fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/slab.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/nfs.h>
  45 #include <linux/nfs4.h>
  46 #include <linux/nfs_fs.h>
  47 #include <linux/nfs_page.h>
  48 #include <linux/namei.h>
  49 #include <linux/mount.h>
  50 #include <linux/module.h>
  51 #include <linux/sunrpc/bc_xprt.h>
  52 #include <linux/xattr.h>
  53 #include <linux/utsname.h>
  54
  55 #include "nfs4_fs.h"
  56 #include "delegation.h"
  57 #include "internal.h"
  58 #include "iostat.h"
  59 #include "callback.h"
  60 #include "pnfs.h"
  61
  62 #define NFSDBG_FACILITY         NFSDBG_PROC
  63
  64 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  65 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  66
  67 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  68
  69 struct nfs4_opendata;
  70 static int _nfs4_proc_open(struct nfs4_opendata *data);
  71 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  72 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  73 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  74 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  75 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  76 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  77                             struct nfs_fattr *fattr, struct iattr *sattr,
  78                             struct nfs4_state *state);
  79
  80 /* Prevent leaks of NFSv4 errors into userland */
  81 static int nfs4_map_errors(int err)
  82 {
  83         if (err >= -1000)
  84                 return err;
  85         switch (err) {
  86         case -NFS4ERR_RESOURCE:
  87                 return -EREMOTEIO;
  88         case -NFS4ERR_BADOWNER:
  89         case -NFS4ERR_BADNAME:
  90                 return -EINVAL;
  91         default:
  92                 dprintk("%s could not handle NFSv4 error %d\n",
  93                                 __func__, -err);
  94                 break;
  95         }
  96         return -EIO;
  97 }
  98
  99 /*
 100  * This is our standard bitmap for GETATTR requests.
 101  */
 102 const u32 nfs4_fattr_bitmap[2] = {
 103         FATTR4_WORD0_TYPE
 104         | FATTR4_WORD0_CHANGE
 105         | FATTR4_WORD0_SIZE
 106         | FATTR4_WORD0_FSID
 107         | FATTR4_WORD0_FILEID,
 108         FATTR4_WORD1_MODE
 109         | FATTR4_WORD1_NUMLINKS
 110         | FATTR4_WORD1_OWNER
 111         | FATTR4_WORD1_OWNER_GROUP
 112         | FATTR4_WORD1_RAWDEV
 113         | FATTR4_WORD1_SPACE_USED
 114         | FATTR4_WORD1_TIME_ACCESS
 115         | FATTR4_WORD1_TIME_METADATA
 116         | FATTR4_WORD1_TIME_MODIFY
 117 };
 118
 119 const u32 nfs4_statfs_bitmap[2] = {
 120         FATTR4_WORD0_FILES_AVAIL
 121         | FATTR4_WORD0_FILES_FREE
 122         | FATTR4_WORD0_FILES_TOTAL,
 123         FATTR4_WORD1_SPACE_AVAIL
 124         | FATTR4_WORD1_SPACE_FREE
 125         | FATTR4_WORD1_SPACE_TOTAL
 126 };
 127
 128 const u32 nfs4_pathconf_bitmap[2] = {
 129         FATTR4_WORD0_MAXLINK
 130         | FATTR4_WORD0_MAXNAME,
 131         0
 132 };
 133
 134 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 135                         | FATTR4_WORD0_MAXREAD
 136                         | FATTR4_WORD0_MAXWRITE
 137                         | FATTR4_WORD0_LEASE_TIME,
 138                         FATTR4_WORD1_TIME_DELTA
 139                         | FATTR4_WORD1_FS_LAYOUT_TYPES
 140 };
 141
 142 const u32 nfs4_fs_locations_bitmap[2] = {
 143         FATTR4_WORD0_TYPE
 144         | FATTR4_WORD0_CHANGE
 145         | FATTR4_WORD0_SIZE
 146         | FATTR4_WORD0_FSID
 147         | FATTR4_WORD0_FILEID
 148         | FATTR4_WORD0_FS_LOCATIONS,
 149         FATTR4_WORD1_MODE
 150         | FATTR4_WORD1_NUMLINKS
 151         | FATTR4_WORD1_OWNER
 152         | FATTR4_WORD1_OWNER_GROUP
 153         | FATTR4_WORD1_RAWDEV
 154         | FATTR4_WORD1_SPACE_USED
 155         | FATTR4_WORD1_TIME_ACCESS
 156         | FATTR4_WORD1_TIME_METADATA
 157         | FATTR4_WORD1_TIME_MODIFY
 158         | FATTR4_WORD1_MOUNTED_ON_FILEID
 159 };
 160
 161 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 162                 struct nfs4_readdir_arg *readdir)
 163 {
 164         __be32 *start, *p;
 165
 166         BUG_ON(readdir->count < 80);
 167         if (cookie > 2) {
 168                 readdir->cookie = cookie;
 169                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 170                 return;
 171         }
 172
 173         readdir->cookie = 0;
 174         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 175         if (cookie == 2)
 176                 return;
 177
 178         /*
 179          * NFSv4 servers do not return entries for '.' and '..'
 180          * Therefore, we fake these entries here.  We let '.'
 181          * have cookie 0 and '..' have cookie 1.  Note that
 182          * when talking to the server, we always send cookie 0
 183          * instead of 1 or 2.
 184          */
 185         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 186
 187         if (cookie == 0) {
 188                 *p++ = xdr_one;                                  /* next */
 189                 *p++ = xdr_zero;                   /* cookie, first word */
 190                 *p++ = xdr_one;                   /* cookie, second word */
 191                 *p++ = xdr_one;                             /* entry len */
 192                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 193                 p++;
 194                 *p++ = xdr_one;                         /* bitmap length */
 195                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 196                 *p++ = htonl(8);              /* attribute buffer length */
 197                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 198         }
 199
 200         *p++ = xdr_one;                                  /* next */
 201         *p++ = xdr_zero;                   /* cookie, first word */
 202         *p++ = xdr_two;                   /* cookie, second word */
 203         *p++ = xdr_two;                             /* entry len */
 204         memcpy(p, "..\0\0", 4);                         /* entry */
 205         p++;
 206         *p++ = xdr_one;                         /* bitmap length */
 207         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 208         *p++ = htonl(8);              /* attribute buffer length */
 209         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 210
 211         readdir->pgbase = (char *)p - (char *)start;
 212         readdir->count -= readdir->pgbase;
 213         kunmap_atomic(start, KM_USER0);
 214 }
 215
 216 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 217 {
 218         int res;
 219
 220         might_sleep();
 221
 222         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 223                         nfs_wait_bit_killable, TASK_KILLABLE);
 224         return res;
 225 }
 226
 227 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 228 {
 229         int res = 0;
 230
 231         might_sleep();
 232
 233         if (*timeout <= 0)
 234                 *timeout = NFS4_POLL_RETRY_MIN;
 235         if (*timeout > NFS4_POLL_RETRY_MAX)
 236                 *timeout = NFS4_POLL_RETRY_MAX;
 237         schedule_timeout_killable(*timeout);
 238         if (fatal_signal_pending(current))
 239                 res = -ERESTARTSYS;
 240         *timeout <<= 1;
 241         return res;
 242 }
 243
 244 /* This is the error handling routine for processes that are allowed
 245  * to sleep.
 246  */
 247 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 248 {
 249         struct nfs_client *clp = server->nfs_client;
 250         struct nfs4_state *state = exception->state;
 251         int ret = errorcode;
 252
 253         exception->retry = 0;
 254         switch(errorcode) {
 255                 case 0:
 256                         return 0;
 257                 case -NFS4ERR_ADMIN_REVOKED:
 258                 case -NFS4ERR_BAD_STATEID:
 259                 case -NFS4ERR_OPENMODE:
 260                         if (state == NULL)
 261                                 break;
 262                         nfs4_schedule_stateid_recovery(server, state);
 263                         goto wait_on_recovery;
 264                 case -NFS4ERR_STALE_STATEID:
 265                 case -NFS4ERR_STALE_CLIENTID:
 266                 case -NFS4ERR_EXPIRED:
 267                         nfs4_schedule_lease_recovery(clp);
 268                         goto wait_on_recovery;
 269 #if defined(CONFIG_NFS_V4_1)
 270                 case -NFS4ERR_BADSESSION:
 271                 case -NFS4ERR_BADSLOT:
 272                 case -NFS4ERR_BAD_HIGH_SLOT:
 273                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 274                 case -NFS4ERR_DEADSESSION:
 275                 case -NFS4ERR_SEQ_FALSE_RETRY:
 276                 case -NFS4ERR_SEQ_MISORDERED:
 277                         dprintk("%s ERROR: %d Reset session\n", __func__,
 278                                 errorcode);
 279                         nfs4_schedule_session_recovery(clp->cl_session);
 280                         exception->retry = 1;
 281                         break;
 282 #endif /* defined(CONFIG_NFS_V4_1) */
 283                 case -NFS4ERR_FILE_OPEN:
 284                         if (exception->timeout > HZ) {
 285                                 /* We have retried a decent amount, time to
 286                                  * fail
 287                                  */
 288                                 ret = -EBUSY;
 289                                 break;
 290                         }
 291                 case -NFS4ERR_GRACE:
 292                 case -NFS4ERR_DELAY:
 293                 case -EKEYEXPIRED:
 294                         ret = nfs4_delay(server->client, &exception->timeout);
 295                         if (ret != 0)
 296                                 break;
 297                 case -NFS4ERR_OLD_STATEID:
 298                         exception->retry = 1;
 299                         break;
 300                 case -NFS4ERR_BADOWNER:
 301                         /* The following works around a Linux server bug! */
 302                 case -NFS4ERR_BADNAME:
 303                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 304                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 305                                 exception->retry = 1;
 306                                 printk(KERN_WARNING "NFS: v4 server %s "
 307                                                 "does not accept raw "
 308                                                 "uid/gids. "
 309                                                 "Reenabling the idmapper.\n",
 310                                                 server->nfs_client->cl_hostname);
 311                         }
 312         }
 313         /* We failed to handle the error */
 314         return nfs4_map_errors(ret);
 315 wait_on_recovery:
 316         ret = nfs4_wait_clnt_recover(clp);
 317         if (ret == 0)
 318                 exception->retry = 1;
 319         return ret;
 320 }
 321
 322
 323 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 324 {
 325         spin_lock(&clp->cl_lock);
 326         if (time_before(clp->cl_last_renewal,timestamp))
 327                 clp->cl_last_renewal = timestamp;
 328         spin_unlock(&clp->cl_lock);
 329 }
 330
 331 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 332 {
 333         do_renew_lease(server->nfs_client, timestamp);
 334 }
 335
 336 #if defined(CONFIG_NFS_V4_1)
 337
 338 /*
 339  * nfs4_free_slot - free a slot and efficiently update slot table.
 340  *
 341  * freeing a slot is trivially done by clearing its respective bit
 342  * in the bitmap.
 343  * If the freed slotid equals highest_used_slotid we want to update it
 344  * so that the server would be able to size down the slot table if needed,
 345  * otherwise we know that the highest_used_slotid is still in use.
 346  * When updating highest_used_slotid there may be "holes" in the bitmap
 347  * so we need to scan down from highest_used_slotid to 0 looking for the now
 348  * highest slotid in use.
 349  * If none found, highest_used_slotid is set to -1.
 350  *
 351  * Must be called while holding tbl->slot_tbl_lock
 352  */
 353 static void
 354 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 355 {
 356         int free_slotid = free_slot - tbl->slots;
 357         int slotid = free_slotid;
 358
 359         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 360         /* clear used bit in bitmap */
 361         __clear_bit(slotid, tbl->used_slots);
 362
 363         /* update highest_used_slotid when it is freed */
 364         if (slotid == tbl->highest_used_slotid) {
 365                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 366                 if (slotid < tbl->max_slots)
 367                         tbl->highest_used_slotid = slotid;
 368                 else
 369                         tbl->highest_used_slotid = -1;
 370         }
 371         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 372                 free_slotid, tbl->highest_used_slotid);
 373 }
 374
 375 /*
 376  * Signal state manager thread if session fore channel is drained
 377  */
 378 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 379 {
 380         struct rpc_task *task;
 381
 382         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 383                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 384                 if (task)
 385                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 386                 return;
 387         }
 388
 389         if (ses->fc_slot_table.highest_used_slotid != -1)
 390                 return;
 391
 392         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 393         complete(&ses->fc_slot_table.complete);
 394 }
 395
 396 /*
 397  * Signal state manager thread if session back channel is drained
 398  */
 399 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 400 {
 401         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 402             ses->bc_slot_table.highest_used_slotid != -1)
 403                 return;
 404         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 405         complete(&ses->bc_slot_table.complete);
 406 }
 407
 408 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 409 {
 410         struct nfs4_slot_table *tbl;
 411
 412         tbl = &res->sr_session->fc_slot_table;
 413         if (!res->sr_slot) {
 414                 /* just wake up the next guy waiting since
 415                  * we may have not consumed a slot after all */
 416                 dprintk("%s: No slot\n", __func__);
 417                 return;
 418         }
 419
 420         spin_lock(&tbl->slot_tbl_lock);
 421         nfs4_free_slot(tbl, res->sr_slot);
 422         nfs4_check_drain_fc_complete(res->sr_session);
 423         spin_unlock(&tbl->slot_tbl_lock);
 424         res->sr_slot = NULL;
 425 }
 426
 427 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 428 {
 429         unsigned long timestamp;
 430         struct nfs_client *clp;
 431
 432         /*
 433          * sr_status remains 1 if an RPC level error occurred. The server
 434          * may or may not have processed the sequence operation..
