *name, struct iattr *sattr, u32 ftype)
2833 {
2834         struct nfs4_createdata *data;
2835
2836         data = kzalloc(sizeof(*data), GFP_KERNEL);
2837         if (data != NULL) {
2838                 struct nfs_server *server = NFS_SERVER(dir);
2839
2840                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2841                 data->msg.rpc_argp = &data->arg;
2842                 data->msg.rpc_resp = &data->res;
2843                 data->arg.dir_fh = NFS_FH(dir);
2844                 data->arg.server = server;
2845                 data->arg.name = name;
2846                 data->arg.attrs = sattr;
2847                 data->arg.ftype = ftype;
2848                 data->arg.bitmask = server->attr_bitmask;
2849                 data->res.server = server;
2850                 data->res.fh = &data->fh;
2851                 data->res.fattr = &data->fattr;
2852                 data->res.dir_fattr = &data->dir_fattr;
2853                 nfs_fattr_init(data->res.fattr);
2854                 nfs_fattr_init(data->res.dir_fattr);
2855         }
2856         return data;
2857 }
2858
2859 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2860 {
2861         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2862                                     &data->arg.seq_args, &data->res.seq_res, 1);
2863         if (status == 0) {
2864                 update_changeattr(dir, &data->res.dir_cinfo);
2865                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2866                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2867         }
2868         return status;
2869 }
2870
2871 static void nfs4_free_createdata(struct nfs4_createdata *data)
2872 {
2873         kfree(data);
2874 }
2875
2876 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2877                 struct page *page, unsigned int len, struct iattr *sattr)
2878 {
2879         struct nfs4_createdata *data;
2880         int status = -ENAMETOOLONG;
2881
2882         if (len > NFS4_MAXPATHLEN)
2883                 goto out;
2884
2885         status = -ENOMEM;
2886         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2887         if (data == NULL)
2888                 goto out;
2889
2890         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2891         data->arg.u.symlink.pages = &page;
2892         data->arg.u.symlink.len = len;
2893
2894         status = nfs4_do_create(dir, dentry, data);
2895
2896         nfs4_free_createdata(data);
2897 out:
2898         return status;
2899 }
2900
2901 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2902                 struct page *page, unsigned int len, struct iattr *sattr)
2903 {
2904         struct nfs4_exception exception = { };
2905         int err;
2906         do {
2907                 err = nfs4_handle_exception(NFS_SERVER(dir),
2908                                 _nfs4_proc_symlink(dir, dentry, page,
2909                                                         len, sattr),
2910                                 &exception);
2911         } while (exception.retry);
2912         return err;
2913 }
2914
2915 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2916                 struct iattr *sattr)
2917 {
2918         struct nfs4_createdata *data;
2919         int status = -ENOMEM;
2920
2921         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2922         if (data == NULL)
2923                 goto out;
2924
2925         status = nfs4_do_create(dir, dentry, data);
2926
2927         nfs4_free_createdata(data);
2928 out:
2929         return status;
2930 }
2931
2932 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2933                 struct iattr *sattr)
2934 {
2935         struct nfs4_exception exception = { };
2936         int err;
2937
2938         sattr->ia_mode &= ~current_umask();
2939         do {
2940                 err = nfs4_handle_exception(NFS_SERVER(dir),
2941                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2942                                 &exception);
2943         } while (exception.retry);
2944         return err;
2945 }
2946
2947 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2948                 u64 cookie, struct page **pages, unsigned int count, int plus)
2949 {
2950         struct inode            *dir = dentry->d_inode;
2951         struct nfs4_readdir_arg args = {
2952                 .fh = NFS_FH(dir),
2953                 .pages = pages,
2954                 .pgbase = 0,
2955                 .count = count,
2956                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2957                 .plus = plus,
2958         };
2959         struct nfs4_readdir_res res;
2960         struct rpc_message msg = {
2961                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2962                 .rpc_argp = &args,
2963                 .rpc_resp = &res,
2964                 .rpc_cred = cred,
2965         };
2966         int                     status;
2967
2968         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2969                         dentry->d_parent->d_name.name,
2970                         dentry->d_name.name,
2971                         (unsigned long long)cookie);
2972         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2973         res.pgbase = args.pgbase;
2974         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
2975         if (status >= 0) {
2976                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2977                 status += args.pgbase;
2978         }
2979
2980         nfs_invalidate_atime(dir);
2981
2982         dprintk("%s: returns %d\n", __func__, status);
2983         return status;
2984 }
2985
2986 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2987                 u64 cookie, struct page **pages, unsigned int count, int plus)
2988 {
2989         struct nfs4_exception exception = { };
2990         int err;
2991         do {
2992                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2993                                 _nfs4_proc_readdir(dentry, cred, cookie,
2994                                         pages, count, plus),
2995                                 &exception);
2996         } while (exception.retry);
2997         return err;
2998 }
2999
3000 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3001                 struct iattr *sattr, dev_t rdev)
3002 {
3003         struct nfs4_createdata *data;
3004         int mode = sattr->ia_mode;
3005         int status = -ENOMEM;
3006
3007         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3008         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3009
3010         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3011         if (data == NULL)
3012                 goto out;
3013
3014         if (S_ISFIFO(mode))
3015                 data->arg.ftype = NF4FIFO;
3016         else if (S_ISBLK(mode)) {
3017                 data->arg.ftype = NF4BLK;
3018                 data->arg.u.device.specdata1 = MAJOR(rdev);
3019                 data->arg.u.device.specdata2 = MINOR(rdev);
3020         }
3021         else if (S_ISCHR(mode)) {
3022                 data->arg.ftype = NF4CHR;
3023                 data->arg.u.device.specdata1 = MAJOR(rdev);
3024                 data->arg.u.device.specdata2 = MINOR(rdev);
3025         }
3026
3027         status = nfs4_do_create(dir, dentry, data);
3028
3029         nfs4_free_createdata(data);
3030 out:
3031         return status;
3032 }
3033
3034 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3035                 struct iattr *sattr, dev_t rdev)
3036 {
3037         struct nfs4_exception exception = { };
3038         int err;
3039
3040         sattr->ia_mode &= ~current_umask();
3041         do {
3042                 err = nfs4_handle_exception(NFS_SERVER(dir),
3043                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3044                                 &exception);
3045         } while (exception.retry);
3046         return err;
3047 }
3048
3049 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3050                  struct nfs_fsstat *fsstat)
3051 {
3052         struct nfs4_statfs_arg args = {
3053                 .fh = fhandle,
3054                 .bitmask = server->attr_bitmask,
3055         };
3056         struct nfs4_statfs_res res = {
3057                 .fsstat = fsstat,
3058         };
3059         struct rpc_message msg = {
3060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3061                 .rpc_argp = &args,
3062                 .rpc_resp = &res,
3063         };
3064
3065         nfs_fattr_init(fsstat->fattr);
3066         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3067 }
3068
3069 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3070 {
3071         struct nfs4_exception exception = { };
3072         int err;
3073         do {
3074                 err = nfs4_handle_exception(server,
3075                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3076                                 &exception);
3077         } while (exception.retry);
3078         return err;
3079 }
3080
3081 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3082                 struct nfs_fsinfo *fsinfo)
3083 {
3084         struct nfs4_fsinfo_arg args = {
3085                 .fh = fhandle,
3086                 .bitmask = server->attr_bitmask,
3087         };
3088         struct nfs4_fsinfo_res res = {
3089                 .fsinfo = fsinfo,
3090         };
3091         struct rpc_message msg = {
3092                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3093                 .rpc_argp = &args,
3094                 .rpc_resp = &res,
3095         };
3096
3097         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3098 }
3099
3100 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3101 {
3102         struct nfs4_exception exception = { };
3103         int err;
3104
3105         do {
3106                 err = nfs4_handle_exception(server,
3107                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3108                                 &exception);
3109         } while (exception.retry);
3110         return err;
3111 }
3112
3113 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3114 {
3115         nfs_fattr_init(fsinfo->fattr);
3116         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3117 }
3118
3119 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3120                 struct nfs_pathconf *pathconf)
3121 {
3122         struct nfs4_pathconf_arg args = {
3123                 .fh = fhandle,
3124                 .bitmask = server->attr_bitmask,
3125         };
3126         struct nfs4_pathconf_res res = {
3127                 .pathconf = pathconf,
3128         };
3129         struct rpc_message msg = {
3130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3131                 .rpc_argp = &args,
3132                 .rpc_resp = &res,
3133         };
3134
3135         /* None of the pathconf attributes are mandatory to implement */
3136         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3137                 memset(pathconf, 0, sizeof(*pathconf));
3138                 return 0;
3139         }
3140
3141         nfs_fattr_init(pathconf->fattr);
3142         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3143 }
3144
3145 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3146                 struct nfs_pathconf *pathconf)
3147 {
3148         struct nfs4_exception exception = { };
3149         int err;
3150
3151         do {
3152                 err = nfs4_handle_exception(server,
3153                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3154                                 &exception);
3155         } while (exception.retry);
3156         return err;
3157 }
3158
3159 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3160 {
3161         struct nfs_server *server = NFS_SERVER(data->inode);
3162
3163         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3164                 nfs_restart_rpc(task, server->nfs_client);
3165                 return -EAGAIN;
3166         }
3167
3168         nfs_invalidate_atime(data->inode);
3169         if (task->tk_status > 0)
3170                 renew_lease(server, data->timestamp);
3171         return 0;
3172 }
3173
3174 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3175 {
3176
3177         dprintk("--> %s\n", __func__);
3178
3179         if (!nfs4_sequence_done(task, &data->res.seq_res))
3180                 return -EAGAIN;
3181
3182         return data->read_done_cb(task, data);
3183 }
3184
3185 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3186 {
3187         data->timestamp   = jiffies;
3188         data->read_done_cb = nfs4_read_done_cb;
3189         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3190 }
3191
3192 /* Reset the the nfs_read_data to send the read to the MDS. */
3193 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3194 {
3195         dprintk("%s Reset task for i/o through\n", __func__);
3196         put_lseg(data->lseg);
3197         data->lseg = NULL;
3198         /* offsets will differ in the dense stripe case */
3199         data->args.offset = data->mds_offset;
3200         data->ds_clp = NULL;
3201         data->args.fh     = NFS_FH(data->inode);
3202         data->read_done_cb = nfs4_read_done_cb;
3203         task->tk_ops = data->mds_ops;
3204         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3205 }
3206 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3207
3208 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3209 {
3210         struct inode *inode = data->inode;
3211
3212         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3213                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3214                 return -EAGAIN;
3215         }
3216         if (task->tk_status >= 0) {
3217                 renew_lease(NFS_SERVER(inode), data->timestamp);
3218                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3219         }
3220         return 0;
3221 }
3222
3223 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3224 {
3225         if (!nfs4_sequence_done(task, &data->res.seq_res))
3226                 return -EAGAIN;
3227         return data->write_done_cb(task, data);
3228 }
3229
3230 /* Reset the the nfs_write_data to send the write to the MDS. */
3231 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3232 {
3233         dprintk("%s Reset task for i/o through\n", __func__);
3234         put_lseg(data->lseg);
3235         data->lseg          = NULL;
3236         data->ds_clp        = NULL;
3237         data->write_done_cb = nfs4_write_done_cb;
3238         data->args.fh       = NFS_FH(data->inode);
3239         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3240         data->args.offset   = data->mds_offset;
3241         data->res.fattr     = &data->fattr;
3242         task->tk_ops        = data->mds_ops;
3243         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3244 }
3245 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3246
3247 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3248 {
3249         struct nfs_server *server = NFS_SERVER(data->inode);
3250
3251         if (data->lseg) {
3252                 data->args.bitmask = NULL;
3253                 data->res.fattr = NULL;
3254         } else
3255                 data->args.bitmask = server->cache_consistency_bitmask;
3256         if (!data->write_done_cb)
3257                 data->write_done_cb = nfs4_write_done_cb;
3258         data->res.server = server;
3259         data->timestamp   = jiffies;
3260
3261         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3262 }
3263
3264 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3265 {
3266         struct inode *inode = data->inode;
3267
3268         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3269                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3270                 return -EAGAIN;
3271         }
3272         nfs_refresh_inode(inode, data->res.fattr);
3273         return 0;
3274 }
3275
3276 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3277 {
3278         if (!nfs4_sequence_done(task, &data->res.seq_res))
3279                 return -EAGAIN;
3280         return data->write_done_cb(task, data);
3281 }
3282
3283 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3284 {
3285         struct nfs_server *server = NFS_SERVER(data->inode);
3286
3287         if (data->lseg) {
3288                 data->args.bitmask = NULL;
3289                 data->res.fattr = NULL;
3290         } else
3291                 data->args.bitmask = server->cache_consistency_bitmask;
3292         if (!data->write_done_cb)
3293                 data->write_done_cb = nfs4_commit_done_cb;
3294         data->res.server = server;
3295         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3296 }
3297
3298 struct nfs4_renewdata {
3299         struct nfs_client       *client;
3300         unsigned long           timestamp;
3301 };
3302
3303 /*
3304  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3305  * standalone procedure for queueing an asynchronous RENEW.