 435          * Proceed as if the server received and processed the sequence
 436          * operation.
 437          */
 438         if (res->sr_status == 1)
 439                 res->sr_status = NFS_OK;
 440
 441         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
 442         if (!res->sr_slot)
 443                 goto out;
 444
 445         /* Check the SEQUENCE operation status */
 446         switch (res->sr_status) {
 447         case 0:
 448                 /* Update the slot's sequence and clientid lease timer */
 449                 ++res->sr_slot->seq_nr;
 450                 timestamp = res->sr_renewal_time;
 451                 clp = res->sr_session->clp;
 452                 do_renew_lease(clp, timestamp);
 453                 /* Check sequence flags */
 454                 if (res->sr_status_flags != 0)
 455                         nfs4_schedule_lease_recovery(clp);
 456                 break;
 457         case -NFS4ERR_DELAY:
 458                 /* The server detected a resend of the RPC call and
 459                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 460                  * of RFC5661.
 461                  */
 462                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 463                         __func__,
 464                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 465                         res->sr_slot->seq_nr);
 466                 goto out_retry;
 467         default:
 468                 /* Just update the slot sequence no. */
 469                 ++res->sr_slot->seq_nr;
 470         }
 471 out:
 472         /* The session may be reset by one of the error handlers. */
 473         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 474         nfs41_sequence_free_slot(res);
 475         return 1;
 476 out_retry:
 477         if (!rpc_restart_call(task))
 478                 goto out;
 479         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 480         return 0;
 481 }
 482
 483 static int nfs4_sequence_done(struct rpc_task *task,
 484                                struct nfs4_sequence_res *res)
 485 {
 486         if (res->sr_session == NULL)
 487                 return 1;
 488         return nfs41_sequence_done(task, res);
 489 }
 490
 491 /*
 492  * nfs4_find_slot - efficiently look for a free slot
 493  *
 494  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 495  * If found, we mark the slot as used, update the highest_used_slotid,
 496  * and respectively set up the sequence operation args.
 497  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 498  *
 499  * Note: must be called with under the slot_tbl_lock.
 500  */
 501 static u8
 502 nfs4_find_slot(struct nfs4_slot_table *tbl)
 503 {
 504         int slotid;
 505         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 506         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 507
 508         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 509                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 510                 tbl->max_slots);
 511         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 512         if (slotid >= tbl->max_slots)
 513                 goto out;
 514         __set_bit(slotid, tbl->used_slots);
 515         if (slotid > tbl->highest_used_slotid)
 516                 tbl->highest_used_slotid = slotid;
 517         ret_id = slotid;
 518 out:
 519         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 520                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 521         return ret_id;
 522 }
 523
 524 int nfs41_setup_sequence(struct nfs4_session *session,
 525                                 struct nfs4_sequence_args *args,
 526                                 struct nfs4_sequence_res *res,
 527                                 int cache_reply,
 528                                 struct rpc_task *task)
 529 {
 530         struct nfs4_slot *slot;
 531         struct nfs4_slot_table *tbl;
 532         u8 slotid;
 533
 534         dprintk("--> %s\n", __func__);
 535         /* slot already allocated? */
 536         if (res->sr_slot != NULL)
 537                 return 0;
 538
 539         tbl = &session->fc_slot_table;
 540
 541         spin_lock(&tbl->slot_tbl_lock);
 542         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 543             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 544                 /*
 545                  * The state manager will wait until the slot table is empty.
 546                  * Schedule the reset thread
 547                  */
 548                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 549                 spin_unlock(&tbl->slot_tbl_lock);
 550                 dprintk("%s Schedule Session Reset\n", __func__);
 551                 return -EAGAIN;
 552         }
 553
 554         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 555             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 556                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 557                 spin_unlock(&tbl->slot_tbl_lock);
 558                 dprintk("%s enforce FIFO order\n", __func__);
 559                 return -EAGAIN;
 560         }
 561
 562         slotid = nfs4_find_slot(tbl);
 563         if (slotid == NFS4_MAX_SLOT_TABLE) {
 564                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 565                 spin_unlock(&tbl->slot_tbl_lock);
 566                 dprintk("<-- %s: no free slots\n", __func__);
 567                 return -EAGAIN;
 568         }
 569         spin_unlock(&tbl->slot_tbl_lock);
 570
 571         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 572         slot = tbl->slots + slotid;
 573         args->sa_session = session;
 574         args->sa_slotid = slotid;
 575         args->sa_cache_this = cache_reply;
 576
 577         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 578
 579         res->sr_session = session;
 580         res->sr_slot = slot;
 581         res->sr_renewal_time = jiffies;
 582         res->sr_status_flags = 0;
 583         /*
 584          * sr_status is only set in decode_sequence, and so will remain
 585          * set to 1 if an rpc level failure occurs.
 586          */
 587         res->sr_status = 1;
 588         return 0;
 589 }
 590 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 591
 592 int nfs4_setup_sequence(const struct nfs_server *server,
 593                         struct nfs4_sequence_args *args,
 594                         struct nfs4_sequence_res *res,
 595                         int cache_reply,
 596                         struct rpc_task *task)
 597 {
 598         struct nfs4_session *session = nfs4_get_session(server);
 599         int ret = 0;
 600
 601         if (session == NULL) {
 602                 args->sa_session = NULL;
 603                 res->sr_session = NULL;
 604                 goto out;
 605         }
 606
 607         dprintk("--> %s clp %p session %p sr_slot %td\n",
 608                 __func__, session->clp, session, res->sr_slot ?
 609                         res->sr_slot - session->fc_slot_table.slots : -1);
 610
 611         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 612                                    task);
 613 out:
 614         dprintk("<-- %s status=%d\n", __func__, ret);
 615         return ret;
 616 }
 617
 618 struct nfs41_call_sync_data {
 619         const struct nfs_server *seq_server;
 620         struct nfs4_sequence_args *seq_args;
 621         struct nfs4_sequence_res *seq_res;
 622         int cache_reply;
 623 };
 624
 625 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 626 {
 627         struct nfs41_call_sync_data *data = calldata;
 628
 629         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 630
 631         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 632                                 data->seq_res, data->cache_reply, task))
 633                 return;
 634         rpc_call_start(task);
 635 }
 636
 637 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 638 {
 639         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 640         nfs41_call_sync_prepare(task, calldata);
 641 }
 642
 643 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 644 {
 645         struct nfs41_call_sync_data *data = calldata;
 646
 647         nfs41_sequence_done(task, data->seq_res);
 648 }
 649
 650 struct rpc_call_ops nfs41_call_sync_ops = {
 651         .rpc_call_prepare = nfs41_call_sync_prepare,
 652         .rpc_call_done = nfs41_call_sync_done,
 653 };
 654
 655 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 656         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 657         .rpc_call_done = nfs41_call_sync_done,
 658 };
 659
 660 static int nfs4_call_sync_sequence(struct nfs_server *server,
 661                                    struct rpc_message *msg,
 662                                    struct nfs4_sequence_args *args,
 663                                    struct nfs4_sequence_res *res,
 664                                    int cache_reply,
 665                                    int privileged)
 666 {
 667         int ret;
 668         struct rpc_task *task;
 669         struct nfs41_call_sync_data data = {
 670                 .seq_server = server,
 671                 .seq_args = args,
 672                 .seq_res = res,
 673                 .cache_reply = cache_reply,
 674         };
 675         struct rpc_task_setup task_setup = {
 676                 .rpc_client = server->client,
 677                 .rpc_message = msg,
 678                 .callback_ops = &nfs41_call_sync_ops,
 679                 .callback_data = &data
 680         };
 681
 682         res->sr_slot = NULL;
 683         if (privileged)
 684                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 685         task = rpc_run_task(&task_setup);
 686         if (IS_ERR(task))
 687                 ret = PTR_ERR(task);
 688         else {
 689                 ret = task->tk_status;
 690                 rpc_put_task(task);
 691         }
 692         return ret;
 693 }
 694
 695 int _nfs4_call_sync_session(struct nfs_server *server,
 696                             struct rpc_message *msg,
 697                             struct nfs4_sequence_args *args,
 698                             struct nfs4_sequence_res *res,
 699                             int cache_reply)
 700 {
 701         return nfs4_call_sync_sequence(server, msg, args, res, cache_reply, 0);
 702 }
 703
 704 #else
 705 static int nfs4_sequence_done(struct rpc_task *task,
 706                                struct nfs4_sequence_res *res)
 707 {
 708         return 1;
 709 }
 710 #endif /* CONFIG_NFS_V4_1 */
 711
 712 int _nfs4_call_sync(struct nfs_server *server,
 713                     struct rpc_message *msg,
 714                     struct nfs4_sequence_args *args,
 715                     struct nfs4_sequence_res *res,
 716                     int cache_reply)
 717 {
 718         args->sa_session = res->sr_session = NULL;
 719         return rpc_call_sync(server->client, msg, 0);
 720 }
 721
 722 #define nfs4_call_sync(server, msg, args, res, cache_reply) \
 723         (server)->nfs_client->cl_mvops->call_sync((server), (msg), &(args)->seq_args, \
 724                         &(res)->seq_res, (cache_reply))
 725
 726 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 727 {
 728         struct nfs_inode *nfsi = NFS_I(dir);
 729
 730         spin_lock(&dir->i_lock);
 731         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 732         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 733                 nfs_force_lookup_revalidate(dir);
 734         nfsi->change_attr = cinfo->after;
 735         spin_unlock(&dir->i_lock);
 736 }
 737
 738 struct nfs4_opendata {
 739         struct kref kref;
 740         struct nfs_openargs o_arg;
 741         struct nfs_openres o_res;
 742         struct nfs_open_confirmargs c_arg;
 743         struct nfs_open_confirmres c_res;
 744         struct nfs_fattr f_attr;
 745         struct nfs_fattr dir_attr;
 746         struct path path;
 747         struct dentry *dir;
 748         struct nfs4_state_owner *owner;
 749         struct nfs4_state *state;
 750         struct iattr attrs;
 751         unsigned long timestamp;
 752         unsigned int rpc_done : 1;
 753         int rpc_status;
 754         int cancelled;
 755 };
 756
 757
 758 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 759 {
 760         p->o_res.f_attr = &p->f_attr;
 761         p->o_res.dir_attr = &p->dir_attr;
 762         p->o_res.seqid = p->o_arg.seqid;
 763         p->c_res.seqid = p->c_arg.seqid;
 764         p->o_res.server = p->o_arg.server;
 765         nfs_fattr_init(&p->f_attr);
 766         nfs_fattr_init(&p->dir_attr);
 767 }
 768
 769 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 770                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 771                 const struct iattr *attrs,
 772                 gfp_t gfp_mask)
 773 {
 774         struct dentry *parent = dget_parent(path->dentry);
 775         struct inode *dir = parent->d_inode;
 776         struct nfs_server *server = NFS_SERVER(dir);
 777         struct nfs4_opendata *p;
 778
 779         p = kzalloc(sizeof(*p), gfp_mask);
 780         if (p == NULL)
 781                 goto err;
 782         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 783         if (p->o_arg.seqid == NULL)
 784                 goto err_free;
 785         path_get(path);
 786         p->path = *path;
 787         p->dir = parent;
 788         p->owner = sp;
 789         atomic_inc(&sp->so_count);
 790         p->o_arg.fh = NFS_FH(dir);
 791         p->o_arg.open_flags = flags;
 792         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 793         p->o_arg.clientid = server->nfs_client->cl_clientid;
 794         p->o_arg.id = sp->so_owner_id.id;
 795         p->o_arg.name = &p->path.dentry->d_name;
 796         p->o_arg.server = server;
 797         p->o_arg.bitmask = server->attr_bitmask;
 798         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 799         if (flags & O_CREAT) {
 800                 u32 *s;
 801
 802                 p->o_arg.u.attrs = &p->attrs;
 803                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 804                 s = (u32 *) p->o_arg.u.verifier.data;
 805                 s[0] = jiffies;
 806                 s[1] = current->pid;
 807         }
 808         p->c_arg.fh = &p->o_res.fh;
 809         p->c_arg.stateid = &p->o_res.stateid;
 810         p->c_arg.seqid = p->o_arg.seqid;
 811         nfs4_init_opendata_res(p);
 812         kref_init(&p->kref);
 813         return p;
 814 err_free:
 815         kfree(p);
 816 err:
 817         dput(parent);
 818         return NULL;
 819 }
 820
 821 static void nfs4_opendata_free(struct kref *kref)
 822 {
 823         struct nfs4_opendata *p = container_of(kref,
 824                         struct nfs4_opendata, kref);
 825
 826         nfs_free_seqid(p->o_arg.seqid);
 827         if (p->state != NULL)
 828                 nfs4_put_open_state(p->state);
 829         nfs4_put_state_owner(p->owner);
 830         dput(p->dir);
 831         path_put(&p->path);
 832         kfree(p);
 833 }
 834
 835 static void nfs4_opendata_put(struct nfs4_opendata *p)
 836 {
 837         if (p != NULL)
 838                 kref_put(&p->kref, nfs4_opendata_free);
 839 }
 840
 841 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 842 {
 843         int ret;
 844
 845         ret = rpc_wait_for_completion_task(task);
 846         return ret;
 847 }
 848
 849 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 850 {
 851         int ret = 0;
 852
 853         if (open_mode & O_EXCL)
 854                 goto out;
 855         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 856                 case FMODE_READ:
 857                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 858                                 && state->n_rdonly != 0;
 859                         break;
 860                 case FMODE_WRITE:
 861                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 862                                 && state->n_wronly != 0;
 863                         break;
 864                 case FMODE_READ|FMODE_WRITE:
 865                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 866                                 && state->n_rdwr != 0;