3306  */
3307 static void nfs4_renew_release(void *calldata)
3308 {
3309         struct nfs4_renewdata *data = calldata;
3310         struct nfs_client *clp = data->client;
3311
3312         if (atomic_read(&clp->cl_count) > 1)
3313                 nfs4_schedule_state_renewal(clp);
3314         nfs_put_client(clp);
3315         kfree(data);
3316 }
3317
3318 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3319 {
3320         struct nfs4_renewdata *data = calldata;
3321         struct nfs_client *clp = data->client;
3322         unsigned long timestamp = data->timestamp;
3323
3324         if (task->tk_status < 0) {
3325                 /* Unless we're shutting down, schedule state recovery! */
3326                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3327                         nfs4_schedule_lease_recovery(clp);
3328                 return;
3329         }
3330         do_renew_lease(clp, timestamp);
3331 }
3332
3333 static const struct rpc_call_ops nfs4_renew_ops = {
3334         .rpc_call_done = nfs4_renew_done,
3335         .rpc_release = nfs4_renew_release,
3336 };
3337
3338 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3339 {
3340         struct rpc_message msg = {
3341                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3342                 .rpc_argp       = clp,
3343                 .rpc_cred       = cred,
3344         };
3345         struct nfs4_renewdata *data;
3346
3347         if (!atomic_inc_not_zero(&clp->cl_count))
3348                 return -EIO;
3349         data = kmalloc(sizeof(*data), GFP_KERNEL);
3350         if (data == NULL)
3351                 return -ENOMEM;
3352         data->client = clp;
3353         data->timestamp = jiffies;
3354         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3355                         &nfs4_renew_ops, data);
3356 }
3357
3358 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3359 {
3360         struct rpc_message msg = {
3361                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3362                 .rpc_argp       = clp,
3363                 .rpc_cred       = cred,
3364         };
3365         unsigned long now = jiffies;
3366         int status;
3367
3368         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3369         if (status < 0)
3370                 return status;
3371         do_renew_lease(clp, now);
3372         return 0;
3373 }
3374
3375 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3376 {
3377         return (server->caps & NFS_CAP_ACLS)
3378                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3379                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3380 }
3381
3382 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3383  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3384  * the stack.
3385  */
3386 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3387
3388 static void buf_to_pages(const void *buf, size_t buflen,
3389                 struct page **pages, unsigned int *pgbase)
3390 {
3391         const void *p = buf;
3392
3393         *pgbase = offset_in_page(buf);
3394         p -= *pgbase;
3395         while (p < buf + buflen) {
3396                 *(pages++) = virt_to_page(p);
3397                 p += PAGE_CACHE_SIZE;
3398         }
3399 }
3400
3401 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3402                 struct page **pages, unsigned int *pgbase)
3403 {
3404         struct page *newpage, **spages;
3405         int rc = 0;
3406         size_t len;
3407         spages = pages;
3408
3409         do {
3410                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3411                 newpage = alloc_page(GFP_KERNEL);
3412
3413                 if (newpage == NULL)
3414                         goto unwind;
3415                 memcpy(page_address(newpage), buf, len);
3416                 buf += len;
3417                 buflen -= len;
3418                 *pages++ = newpage;
3419                 rc++;
3420         } while (buflen != 0);
3421
3422         return rc;
3423
3424 unwind:
3425         for(; rc > 0; rc--)
3426                 __free_page(spages[rc-1]);
3427         return -ENOMEM;
3428 }
3429
3430 struct nfs4_cached_acl {
3431         int cached;
3432         size_t len;
3433         char data[0];
3434 };
3435
3436 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3437 {
3438         struct nfs_inode *nfsi = NFS_I(inode);
3439
3440         spin_lock(&inode->i_lock);
3441         kfree(nfsi->nfs4_acl);
3442         nfsi->nfs4_acl = acl;
3443         spin_unlock(&inode->i_lock);
3444 }
3445
3446 static void nfs4_zap_acl_attr(struct inode *inode)
3447 {
3448         nfs4_set_cached_acl(inode, NULL);
3449 }
3450
3451 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3452 {
3453         struct nfs_inode *nfsi = NFS_I(inode);
3454         struct nfs4_cached_acl *acl;
3455         int ret = -ENOENT;
3456
3457         spin_lock(&inode->i_lock);
3458         acl = nfsi->nfs4_acl;
3459         if (acl == NULL)
3460                 goto out;
3461         if (buf == NULL) /* user is just asking for length */
3462                 goto out_len;
3463         if (acl->cached == 0)
3464                 goto out;
3465         ret = -ERANGE; /* see getxattr(2) man page */
3466         if (acl->len > buflen)
3467                 goto out;
3468         memcpy(buf, acl->data, acl->len);
3469 out_len:
3470         ret = acl->len;
3471 out:
3472         spin_unlock(&inode->i_lock);
3473         return ret;
3474 }
3475
3476 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3477 {
3478         struct nfs4_cached_acl *acl;
3479
3480         if (buf && acl_len <= PAGE_SIZE) {
3481                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3482                 if (acl == NULL)
3483                         goto out;
3484                 acl->cached = 1;
3485                 memcpy(acl->data, buf, acl_len);
3486         } else {
3487                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3488                 if (acl == NULL)
3489                         goto out;
3490                 acl->cached = 0;
3491         }
3492         acl->len = acl_len;
3493 out:
3494         nfs4_set_cached_acl(inode, acl);
3495 }
3496
3497 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3498 {
3499         struct page *pages[NFS4ACL_MAXPAGES];
3500         struct nfs_getaclargs args = {
3501                 .fh = NFS_FH(inode),
3502                 .acl_pages = pages,
3503                 .acl_len = buflen,
3504         };
3505         struct nfs_getaclres res = {
3506                 .acl_len = buflen,
3507         };
3508         void *resp_buf;
3509         struct rpc_message msg = {
3510                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3511                 .rpc_argp = &args,
3512                 .rpc_resp = &res,
3513         };
3514         struct page *localpage = NULL;
3515         int ret;
3516
3517         if (buflen < PAGE_SIZE) {
3518                 /* As long as we're doing a round trip to the server anyway,
3519                  * let's be prepared for a page of acl data. */
3520                 localpage = alloc_page(GFP_KERNEL);
3521                 resp_buf = page_address(localpage);
3522                 if (localpage == NULL)
3523                         return -ENOMEM;
3524                 args.acl_pages[0] = localpage;
3525                 args.acl_pgbase = 0;
3526                 args.acl_len = PAGE_SIZE;
3527         } else {
3528                 resp_buf = buf;
3529                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3530         }
3531         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3532         if (ret)
3533                 goto out_free;
3534         if (res.acl_len > args.acl_len)
3535                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3536         else
3537                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3538         if (buf) {
3539                 ret = -ERANGE;
3540                 if (res.acl_len > buflen)
3541                         goto out_free;
3542                 if (localpage)
3543                         memcpy(buf, resp_buf, res.acl_len);
3544         }
3545         ret = res.acl_len;
3546 out_free:
3547         if (localpage)
3548                 __free_page(localpage);
3549         return ret;
3550 }
3551
3552 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3553 {
3554         struct nfs4_exception exception = { };
3555         ssize_t ret;
3556         do {
3557                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3558                 if (ret >= 0)
3559                         break;
3560                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3561         } while (exception.retry);
3562         return ret;
3563 }
3564
3565 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3566 {
3567         struct nfs_server *server = NFS_SERVER(inode);
3568         int ret;
3569
3570         if (!nfs4_server_supports_acls(server))
3571                 return -EOPNOTSUPP;
3572         ret = nfs_revalidate_inode(server, inode);
3573         if (ret < 0)
3574                 return ret;
3575         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3576                 nfs_zap_acl_cache(inode);
3577         ret = nfs4_read_cached_acl(inode, buf, buflen);
3578         if (ret != -ENOENT)
3579                 return ret;
3580         return nfs4_get_acl_uncached(inode, buf, buflen);
3581 }
3582
3583 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3584 {
3585         struct nfs_server *server = NFS_SERVER(inode);
3586         struct page *pages[NFS4ACL_MAXPAGES];
3587         struct nfs_setaclargs arg = {
3588                 .fh             = NFS_FH(inode),
3589                 .acl_pages      = pages,
3590                 .acl_len        = buflen,
3591         };
3592         struct nfs_setaclres res;
3593         struct rpc_message msg = {
3594                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3595                 .rpc_argp       = &arg,
3596                 .rpc_resp       = &res,
3597         };
3598         int ret, i;
3599
3600         if (!nfs4_server_supports_acls(server))
3601                 return -EOPNOTSUPP;
3602         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3603         if (i < 0)
3604                 return i;
3605         nfs_inode_return_delegation(inode);
3606         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3607
3608         /*
3609          * Free each page after tx, so the only ref left is
3610          * held by the network stack
3611          */
3612         for (; i > 0; i--)
3613                 put_page(pages[i-1]);
3614
3615         /*
3616          * Acl update can result in inode attribute update.
3617          * so mark the attribute cache invalid.