 867         }
 868 out:
 869         return ret;
 870 }
 871
 872 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 873 {
 874         if ((delegation->type & fmode) != fmode)
 875                 return 0;
 876         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 877                 return 0;
 878         nfs_mark_delegation_referenced(delegation);
 879         return 1;
 880 }
 881
 882 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 883 {
 884         switch (fmode) {
 885                 case FMODE_WRITE:
 886                         state->n_wronly++;
 887                         break;
 888                 case FMODE_READ:
 889                         state->n_rdonly++;
 890                         break;
 891                 case FMODE_READ|FMODE_WRITE:
 892                         state->n_rdwr++;
 893         }
 894         nfs4_state_set_mode_locked(state, state->state | fmode);
 895 }
 896
 897 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 898 {
 899         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 900                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 901         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 902         switch (fmode) {
 903                 case FMODE_READ:
 904                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 905                         break;
 906                 case FMODE_WRITE:
 907                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 908                         break;
 909                 case FMODE_READ|FMODE_WRITE:
 910                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 911         }
 912 }
 913
 914 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 915 {
 916         write_seqlock(&state->seqlock);
 917         nfs_set_open_stateid_locked(state, stateid, fmode);
 918         write_sequnlock(&state->seqlock);
 919 }
 920
 921 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 922 {
 923         /*
 924          * Protect the call to nfs4_state_set_mode_locked and
 925          * serialise the stateid update
 926          */
 927         write_seqlock(&state->seqlock);
 928         if (deleg_stateid != NULL) {
 929                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 930                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 931         }
 932         if (open_stateid != NULL)
 933                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 934         write_sequnlock(&state->seqlock);
 935         spin_lock(&state->owner->so_lock);
 936         update_open_stateflags(state, fmode);
 937         spin_unlock(&state->owner->so_lock);
 938 }
 939
 940 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 941 {
 942         struct nfs_inode *nfsi = NFS_I(state->inode);
 943         struct nfs_delegation *deleg_cur;
 944         int ret = 0;
 945
 946         fmode &= (FMODE_READ|FMODE_WRITE);
 947
 948         rcu_read_lock();
 949         deleg_cur = rcu_dereference(nfsi->delegation);
 950         if (deleg_cur == NULL)
 951                 goto no_delegation;
 952
 953         spin_lock(&deleg_cur->lock);
 954         if (nfsi->delegation != deleg_cur ||
 955             (deleg_cur->type & fmode) != fmode)
 956                 goto no_delegation_unlock;
 957
 958         if (delegation == NULL)
 959                 delegation = &deleg_cur->stateid;
 960         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 961                 goto no_delegation_unlock;
 962
 963         nfs_mark_delegation_referenced(deleg_cur);
 964         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 965         ret = 1;
 966 no_delegation_unlock:
 967         spin_unlock(&deleg_cur->lock);
 968 no_delegation:
 969         rcu_read_unlock();
 970
 971         if (!ret && open_stateid != NULL) {
 972                 __update_open_stateid(state, open_stateid, NULL, fmode);
 973                 ret = 1;
 974         }
 975
 976         return ret;
 977 }
 978
 979
 980 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 981 {
 982         struct nfs_delegation *delegation;
 983
 984         rcu_read_lock();
 985         delegation = rcu_dereference(NFS_I(inode)->delegation);
 986         if (delegation == NULL || (delegation->type & fmode) == fmode) {
 987                 rcu_read_unlock();
 988                 return;
 989         }
 990         rcu_read_unlock();
 991         nfs_inode_return_delegation(inode);
 992 }
 993
 994 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
 995 {
 996         struct nfs4_state *state = opendata->state;
 997         struct nfs_inode *nfsi = NFS_I(state->inode);
 998         struct nfs_delegation *delegation;
 999         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1000         fmode_t fmode = opendata->o_arg.fmode;
1001         nfs4_stateid stateid;
1002         int ret = -EAGAIN;
1003
1004         for (;;) {
1005                 if (can_open_cached(state, fmode, open_mode)) {
1006                         spin_lock(&state->owner->so_lock);
1007                         if (can_open_cached(state, fmode, open_mode)) {
1008                                 update_open_stateflags(state, fmode);
1009                                 spin_unlock(&state->owner->so_lock);
1010                                 goto out_return_state;
1011                         }
1012                         spin_unlock(&state->owner->so_lock);
1013                 }
1014                 rcu_read_lock();
1015                 delegation = rcu_dereference(nfsi->delegation);
1016                 if (delegation == NULL ||
1017                     !can_open_delegated(delegation, fmode)) {
1018                         rcu_read_unlock();
1019                         break;
1020                 }
1021                 /* Save the delegation */
1022                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1023                 rcu_read_unlock();
1024                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1025                 if (ret != 0)
1026                         goto out;
1027                 ret = -EAGAIN;
1028
1029                 /* Try to update the stateid using the delegation */
1030                 if (update_open_stateid(state, NULL, &stateid, fmode))
1031                         goto out_return_state;
1032         }
1033 out:
1034         return ERR_PTR(ret);
1035 out_return_state:
1036         atomic_inc(&state->count);
1037         return state;
1038 }
1039
1040 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1041 {
1042         struct inode *inode;
1043         struct nfs4_state *state = NULL;
1044         struct nfs_delegation *delegation;
1045         int ret;
1046
1047         if (!data->rpc_done) {
1048                 state = nfs4_try_open_cached(data);
1049                 goto out;
1050         }
1051
1052         ret = -EAGAIN;
1053         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1054                 goto err;
1055         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1056         ret = PTR_ERR(inode);
1057         if (IS_ERR(inode))
1058                 goto err;
1059         ret = -ENOMEM;
1060         state = nfs4_get_open_state(inode, data->owner);
1061         if (state == NULL)
1062                 goto err_put_inode;
1063         if (data->o_res.delegation_type != 0) {
1064                 int delegation_flags = 0;
1065
1066                 rcu_read_lock();
1067                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1068                 if (delegation)
1069                         delegation_flags = delegation->flags;
1070                 rcu_read_unlock();
1071                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1072                         nfs_inode_set_delegation(state->inode,
1073                                         data->owner->so_cred,
1074                                         &data->o_res);
1075                 else
1076                         nfs_inode_reclaim_delegation(state->inode,
1077                                         data->owner->so_cred,
1078                                         &data->o_res);
1079         }
1080
1081         update_open_stateid(state, &data->o_res.stateid, NULL,
1082                         data->o_arg.fmode);
1083         iput(inode);
1084 out:
1085         return state;
1086 err_put_inode:
1087         iput(inode);
1088 err:
1089         return ERR_PTR(ret);
1090 }
1091
1092 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1093 {
1094         struct nfs_inode *nfsi = NFS_I(state->inode);
1095         struct nfs_open_context *ctx;
1096
1097         spin_lock(&state->inode->i_lock);
1098         list_for_each_entry(ctx, &nfsi->open_files, list) {
1099                 if (ctx->state != state)
1100                         continue;
1101                 get_nfs_open_context(ctx);
1102                 spin_unlock(&state->inode->i_lock);
1103                 return ctx;
1104         }
1105         spin_unlock(&state->inode->i_lock);
1106         return ERR_PTR(-ENOENT);
1107 }
1108
1109 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1110 {
1111         struct nfs4_opendata *opendata;
1112
1113         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL, GFP_NOFS);
1114         if (opendata == NULL)
1115                 return ERR_PTR(-ENOMEM);
1116         opendata->state = state;
1117         atomic_inc(&state->count);
1118         return opendata;
1119 }
1120
1121 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1122 {
1123         struct nfs4_state *newstate;
1124         int ret;
1125
1126         opendata->o_arg.open_flags = 0;
1127         opendata->o_arg.fmode = fmode;
1128         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1129         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1130         nfs4_init_opendata_res(opendata);
1131         ret = _nfs4_recover_proc_open(opendata);
1132         if (ret != 0)
1133                 return ret;
1134         newstate = nfs4_opendata_to_nfs4_state(opendata);
1135         if (IS_ERR(newstate))
1136                 return PTR_ERR(newstate);
1137         nfs4_close_state(&opendata->path, newstate, fmode);
1138         *res = newstate;
1139         return 0;
1140 }
1141
1142 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1143 {
1144         struct nfs4_state *newstate;
1145         int ret;
1146
1147         /* memory barrier prior to reading state->n_* */
1148         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1149         smp_rmb();
1150         if (state->n_rdwr != 0) {
1151                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1152                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1153                 if (ret != 0)
1154                         return ret;
1155                 if (newstate != state)
1156                         return -ESTALE;
1157         }
1158         if (state->n_wronly != 0) {
1159                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1160                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1161                 if (ret != 0)
1162                         return ret;
1163                 if (newstate != state)
1164                         return -ESTALE;
1165         }
1166         if (state->n_rdonly != 0) {
1167                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1168                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1169                 if (ret != 0)
1170                         return ret;
1171                 if (newstate != state)
1172                         return -ESTALE;
1173         }
1174         /*
1175          * We may have performed cached opens for all three recoveries.
1176          * Check if we need to update the current stateid.
1177          */
1178         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1179             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1180                 write_seqlock(&state->seqlock);
1181                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1182                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1183                 write_sequnlock(&state->seqlock);
1184         }
1185         return 0;
1186 }
1187
1188 /*
1189  * OPEN_RECLAIM:
1190  *      reclaim state on the server after a reboot.
1191  */
1192 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1193 {
1194         struct nfs_delegation *delegation;
1195         struct nfs4_opendata *opendata;
1196         fmode_t delegation_type = 0;
1197         int status;
1198
1199         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1200         if (IS_ERR(opendata))
1201                 return PTR_ERR(opendata);
1202         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1203         opendata->o_arg.fh = NFS_FH(state->inode);
1204         rcu_read_lock();
1205         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1206         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1207                 delegation_type = delegation->type;
1208         rcu_read_unlock();
1209         opendata->o_arg.u.delegation_type = delegation_type;
1210         status = nfs4_open_recover(opendata, state);
1211         nfs4_opendata_put(opendata);
1212         return status;
1213 }
1214
1215 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1216 {
1217         struct nfs_server *server = NFS_SERVER(state->inode);
1218         struct nfs4_exception exception = { };
1219         int err;
1220         do {
1221                 err = _nfs4_do_open_reclaim(ctx, state);
1222                 if (err != -NFS4ERR_DELAY)
1223                         break;
1224                 nfs4_handle_exception(server, err, &exception);
1225         } while (exception.retry);
1226         return err;
1227 }
1228
1229 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1230 {
1231         struct nfs_open_context *ctx;
1232         int ret;
1233
1234         ctx = nfs4_state_find_open_context(state);
1235         if (IS_ERR(ctx))
1236                 return PTR_ERR(ctx);
1237         ret = nfs4_do_open_reclaim(ctx, state);
1238         put_nfs_open_context(ctx);
1239         return ret;
1240 }
1241
1242 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1243 {
1244         struct nfs4_opendata *opendata;
1245         int ret;
1246
1247         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1248         if (IS_ERR(opendata))
1249                 return PTR_ERR(opendata);
1250         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1251         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1252                         sizeof(opendata->o_arg.u.delegation.data));
1253         ret = nfs4_open_recover(opendata, state);
1254         nfs4_opendata_put(opendata);
1255         return ret;
1256 }
1257
1258 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1259 {
1260         struct nfs4_exception exception = { };
1261         struct nfs_server *server = NFS_SERVER(state->inode);
1262         int err;
1263         do {
1264                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1265                 switch (err) {
1266                         case 0:
1267                         case -ENOENT:
1268                         case -ESTALE:
1269                                 goto out;
1270                         case -NFS4ERR_BADSESSION:
1271                         case -NFS4ERR_BADSLOT:
1272                         case -NFS4ERR_BAD_HIGH_SLOT:
1273                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1274                         case -NFS4ERR_DEADSESSION:
1275                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1276                                 goto out;
1277                         case -NFS4ERR_STALE_CLIENTID:
1278                         case -NFS4ERR_STALE_STATEID:
1279                         case -NFS4ERR_EXPIRED:
1280                                 /* Don't recall a delegation if it was lost */
1281                                 nfs4_schedule_lease_recovery(server->nfs_client);
1282                                 goto out;
1283                         case -ERESTARTSYS:
1284                                 /*
1285                                  * The show must go on: exit, but mark the
1286                                  * stateid as needing recovery.
1287                                  */
1288                         case -NFS4ERR_ADMIN_REVOKED:
1289                         case -NFS4ERR_BAD_STATEID:
1290                                 nfs4_schedule_stateid_recovery(server, state);
1291                         case -EKEYEXPIRED:
1292                                 /*
1293                                  * User RPCSEC_GSS context has expired.
1294                                  * We cannot recover this stateid now, so
1295                                  * skip it and allow recovery thread to
1296                                  * proceed.
1297                                  */
1298                         case -ENOMEM:
1299                                 err = 0;
1300                                 goto out;
1301                 }
1302                 err = nfs4_handle_exception(server, err, &exception);
1303         } while (exception.retry);
1304 out:
1305         return err;
1306 }
1307
1308 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1309 {
1310         struct nfs4_opendata *data = calldata;
1311
1312         data->rpc_status = task->tk_status;
1313         if (data->rpc_status == 0) {
1314                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1315                                 sizeof(data->o_res.stateid.data));
1316                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1317                 renew_lease(data->o_res.server, data->timestamp);
1318                 data->rpc_done = 1;
1319         }
1320 }
1321
1322 static void nfs4_open_confirm_release(void *calldata)
1323 {
1324         struct nfs4_opendata *data = calldata;
1325         struct nfs4_state *state = NULL;
1326
1327         /* If this request hasn't been cancelled, do nothing */
1328         if (data->cancelled == 0)
1329                 goto out_free;
1330         /* In case of error, no cleanup! */
1331         if (!data->rpc_done)
1332                 goto out_free;
1333         state = nfs4_opendata_to_nfs4_state(data);
1334         if (!IS_ERR(state))
1335                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1336 out_free:
1337         nfs4_opendata_put(data);
1338 }
1339
1340 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1341         .rpc_call_done = nfs4_open_confirm_done,
1342         .rpc_release = nfs4_open_confirm_release,
1343 };
1344
1345 /*
1346  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1347  */
1348 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1349 {
1350         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1351         struct rpc_task *task;
1352         struct  rpc_message msg = {
1353                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1354                 .rpc_argp = &data->c_arg,
1355                 .rpc_resp = &data->c_res,
1356                 .rpc_cred = data->owner->so_cred,
1357         };
1358         struct rpc_task_setup task_setup_data = {
1359                 .rpc_client = server->client,
1360                 .rpc_message = &msg,
1361                 .callback_ops = &nfs4_open_confirm_ops,
1362                 .callback_data = data,
1363                 .workqueue = nfsiod_workqueue,
1364                 .flags = RPC_TASK_ASYNC,
1365         };
1366         int status;
1367
1368         kref_get(&data->kref);
1369         data->rpc_done = 0;
1370         data->rpc_status = 0;
1371         data->timestamp = jiffies;
1372         task = rpc_run_task(&task_setup_data);
1373         if (IS_ERR(task))
1374                 return PTR_ERR(task);
1375         status = nfs4_wait_for_completion_rpc_task(task);
1376         if (status != 0) {
1377                 data->cancelled = 1;
1378                 smp_wmb();
1379         } else
1380                 status = data->rpc_status;
1381         rpc_put_task(task);
1382         return status;
1383 }
1384
1385 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1386 {
1387         struct nfs4_opendata *data = calldata;
1388         struct nfs4_state_owner *sp = data->owner;
1389
1390         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1391                 return;
1392         /*
1393          * Check if we still need to send an OPEN call, or if we can use
1394          * a delegation instead.
1395          */
1396         if (data->state != NULL) {
1397                 struct nfs_delegation *delegation;
1398
1399                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1400                         goto out_no_action;
1401                 rcu_read_lock();
1402                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1403                 if (delegation != NULL &&
1404                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1405                         rcu_read_unlock();
1406                         goto out_no_action;
1407                 }
1408                 rcu_read_unlock();
1409         }
1410         /* Update sequence id. */
1411         data->o_arg.id = sp->so_owner_id.id;
1412         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1413         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1414                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1415                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1416         }
1417         data->timestamp = jiffies;
1418         if (nfs4_setup_sequence(data->o_arg.server,
1419                                 &data->o_arg.seq_args,
1420                                 &data->o_res.seq_res, 1, task))
1421                 return;
1422         rpc_call_start(task);
1423         return;
1424 out_no_action:
1425         task->tk_action = NULL;
1426
1427 }
1428
1429 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1430 {
1431         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1432         nfs4_open_prepare(task, calldata);
1433 }
1434
1435 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1436 {
1437         struct nfs4_opendata *data = calldata;
1438
1439         data->rpc_status = task->tk_status;
1440
1441         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1442                 return;
1443
1444         if (task->tk_status == 0) {
1445                 switch (data->o_res.f_attr->mode & S_IFMT) {
1446                         case S_IFREG:
1447                                 break;
1448                         case S_IFLNK:
1449                                 data->rpc_status = -ELOOP;
1450                                 break;
1451                         case S_IFDIR:
1452                                 data->rpc_status = -EISDIR;
1453                                 break;
1454                         default:
1455                                 data->rpc_status = -ENOTDIR;
1456                 }
1457                 renew_lease(data->o_res.server, data->timestamp);
1458                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1459                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1460         }
1461         data->rpc_done = 1;
1462 }
1463
1464 static void nfs4_open_release(void *calldata)
1465 {
1466         struct nfs4_opendata *data = calldata;
1467         struct nfs4_state *state = NULL;
1468
1469         /* If this request hasn't been cancelled, do nothing */
1470         if (data->cancelled == 0)
1471                 goto out_free;
1472         /* In case of error, no cleanup! */
1473         if (data->rpc_status != 0 || !data->rpc_done)
1474                 goto out_free;
1475         /* In case we need an open_confirm, no cleanup! */
1476         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1477                 goto out_free;
1478         state = nfs4_opendata_to_nfs4_state(data);
1479         if (!IS_ERR(state))
1480                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1481 out_free:
1482         nfs4_opendata_put(data);
1483 }
1484
1485 static const struct rpc_call_ops nfs4_open_ops = {
1486         .rpc_call_prepare = nfs4_open_prepare,
1487         .rpc_call_done = nfs4_open_done,
1488         .rpc_release = nfs4_open_release,
1489 };
1490
1491 static const struct rpc_call_ops nfs4_recover_open_ops = {
1492         .rpc_call_prepare = nfs4_recover_open_prepare,
1493         .rpc_call_done = nfs4_open_done,
1494         .rpc_release = nfs4_open_release,
1495 };
1496
1497 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1498 {
1499         struct inode *dir = data->dir->d_inode;
1500         struct nfs_server *server = NFS_SERVER(dir);
1501         struct nfs_openargs *o_arg = &data->o_arg;
1502         struct nfs_openres *o_res = &data->o_res;
1503         struct rpc_task *task;
1504         struct rpc_message msg = {
1505                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1506                 .rpc_argp = o_arg,
1507                 .rpc_resp = o_res,
1508                 .rpc_cred = data->owner->so_cred,
1509         };
1510         struct rpc_task_setup task_setup_data = {
1511                 .rpc_client = server->client,
1512                 .rpc_message = &msg,
1513                 .callback_ops = &nfs4_open_ops,
1514                 .callback_data = data,
1515                 .workqueue = nfsiod_workqueue,
1516                 .flags = RPC_TASK_ASYNC,
1517         };
1518         int status;
1519
1520         kref_get(&data->kref);
1521         data->rpc_done = 0;
1522         data->rpc_status = 0;
1523         data->cancelled = 0;
1524         if (isrecover)
1525                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1526         task = rpc_run_task(&task_setup_data);
1527         if (IS_ERR(task))
1528                 return PTR_ERR(task);
1529         status = nfs4_wait_for_completion_rpc_task(task);
1530         if (status != 0) {
1531                 data->cancelled = 1;
1532                 smp_wmb();
1533         } else
1534                 status = data->rpc_status;
1535         rpc_put_task(task);
1536
1537         return status;
1538 }
1539
1540 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1541 {
1542         struct inode *dir = data->dir->d_inode;
1543         struct nfs_openres *o_res = &data->o_res;
1544         int status;
1545
1546         status = nfs4_run_open_task(data, 1);
1547         if (status != 0 || !data->rpc_done)
1548                 return status;
1549
1550         nfs_refresh_inode(dir, o_res->dir_attr);
1551
1552         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1553                 status = _nfs4_proc_open_confirm(data);
1554                 if (status != 0)
1555                         return status;
1556         }
1557
1558         return status;
1559 }
1560
1561 /*
1562  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1563  */
1564 static int _nfs4_proc_open(struct nfs4_opendata *data)
1565 {
1566         struct inode *dir = data->dir->d_inode;
1567         struct nfs_server *server = NFS_SERVER(dir);
1568         struct nfs_openargs *o_arg = &data->o_arg;
1569         struct nfs_openres *o_res = &data->o_res;
1570         int status;
1571
1572         status = nfs4_run_open_task(data, 0);
1573         if (status != 0 || !data->rpc_done)
1574                 return status;
1575
1576         if (o_arg->open_flags & O_CREAT) {
1577                 update_changeattr(dir, &o_res->cinfo);
1578                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1579         } else
1580                 nfs_refresh_inode(dir, o_res->dir_attr);
1581         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1582                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1583         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1584                 status = _nfs4_proc_open_confirm(data);
1585                 if (status != 0)
1586                         return status;
1587         }
1588         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1589                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1590         return 0;
1591 }
1592
1593 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1594 {
1595         unsigned int loop;
1596         int ret;
1597
1598         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1599                 ret = nfs4_wait_clnt_recover(clp);
1600                 if (ret != 0)
1601                         break;
1602                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1603                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1604                         break;
1605                 nfs4_schedule_state_manager(clp);
1606                 ret = -EIO;
1607         }
1608         return ret;
1609 }
1610
1611 static int nfs4_recover_expired_lease(struct nfs_server *server)
1612 {
1613         return nfs4_client_recover_expired_lease(server->nfs_client);
1614 }
1615
1616 /*
1617  * OPEN_EXPIRED:
1618  *      reclaim state on the server after a network partition.
1619  *      Assumes caller holds the appropriate lock
1620  */
1621 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1622 {
1623         struct nfs4_opendata *opendata;
1624         int ret;
1625
1626         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1627         if (IS_ERR(opendata))
1628                 return PTR_ERR(opendata);
1629         ret = nfs4_open_recover(opendata, state);
1630         if (ret == -ESTALE)
1631                 d_drop(ctx->path.dentry);
1632         nfs4_opendata_put(opendata);
1633         return ret;
1634 }
1635
1636 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1637 {
1638         struct nfs_server *server = NFS_SERVER(state->inode);
1639         struct nfs4_exception exception = { };
1640         int err;
1641
1642         do {
1643                 err = _nfs4_open_expired(ctx, state);
1644                 switch (err) {
1645                 default:
1646                         goto out;
1647                 case -NFS4ERR_GRACE:
1648                 case -NFS4ERR_DELAY:
1649                         nfs4_handle_exception(server, err, &exception);
1650                         err = 0;
1651                 }
1652         } while (exception.retry);
1653 out:
1654         return err;
1655 }
1656
1657 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1658 {
1659         struct nfs_open_context *ctx;
1660         int ret;
1661
1662         ctx = nfs4_state_find_open_context(state);
1663         if (IS_ERR(ctx))
1664                 return PTR_ERR(ctx);
1665         ret = nfs4_do_open_expired(ctx, state);
1666         put_nfs_open_context(ctx);
1667         return ret;
1668 }
1669
1670 /*
1671  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1672  * fields corresponding to attributes that were used to store the verifier.