3618          */
3619         spin_lock(&inode->i_lock);
3620         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3621         spin_unlock(&inode->i_lock);
3622         nfs_access_zap_cache(inode);
3623         nfs_zap_acl_cache(inode);
3624         return ret;
3625 }
3626
3627 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3628 {
3629         struct nfs4_exception exception = { };
3630         int err;
3631         do {
3632                 err = nfs4_handle_exception(NFS_SERVER(inode),
3633                                 __nfs4_proc_set_acl(inode, buf, buflen),
3634                                 &exception);
3635         } while (exception.retry);
3636         return err;
3637 }
3638
3639 static int
3640 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3641 {
3642         struct nfs_client *clp = server->nfs_client;
3643
3644         if (task->tk_status >= 0)
3645                 return 0;
3646         switch(task->tk_status) {
3647                 case -NFS4ERR_ADMIN_REVOKED:
3648                 case -NFS4ERR_BAD_STATEID:
3649                 case -NFS4ERR_OPENMODE:
3650                         if (state == NULL)
3651                                 break;
3652                         nfs4_schedule_stateid_recovery(server, state);
3653                         goto wait_on_recovery;
3654                 case -NFS4ERR_STALE_STATEID:
3655                 case -NFS4ERR_STALE_CLIENTID:
3656                 case -NFS4ERR_EXPIRED:
3657                         nfs4_schedule_lease_recovery(clp);
3658                         goto wait_on_recovery;
3659 #if defined(CONFIG_NFS_V4_1)
3660                 case -NFS4ERR_BADSESSION:
3661                 case -NFS4ERR_BADSLOT:
3662                 case -NFS4ERR_BAD_HIGH_SLOT:
3663                 case -NFS4ERR_DEADSESSION:
3664                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3665                 case -NFS4ERR_SEQ_FALSE_RETRY:
3666                 case -NFS4ERR_SEQ_MISORDERED:
3667                         dprintk("%s ERROR %d, Reset session\n", __func__,
3668                                 task->tk_status);
3669                         nfs4_schedule_session_recovery(clp->cl_session);
3670                         task->tk_status = 0;
3671                         return -EAGAIN;
3672 #endif /* CONFIG_NFS_V4_1 */
3673                 case -NFS4ERR_DELAY:
3674                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3675                 case -NFS4ERR_GRACE:
3676                 case -EKEYEXPIRED:
3677                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3678                         task->tk_status = 0;
3679                         return -EAGAIN;
3680                 case -NFS4ERR_OLD_STATEID:
3681                         task->tk_status = 0;
3682                         return -EAGAIN;
3683         }
3684         task->tk_status = nfs4_map_errors(task->tk_status);
3685         return 0;
3686 wait_on_recovery:
3687         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3688         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3689                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3690         task->tk_status = 0;
3691         return -EAGAIN;
3692 }
3693
3694 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3695                 unsigned short port, struct rpc_cred *cred,
3696                 struct nfs4_setclientid_res *res)
3697 {
3698         nfs4_verifier sc_verifier;
3699         struct nfs4_setclientid setclientid = {
3700                 .sc_verifier = &sc_verifier,
3701                 .sc_prog = program,
3702                 .sc_cb_ident = clp->cl_cb_ident,
3703         };
3704         struct rpc_message msg = {
3705                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3706                 .rpc_argp = &setclientid,
3707                 .rpc_resp = res,
3708                 .rpc_cred = cred,
3709         };
3710         __be32 *p;
3711         int loop = 0;
3712         int status;
3713
3714         p = (__be32*)sc_verifier.data;
3715         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3716         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3717
3718         for(;;) {
3719                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3720                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3721                                 clp->cl_ipaddr,
3722                                 rpc_peeraddr2str(clp->cl_rpcclient,
3723                                                         RPC_DISPLAY_ADDR),
3724                                 rpc_peeraddr2str(clp->cl_rpcclient,
3725                                                         RPC_DISPLAY_PROTO),
3726                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3727                                 clp->cl_id_uniquifier);
3728                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3729                                 sizeof(setclientid.sc_netid),
3730                                 rpc_peeraddr2str(clp->cl_rpcclient,
3731                                                         RPC_DISPLAY_NETID));
3732                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3733                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3734                                 clp->cl_ipaddr, port >> 8, port & 255);
3735
3736                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3737                 if (status != -NFS4ERR_CLID_INUSE)
3738                         break;
3739                 if (signalled())
3740                         break;
3741                 if (loop++ & 1)
3742                         ssleep(clp->cl_lease_time / HZ + 1);
3743                 else
3744                         if (++clp->cl_id_uniquifier == 0)
3745                                 break;
3746         }
3747         return status;
3748 }
3749
3750 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3751                 struct nfs4_setclientid_res *arg,
3752                 struct rpc_cred *cred)
3753 {
3754         struct nfs_fsinfo fsinfo;
3755         struct rpc_message msg = {
3756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3757                 .rpc_argp = arg,
3758                 .rpc_resp = &fsinfo,
3759                 .rpc_cred = cred,
3760         };
3761         unsigned long now;
3762         int status;
3763
3764         now = jiffies;
3765         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3766         if (status == 0) {
3767                 spin_lock(&clp->cl_lock);
3768                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3769                 clp->cl_last_renewal = now;
3770                 spin_unlock(&clp->cl_lock);
3771         }
3772         return status;
3773 }
3774
3775 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3776                 struct nfs4_setclientid_res *arg,
3777                 struct rpc_cred *cred)
3778 {
3779         long timeout = 0;
3780         int err;
3781         do {
3782                 err = _nfs4_proc_setclientid_confirm(clp, arg, cred);
3783                 switch (err) {
3784                         case 0:
3785                                 return err;
3786                         case -NFS4ERR_RESOURCE:
3787                                 /* The IBM lawyers misread another document! */
3788                         case -NFS4ERR_DELAY:
3789                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3790                 }
3791         } while (err == 0);
3792         return err;
3793 }
3794
3795 struct nfs4_delegreturndata {
3796         struct nfs4_delegreturnargs args;
3797         struct nfs4_delegreturnres res;
3798         struct nfs_fh fh;
3799         nfs4_stateid stateid;
3800         unsigned long timestamp;
3801         struct nfs_fattr fattr;
3802         int rpc_status;
3803 };
3804
3805 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3806 {
3807         struct nfs4_delegreturndata *data = calldata;
3808
3809         if (!nfs4_sequence_done(task, &data->res.seq_res))
3810                 return;
3811
3812         switch (task->tk_status) {
3813         case -NFS4ERR_STALE_STATEID:
3814         case -NFS4ERR_EXPIRED:
3815         case 0:
3816                 renew_lease(data->res.server, data->timestamp);
3817                 break;
3818         default:
3819                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3820                                 -EAGAIN) {
3821                         nfs_restart_rpc(task, data->res.server->nfs_client);
3822                         return;
3823                 }
3824         }
3825         data->rpc_status = task->tk_status;
3826 }
3827
3828 static void nfs4_delegreturn_release(void *calldata)
3829 {
3830         kfree(calldata);
3831 }
3832
3833 #if defined(CONFIG_NFS_V4_1)
3834 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3835 {
3836         struct nfs4_delegreturndata *d_data;
3837
3838         d_data = (struct nfs4_delegreturndata *)data;
3839
3840         if (nfs4_setup_sequence(d_data->res.server,
3841                                 &d_data->args.seq_args,
3842                                 &d_data->res.seq_res, 1, task))
3843                 return;
3844         rpc_call_start(task);
3845 }
3846 #endif /* CONFIG_NFS_V4_1 */
3847
3848 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3849 #if defined(CONFIG_NFS_V4_1)
3850         .rpc_call_prepare = nfs4_delegreturn_prepare,
3851 #endif /* CONFIG_NFS_V4_1 */
3852         .rpc_call_done = nfs4_delegreturn_done,
3853         .rpc_release = nfs4_delegreturn_release,
3854 };
3855
3856 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3857 {
3858         struct nfs4_delegreturndata *data;
3859         struct nfs_server *server = NFS_SERVER(inode);
3860         struct rpc_task *task;
3861         struct rpc_message msg = {
3862                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3863                 .rpc_cred = cred,
3864         };
3865         struct rpc_task_setup task_setup_data = {
3866                 .rpc_client = server->client,
3867                 .rpc_message = &msg,
3868                 .callback_ops = &nfs4_delegreturn_ops,
3869                 .flags = RPC_TASK_ASYNC,
3870         };
3871         int status = 0;
3872
3873         data = kzalloc(sizeof(*data), GFP_NOFS);
3874         if (data == NULL)
3875                 return -ENOMEM;
3876         data->args.fhandle = &data->fh;
3877         data->args.stateid = &data->stateid;
3878         data->args.bitmask = server->attr_bitmask;
3879         nfs_copy_fh(&data->fh, NFS_FH(inode));
3880         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3881         data->res.fattr = &data->fattr;
3882         data->res.server = server;
3883         nfs_fattr_init(data->res.fattr);
3884         data->timestamp = jiffies;
3885         data->rpc_status = 0;
3886
3887         task_setup_data.callback_data = data;
3888         msg.rpc_argp = &data->args;
3889         msg.rpc_resp = &data->res;
3890         task = rpc_run_task(&task_setup_data);
3891         if (IS_ERR(task))
3892                 return PTR_ERR(task);
3893         if (!issync)
3894                 goto out;
3895         status = nfs4_wait_for_completion_rpc_task(task);
3896         if (status != 0)
3897                 goto out;
3898         status = data->rpc_status;
3899         if (status != 0)
3900                 goto out;
3901         nfs_refresh_inode(inode, &data->fattr);
3902 out:
3903         rpc_put_task(task);
3904         return status;
3905 }
3906
3907 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3908 {
3909         struct nfs_server *server = NFS_SERVER(inode);
3910         struct nfs4_exception exception = { };
3911         int err;
3912         do {
3913                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3914                 switch (err) {
3915                         case -NFS4ERR_STALE_STATEID:
3916                         case -NFS4ERR_EXPIRED:
3917                         case 0:
3918                                 return 0;
3919                 }
3920                 err = nfs4_handle_exception(server, err, &exception);
3921         } while (exception.retry);
3922         return err;
3923 }
3924
3925 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3926 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3927
3928 /*
3929  * sleep, with exponential backoff, and retry the LOCK operation.
3930  */
3931 static unsigned long
3932 nfs4_set_lock_task_retry(unsigned long timeout)
3933 {
3934         schedule_timeout_killable(timeout);
3935         timeout <<= 1;
3936         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3937                 return NFS4_LOCK_MAXTIMEOUT;
3938         return timeout;
3939 }
3940
3941 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3942 {
3943         struct inode *inode = state->inode;
3944         struct nfs_server *server = NFS_SERVER(inode);
3945         struct nfs_client *clp = server->nfs_client;
3946         struct nfs_lockt_args arg = {
3947                 .fh = NFS_FH(inode),
3948                 .fl = request,
3949         };
3950         struct nfs_lockt_res res = {
3951                 .denied = request,
3952         };
3953         struct rpc_message msg = {
3954                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3955                 .rpc_argp       = &arg,
3956                 .rpc_resp       = &res,
3957                 .rpc_cred       = state->owner->so_cred,
3958         };
3959         struct nfs4_lock_state *lsp;
3960         int status;
3961
3962         arg.lock_owner.clientid = clp->cl_clientid;
3963         status = nfs4_set_lock_state(state, request);
3964         if (status != 0)
3965                 goto out;
3966         lsp = request->fl_u.nfs4_fl.owner;
3967         arg.lock_owner.id = lsp->ls_id.id;
3968         arg.lock_owner.s_dev = server->s_dev;
3969         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3970         switch (status) {
3971                 case 0:
3972                         request->fl_type = F_UNLCK;
3973                         break;
3974                 case -NFS4ERR_DENIED:
3975                         status = 0;
3976         }
3977         request->fl_ops->fl_release_private(request);
3978 out:
3979         return status;
3980 }
3981
3982 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3983 {
3984         struct nfs4_exception exception = { };
3985         int err;
3986
3987         do {
3988                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3989                                 _nfs4_proc_getlk(state, cmd, request),
3990                                 &exception);
3991         } while (exception.retry);
3992         return err;
3993 }
3994
3995 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3996 {
3997         int res = 0;
3998         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3999                 case FL_POSIX:
4000                         res = posix_lock_file_wait(file, fl);
4001                         break;
4002                 case FL_FLOCK:
4003                         res = flock_lock_file_wait(file, fl);
4004                         break;
4005                 default:
4006                         BUG();
4007         }
4008         return res;
4009 }
4010
4011 struct nfs4_unlockdata {
4012         struct nfs_locku_args arg;
4013         struct nfs_locku_res res;
4014         struct nfs4_lock_state *lsp;
4015         struct nfs_open_context *ctx;
4016         struct file_lock fl;
4017         const struct nfs_server *server;
4018         unsigned long timestamp;
4019 };
4020
4021 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4022                 struct nfs_open_context *ctx,
4023                 struct nfs4_lock_state *lsp,
4024                 struct nfs_seqid *seqid)
4025 {
4026         struct nfs4_unlockdata *p;
4027         struct inode *inode = lsp->ls_state->inode;
4028
4029         p = kzalloc(sizeof(*p), GFP_NOFS);
4030         if (p == NULL)
4031                 return NULL;
4032         p->arg.fh = NFS_FH(inode);
4033         p->arg.fl = &p->fl;
4034         p->arg.seqid = seqid;
4035         p->res.seqid = seqid;
4036         p->arg.stateid = &lsp->ls_stateid;
4037         p->lsp = lsp;
4038         atomic_inc(&lsp->ls_count);
4039         /* Ensure we don't close file until we're done freeing locks! */
4040         p->ctx = get_nfs_open_context(ctx);
4041         memcpy(&p->fl, fl, sizeof(p->fl));
4042         p->server = NFS_SERVER(inode);
4043         return p;
4044 }
4045
4046 static void nfs4_locku_release_calldata(void *data)
4047 {
4048         struct nfs4_unlockdata *calldata = data;
4049         nfs_free_seqid(calldata->arg.seqid);
4050         nfs4_put_lock_state(calldata->lsp);
4051         put_nfs_open_context(calldata->ctx);
4052         kfree(calldata);
4053 }
4054
4055 static void nfs4_locku_done(struct rpc_task *task, void *data)
4056 {
4057         struct nfs4_unlockdata *calldata = data;
4058
4059         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4060                 return;
4061         switch (task->tk_status) {
4062                 case 0:
4063                         memcpy(calldata->lsp->ls_stateid.data,
4064                                         calldata->res.stateid.data,
4065                                         sizeof(calldata->lsp->ls_stateid.data));
4066                         renew_lease(calldata->server, calldata->timestamp);
4067                         break;
4068                 case -NFS4ERR_BAD_STATEID:
4069                 case -NFS4ERR_OLD_STATEID:
4070                 case -NFS4ERR_STALE_STATEID:
4071                 case -NFS4ERR_EXPIRED:
4072                         break;
4073                 default:
4074                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4075                                 nfs_restart_rpc(task,
4076                                                  calldata->server->nfs_client);
4077         }
4078 }
4079
4080 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4081 {
4082         struct nfs4_unlockdata *calldata = data;
4083
4084         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4085                 return;
4086         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4087                 /* Note: exit _without_ running nfs4_locku_done */
4088                 task->tk_action = NULL;
4089                 return;
4090         }
4091         calldata->timestamp = jiffies;
4092         if (nfs4_setup_sequence(calldata->server,
4093                                 &calldata->arg.seq_args,
4094                                 &calldata->res.seq_res, 1, task))
4095                 return;
4096         rpc_call_start(task);
4097 }
4098
4099 static const struct rpc_call_ops nfs4_locku_ops = {
4100         .rpc_call_prepare = nfs4_locku_prepare,
4101         .rpc_call_done = nfs4_locku_done,
4102         .rpc_release = nfs4_locku_release_calldata,
4103 };
4104
4105 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4106                 struct nfs_open_context *ctx,
4107                 struct nfs4_lock_state *lsp,
4108                 struct nfs_seqid *seqid)
4109 {
4110         struct nfs4_unlockdata *data;
4111         struct rpc_message msg = {
4112                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4113                 .rpc_cred = ctx->cred,
4114         };
4115         struct rpc_task_setup task_setup_data = {
4116                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4117                 .rpc_message = &msg,
4118                 .callback_ops = &nfs4_locku_ops,
4119                 .workqueue = nfsiod_workqueue,
4120                 .flags = RPC_TASK_ASYNC,
4121         };
4122
4123         /* Ensure this is an unlock - when canceling a lock, the
4124          * canceled lock is passed in, and it won't be an unlock.