1673  * Make sure we clobber those fields in the later setattr call
1674  */
1675 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1676 {
1677         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1678             !(sattr->ia_valid & ATTR_ATIME_SET))
1679                 sattr->ia_valid |= ATTR_ATIME;
1680
1681         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1682             !(sattr->ia_valid & ATTR_MTIME_SET))
1683                 sattr->ia_valid |= ATTR_MTIME;
1684 }
1685
1686 /*
1687  * Returns a referenced nfs4_state
1688  */
1689 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)
1690 {
1691         struct nfs4_state_owner  *sp;
1692         struct nfs4_state     *state = NULL;
1693         struct nfs_server       *server = NFS_SERVER(dir);
1694         struct nfs4_opendata *opendata;
1695         int status;
1696
1697         /* Protect against reboot recovery conflicts */
1698         status = -ENOMEM;
1699         if (!(sp = nfs4_get_state_owner(server, cred))) {
1700                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1701                 goto out_err;
1702         }
1703         status = nfs4_recover_expired_lease(server);
1704         if (status != 0)
1705                 goto err_put_state_owner;
1706         if (path->dentry->d_inode != NULL)
1707                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1708         status = -ENOMEM;
1709         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr, GFP_KERNEL);
1710         if (opendata == NULL)
1711                 goto err_put_state_owner;
1712
1713         if (path->dentry->d_inode != NULL)
1714                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1715
1716         status = _nfs4_proc_open(opendata);
1717         if (status != 0)
1718                 goto err_opendata_put;
1719
1720         state = nfs4_opendata_to_nfs4_state(opendata);
1721         status = PTR_ERR(state);
1722         if (IS_ERR(state))
1723                 goto err_opendata_put;
1724         if (server->caps & NFS_CAP_POSIX_LOCK)
1725                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1726
1727         if (opendata->o_arg.open_flags & O_EXCL) {
1728                 nfs4_exclusive_attrset(opendata, sattr);
1729
1730                 nfs_fattr_init(opendata->o_res.f_attr);
1731                 status = nfs4_do_setattr(state->inode, cred,
1732                                 opendata->o_res.f_attr, sattr,
1733                                 state);
1734                 if (status == 0)
1735                         nfs_setattr_update_inode(state->inode, sattr);
1736                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1737         }
1738         nfs4_opendata_put(opendata);
1739         nfs4_put_state_owner(sp);
1740         *res = state;
1741         return 0;
1742 err_opendata_put:
1743         nfs4_opendata_put(opendata);
1744 err_put_state_owner:
1745         nfs4_put_state_owner(sp);
1746 out_err:
1747         *res = NULL;
1748         return status;
1749 }
1750
1751
1752 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)
1753 {
1754         struct nfs4_exception exception = { };
1755         struct nfs4_state *res;
1756         int status;
1757
1758         do {
1759                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1760                 if (status == 0)
1761                         break;
1762                 /* NOTE: BAD_SEQID means the server and client disagree about the
1763                  * book-keeping w.r.t. state-changing operations
1764                  * (OPEN/CLOSE/LOCK/LOCKU...)
1765                  * It is actually a sign of a bug on the client or on the server.
1766                  *
1767                  * If we receive a BAD_SEQID error in the particular case of
1768                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1769                  * have unhashed the old state_owner for us, and that we can
1770                  * therefore safely retry using a new one. We should still warn
1771                  * the user though...
1772                  */
1773                 if (status == -NFS4ERR_BAD_SEQID) {
1774                         printk(KERN_WARNING "NFS: v4 server %s "
1775                                         " returned a bad sequence-id error!\n",
1776                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1777                         exception.retry = 1;
1778                         continue;
1779                 }
1780                 /*
1781                  * BAD_STATEID on OPEN means that the server cancelled our
1782                  * state before it received the OPEN_CONFIRM.
1783                  * Recover by retrying the request as per the discussion
1784                  * on Page 181 of RFC3530.
1785                  */
1786                 if (status == -NFS4ERR_BAD_STATEID) {
1787                         exception.retry = 1;
1788                         continue;
1789                 }
1790                 if (status == -EAGAIN) {
1791                         /* We must have found a delegation */
1792                         exception.retry = 1;
1793                         continue;
1794                 }
1795                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1796                                         status, &exception));
1797         } while (exception.retry);
1798         return res;
1799 }
1800
1801 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1802                             struct nfs_fattr *fattr, struct iattr *sattr,
1803                             struct nfs4_state *state)
1804 {
1805         struct nfs_server *server = NFS_SERVER(inode);
1806         struct nfs_setattrargs  arg = {
1807                 .fh             = NFS_FH(inode),
1808                 .iap            = sattr,
1809                 .server         = server,
1810                 .bitmask = server->attr_bitmask,
1811         };
1812         struct nfs_setattrres  res = {
1813                 .fattr          = fattr,
1814                 .server         = server,
1815         };
1816         struct rpc_message msg = {
1817                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1818                 .rpc_argp       = &arg,
1819                 .rpc_resp       = &res,
1820                 .rpc_cred       = cred,
1821         };
1822         unsigned long timestamp = jiffies;
1823         int status;
1824
1825         nfs_fattr_init(fattr);
1826
1827         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1828                 /* Use that stateid */
1829         } else if (state != NULL) {
1830                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1831         } else
1832                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1833
1834         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
1835         if (status == 0 && state != NULL)
1836                 renew_lease(server, timestamp);
1837         return status;
1838 }
1839
1840 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1841                            struct nfs_fattr *fattr, struct iattr *sattr,
1842                            struct nfs4_state *state)
1843 {
1844         struct nfs_server *server = NFS_SERVER(inode);
1845         struct nfs4_exception exception = { };
1846         int err;
1847         do {
1848                 err = nfs4_handle_exception(server,
1849                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1850                                 &exception);
1851         } while (exception.retry);
1852         return err;
1853 }
1854
1855 struct nfs4_closedata {
1856         struct path path;
1857         struct inode *inode;
1858         struct nfs4_state *state;
1859         struct nfs_closeargs arg;
1860         struct nfs_closeres res;
1861         struct nfs_fattr fattr;
1862         unsigned long timestamp;
1863         bool roc;
1864         u32 roc_barrier;
1865 };
1866
1867 static void nfs4_free_closedata(void *data)
1868 {
1869         struct nfs4_closedata *calldata = data;
1870         struct nfs4_state_owner *sp = calldata->state->owner;
1871
1872         if (calldata->roc)
1873                 pnfs_roc_release(calldata->state->inode);
1874         nfs4_put_open_state(calldata->state);
1875         nfs_free_seqid(calldata->arg.seqid);
1876         nfs4_put_state_owner(sp);
1877         path_put(&calldata->path);
1878         kfree(calldata);
1879 }
1880
1881 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1882                 fmode_t fmode)
1883 {
1884         spin_lock(&state->owner->so_lock);
1885         if (!(fmode & FMODE_READ))
1886                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1887         if (!(fmode & FMODE_WRITE))
1888                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1889         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1890         spin_unlock(&state->owner->so_lock);
1891 }
1892
1893 static void nfs4_close_done(struct rpc_task *task, void *data)
1894 {
1895         struct nfs4_closedata *calldata = data;
1896         struct nfs4_state *state = calldata->state;
1897         struct nfs_server *server = NFS_SERVER(calldata->inode);
1898
1899         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1900                 return;
1901         /* hmm. we are done with the inode, and in the process of freeing
1902          * the state_owner. we keep this around to process errors
1903          */
1904         switch (task->tk_status) {
1905                 case 0:
1906                         if (calldata->roc)
1907                                 pnfs_roc_set_barrier(state->inode,
1908                                                      calldata->roc_barrier);
1909                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1910                         renew_lease(server, calldata->timestamp);
1911                         nfs4_close_clear_stateid_flags(state,
1912                                         calldata->arg.fmode);
1913                         break;
1914                 case -NFS4ERR_STALE_STATEID:
1915                 case -NFS4ERR_OLD_STATEID:
1916                 case -NFS4ERR_BAD_STATEID:
1917                 case -NFS4ERR_EXPIRED:
1918                         if (calldata->arg.fmode == 0)
1919                                 break;
1920                 default:
1921                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1922                                 rpc_restart_call_prepare(task);
1923         }
1924         nfs_release_seqid(calldata->arg.seqid);
1925         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1926 }
1927
1928 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1929 {
1930         struct nfs4_closedata *calldata = data;
1931         struct nfs4_state *state = calldata->state;
1932         int call_close = 0;
1933
1934         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1935                 return;
1936
1937         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1938         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1939         spin_lock(&state->owner->so_lock);
1940         /* Calculate the change in open mode */
1941         if (state->n_rdwr == 0) {
1942                 if (state->n_rdonly == 0) {
1943                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1944                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1945                         calldata->arg.fmode &= ~FMODE_READ;
1946                 }
1947                 if (state->n_wronly == 0) {
1948                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1949                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1950                         calldata->arg.fmode &= ~FMODE_WRITE;
1951                 }
1952         }
1953         spin_unlock(&state->owner->so_lock);
1954
1955         if (!call_close) {
1956                 /* Note: exit _without_ calling nfs4_close_done */
1957                 task->tk_action = NULL;
1958                 return;
1959         }
1960
1961         if (calldata->arg.fmode == 0) {
1962                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1963                 if (calldata->roc &&
1964                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
1965                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
1966                                      task, NULL);
1967                         return;
1968                 }
1969         }
1970
1971         nfs_fattr_init(calldata->res.fattr);
1972         calldata->timestamp = jiffies;
1973         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
1974                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1975                                 1, task))
1976                 return;
1977         rpc_call_start(task);
1978 }
1979
1980 static const struct rpc_call_ops nfs4_close_ops = {
1981         .rpc_call_prepare = nfs4_close_prepare,
1982         .rpc_call_done = nfs4_close_done,
1983         .rpc_release = nfs4_free_closedata,
1984 };
1985
1986 /*
1987  * It is possible for data to be read/written from a mem-mapped file
1988  * after the sys_close call (which hits the vfs layer as a flush).
1989  * This means that we can't safely call nfsv4 close on a file until
1990  * the inode is cleared. This in turn means that we are not good
1991  * NFSv4 citizens - we do not indicate to the server to update the file's
1992  * share state even when we are done with one of the three share
1993  * stateid's in the inode.
1994  *
1995  * NOTE: Caller must be holding the sp->so_owner semaphore!