4125          */
4126         fl->fl_type = F_UNLCK;
4127
4128         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4129         if (data == NULL) {
4130                 nfs_free_seqid(seqid);
4131                 return ERR_PTR(-ENOMEM);
4132         }
4133
4134         msg.rpc_argp = &data->arg;
4135         msg.rpc_resp = &data->res;
4136         task_setup_data.callback_data = data;
4137         return rpc_run_task(&task_setup_data);
4138 }
4139
4140 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4141 {
4142         struct nfs_inode *nfsi = NFS_I(state->inode);
4143         struct nfs_seqid *seqid;
4144         struct nfs4_lock_state *lsp;
4145         struct rpc_task *task;
4146         int status = 0;
4147         unsigned char fl_flags = request->fl_flags;
4148
4149         status = nfs4_set_lock_state(state, request);
4150         /* Unlock _before_ we do the RPC call */
4151         request->fl_flags |= FL_EXISTS;
4152         down_read(&nfsi->rwsem);
4153         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4154                 up_read(&nfsi->rwsem);
4155                 goto out;
4156         }
4157         up_read(&nfsi->rwsem);
4158         if (status != 0)
4159                 goto out;
4160         /* Is this a delegated lock? */
4161         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4162                 goto out;
4163         lsp = request->fl_u.nfs4_fl.owner;
4164         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4165         status = -ENOMEM;
4166         if (seqid == NULL)
4167                 goto out;
4168         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4169         status = PTR_ERR(task);
4170         if (IS_ERR(task))
4171                 goto out;
4172         status = nfs4_wait_for_completion_rpc_task(task);
4173         rpc_put_task(task);
4174 out:
4175         request->fl_flags = fl_flags;
4176         return status;
4177 }
4178
4179 struct nfs4_lockdata {
4180         struct nfs_lock_args arg;
4181         struct nfs_lock_res res;
4182         struct nfs4_lock_state *lsp;
4183         struct nfs_open_context *ctx;
4184         struct file_lock fl;
4185         unsigned long timestamp;
4186         int rpc_status;
4187         int cancelled;
4188         struct nfs_server *server;
4189 };
4190
4191 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4192                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4193                 gfp_t gfp_mask)
4194 {
4195         struct nfs4_lockdata *p;
4196         struct inode *inode = lsp->ls_state->inode;
4197         struct nfs_server *server = NFS_SERVER(inode);
4198
4199         p = kzalloc(sizeof(*p), gfp_mask);
4200         if (p == NULL)
4201                 return NULL;
4202
4203         p->arg.fh = NFS_FH(inode);
4204         p->arg.fl = &p->fl;
4205         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4206         if (p->arg.open_seqid == NULL)
4207                 goto out_free;
4208         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4209         if (p->arg.lock_seqid == NULL)
4210                 goto out_free_seqid;
4211         p->arg.lock_stateid = &lsp->ls_stateid;
4212         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4213         p->arg.lock_owner.id = lsp->ls_id.id;
4214         p->arg.lock_owner.s_dev = server->s_dev;
4215         p->res.lock_seqid = p->arg.lock_seqid;
4216         p->lsp = lsp;
4217         p->server = server;
4218         atomic_inc(&lsp->ls_count);
4219         p->ctx = get_nfs_open_context(ctx);
4220         memcpy(&p->fl, fl, sizeof(p->fl));
4221         return p;
4222 out_free_seqid:
4223         nfs_free_seqid(p->arg.open_seqid);
4224 out_free:
4225         kfree(p);
4226         return NULL;
4227 }
4228
4229 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4230 {
4231         struct nfs4_lockdata *data = calldata;
4232         struct nfs4_state *state = data->lsp->ls_state;
4233
4234         dprintk("%s: begin!\n", __func__);
4235         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4236                 return;
4237         /* Do we need to do an open_to_lock_owner? */
4238         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4239                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4240                         return;
4241                 data->arg.open_stateid = &state->stateid;
4242                 data->arg.new_lock_owner = 1;
4243                 data->res.open_seqid = data->arg.open_seqid;
4244         } else
4245                 data->arg.new_lock_owner = 0;
4246         data->timestamp = jiffies;
4247         if (nfs4_setup_sequence(data->server,
4248                                 &data->arg.seq_args,
4249                                 &data->res.seq_res, 1, task))
4250                 return;
4251         rpc_call_start(task);
4252         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4253 }
4254
4255 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4256 {
4257         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4258         nfs4_lock_prepare(task, calldata);
4259 }
4260
4261 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4262 {
4263         struct nfs4_lockdata *data = calldata;
4264
4265         dprintk("%s: begin!\n", __func__);
4266
4267         if (!nfs4_sequence_done(task, &data->res.seq_res))
4268                 return;
4269
4270         data->rpc_status = task->tk_status;
4271         if (data->arg.new_lock_owner != 0) {
4272                 if (data->rpc_status == 0)
4273                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4274                 else
4275                         goto out;
4276         }
4277         if (data->rpc_status == 0) {
4278                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4279                                         sizeof(data->lsp->ls_stateid.data));
4280                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4281                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4282         }
4283 out:
4284         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4285 }
4286
4287 static void nfs4_lock_release(void *calldata)
4288 {
4289         struct nfs4_lockdata *data = calldata;
4290
4291         dprintk("%s: begin!\n", __func__);
4292         nfs_free_seqid(data->arg.open_seqid);
4293         if (data->cancelled != 0) {
4294                 struct rpc_task *task;
4295                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4296                                 data->arg.lock_seqid);
4297                 if (!IS_ERR(task))
4298                         rpc_put_task_async(task);
4299                 dprintk("%s: cancelling lock!\n", __func__);
4300         } else
4301                 nfs_free_seqid(data->arg.lock_seqid);
4302         nfs4_put_lock_state(data->lsp);
4303         put_nfs_open_context(data->ctx);
4304         kfree(data);
4305         dprintk("%s: done!\n", __func__);
4306 }
4307
4308 static const struct rpc_call_ops nfs4_lock_ops = {
4309         .rpc_call_prepare = nfs4_lock_prepare,
4310         .rpc_call_done = nfs4_lock_done,
4311         .rpc_release = nfs4_lock_release,
4312 };
4313
4314 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4315         .rpc_call_prepare = nfs4_recover_lock_prepare,
4316         .rpc_call_done = nfs4_lock_done,
4317         .rpc_release = nfs4_lock_release,
4318 };
4319
4320 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4321 {
4322         switch (error) {
4323         case -NFS4ERR_ADMIN_REVOKED:
4324         case -NFS4ERR_BAD_STATEID:
4325                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4326                 if (new_lock_owner != 0 ||
4327                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4328                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4329                 break;
4330         case -NFS4ERR_STALE_STATEID:
4331                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4332         case -NFS4ERR_EXPIRED:
4333                 nfs4_schedule_lease_recovery(server->nfs_client);
4334         };
4335 }
4336
4337 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4338 {
4339         struct nfs4_lockdata *data;
4340         struct rpc_task *task;
4341         struct rpc_message msg = {
4342                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4343                 .rpc_cred = state->owner->so_cred,
4344         };
4345         struct rpc_task_setup task_setup_data = {
4346                 .rpc_client = NFS_CLIENT(state->inode),
4347                 .rpc_message = &msg,
4348                 .callback_ops = &nfs4_lock_ops,
4349                 .workqueue = nfsiod_workqueue,
4350                 .flags = RPC_TASK_ASYNC,
4351         };
4352         int ret;
4353
4354         dprintk("%s: begin!\n", __func__);
4355         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4356                         fl->fl_u.nfs4_fl.owner,
4357                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4358         if (data == NULL)
4359                 return -ENOMEM;
4360         if (IS_SETLKW(cmd))
4361                 data->arg.block = 1;
4362         if (recovery_type > NFS_LOCK_NEW) {
4363                 if (recovery_type == NFS_LOCK_RECLAIM)
4364                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4365                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4366         }
4367         msg.rpc_argp = &data->arg;
4368         msg.rpc_resp = &data->res;
4369         task_setup_data.callback_data = data;
4370         task = rpc_run_task(&task_setup_data);
4371         if (IS_ERR(task))
4372                 return PTR_ERR(task);
4373         ret = nfs4_wait_for_completion_rpc_task(task);
4374         if (ret == 0) {
4375                 ret = data->rpc_status;
4376                 if (ret)
4377                         nfs4_handle_setlk_error(data->server, data->lsp,
4378                                         data->arg.new_lock_owner, ret);
4379         } else
4380                 data->cancelled = 1;
4381         rpc_put_task(task);
4382         dprintk("%s: done, ret = %d!\n", __func__, ret);
4383         return ret;
4384 }
4385
4386 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4387 {
4388         struct nfs_server *server = NFS_SERVER(state->inode);
4389         struct nfs4_exception exception = { };
4390         int err;
4391
4392         do {
4393                 /* Cache the lock if possible... */
4394                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4395                         return 0;
4396                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4397                 if (err != -NFS4ERR_DELAY)
4398                         break;
4399                 nfs4_handle_exception(server, err, &exception);
4400         } while (exception.retry);
4401         return err;
4402 }
4403
4404 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4405 {
4406         struct nfs_server *server = NFS_SERVER(state->inode);
4407         struct nfs4_exception exception = { };
4408         int err;
4409
4410         err = nfs4_set_lock_state(state, request);
4411         if (err != 0)
4412                 return err;
4413         do {
4414                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4415                         return 0;
4416                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4417                 switch (err) {
4418                 default:
4419                         goto out;
4420                 case -NFS4ERR_GRACE:
4421                 case -NFS4ERR_DELAY:
4422                         nfs4_handle_exception(server, err, &exception);
4423                         err = 0;
4424                 }
4425         } while (exception.retry);
4426 out:
4427         return err;
4428 }
4429
4430 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4431 {
4432         struct nfs_inode *nfsi = NFS_I(state->inode);
4433         unsigned char fl_flags = request->fl_flags;
4434         int status = -ENOLCK;
4435
4436         if ((fl_flags & FL_POSIX) &&
4437                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4438                 goto out;
4439         /* Is this a delegated open? */
4440         status = nfs4_set_lock_state(state, request);
4441         if (status != 0)
4442                 goto out;
4443         request->fl_flags |= FL_ACCESS;
4444         status = do_vfs_lock(request->fl_file, request);
4445         if (status < 0)
4446                 goto out;
4447         down_read(&nfsi->rwsem);
4448         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4449                 /* Yes: cache locks! */
4450                 /* ...but avoid races with delegation recall... */
4451                 request->fl_flags = fl_flags & ~FL_SLEEP;
4452                 status = do_vfs_lock(request->fl_file, request);
4453                 goto out_unlock;
4454         }
4455         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4456         if (status != 0)
4457                 goto out_unlock;
4458         /* Note: we always want to sleep here! */
4459         request->fl_flags = fl_flags | FL_SLEEP;
4460         if (do_vfs_lock(request->fl_file, request) < 0)
4461                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4462 out_unlock:
4463         up_read(&nfsi->rwsem);
4464 out:
4465         request->fl_flags = fl_flags;
4466         return status;
4467 }
4468
4469 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4470 {
4471         struct nfs4_exception exception = { };
4472         int err;
4473
4474         do {
4475                 err = _nfs4_proc_setlk(state, cmd, request);
4476                 if (err == -NFS4ERR_DENIED)
4477                         err = -EAGAIN;
4478                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4479                                 err, &exception);