1996  */
1997 int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
1998 {
1999         struct nfs_server *server = NFS_SERVER(state->inode);
2000         struct nfs4_closedata *calldata;
2001         struct nfs4_state_owner *sp = state->owner;
2002         struct rpc_task *task;
2003         struct rpc_message msg = {
2004                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2005                 .rpc_cred = state->owner->so_cred,
2006         };
2007         struct rpc_task_setup task_setup_data = {
2008                 .rpc_client = server->client,
2009                 .rpc_message = &msg,
2010                 .callback_ops = &nfs4_close_ops,
2011                 .workqueue = nfsiod_workqueue,
2012                 .flags = RPC_TASK_ASYNC,
2013         };
2014         int status = -ENOMEM;
2015
2016         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2017         if (calldata == NULL)
2018                 goto out;
2019         calldata->inode = state->inode;
2020         calldata->state = state;
2021         calldata->arg.fh = NFS_FH(state->inode);
2022         calldata->arg.stateid = &state->open_stateid;
2023         /* Serialization for the sequence id */
2024         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2025         if (calldata->arg.seqid == NULL)
2026                 goto out_free_calldata;
2027         calldata->arg.fmode = 0;
2028         calldata->arg.bitmask = server->cache_consistency_bitmask;
2029         calldata->res.fattr = &calldata->fattr;
2030         calldata->res.seqid = calldata->arg.seqid;
2031         calldata->res.server = server;
2032         calldata->roc = roc;
2033         path_get(path);
2034         calldata->path = *path;
2035
2036         msg.rpc_argp = &calldata->arg;
2037         msg.rpc_resp = &calldata->res;
2038         task_setup_data.callback_data = calldata;
2039         task = rpc_run_task(&task_setup_data);
2040         if (IS_ERR(task))
2041                 return PTR_ERR(task);
2042         status = 0;
2043         if (wait)
2044                 status = rpc_wait_for_completion_task(task);
2045         rpc_put_task(task);
2046         return status;
2047 out_free_calldata:
2048         kfree(calldata);
2049 out:
2050         if (roc)
2051                 pnfs_roc_release(state->inode);
2052         nfs4_put_open_state(state);
2053         nfs4_put_state_owner(sp);
2054         return status;
2055 }
2056
2057 static struct inode *
2058 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2059 {
2060         struct nfs4_state *state;
2061
2062         /* Protect against concurrent sillydeletes */
2063         state = nfs4_do_open(dir, &ctx->path, ctx->mode, open_flags, attr, ctx->cred);
2064         if (IS_ERR(state))
2065                 return ERR_CAST(state);
2066         ctx->state = state;
2067         return igrab(state->inode);
2068 }
2069
2070 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2071 {
2072         if (ctx->state == NULL)
2073                 return;
2074         if (is_sync)
2075                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2076         else
2077                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2078 }
2079
2080 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2081 {
2082         struct nfs4_server_caps_arg args = {
2083                 .fhandle = fhandle,
2084         };
2085         struct nfs4_server_caps_res res = {};
2086         struct rpc_message msg = {
2087                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2088                 .rpc_argp = &args,
2089                 .rpc_resp = &res,
2090         };
2091         int status;
2092
2093         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2094         if (status == 0) {
2095                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2096                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2097                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2098                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2099                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2100                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2101                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2102                         server->caps |= NFS_CAP_ACLS;
2103                 if (res.has_links != 0)
2104                         server->caps |= NFS_CAP_HARDLINKS;
2105                 if (res.has_symlinks != 0)
2106                         server->caps |= NFS_CAP_SYMLINKS;
2107                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2108                         server->caps |= NFS_CAP_FILEID;
2109                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2110                         server->caps |= NFS_CAP_MODE;
2111                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2112                         server->caps |= NFS_CAP_NLINK;
2113                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2114                         server->caps |= NFS_CAP_OWNER;
2115                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2116                         server->caps |= NFS_CAP_OWNER_GROUP;
2117                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2118                         server->caps |= NFS_CAP_ATIME;
2119                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2120                         server->caps |= NFS_CAP_CTIME;
2121                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2122                         server->caps |= NFS_CAP_MTIME;
2123
2124                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2125                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2126                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2127                 server->acl_bitmask = res.acl_bitmask;
2128         }
2129
2130         return status;
2131 }
2132
2133 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2134 {
2135         struct nfs4_exception exception = { };
2136         int err;
2137         do {
2138                 err = nfs4_handle_exception(server,
2139                                 _nfs4_server_capabilities(server, fhandle),
2140                                 &exception);
2141         } while (exception.retry);
2142         return err;
2143 }
2144
2145 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2146                 struct nfs_fsinfo *info)
2147 {
2148         struct nfs4_lookup_root_arg args = {
2149                 .bitmask = nfs4_fattr_bitmap,
2150         };
2151         struct nfs4_lookup_res res = {
2152                 .server = server,
2153                 .fattr = info->fattr,
2154                 .fh = fhandle,
2155         };
2156         struct rpc_message msg = {
2157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2158                 .rpc_argp = &args,
2159                 .rpc_resp = &res,
2160         };
2161
2162         nfs_fattr_init(info->fattr);
2163         return nfs4_call_sync(server, &msg, &args, &res, 0);
2164 }
2165
2166 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2167                 struct nfs_fsinfo *info)
2168 {
2169         struct nfs4_exception exception = { };
2170         int err;
2171         do {
2172                 err = nfs4_handle_exception(server,
2173                                 _nfs4_lookup_root(server, fhandle, info),
2174                                 &exception);
2175         } while (exception.retry);
2176         return err;
2177 }
2178
2179 /*
2180  * get the file handle for the "/" directory on the server
2181  */
2182 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2183                               struct nfs_fsinfo *info)
2184 {
2185         int status;
2186
2187         status = nfs4_lookup_root(server, fhandle, info);
2188         if (status == 0)
2189                 status = nfs4_server_capabilities(server, fhandle);
2190         if (status == 0)
2191                 status = nfs4_do_fsinfo(server, fhandle, info);
2192         return nfs4_map_errors(status);
2193 }
2194
2195 /*
2196  * Get locations and (maybe) other attributes of a referral.
2197  * Note that we'll actually follow the referral later when
2198  * we detect fsid mismatch in inode revalidation
2199  */
2200 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2201 {
2202         int status = -ENOMEM;
2203         struct page *page = NULL;
2204         struct nfs4_fs_locations *locations = NULL;
2205
2206         page = alloc_page(GFP_KERNEL);
2207         if (page == NULL)
2208                 goto out;
2209         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2210         if (locations == NULL)
2211                 goto out;
2212
2213         status = nfs4_proc_fs_locations(dir, name, locations, page);
2214         if (status != 0)
2215                 goto out;
2216         /* Make sure server returned a different fsid for the referral */
2217         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2218                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2219                 status = -EIO;
2220                 goto out;
2221         }
2222
2223         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2224         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2225         if (!fattr->mode)
2226                 fattr->mode = S_IFDIR;
2227         memset(fhandle, 0, sizeof(struct nfs_fh));
2228 out:
2229         if (page)
2230                 __free_page(page);
2231         kfree(locations);
2232         return status;
2233 }
2234
2235 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2236 {
2237         struct nfs4_getattr_arg args = {
2238                 .fh = fhandle,
2239                 .bitmask = server->attr_bitmask,
2240         };
2241         struct nfs4_getattr_res res = {
2242                 .fattr = fattr,
2243                 .server = server,
2244         };
2245         struct rpc_message msg = {
2246                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2247                 .rpc_argp = &args,
2248                 .rpc_resp = &res,
2249         };
2250
2251         nfs_fattr_init(fattr);
2252         return nfs4_call_sync(server, &msg, &args, &res, 0);
2253 }
2254
2255 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2256 {
2257         struct nfs4_exception exception = { };
2258         int err;
2259         do {
2260                 err = nfs4_handle_exception(server,
2261                                 _nfs4_proc_getattr(server, fhandle, fattr),
2262                                 &exception);
2263         } while (exception.retry);
2264         return err;
2265 }
2266
2267 /*
2268  * The file is not closed if it is opened due to the a request to change
2269  * the size of the file. The open call will not be needed once the
2270  * VFS layer lookup-intents are implemented.
2271  *
2272  * Close is called when the inode is destroyed.
2273  * If we haven't opened the file for O_WRONLY, we
2274  * need to in the size_change case to obtain a stateid.
2275  *
2276  * Got race?
2277  * Because OPEN is always done by name in nfsv4, it is
2278  * possible that we opened a different file by the same
2279  * name.  We can recognize this race condition, but we
2280  * can't do anything about it besides returning an error.
2281  *
2282  * This will be fixed with VFS changes (lookup-intent).
2283  */
2284 static int
2285 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2286                   struct iattr *sattr)
2287 {
2288         struct inode *inode = dentry->d_inode;
2289         struct rpc_cred *cred = NULL;
2290         struct nfs4_state *state = NULL;
2291         int status;
2292
2293         nfs_fattr_init(fattr);
2294
2295         /* Search for an existing open(O_WRITE) file */
2296         if (sattr->ia_valid & ATTR_FILE) {
2297                 struct nfs_open_context *ctx;
2298
2299                 ctx = nfs_file_open_context(sattr->ia_file);
2300                 if (ctx) {
2301                         cred = ctx->cred;
2302                         state = ctx->state;
2303                 }
2304         }
2305
2306         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2307         if (status == 0)
2308                 nfs_setattr_update_inode(inode, sattr);
2309         return status;
2310 }
2311
2312 static int _nfs4_proc_lookupfh(struct nfs_server *server, const struct nfs_fh *dirfh,
2313                 const struct qstr *name, struct nfs_fh *fhandle,
2314                 struct nfs_fattr *fattr)
2315 {
2316         int                    status;
2317         struct nfs4_lookup_arg args = {
2318                 .bitmask = server->attr_bitmask,
2319                 .dir_fh = dirfh,
2320                 .name = name,
2321         };
2322         struct nfs4_lookup_res res = {
2323                 .server = server,
2324                 .fattr = fattr,
2325                 .fh = fhandle,
2326         };
2327         struct rpc_message msg = {
2328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2329                 .rpc_argp = &args,
2330                 .rpc_resp = &res,
2331         };
2332
2333         nfs_fattr_init(fattr);
2334
2335         dprintk("NFS call  lookupfh %s\n", name->name);
2336         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2337         dprintk("NFS reply lookupfh: %d\n", status);
2338         return status;
2339 }
2340
2341 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2342                               struct qstr *name, struct nfs_fh *fhandle,
2343                               struct nfs_fattr *fattr)
2344 {
2345         struct nfs4_exception exception = { };
2346         int err;
2347         do {
2348                 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
2349                 /* FIXME: !!!! */
2350                 if (err == -NFS4ERR_MOVED) {
2351                         err = -EREMOTE;
2352                         break;
2353                 }
2354                 err = nfs4_handle_exception(server, err, &exception);
2355         } while (exception.retry);
2356         return err;
2357 }
2358
2359 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
2360                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2361 {
2362         int status;
2363
2364         dprintk("NFS call  lookup %s\n", name->name);
2365         status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2366         if (status == -NFS4ERR_MOVED)
2367                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2368         dprintk("NFS reply lookup: %d\n", status);
2369         return status;
2370 }
2371
2372 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2373 {
2374         struct nfs4_exception exception = { };
2375         int err;
2376         do {
2377                 err = nfs4_handle_exception(NFS_SERVER(dir),
2378                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
2379                                 &exception);
2380         } while (exception.retry);
2381         return err;
2382 }
2383
2384 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2385 {
2386         struct nfs_server *server = NFS_SERVER(inode);
2387         struct nfs4_accessargs args = {
2388                 .fh = NFS_FH(inode),
2389                 .bitmask = server->attr_bitmask,
2390         };
2391         struct nfs4_accessres res = {
2392                 .server = server,
2393         };
2394         struct rpc_message msg = {
2395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2396                 .rpc_argp = &args,
2397                 .rpc_resp = &res,
2398                 .rpc_cred = entry->cred,
2399         };
2400         int mode = entry->mask;
2401         int status;
2402
2403         /*
2404          * Determine which access bits we want to ask for...
2405          */
2406         if (mode & MAY_READ)
2407                 args.access |= NFS4_ACCESS_READ;
2408         if (S_ISDIR(inode->i_mode)) {
2409                 if (mode & MAY_WRITE)
2410                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2411                 if (mode & MAY_EXEC)
2412                         args.access |= NFS4_ACCESS_LOOKUP;
2413         } else {
2414                 if (mode & MAY_WRITE)
2415                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2416                 if (mode & MAY_EXEC)
2417                         args.access |= NFS4_ACCESS_EXECUTE;
2418         }
2419
2420         res.fattr = nfs_alloc_fattr();
2421         if (res.fattr == NULL)
2422                 return -ENOMEM;
2423
2424         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2425         if (!status) {
2426                 entry->mask = 0;
2427                 if (res.access & NFS4_ACCESS_READ)
2428                         entry->mask |= MAY_READ;
2429                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2430                         entry->mask |= MAY_WRITE;
2431                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2432                         entry->mask |= MAY_EXEC;
2433                 nfs_refresh_inode(inode, res.fattr);
2434         }
2435         nfs_free_fattr(res.fattr);
2436         return status;
2437 }
2438
2439 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2440 {
2441         struct nfs4_exception exception = { };
2442         int err;
2443         do {
2444                 err = nfs4_handle_exception(NFS_SERVER(inode),
2445                                 _nfs4_proc_access(inode, entry),
2446                                 &exception);
2447         } while (exception.retry);
2448         return err;
2449 }
2450
2451 /*
2452  * TODO: For the time being, we don't try to get any attributes
2453  * along with any of the zero-copy operations READ, READDIR,
2454  * READLINK, WRITE.
2455  *
2456  * In the case of the first three, we want to put the GETATTR
2457  * after the read-type operation -- this is because it is hard
2458  * to predict the length of a GETATTR response in v4, and thus
2459  * align the READ data correctly.  This means that the GETATTR
2460  * may end up partially falling into the page cache, and we should
2461  * shift it into the 'tail' of the xdr_buf before processing.
2462  * To do this efficiently, we need to know the total length
2463  * of data received, which doesn't seem to be available outside
2464  * of the RPC layer.
2465  *
2466  * In the case of WRITE, we also want to put the GETATTR after
2467  * the operation -- in this case because we want to make sure
2468  * we get the post-operation mtime and size.  This means that
2469  * we can't use xdr_encode_pages() as written: we need a variant
2470  * of it which would leave room in the 'tail' iovec.
2471  *
2472  * Both of these changes to the XDR layer would in fact be quite
2473  * minor, but I decided to leave them for a subsequent patch.