4480         } while (exception.retry);
4481         return err;
4482 }
4483
4484 static int
4485 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4486 {
4487         struct nfs_open_context *ctx;
4488         struct nfs4_state *state;
4489         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4490         int status;
4491
4492         /* verify open state */
4493         ctx = nfs_file_open_context(filp);
4494         state = ctx->state;
4495
4496         if (request->fl_start < 0 || request->fl_end < 0)
4497                 return -EINVAL;
4498
4499         if (IS_GETLK(cmd)) {
4500                 if (state != NULL)
4501                         return nfs4_proc_getlk(state, F_GETLK, request);
4502                 return 0;
4503         }
4504
4505         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4506                 return -EINVAL;
4507
4508         if (request->fl_type == F_UNLCK) {
4509                 if (state != NULL)
4510                         return nfs4_proc_unlck(state, cmd, request);
4511                 return 0;
4512         }
4513
4514         if (state == NULL)
4515                 return -ENOLCK;
4516         do {
4517                 status = nfs4_proc_setlk(state, cmd, request);
4518                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4519                         break;
4520                 timeout = nfs4_set_lock_task_retry(timeout);
4521                 status = -ERESTARTSYS;
4522                 if (signalled())
4523                         break;
4524         } while(status < 0);
4525         return status;
4526 }
4527
4528 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4529 {
4530         struct nfs_server *server = NFS_SERVER(state->inode);
4531         struct nfs4_exception exception = { };
4532         int err;
4533
4534         err = nfs4_set_lock_state(state, fl);
4535         if (err != 0)
4536                 goto out;
4537         do {
4538                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4539                 switch (err) {
4540                         default:
4541                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4542                                                 __func__, err);
4543                         case 0:
4544                         case -ESTALE:
4545                                 goto out;
4546                         case -NFS4ERR_EXPIRED:
4547                         case -NFS4ERR_STALE_CLIENTID:
4548                         case -NFS4ERR_STALE_STATEID:
4549                                 nfs4_schedule_lease_recovery(server->nfs_client);
4550                                 goto out;
4551                         case -NFS4ERR_BADSESSION:
4552                         case -NFS4ERR_BADSLOT:
4553                         case -NFS4ERR_BAD_HIGH_SLOT:
4554                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4555                         case -NFS4ERR_DEADSESSION:
4556                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4557                                 goto out;
4558                         case -ERESTARTSYS:
4559                                 /*
4560                                  * The show must go on: exit, but mark the
4561                                  * stateid as needing recovery.
4562                                  */
4563                         case -NFS4ERR_ADMIN_REVOKED:
4564                         case -NFS4ERR_BAD_STATEID:
4565                         case -NFS4ERR_OPENMODE:
4566                                 nfs4_schedule_stateid_recovery(server, state);
4567                                 err = 0;
4568                                 goto out;
4569                         case -EKEYEXPIRED:
4570                                 /*
4571                                  * User RPCSEC_GSS context has expired.
4572                                  * We cannot recover this stateid now, so
4573                                  * skip it and allow recovery thread to
4574                                  * proceed.
4575                                  */
4576                                 err = 0;
4577                                 goto out;
4578                         case -ENOMEM:
4579                         case -NFS4ERR_DENIED:
4580                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4581                                 err = 0;
4582                                 goto out;
4583                         case -NFS4ERR_DELAY:
4584                                 break;
4585                 }
4586                 err = nfs4_handle_exception(server, err, &exception);
4587         } while (exception.retry);
4588 out:
4589         return err;
4590 }
4591
4592 static void nfs4_release_lockowner_release(void *calldata)
4593 {
4594         kfree(calldata);
4595 }
4596
4597 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4598         .rpc_release = nfs4_release_lockowner_release,
4599 };
4600
4601 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4602 {
4603         struct nfs_server *server = lsp->ls_state->owner->so_server;
4604         struct nfs_release_lockowner_args *args;
4605         struct rpc_message msg = {
4606                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4607         };
4608
4609         if (server->nfs_client->cl_mvops->minor_version != 0)
4610                 return;
4611         args = kmalloc(sizeof(*args), GFP_NOFS);
4612         if (!args)
4613                 return;
4614         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4615         args->lock_owner.id = lsp->ls_id.id;
4616         args->lock_owner.s_dev = server->s_dev;
4617         msg.rpc_argp = args;
4618         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4619 }
4620
4621 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4622
4623 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4624                                    const void *buf, size_t buflen,
4625                                    int flags, int type)
4626 {
4627         if (strcmp(key, "") != 0)
4628                 return -EINVAL;
4629
4630         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4631 }
4632
4633 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4634                                    void *buf, size_t buflen, int type)
4635 {
4636         if (strcmp(key, "") != 0)
4637                 return -EINVAL;
4638
4639         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4640 }
4641
4642 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4643                                        size_t list_len, const char *name,
4644                                        size_t name_len, int type)
4645 {
4646         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4647
4648         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4649                 return 0;
4650
4651         if (list && len <= list_len)
4652                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4653         return len;
4654 }
4655
4656 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4657 {
4658         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4659                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4660                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4661                 return;
4662
4663         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4664                 NFS_ATTR_FATTR_NLINK;
4665         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4666         fattr->nlink = 2;
4667 }
4668
4669 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4670                 struct nfs4_fs_locations *fs_locations, struct page *page)
4671 {
4672         struct nfs_server *server = NFS_SERVER(dir);
4673         u32 bitmask[2] = {
4674                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4675                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4676         };
4677         struct nfs4_fs_locations_arg args = {
4678                 .dir_fh = NFS_FH(dir),
4679                 .name = name,
4680                 .page = page,
4681                 .bitmask = bitmask,
4682         };
4683         struct nfs4_fs_locations_res res = {
4684                 .fs_locations = fs_locations,
4685         };
4686         struct rpc_message msg = {
4687                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4688                 .rpc_argp = &args,
4689                 .rpc_resp = &res,
4690         };
4691         int status;
4692
4693         dprintk("%s: start\n", __func__);
4694         nfs_fattr_init(&fs_locations->fattr);
4695         fs_locations->server = server;
4696         fs_locations->nlocations = 0;
4697         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4698         nfs_fixup_referral_attributes(&fs_locations->fattr);
4699         dprintk("%s: returned status = %d\n", __func__, status);
4700         return status;
4701 }
4702
4703 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4704 {
4705         int status;
4706         struct nfs4_secinfo_arg args = {
4707                 .dir_fh = NFS_FH(dir),
4708                 .name   = name,
4709         };
4710         struct nfs4_secinfo_res res = {
4711                 .flavors     = flavors,
4712         };
4713         struct rpc_message msg = {
4714                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4715                 .rpc_argp = &args,
4716                 .rpc_resp = &res,
4717         };
4718
4719         dprintk("NFS call  secinfo %s\n", name->name);
4720         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4721         dprintk("NFS reply  secinfo: %d\n", status);
4722         return status;
4723 }
4724
4725 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4726 {
4727         struct nfs4_exception exception = { };
4728         int err;
4729         do {
4730                 err = nfs4_handle_exception(NFS_SERVER(dir),
4731                                 _nfs4_proc_secinfo(dir, name, flavors),
4732                                 &exception);
4733         } while (exception.retry);
4734         return err;
4735 }
4736
4737 #ifdef CONFIG_NFS_V4_1
4738 /*
4739  * Check the exchange flags returned by the server for invalid flags, having
4740  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4741  * DS flags set.
4742  */
4743 static int nfs4_check_cl_exchange_flags(u32 flags)
4744 {
4745         if (flags & ~EXCHGID4_FLAG_MASK_R)
4746                 goto out_inval;
4747         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4748             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4749                 goto out_inval;
4750         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4751                 goto out_inval;
4752         return NFS_OK;
4753 out_inval:
4754         return -NFS4ERR_INVAL;
4755 }
4756
4757 /*
4758  * nfs4_proc_exchange_id()
4759  *
4760  * Since the clientid has expired, all compounds using sessions
4761  * associated with the stale clientid will be returning
4762  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4763  * be in some phase of session reset.
4764  */
4765 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4766 {
4767         nfs4_verifier verifier;
4768         struct nfs41_exchange_id_args args = {
4769                 .client = clp,
4770                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4771         };
4772         struct nfs41_exchange_id_res res = {
4773                 .client = clp,
4774         };
4775         int status;
4776         struct rpc_message msg = {
4777                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4778                 .rpc_argp = &args,
4779                 .rpc_resp = &res,
4780                 .rpc_cred = cred,
4781         };
4782         __be32 *p;
4783
4784         dprintk("--> %s\n", __func__);
4785         BUG_ON(clp == NULL);
4786
4787         p = (u32 *)verifier.data;
4788         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4789         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4790         args.verifier = &verifier;
4791
4792         args.id_len = scnprintf(args.id, sizeof(args.id),
4793                                 "%s/%s.%s/%u",
4794                                 clp->cl_ipaddr,
4795                                 init_utsname()->nodename,
4796                                 init_utsname()->domainname,
4797                                 clp->cl_rpcclient->cl_auth->au_flavor);
4798
4799         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4800         if (!status)
4801                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4802         dprintk("<-- %s status= %d\n", __func__, status);
4803         return status;
4804 }
4805
4806 struct nfs4_get_lease_time_data {
4807         struct nfs4_get_lease_time_args *args;
4808         struct nfs4_get_lease_time_res *res;
4809         struct nfs_client *clp;
4810 };
4811
4812 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4813                                         void *calldata)
4814 {
4815         int ret;
4816         struct nfs4_get_lease_time_data *data =
4817                         (struct nfs4_get_lease_time_data *)calldata;
4818
4819         dprintk("--> %s\n", __func__);
4820         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4821         /* just setup sequence, do not trigger session recovery
4822            since we're invoked within one */
4823         ret = nfs41_setup_sequence(data->clp->cl_session,
4824                                    &data->args->la_seq_args,
4825                                    &data->res->lr_seq_res, 0, task);
4826
4827         BUG_ON(ret == -EAGAIN);
4828         rpc_call_start(task);
4829         dprintk("<-- %s\n", __func__);
4830 }
4831
4832 /*
4833  * Called from nfs4_state_manager thread for session setup, so don't recover
4834  * from sequence operation or clientid errors.
4835  */
4836 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4837 {
4838         struct nfs4_get_lease_time_data *data =
4839                         (struct nfs4_get_lease_time_data *)calldata;
4840
4841         dprintk("--> %s\n", __func__);
4842         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4843                 return;
4844         switch (task->tk_status) {
4845         case -NFS4ERR_DELAY:
4846         case -NFS4ERR_GRACE:
4847                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4848                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4849                 task->tk_status = 0;
4850                 nfs_restart_rpc(task, data->clp);
4851                 return;
4852         }
4853         dprintk("<-- %s\n", __func__);
4854 }
4855
4856 struct rpc_call_ops nfs4_get_lease_time_ops = {
4857         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4858         .rpc_call_done = nfs4_get_lease_time_done,
4859 };
4860
4861 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4862 {
4863         struct rpc_task *task;
4864         struct nfs4_get_lease_time_args args;
4865         struct nfs4_get_lease_time_res res = {
4866                 .lr_fsinfo = fsinfo,
4867         };
4868         struct nfs4_get_lease_time_data data = {
4869                 .args = &args,
4870                 .res = &res,
4871                 .clp = clp,
4872         };
4873         struct rpc_message msg = {
4874                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4875                 .rpc_argp = &args,
4876                 .rpc_resp = &res,
4877         };
4878         struct rpc_task_setup task_setup = {
4879                 .rpc_client = clp->cl_rpcclient,
4880                 .rpc_message = &msg,
4881                 .callback_ops = &nfs4_get_lease_time_ops,
4882                 .callback_data = &data
4883         };
4884         int status;
4885
4886         dprintk("--> %s\n", __func__);
4887         task = rpc_run_task(&task_setup);
4888
4889         if (IS_ERR(task))
4890                 status = PTR_ERR(task);
4891         else {
4892                 status = task->tk_status;
4893                 rpc_put_task(task);
4894         }
4895         dprintk("<-- %s return %d\n", __func__, status);
4896
4897         return status;
4898 }
4899
4900 /*
4901  * Reset a slot table
4902  */
4903 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4904                                  int ivalue)
4905 {
4906         struct nfs4_slot *new = NULL;
4907         int i;
4908         int ret = 0;
4909
4910         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4911                 max_reqs, tbl->max_slots);
4912
4913         /* Does the newly negotiated max_reqs match the existing slot table? */
4914         if (max_reqs != tbl->max_slots) {
4915                 ret = -ENOMEM;
4916                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4917                               GFP_NOFS);
4918                 if (!new)
4919                         goto out;
4920                 ret = 0;
4921                 kfree(tbl->slots);
4922         }
4923         spin_lock(&tbl->slot_tbl_lock);
4924         if (new) {
4925                 tbl->slots = new;
4926                 tbl->max_slots = max_reqs;
4927         }
4928         for (i = 0; i < tbl->max_slots; ++i)
4929                 tbl->slots[i].seq_nr = ivalue;
4930         spin_unlock(&tbl->slot_tbl_lock);
4931         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4932                 tbl, tbl->slots, tbl->max_slots);
4933 out:
4934         dprintk("<-- %s: return %d\n", __func__, ret);
4935         return ret;
4936 }
4937
4938 /*
4939  * Reset the forechannel and backchannel slot tables
4940  */
4941 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4942 {
4943         int status;
4944
4945         status = nfs4_reset_slot_table(&session->fc_slot_table,
4946                         session->fc_attrs.max_reqs, 1);
4947         if (status)
4948                 return status;
4949
4950         status = nfs4_reset_slot_table(&session->bc_slot_table,
4951                         session->bc_attrs.max_reqs, 0);
4952         return status;
4953 }
4954
4955 /* Destroy the slot table */
4956 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4957 {
4958         if (session->fc_slot_table.slots != NULL) {
4959                 kfree(session->fc_slot_table.slots);
4960                 session->fc_slot_table.slots = NULL;
4961         }
4962         if (session->bc_slot_table.slots != NULL) {
4963                 kfree(session->bc_slot_table.slots);
4964                 session->bc_slot_table.slots = NULL;
4965         }
4966         return;
4967 }
4968
4969 /*
4970  * Initialize slot table
4971  */
4972 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4973                 int max_slots, int ivalue)
4974 {
4975         struct nfs4_slot *slot;
4976         int ret = -ENOMEM;
4977
4978         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4979
4980         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4981
4982         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4983         if (!slot)
4984                 goto out;
4985         ret = 0;
4986
4987         spin_lock(&tbl->slot_tbl_lock);
4988         tbl->max_slots = max_slots;
4989         tbl->slots = slot;
4990         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4991         spin_unlock(&tbl->slot_tbl_lock);
4992         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4993                 tbl, tbl->slots, tbl->max_slots);
4994 out:
4995         dprintk("<-- %s: return %d\n", __func__, ret);
4996         return ret;
4997 }
4998
4999 /*
5000  * Initialize the forechannel and backchannel tables
5001  */
5002 static int nfs4_init_slot_tables(struct nfs4_session *session)
5003 {
5004         struct nfs4_slot_table *tbl;
5005         int status = 0;
5006
5007         tbl = &session->fc_slot_table;
5008         if (tbl->slots == NULL) {
5009                 status = nfs4_init_slot_table(tbl,
5010                                 session->fc_attrs.max_reqs, 1);
5011                 if (status)
5012                         return status;
5013         }
5014
5015         tbl = &session->bc_slot_table;
5016         if (tbl->slots == NULL) {
5017                 status = nfs4_init_slot_table(tbl,
5018                                 session->bc_attrs.max_reqs, 0);
5019                 if (status)
5020                         nfs4_destroy_slot_tables(session);
5021         }
5022
5023         return status;
5024 }
5025
5026 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5027 {
5028         struct nfs4_session *session;
5029         struct nfs4_slot_table *tbl;
5030
5031         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5032         if (!session)
5033                 return NULL;
5034
5035         tbl = &session->fc_slot_table;
5036         tbl->highest_used_slotid = -1;
5037         spin_lock_init(&tbl->slot_tbl_lock);
5038         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5039         init_completion(&tbl->complete);
5040
5041         tbl = &session->bc_slot_table;
5042         tbl->highest_used_slotid = -1;
5043         spin_lock_init(&tbl->slot_tbl_lock);
5044         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5045         init_completion(&tbl->complete);
5046
5047         session->session_state = 1<<NFS4_SESSION_INITING;
5048
5049         session->clp = clp;
5050         return session;
5051 }
5052
5053 void nfs4_destroy_session(struct nfs4_session *session)
5054 {
5055         nfs4_proc_destroy_session(session);
5056         dprintk("%s Destroy backchannel for xprt %p\n",
5057                 __func__, session->clp->cl_rpcclient->cl_xprt);
5058         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5059                                 NFS41_BC_MIN_CALLBACKS);
5060         nfs4_destroy_slot_tables(session);
5061         kfree(session);
5062 }
5063
5064 /*
5065  * Initialize the values to be used by the client in CREATE_SESSION
5066  * If nfs4_init_session set the fore channel request and response sizes,
5067  * use them.
5068  *
5069  * Set the back channel max_resp_sz_cached to zero to force the client to
5070  * always set csa_cachethis to FALSE because the current implementation
5071  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5072  */
5073 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5074 {
5075         struct nfs4_session *session = args->client->cl_session;
5076         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5077                      mxresp_sz = session->fc_attrs.max_resp_sz;
5078
5079         if (mxrqst_sz == 0)
5080                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5081         if (mxresp_sz == 0)
5082                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5083         /* Fore channel attributes */
5084         args->fc_attrs.headerpadsz = 0;
5085         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5086         args->fc_attrs.max_resp_sz = mxresp_sz;
5087         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5088         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5089
5090         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5091                 "max_ops=%u max_reqs=%u\n",
5092                 __func__,
5093                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5094                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5095
5096         /* Back channel attributes */
5097         args->bc_attrs.headerpadsz = 0;
5098         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5099         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5100         args->bc_attrs.max_resp_sz_cached = 0;
5101         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5102         args->bc_attrs.max_reqs = 1;
5103
5104         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5105                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5106                 __func__,
5107                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5108                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5109                 args->bc_attrs.max_reqs);
5110 }
5111
5112 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5113 {
5114         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5115         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5116
5117         if (rcvd->headerpadsz > sent->headerpadsz)
5118                 return -EINVAL;
5119         if (rcvd->max_resp_sz > sent->max_resp_sz)
5120                 return -EINVAL;
5121         /*
5122          * Our requested max_ops is the minimum we need; we're not
5123          * prepared to break up compounds into smaller pieces than that.
5124          * So, no point even trying to continue if the server won't
5125          * cooperate:
5126          */
5127         if (rcvd->max_ops < sent->max_ops)
5128                 return -EINVAL;
5129         if (rcvd->max_reqs == 0)
5130                 return -EINVAL;
5131         return 0;
5132 }
5133
5134 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5135 {
5136         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5137         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5138
5139         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5140                 return -EINVAL;
5141         if (rcvd->max_resp_sz < sent->max_resp_sz)
5142                 return -EINVAL;
5143         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5144                 return -EINVAL;
5145         /* These would render the backchannel useless: */
5146         if (rcvd->max_ops  == 0)
5147                 return -EINVAL;
5148         if (rcvd->max_reqs == 0)
5149                 return -EINVAL;
5150         return 0;
5151 }
5152
5153 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5154                                      struct nfs4_session *session)
5155 {
5156         int ret;
5157
5158         ret = nfs4_verify_fore_channel_attrs(args, session);
5159         if (ret)
5160                 return ret;
5161         return nfs4_verify_back_channel_attrs(args, session);
5162 }
5163
5164 static int _nfs4_proc_create_session(struct nfs_client *clp)
5165 {
5166         struct nfs4_session *session = clp->cl_session;
5167         struct nfs41_create_session_args args = {
5168                 .client = clp,
5169                 .cb_program = NFS4_CALLBACK,
5170         };
5171         struct nfs41_create_session_res res = {
5172                 .client = clp,
5173         };
5174         struct rpc_message msg = {
5175                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5176                 .rpc_argp = &args,
5177                 .rpc_resp = &res,
5178         };
5179         int status;
5180
5181         nfs4_init_channel_attrs(&args);
5182         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5183
5184         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5185
5186         if (!status)
5187                 /* Verify the session's negotiated channel_attrs values */
5188                 status = nfs4_verify_channel_attrs(&args, session);
5189         if (!status) {
5190                 /* Increment the clientid slot sequence id */
5191                 clp->cl_seqid++;
5192         }
5193
5194         return status;
5195 }
5196
5197 /*
5198  * Issues a CREATE_SESSION operation to the server.
5199  * It is the responsibility of the caller to verify the session is
5200  * expired before calling this routine.