2474  */
2475 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2476                 unsigned int pgbase, unsigned int pglen)
2477 {
2478         struct nfs4_readlink args = {
2479                 .fh       = NFS_FH(inode),
2480                 .pgbase   = pgbase,
2481                 .pglen    = pglen,
2482                 .pages    = &page,
2483         };
2484         struct nfs4_readlink_res res;
2485         struct rpc_message msg = {
2486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2487                 .rpc_argp = &args,
2488                 .rpc_resp = &res,
2489         };
2490
2491         return nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
2492 }
2493
2494 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2495                 unsigned int pgbase, unsigned int pglen)
2496 {
2497         struct nfs4_exception exception = { };
2498         int err;
2499         do {
2500                 err = nfs4_handle_exception(NFS_SERVER(inode),
2501                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2502                                 &exception);
2503         } while (exception.retry);
2504         return err;
2505 }
2506
2507 /*
2508  * Got race?
2509  * We will need to arrange for the VFS layer to provide an atomic open.
2510  * Until then, this create/open method is prone to inefficiency and race
2511  * conditions due to the lookup, create, and open VFS calls from sys_open()
2512  * placed on the wire.
2513  *
2514  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2515  * The file will be opened again in the subsequent VFS open call
2516  * (nfs4_proc_file_open).
2517  *
2518  * The open for read will just hang around to be used by any process that
2519  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2520  */
2521
2522 static int
2523 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2524                  int flags, struct nfs_open_context *ctx)
2525 {
2526         struct path my_path = {
2527                 .dentry = dentry,
2528         };
2529         struct path *path = &my_path;
2530         struct nfs4_state *state;
2531         struct rpc_cred *cred = NULL;
2532         fmode_t fmode = 0;
2533         int status = 0;
2534
2535         if (ctx != NULL) {
2536                 cred = ctx->cred;
2537                 path = &ctx->path;
2538                 fmode = ctx->mode;
2539         }
2540         sattr->ia_mode &= ~current_umask();
2541         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2542         d_drop(dentry);
2543         if (IS_ERR(state)) {
2544                 status = PTR_ERR(state);
2545                 goto out;
2546         }
2547         d_add(dentry, igrab(state->inode));
2548         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2549         if (ctx != NULL)
2550                 ctx->state = state;
2551         else
2552                 nfs4_close_sync(path, state, fmode);
2553 out:
2554         return status;
2555 }
2556
2557 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2558 {
2559         struct nfs_server *server = NFS_SERVER(dir);
2560         struct nfs_removeargs args = {
2561                 .fh = NFS_FH(dir),
2562                 .name.len = name->len,
2563                 .name.name = name->name,
2564                 .bitmask = server->attr_bitmask,
2565         };
2566         struct nfs_removeres res = {
2567                 .server = server,
2568         };
2569         struct rpc_message msg = {
2570                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2571                 .rpc_argp = &args,
2572                 .rpc_resp = &res,
2573         };
2574         int status = -ENOMEM;
2575
2576         res.dir_attr = nfs_alloc_fattr();
2577         if (res.dir_attr == NULL)
2578                 goto out;
2579
2580         status = nfs4_call_sync(server, &msg, &args, &res, 1);
2581         if (status == 0) {
2582                 update_changeattr(dir, &res.cinfo);
2583                 nfs_post_op_update_inode(dir, res.dir_attr);
2584         }
2585         nfs_free_fattr(res.dir_attr);
2586 out:
2587         return status;
2588 }
2589
2590 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2591 {
2592         struct nfs4_exception exception = { };
2593         int err;
2594         do {
2595                 err = nfs4_handle_exception(NFS_SERVER(dir),
2596                                 _nfs4_proc_remove(dir, name),
2597                                 &exception);
2598         } while (exception.retry);
2599         return err;
2600 }
2601
2602 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2603 {
2604         struct nfs_server *server = NFS_SERVER(dir);
2605         struct nfs_removeargs *args = msg->rpc_argp;
2606         struct nfs_removeres *res = msg->rpc_resp;
2607
2608         args->bitmask = server->cache_consistency_bitmask;
2609         res->server = server;
2610         res->seq_res.sr_slot = NULL;
2611         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2612 }
2613
2614 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2615 {
2616         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2617
2618         if (!nfs4_sequence_done(task, &res->seq_res))
2619                 return 0;
2620         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2621                 return 0;
2622         update_changeattr(dir, &res->cinfo);
2623         nfs_post_op_update_inode(dir, res->dir_attr);
2624         return 1;
2625 }
2626
2627 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2628 {
2629         struct nfs_server *server = NFS_SERVER(dir);
2630         struct nfs_renameargs *arg = msg->rpc_argp;
2631         struct nfs_renameres *res = msg->rpc_resp;
2632
2633         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2634         arg->bitmask = server->attr_bitmask;
2635         res->server = server;
2636 }
2637
2638 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2639                                  struct inode *new_dir)
2640 {
2641         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2642
2643         if (!nfs4_sequence_done(task, &res->seq_res))
2644                 return 0;
2645         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2646                 return 0;
2647
2648         update_changeattr(old_dir, &res->old_cinfo);
2649         nfs_post_op_update_inode(old_dir, res->old_fattr);
2650         update_changeattr(new_dir, &res->new_cinfo);
2651         nfs_post_op_update_inode(new_dir, res->new_fattr);
2652         return 1;
2653 }
2654
2655 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2656                 struct inode *new_dir, struct qstr *new_name)
2657 {
2658         struct nfs_server *server = NFS_SERVER(old_dir);
2659         struct nfs_renameargs arg = {
2660                 .old_dir = NFS_FH(old_dir),
2661                 .new_dir = NFS_FH(new_dir),
2662                 .old_name = old_name,
2663                 .new_name = new_name,
2664                 .bitmask = server->attr_bitmask,
2665         };
2666         struct nfs_renameres res = {
2667                 .server = server,
2668         };
2669         struct rpc_message msg = {
2670                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2671                 .rpc_argp = &arg,
2672                 .rpc_resp = &res,
2673         };
2674         int status = -ENOMEM;
2675
2676         res.old_fattr = nfs_alloc_fattr();
2677         res.new_fattr = nfs_alloc_fattr();
2678         if (res.old_fattr == NULL || res.new_fattr == NULL)
2679                 goto out;
2680
2681         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2682         if (!status) {
2683                 update_changeattr(old_dir, &res.old_cinfo);
2684                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2685                 update_changeattr(new_dir, &res.new_cinfo);
2686                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2687         }
2688 out:
2689         nfs_free_fattr(res.new_fattr);
2690         nfs_free_fattr(res.old_fattr);
2691         return status;
2692 }
2693
2694 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2695                 struct inode *new_dir, struct qstr *new_name)
2696 {
2697         struct nfs4_exception exception = { };
2698         int err;
2699         do {
2700                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2701                                 _nfs4_proc_rename(old_dir, old_name,
2702                                         new_dir, new_name),
2703                                 &exception);
2704         } while (exception.retry);
2705         return err;
2706 }
2707
2708 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2709 {
2710         struct nfs_server *server = NFS_SERVER(inode);
2711         struct nfs4_link_arg arg = {
2712                 .fh     = NFS_FH(inode),
2713                 .dir_fh = NFS_FH(dir),
2714                 .name   = name,
2715                 .bitmask = server->attr_bitmask,
2716         };
2717         struct nfs4_link_res res = {
2718                 .server = server,
2719         };
2720         struct rpc_message msg = {
2721                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2722                 .rpc_argp = &arg,
2723                 .rpc_resp = &res,
2724         };
2725         int status = -ENOMEM;
2726
2727         res.fattr = nfs_alloc_fattr();
2728         res.dir_attr = nfs_alloc_fattr();
2729         if (res.fattr == NULL || res.dir_attr == NULL)
2730                 goto out;
2731
2732         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2733         if (!status) {
2734                 update_changeattr(dir, &res.cinfo);
2735                 nfs_post_op_update_inode(dir, res.dir_attr);
2736                 nfs_post_op_update_inode(inode, res.fattr);
2737         }
2738 out:
2739         nfs_free_fattr(res.dir_attr);
2740         nfs_free_fattr(res.fattr);
2741         return status;
2742 }
2743
2744 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2745 {
2746         struct nfs4_exception exception = { };
2747         int err;
2748         do {
2749                 err = nfs4_handle_exception(NFS_SERVER(inode),
2750                                 _nfs4_proc_link(inode, dir, name),
2751                                 &exception);
2752         } while (exception.retry);
2753         return err;
2754 }
2755
2756 struct nfs4_createdata {
2757         struct rpc_message msg;
2758         struct nfs4_create_arg arg;
2759         struct nfs4_create_res res;
2760         struct nfs_fh fh;
2761         struct nfs_fattr fattr;
2762         struct nfs_fattr dir_fattr;
2763 };
2764
2765 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2766                 struct qstr *name, struct iattr *sattr, u32 ftype)
2767 {
2768         struct nfs4_createdata *data;
2769
2770         data = kzalloc(sizeof(*data), GFP_KERNEL);
2771         if (data != NULL) {
2772                 struct nfs_server *server = NFS_SERVER(dir);
2773
2774                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2775                 data->msg.rpc_argp = &data->arg;
2776                 data->msg.rpc_resp = &data->res;
2777                 data->arg.dir_fh = NFS_FH(dir);
2778                 data->arg.server = server;
2779                 data->arg.name = name;
2780                 data->arg.attrs = sattr;
2781                 data->arg.ftype = ftype;
2782                 data->arg.bitmask = server->attr_bitmask;
2783                 data->res.server = server;
2784                 data->res.fh = &data->fh;
2785                 data->res.fattr = &data->fattr;
2786                 data->res.dir_fattr = &data->dir_fattr;
2787                 nfs_fattr_init(data->res.fattr);
2788                 nfs_fattr_init(data->res.dir_fattr);
2789         }
2790         return data;
2791 }
2792
2793 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2794 {
2795         int status = nfs4_call_sync(NFS_SERVER(dir), &data->msg,
2796                                     &data->arg, &data->res, 1);
2797         if (status == 0) {
2798                 update_changeattr(dir, &data->res.dir_cinfo);
2799                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2800                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2801         }
2802         return status;
2803 }
2804
2805 static void nfs4_free_createdata(struct nfs4_createdata *data)
2806 {
2807         kfree(data);
2808 }
2809
2810 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2811                 struct page *page, unsigned int len, struct iattr *sattr)
2812 {
2813         struct nfs4_createdata *data;
2814         int status = -ENAMETOOLONG;
2815
2816         if (len > NFS4_MAXPATHLEN)
2817                 goto out;
2818
2819         status = -ENOMEM;
2820         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2821         if (data == NULL)
2822                 goto out;
2823
2824         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2825         data->arg.u.symlink.pages = &page;
2826         data->arg.u.symlink.len = len;
2827
2828         status = nfs4_do_create(dir, dentry, data);
2829
2830         nfs4_free_createdata(data);
2831 out:
2832         return status;
2833 }
2834
2835 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2836                 struct page *page, unsigned int len, struct iattr *sattr)
2837 {
2838         struct nfs4_exception exception = { };
2839         int err;
2840         do {
2841                 err = nfs4_handle_exception(NFS_SERVER(dir),
2842                                 _nfs4_proc_symlink(dir, dentry, page,
2843                                                         len, sattr),
2844                                 &exception);
2845         } while (exception.retry);
2846         return err;
2847 }
2848
2849 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2850                 struct iattr *sattr)
2851 {
2852         struct nfs4_createdata *data;
2853         int status = -ENOMEM;
2854
2855         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2856         if (data == NULL)
2857                 goto out;
2858
2859         status = nfs4_do_create(dir, dentry, data);
2860
2861         nfs4_free_createdata(data);
2862 out:
2863         return status;
2864 }
2865
2866 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2867                 struct iattr *sattr)
2868 {
2869         struct nfs4_exception exception = { };
2870         int err;
2871
2872         sattr->ia_mode &= ~current_umask();
2873         do {
2874                 err = nfs4_handle_exception(NFS_SERVER(dir),
2875                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2876                                 &exception);
2877         } while (exception.retry);
2878         return err;
2879 }
2880
2881 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2882                 u64 cookie, struct page **pages, unsigned int count, int plus)
2883 {
2884         struct inode            *dir = dentry->d_inode;
2885         struct nfs4_readdir_arg args = {
2886                 .fh = NFS_FH(dir),
2887                 .pages = pages,
2888                 .pgbase = 0,
2889                 .count = count,
2890                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2891                 .plus = plus,
2892         };
2893         struct nfs4_readdir_res res;
2894         struct rpc_message msg = {
2895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2896                 .rpc_argp = &args,
2897                 .rpc_resp = &res,
2898                 .rpc_cred = cred,
2899         };
2900         int                     status;
2901
2902         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2903                         dentry->d_parent->d_name.name,
2904                         dentry->d_name.name,
2905                         (unsigned long long)cookie);
2906         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2907         res.pgbase = args.pgbase;
2908         status = nfs4_call_sync(NFS_SERVER(dir), &msg, &args, &res, 0);
2909         if (status >= 0) {
2910                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2911                 status += args.pgbase;