5201  */
5202 int nfs4_proc_create_session(struct nfs_client *clp)
5203 {
5204         int status;
5205         unsigned *ptr;
5206         struct nfs4_session *session = clp->cl_session;
5207         long timeout = 0;
5208         int err;
5209
5210         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5211
5212         do {
5213                 status = _nfs4_proc_create_session(clp);
5214                 if (status == -NFS4ERR_DELAY) {
5215                         err = nfs4_delay(clp->cl_rpcclient, &timeout);
5216                         if (err)
5217                                 status = err;
5218                 }
5219         } while (status == -NFS4ERR_DELAY);
5220
5221         if (status)
5222                 goto out;
5223
5224         /* Init and reset the fore channel */
5225         status = nfs4_init_slot_tables(session);
5226         dprintk("slot table initialization returned %d\n", status);
5227         if (status)
5228                 goto out;
5229         status = nfs4_reset_slot_tables(session);
5230         dprintk("slot table reset returned %d\n", status);
5231         if (status)
5232                 goto out;
5233
5234         ptr = (unsigned *)&session->sess_id.data[0];
5235         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5236                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5237 out:
5238         dprintk("<-- %s\n", __func__);
5239         return status;
5240 }
5241
5242 /*
5243  * Issue the over-the-wire RPC DESTROY_SESSION.
5244  * The caller must serialize access to this routine.
5245  */
5246 int nfs4_proc_destroy_session(struct nfs4_session *session)
5247 {
5248         int status = 0;
5249         struct rpc_message msg;
5250
5251         dprintk("--> nfs4_proc_destroy_session\n");
5252
5253         /* session is still being setup */
5254         if (session->clp->cl_cons_state != NFS_CS_READY)
5255                 return status;
5256
5257         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5258         msg.rpc_argp = session;
5259         msg.rpc_resp = NULL;
5260         msg.rpc_cred = NULL;
5261         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5262
5263         if (status)
5264                 printk(KERN_WARNING
5265                         "Got error %d from the server on DESTROY_SESSION. "
5266                         "Session has been destroyed regardless...\n", status);
5267
5268         dprintk("<-- nfs4_proc_destroy_session\n");
5269         return status;
5270 }
5271
5272 int nfs4_init_session(struct nfs_server *server)
5273 {
5274         struct nfs_client *clp = server->nfs_client;
5275         struct nfs4_session *session;
5276         unsigned int rsize, wsize;
5277         int ret;
5278
5279         if (!nfs4_has_session(clp))
5280                 return 0;
5281
5282         session = clp->cl_session;
5283         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5284                 return 0;
5285
5286         rsize = server->rsize;
5287         if (rsize == 0)
5288                 rsize = NFS_MAX_FILE_IO_SIZE;
5289         wsize = server->wsize;
5290         if (wsize == 0)
5291                 wsize = NFS_MAX_FILE_IO_SIZE;
5292
5293         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5294         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5295
5296         ret = nfs4_recover_expired_lease(server);
5297         if (!ret)
5298                 ret = nfs4_check_client_ready(clp);
5299         return ret;
5300 }
5301
5302 int nfs4_init_ds_session(struct nfs_client *clp)
5303 {
5304         struct nfs4_session *session = clp->cl_session;
5305         int ret;
5306
5307         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5308                 return 0;
5309
5310         ret = nfs4_client_recover_expired_lease(clp);
5311         if (!ret)
5312                 /* Test for the DS role */
5313                 if (!is_ds_client(clp))
5314                         ret = -ENODEV;
5315         if (!ret)
5316                 ret = nfs4_check_client_ready(clp);
5317         return ret;
5318
5319 }
5320 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5321
5322
5323 /*
5324  * Renew the cl_session lease.
5325  */
5326 struct nfs4_sequence_data {
5327         struct nfs_client *clp;
5328         struct nfs4_sequence_args args;
5329         struct nfs4_sequence_res res;
5330 };
5331
5332 static void nfs41_sequence_release(void *data)
5333 {
5334         struct nfs4_sequence_data *calldata = data;
5335         struct nfs_client *clp = calldata->clp;
5336
5337         if (atomic_read(&clp->cl_count) > 1)
5338                 nfs4_schedule_state_renewal(clp);
5339         nfs_put_client(clp);
5340         kfree(calldata);
5341 }
5342
5343 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5344 {
5345         switch(task->tk_status) {
5346         case -NFS4ERR_DELAY:
5347                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5348                 return -EAGAIN;
5349         default:
5350                 nfs4_schedule_lease_recovery(clp);
5351         }
5352         return 0;
5353 }
5354
5355 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5356 {
5357         struct nfs4_sequence_data *calldata = data;
5358         struct nfs_client *clp = calldata->clp;
5359
5360         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5361                 return;
5362
5363         if (task->tk_status < 0) {
5364                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5365                 if (atomic_read(&clp->cl_count) == 1)
5366                         goto out;
5367
5368                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5369                         rpc_restart_call_prepare(task);
5370                         return;
5371                 }
5372         }
5373         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5374 out:
5375         dprintk("<-- %s\n", __func__);
5376 }
5377
5378 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5379 {
5380         struct nfs4_sequence_data *calldata = data;
5381         struct nfs_client *clp = calldata->clp;
5382         struct nfs4_sequence_args *args;
5383         struct nfs4_sequence_res *res;
5384
5385         args = task->tk_msg.rpc_argp;
5386         res = task->tk_msg.rpc_resp;
5387
5388         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5389                 return;
5390         rpc_call_start(task);
5391 }
5392
5393 static const struct rpc_call_ops nfs41_sequence_ops = {
5394         .rpc_call_done = nfs41_sequence_call_done,
5395         .rpc_call_prepare = nfs41_sequence_prepare,
5396         .rpc_release = nfs41_sequence_release,
5397 };
5398
5399 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5400 {
5401         struct nfs4_sequence_data *calldata;
5402         struct rpc_message msg = {
5403                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5404                 .rpc_cred = cred,
5405         };
5406         struct rpc_task_setup task_setup_data = {
5407                 .rpc_client = clp->cl_rpcclient,
5408                 .rpc_message = &msg,
5409                 .callback_ops = &nfs41_sequence_ops,
5410                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5411         };
5412
5413         if (!atomic_inc_not_zero(&clp->cl_count))
5414                 return ERR_PTR(-EIO);
5415         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5416         if (calldata == NULL) {
5417                 nfs_put_client(clp);
5418                 return ERR_PTR(-ENOMEM);
5419         }
5420         msg.rpc_argp = &calldata->args;
5421         msg.rpc_resp = &calldata->res;
5422         calldata->clp = clp;
5423         task_setup_data.callback_data = calldata;
5424
5425         return rpc_run_task(&task_setup_data);
5426 }
5427
5428 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5429 {
5430         struct rpc_task *task;
5431         int ret = 0;
5432
5433         task = _nfs41_proc_sequence(clp, cred);
5434         if (IS_ERR(task))
5435                 ret = PTR_ERR(task);
5436         else
5437                 rpc_put_task_async(task);
5438         dprintk("<-- %s status=%d\n", __func__, ret);
5439         return ret;
5440 }
5441
5442 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5443 {
5444         struct rpc_task *task;
5445         int ret;
5446
5447         task = _nfs41_proc_sequence(clp, cred);
5448         if (IS_ERR(task)) {
5449                 ret = PTR_ERR(task);
5450                 goto out;
5451         }
5452         ret = rpc_wait_for_completion_task(task);
5453         if (!ret) {
5454                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5455
5456                 if (task->tk_status == 0)
5457                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5458                 ret = task->tk_status;
5459         }
5460         rpc_put_task(task);
5461 out:
5462         dprintk("<-- %s status=%d\n", __func__, ret);
5463         return ret;
5464 }
5465
5466 struct nfs4_reclaim_complete_data {
5467         struct nfs_client *clp;
5468         struct nfs41_reclaim_complete_args arg;
5469         struct nfs41_reclaim_complete_res res;
5470 };
5471
5472 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5473 {
5474         struct nfs4_reclaim_complete_data *calldata = data;
5475
5476         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5477         if (nfs41_setup_sequence(calldata->clp->cl_session,
5478                                 &calldata->arg.seq_args,
5479                                 &calldata->res.seq_res, 0, task))
5480                 return;
5481
5482         rpc_call_start(task);
5483 }
5484
5485 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5486 {
5487         switch(task->tk_status) {
5488         case 0:
5489         case -NFS4ERR_COMPLETE_ALREADY:
5490         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5491                 break;
5492         case -NFS4ERR_DELAY:
5493                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5494                 return -EAGAIN;
5495         default:
5496                 nfs4_schedule_lease_recovery(clp);
5497         }
5498         return 0;
5499 }
5500
5501 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5502 {
5503         struct nfs4_reclaim_complete_data *calldata = data;
5504         struct nfs_client *clp = calldata->clp;
5505         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5506
5507         dprintk("--> %s\n", __func__);
5508         if (!nfs41_sequence_done(task, res))
5509                 return;
5510
5511         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5512                 rpc_restart_call_prepare(task);
5513                 return;
5514         }
5515         dprintk("<-- %s\n", __func__);
5516 }
5517
5518 static void nfs4_free_reclaim_complete_data(void *data)
5519 {
5520         struct nfs4_reclaim_complete_data *calldata = data;
5521
5522         kfree(calldata);
5523 }
5524
5525 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5526         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5527         .rpc_call_done = nfs4_reclaim_complete_done,
5528         .rpc_release = nfs4_free_reclaim_complete_data,
5529 };
5530
5531 /*
5532  * Issue a global reclaim complete.
5533  */
5534 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5535 {
5536         struct nfs4_reclaim_complete_data *calldata;
5537         struct rpc_task *task;
5538         struct rpc_message msg = {
5539                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5540         };
5541         struct rpc_task_setup task_setup_data = {
5542                 .rpc_client = clp->cl_rpcclient,
5543                 .rpc_message = &msg,
5544                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5545                 .flags = RPC_TASK_ASYNC,
5546         };
5547         int status = -ENOMEM;
5548
5549         dprintk("--> %s\n", __func__);
5550         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5551         if (calldata == NULL)
5552                 goto out;
5553         calldata->clp = clp;
5554         calldata->arg.one_fs = 0;
5555
5556         msg.rpc_argp = &calldata->arg;
5557         msg.rpc_resp = &calldata->res;
5558         task_setup_data.callback_data = calldata;
5559         task = rpc_run_task(&task_setup_data);
5560         if (IS_ERR(task)) {
5561                 status = PTR_ERR(task);
5562                 goto out;
5563         }
5564         status = nfs4_wait_for_completion_rpc_task(task);
5565         if (status == 0)
5566                 status = task->tk_status;
5567         rpc_put_task(task);
5568         return 0;
5569 out:
5570         dprintk("<-- %s status=%d\n", __func__, status);
5571         return status;
5572 }
5573
5574 static void
5575 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5576 {
5577         struct nfs4_layoutget *lgp = calldata;
5578         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5579
5580         dprintk("--> %s\n", __func__);
5581         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5582          * right now covering the LAYOUTGET we are about to send.
5583          * However, that is not so catastrophic, and there seems
5584          * to be no way to prevent it completely.
5585          */
5586         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5587                                 &lgp->res.seq_res, 0, task))
5588                 return;
5589         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5590                                           NFS_I(lgp->args.inode)->layout,
5591                                           lgp->args.ctx->state)) {
5592                 rpc_exit(task, NFS4_OK);
5593                 return;
5594         }
5595         rpc_call_start(task);
5596 }
5597
5598 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5599 {
5600         struct nfs4_layoutget *lgp = calldata;
5601         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5602
5603         dprintk("--> %s\n", __func__);
5604
5605         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5606                 return;
5607
5608         switch (task->tk_status) {
5609         case 0:
5610                 break;
5611         case -NFS4ERR_LAYOUTTRYLATER:
5612         case -NFS4ERR_RECALLCONFLICT:
5613                 task->tk_status = -NFS4ERR_DELAY;
5614                 /* Fall through */
5615         default:
5616                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5617                         rpc_restart_call_prepare(task);
5618                         return;
5619                 }
5620         }
5621         dprintk("<-- %s\n", __func__);
5622 }
5623
5624 static void nfs4_layoutget_release(void *calldata)
5625 {
5626         struct nfs4_layoutget *lgp = calldata;
5627
5628         dprintk("--> %s\n", __func__);
5629         put_nfs_open_context(lgp->args.ctx);
5630         kfree(calldata);
5631         dprintk("<-- %s\n", __func__);
5632 }
5633
5634 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5635         .rpc_call_prepare = nfs4_layoutget_prepare,
5636         .rpc_call_done = nfs4_layoutget_done,
5637         .rpc_release = nfs4_layoutget_release,
5638 };
5639
5640 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5641 {
5642         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5643         struct rpc_task *task;
5644         struct rpc_message msg = {
5645                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5646                 .rpc_argp = &lgp->args,
5647                 .rpc_resp = &lgp->res,
5648         };
5649         struct rpc_task_setup task_setup_data = {
5650                 .rpc_client = server->client,
5651                 .rpc_message = &msg,
5652                 .callback_ops = &nfs4_layoutget_call_ops,
5653                 .callback_data = lgp,
5654                 .flags = RPC_TASK_ASYNC,
5655         };
5656         int status = 0;
5657
5658         dprintk("--> %s\n", __func__);
5659
5660         lgp->res.layoutp = &lgp->args.layout;
5661         lgp->res.seq_res.sr_slot = NULL;
5662         task = rpc_run_task(&task_setup_data);
5663         if (IS_ERR(task))
5664                 return PTR_ERR(task);
5665         status = nfs4_wait_for_completion_rpc_task(task);
5666         if (status == 0)
5667                 status = task->tk_status;
5668         if (status == 0)
5669                 status = pnfs_layout_process(lgp);
5670         rpc_put_task(task);
5671         dprintk("<-- %s status=%d\n", __func__, status);
5672         return status;
5673 }
5674
5675 static int
5676 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5677 {
5678         struct nfs4_getdeviceinfo_args args = {
5679                 .pdev = pdev,
5680         };
5681         struct nfs4_getdeviceinfo_res res = {
5682                 .pdev = pdev,
5683         };
5684         struct rpc_message msg = {
5685                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5686                 .rpc_argp = &args,
5687                 .rpc_resp = &res,
5688         };
5689         int status;
5690
5691         dprintk("--> %s\n", __func__);
5692         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5693         dprintk("<-- %s status=%d\n", __func__, status);
5694
5695         return status;
5696 }
5697
5698 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5699 {
5700         struct nfs4_exception exception = { };
5701         int err;
5702
5703         do {
5704                 err = nfs4_handle_exception(server,
5705                                         _nfs4_proc_getdeviceinfo(server, pdev),
5706                                         &exception);
5707         } while (exception.retry);
5708         return err;
5709 }
5710 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5711
5712 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5713 {
5714         struct nfs4_layoutcommit_data *data = calldata;
5715         struct nfs_server *server = NFS_SERVER(data->args.inode);
5716
5717         if (nfs4_setup_sequence(server, &data->args.seq_args,
5718                                 &data->res.seq_res, 1, task))
5719                 return;
5720         rpc_call_start(task);
5721 }
5722
5723 static void
5724 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5725 {
5726         struct nfs4_layoutcommit_data *data = calldata;
5727         struct nfs_server *server = NFS_SERVER(data->args.inode);
5728
5729         if (!nfs4_sequence_done(task, &data->res.seq_res))
5730                 return;
5731
5732         switch (task->tk_status) { /* Just ignore these failures */
5733         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5734         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5735         case NFS4ERR_BADLAYOUT:     /* no layout */
5736         case NFS4ERR_GRACE:         /* loca_recalim always false */
5737                 task->tk_status = 0;
5738         }
5739
5740         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5741                 nfs_restart_rpc(task, server->nfs_client);
5742                 return;
5743         }
5744
5745         if (task->tk_status == 0)
5746                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5747                                                    data->res.fattr);
5748 }
5749
5750 static void nfs4_layoutcommit_release(void *calldata)
5751 {
5752         struct nfs4_layoutcommit_data *data = calldata;
5753
5754         /* Matched by references in pnfs_set_layoutcommit */
5755         put_lseg(data->lseg);
5756         put_rpccred(data->cred);
5757         kfree(data);
5758 }
5759
5760 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5761         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5762         .rpc_call_done = nfs4_layoutcommit_done,
5763         .rpc_release = nfs4_layoutcommit_release,
5764 };
5765
5766 int
5767 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5768 {
5769         struct rpc_message msg = {
5770                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5771                 .rpc_argp = &data->args,
5772                 .rpc_resp = &data->res,
5773                 .rpc_cred = data->cred,
5774         };
5775         struct rpc_task_setup task_setup_data = {
5776                 .task = &data->task,
5777                 .rpc_client = NFS_CLIENT(data->args.inode),
5778                 .rpc_message = &msg,
5779                 .callback_ops = &nfs4_layoutcommit_ops,
5780                 .callback_data = data,
5781                 .flags = RPC_TASK_ASYNC,
5782         };
5783         struct rpc_task *task;
5784         int status = 0;
5785
5786         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
5787                 "lbw: %llu inode %lu\n",
5788                 data->task.tk_pid, sync,
5789                 data->args.lastbytewritten,
5790                 data->args.inode->i_ino);
5791
5792         task = rpc_run_task(&task_setup_data);
5793         if (IS_ERR(task))
5794                 return PTR_ERR(task);
5795         if (sync == false)
5796                 goto out;
5797         status = nfs4_wait_for_completion_rpc_task(task);
5798         if (status != 0)
5799                 goto out;
5800         status = task->tk_status;
5801 out:
5802         dprintk("%s: status %d\n", __func__, status);
5803         rpc_put_task(task);
5804         return status;
5805 }
5806 #endif /* CONFIG_NFS_V4_1 */
5807
5808 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5809         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5810         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5811         .recover_open   = nfs4_open_reclaim,
5812         .recover_lock   = nfs4_lock_reclaim,
5813         .establish_clid = nfs4_init_clientid,
5814         .get_clid_cred  = nfs4_get_setclientid_cred,
5815 };
5816
5817 #if defined(CONFIG_NFS_V4_1)
5818 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5819         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5820         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5821         .recover_open   = nfs4_open_reclaim,
5822         .recover_lock   = nfs4_lock_reclaim,
5823         .establish_clid = nfs41_init_clientid,
5824         .get_clid_cred  = nfs4_get_exchange_id_cred,
5825         .reclaim_complete = nfs41_proc_reclaim_complete,
5826 };
5827 #endif /* CONFIG_NFS_V4_1 */
5828
5829 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5830         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5831         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5832         .recover_open   = nfs4_open_expired,
5833         .recover_lock   = nfs4_lock_expired,
5834         .establish_clid = nfs4_init_clientid,
5835         .get_clid_cred  = nfs4_get_setclientid_cred,
5836 };
5837
5838 #if defined(CONFIG_NFS_V4_1)
5839 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5840         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5841         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5842         .recover_open   = nfs4_open_expired,
5843         .recover_lock   = nfs4_lock_expired,
5844         .establish_clid = nfs41_init_clientid,
5845         .get_clid_cred  = nfs4_get_exchange_id_cred,
5846 };
5847 #endif /* CONFIG_NFS_V4_1 */
5848
5849 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5850         .sched_state_renewal = nfs4_proc_async_renew,
5851         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5852         .renew_lease = nfs4_proc_renew,
5853 };
5854
5855 #if defined(CONFIG_NFS_V4_1)
5856 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5857         .sched_state_renewal = nfs41_proc_async_sequence,
5858         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5859         .renew_lease = nfs4_proc_sequence,
5860 };
5861 #endif
5862
5863 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5864         .minor_version = 0,
5865         .call_sync = _nfs4_call_sync,
5866         .validate_stateid = nfs4_validate_delegation_stateid,
5867         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5868         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5869         .state_renewal_ops = &nfs40_state_renewal_ops,
5870 };
5871
5872 #if defined(CONFIG_NFS_V4_1)
5873 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5874         .minor_version = 1,
5875         .call_sync = _nfs4_call_sync_session,
5876         .validate_stateid = nfs41_validate_delegation_stateid,
5877         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5878         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5879         .state_renewal_ops = &nfs41_state_renewal_ops,
5880 };
5881 #endif
5882
5883 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5884         [0] = &nfs_v4_0_minor_ops,
5885 #if defined(CONFIG_NFS_V4_1)
5886         [1] = &nfs_v4_1_minor_ops,
5887 #endif
5888 };
5889
5890 static const struct inode_operations nfs4_file_inode_operations = {
5891         .permission     = nfs_permission,
5892         .getattr        = nfs_getattr,
5893         .setattr        = nfs_setattr,
5894         .getxattr       = generic_getxattr,
5895         .setxattr       = generic_setxattr,
5896         .listxattr      = generic_listxattr,
5897         .removexattr    = generic_removexattr,
5898 };
5899
5900 const struct nfs_rpc_ops nfs_v4_clientops = {
5901         .version        = 4,                    /* protocol version */
5902         .dentry_ops     = &nfs4_dentry_operations,
5903         .dir_inode_ops  = &nfs4_dir_inode_operations,
5904         .file_inode_ops = &nfs4_file_inode_operations,
5905         .getroot        = nfs4_proc_get_root,
5906         .getattr        = nfs4_proc_getattr,
5907         .setattr        = nfs4_proc_setattr,
5908         .lookupfh       = nfs4_proc_lookupfh,
5909         .lookup         = nfs4_proc_lookup,
5910         .access         = nfs4_proc_access,
5911         .readlink       = nfs4_proc_readlink,
5912         .create         = nfs4_proc_create,
5913         .remove         = nfs4_proc_remove,
5914         .unlink_setup   = nfs4_proc_unlink_setup,
5915         .unlink_done    = nfs4_proc_unlink_done,
5916         .rename         = nfs4_proc_rename,
5917         .rename_setup   = nfs4_proc_rename_setup,
5918         .rename_done    = nfs4_proc_rename_done,
5919         .link           = nfs4_proc_link,
5920         .symlink        = nfs4_proc_symlink,
5921         .mkdir          = nfs4_proc_mkdir,
5922         .rmdir          = nfs4_proc_remove,
5923         .readdir        = nfs4_proc_readdir,
5924         .mknod          = nfs4_proc_mknod,
5925         .statfs         = nfs4_proc_statfs,
5926         .fsinfo         = nfs4_proc_fsinfo,
5927         .pathconf       = nfs4_proc_pathconf,
5928         .set_capabilities = nfs4_server_capabilities,
5929         .decode_dirent  = nfs4_decode_dirent,
5930         .read_setup     = nfs4_proc_read_setup,
5931         .read_done      = nfs4_read_done,
5932         .write_setup    = nfs4_proc_write_setup,
5933         .write_done     = nfs4_write_done,
5934         .commit_setup   = nfs4_proc_commit_setup,
5935         .commit_done    = nfs4_commit_done,
5936         .lock           = nfs4_proc_lock,
5937         .clear_acl_cache = nfs4_zap_acl_attr,
5938         .close_context  = nfs4_close_context,
5939         .open_context   = nfs4_atomic_open,
5940         .init_client    = nfs4_init_client,
5941         .secinfo        = nfs4_proc_secinfo,
5942 };
5943
5944 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5945         .prefix = XATTR_NAME_NFSV4_ACL,
5946         .list   = nfs4_xattr_list_nfs4_acl,
5947         .get    = nfs4_xattr_get_nfs4_acl,
5948         .set    = nfs4_xattr_set_nfs4_acl,
5949 };
5950
5951 const struct xattr_handler *nfs4_xattr_handlers[] = {
5952         &nfs4_xattr_nfs4_acl_handler,
5953         NULL
5954 };
5955
5956 /*
5957  * Local variables:
5958  *  c-basic-offset: 8
5959  * End:
5960  */