2912         }
2913
2914         nfs_invalidate_atime(dir);
2915
2916         dprintk("%s: returns %d\n", __func__, status);
2917         return status;
2918 }
2919
2920 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2921                 u64 cookie, struct page **pages, unsigned int count, int plus)
2922 {
2923         struct nfs4_exception exception = { };
2924         int err;
2925         do {
2926                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2927                                 _nfs4_proc_readdir(dentry, cred, cookie,
2928                                         pages, count, plus),
2929                                 &exception);
2930         } while (exception.retry);
2931         return err;
2932 }
2933
2934 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2935                 struct iattr *sattr, dev_t rdev)
2936 {
2937         struct nfs4_createdata *data;
2938         int mode = sattr->ia_mode;
2939         int status = -ENOMEM;
2940
2941         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2942         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2943
2944         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2945         if (data == NULL)
2946                 goto out;
2947
2948         if (S_ISFIFO(mode))
2949                 data->arg.ftype = NF4FIFO;
2950         else if (S_ISBLK(mode)) {
2951                 data->arg.ftype = NF4BLK;
2952                 data->arg.u.device.specdata1 = MAJOR(rdev);
2953                 data->arg.u.device.specdata2 = MINOR(rdev);
2954         }
2955         else if (S_ISCHR(mode)) {
2956                 data->arg.ftype = NF4CHR;
2957                 data->arg.u.device.specdata1 = MAJOR(rdev);
2958                 data->arg.u.device.specdata2 = MINOR(rdev);
2959         }
2960
2961         status = nfs4_do_create(dir, dentry, data);
2962
2963         nfs4_free_createdata(data);
2964 out:
2965         return status;
2966 }
2967
2968 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2969                 struct iattr *sattr, dev_t rdev)
2970 {
2971         struct nfs4_exception exception = { };
2972         int err;
2973
2974         sattr->ia_mode &= ~current_umask();
2975         do {
2976                 err = nfs4_handle_exception(NFS_SERVER(dir),
2977                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
2978                                 &exception);
2979         } while (exception.retry);
2980         return err;
2981 }
2982
2983 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
2984                  struct nfs_fsstat *fsstat)
2985 {
2986         struct nfs4_statfs_arg args = {
2987                 .fh = fhandle,
2988                 .bitmask = server->attr_bitmask,
2989         };
2990         struct nfs4_statfs_res res = {
2991                 .fsstat = fsstat,
2992         };
2993         struct rpc_message msg = {
2994                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
2995                 .rpc_argp = &args,
2996                 .rpc_resp = &res,
2997         };
2998
2999         nfs_fattr_init(fsstat->fattr);
3000         return  nfs4_call_sync(server, &msg, &args, &res, 0);
3001 }
3002
3003 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3004 {
3005         struct nfs4_exception exception = { };
3006         int err;
3007         do {
3008                 err = nfs4_handle_exception(server,
3009                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3010                                 &exception);
3011         } while (exception.retry);
3012         return err;
3013 }
3014
3015 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3016                 struct nfs_fsinfo *fsinfo)
3017 {
3018         struct nfs4_fsinfo_arg args = {
3019                 .fh = fhandle,
3020                 .bitmask = server->attr_bitmask,
3021         };
3022         struct nfs4_fsinfo_res res = {
3023                 .fsinfo = fsinfo,
3024         };
3025         struct rpc_message msg = {
3026                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3027                 .rpc_argp = &args,
3028                 .rpc_resp = &res,
3029         };
3030
3031         return nfs4_call_sync(server, &msg, &args, &res, 0);
3032 }
3033
3034 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3035 {
3036         struct nfs4_exception exception = { };
3037         int err;
3038
3039         do {
3040                 err = nfs4_handle_exception(server,
3041                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3042                                 &exception);
3043         } while (exception.retry);
3044         return err;
3045 }
3046
3047 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3048 {
3049         nfs_fattr_init(fsinfo->fattr);
3050         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3051 }
3052
3053 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3054                 struct nfs_pathconf *pathconf)
3055 {
3056         struct nfs4_pathconf_arg args = {
3057                 .fh = fhandle,
3058                 .bitmask = server->attr_bitmask,
3059         };
3060         struct nfs4_pathconf_res res = {
3061                 .pathconf = pathconf,
3062         };
3063         struct rpc_message msg = {
3064                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3065                 .rpc_argp = &args,
3066                 .rpc_resp = &res,
3067         };
3068
3069         /* None of the pathconf attributes are mandatory to implement */
3070         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3071                 memset(pathconf, 0, sizeof(*pathconf));
3072                 return 0;
3073         }
3074
3075         nfs_fattr_init(pathconf->fattr);
3076         return nfs4_call_sync(server, &msg, &args, &res, 0);
3077 }
3078
3079 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3080                 struct nfs_pathconf *pathconf)
3081 {
3082         struct nfs4_exception exception = { };
3083         int err;
3084
3085         do {
3086                 err = nfs4_handle_exception(server,
3087                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3088                                 &exception);
3089         } while (exception.retry);
3090         return err;
3091 }
3092
3093 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3094 {
3095         struct nfs_server *server = NFS_SERVER(data->inode);
3096
3097         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3098                 nfs_restart_rpc(task, server->nfs_client);
3099                 return -EAGAIN;
3100         }
3101
3102         nfs_invalidate_atime(data->inode);
3103         if (task->tk_status > 0)
3104                 renew_lease(server, data->timestamp);
3105         return 0;
3106 }
3107
3108 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3109 {
3110
3111         dprintk("--> %s\n", __func__);
3112
3113         if (!nfs4_sequence_done(task, &data->res.seq_res))
3114                 return -EAGAIN;
3115
3116         return data->read_done_cb(task, data);
3117 }
3118
3119 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3120 {
3121         data->timestamp   = jiffies;
3122         data->read_done_cb = nfs4_read_done_cb;
3123         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3124 }
3125
3126 /* Reset the the nfs_read_data to send the read to the MDS. */
3127 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3128 {
3129         dprintk("%s Reset task for i/o through\n", __func__);
3130         put_lseg(data->lseg);
3131         data->lseg = NULL;
3132         /* offsets will differ in the dense stripe case */
3133         data->args.offset = data->mds_offset;
3134         data->ds_clp = NULL;
3135         data->args.fh     = NFS_FH(data->inode);
3136         data->read_done_cb = nfs4_read_done_cb;
3137         task->tk_ops = data->mds_ops;
3138         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3139 }
3140 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3141
3142 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3143 {
3144         struct inode *inode = data->inode;
3145
3146         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3147                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3148                 return -EAGAIN;
3149         }
3150         if (task->tk_status >= 0) {
3151                 renew_lease(NFS_SERVER(inode), data->timestamp);
3152                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3153         }
3154         return 0;
3155 }
3156
3157 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3158 {
3159         if (!nfs4_sequence_done(task, &data->res.seq_res))
3160                 return -EAGAIN;
3161         return data->write_done_cb(task, data);
3162 }
3163
3164 /* Reset the the nfs_write_data to send the write to the MDS. */
3165 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3166 {
3167         dprintk("%s Reset task for i/o through\n", __func__);
3168         put_lseg(data->lseg);
3169         data->lseg          = NULL;
3170         data->ds_clp        = NULL;
3171         data->write_done_cb = nfs4_write_done_cb;
3172         data->args.fh       = NFS_FH(data->inode);
3173         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3174         data->args.offset   = data->mds_offset;
3175         data->res.fattr     = &data->fattr;
3176         task->tk_ops        = data->mds_ops;
3177         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3178 }
3179 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3180
3181 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3182 {
3183         struct nfs_server *server = NFS_SERVER(data->inode);
3184
3185         if (data->lseg) {
3186                 data->args.bitmask = NULL;
3187                 data->res.fattr = NULL;
3188         } else
3189                 data->args.bitmask = server->cache_consistency_bitmask;
3190         if (!data->write_done_cb)
3191                 data->write_done_cb = nfs4_write_done_cb;
3192         data->res.server = server;
3193         data->timestamp   = jiffies;
3194
3195         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3196 }
3197
3198 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3199 {
3200         struct inode *inode = data->inode;
3201
3202         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3203                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3204                 return -EAGAIN;
3205         }
3206         nfs_refresh_inode(inode, data->res.fattr);
3207         return 0;
3208 }
3209
3210 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3211 {
3212         if (!nfs4_sequence_done(task, &data->res.seq_res))
3213                 return -EAGAIN;
3214         return data->write_done_cb(task, data);
3215 }
3216
3217 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3218 {
3219         struct nfs_server *server = NFS_SERVER(data->inode);
3220
3221         if (data->lseg) {
3222                 data->args.bitmask = NULL;
3223                 data->res.fattr = NULL;
3224         } else
3225                 data->args.bitmask = server->cache_consistency_bitmask;
3226         if (!data->write_done_cb)
3227                 data->write_done_cb = nfs4_commit_done_cb;
3228         data->res.server = server;
3229         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3230 }
3231
3232 struct nfs4_renewdata {
3233         struct nfs_client       *client;
3234         unsigned long           timestamp;
3235 };
3236
3237 /*
3238  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3239  * standalone procedure for queueing an asynchronous RENEW.
3240  */
3241 static void nfs4_renew_release(void *calldata)
3242 {
3243         struct nfs4_renewdata *data = calldata;
3244         struct nfs_client *clp = data->client;
3245
3246         if (atomic_read(&clp->cl_count) > 1)
3247                 nfs4_schedule_state_renewal(clp);
3248         nfs_put_client(clp);
3249         kfree(data);
3250 }
3251
3252 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3253 {
3254         struct nfs4_renewdata *data = calldata;
3255         struct nfs_client *clp = data->client;
3256         unsigned long timestamp = data->timestamp;
3257
3258         if (task->tk_status < 0) {
3259                 /* Unless we're shutting down, schedule state recovery! */
3260                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3261                         nfs4_schedule_lease_recovery(clp);
3262                 return;
3263         }
3264         do_renew_lease(clp, timestamp);
3265 }
3266
3267 static const struct rpc_call_ops nfs4_renew_ops = {
3268         .rpc_call_done = nfs4_renew_done,
3269         .rpc_release = nfs4_renew_release,
3270 };
3271
3272 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3273 {
3274         struct rpc_message msg = {
3275                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3276                 .rpc_argp       = clp,
3277                 .rpc_cred       = cred,
3278         };
3279         struct nfs4_renewdata *data;
3280
3281         if (!atomic_inc_not_zero(&clp->cl_count))
3282                 return -EIO;
3283         data = kmalloc(sizeof(*data), GFP_KERNEL);
3284         if (data == NULL)
3285                 return -ENOMEM;
3286         data->client = clp;
3287         data->timestamp = jiffies;
3288         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3289                         &nfs4_renew_ops, data);
3290 }
3291
3292 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3293 {
3294         struct rpc_message msg = {
3295                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3296                 .rpc_argp       = clp,
3297                 .rpc_cred       = cred,
3298         };
3299         unsigned long now = jiffies;
3300         int status;
3301
3302         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3303         if (status < 0)
3304                 return status;
3305         do_renew_lease(clp, now);
3306         return 0;
3307 }
3308
3309 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3310 {
3311         return (server->caps & NFS_CAP_ACLS)
3312                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3313                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3314 }
3315
3316 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3317  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3318  * the stack.
3319  */
3320 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3321
3322 static void buf_to_pages(const void *buf, size_t buflen,
3323                 struct page **pages, unsigned int *pgbase)
3324 {
3325         const void *p = buf;
3326
3327         *pgbase = offset_in_page(buf);
3328         p -= *pgbase;
3329         while (p < buf + buflen) {
3330                 *(pages++) = virt_to_page(p);
3331                 p += PAGE_CACHE_SIZE;
3332         }
3333 }
3334
3335 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3336                 struct page **pages, unsigned int *pgbase)
3337 {
3338         struct page *newpage, **spages;
3339         int rc = 0;
3340         size_t len;
3341         spages = pages;
3342
3343         do {
3344                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3345                 newpage = alloc_page(GFP_KERNEL);
3346
3347                 if (newpage == NULL)
3348                         goto unwind;
3349                 memcpy(page_address(newpage), buf, len);
3350                 buf += len;
3351                 buflen -= len;
3352                 *pages++ = newpage;
3353                 rc++;
3354         } while (buflen != 0);
3355
3356         return rc;
3357
3358 unwind:
3359         for(; rc > 0; rc--)
3360                 __free_page(spages[rc-1]);
3361         return -ENOMEM;
3362 }
3363
3364 struct nfs4_cached_acl {
3365         int cached;
3366         size_t len;
3367         char data[0];
3368 };
3369
3370 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3371 {
3372         struct nfs_inode *nfsi = NFS_I(inode);
3373
3374         spin_lock(&inode->i_lock);
3375         kfree(nfsi->nfs4_acl);
3376         nfsi->nfs4_acl = acl;
3377         spin_unlock(&inode->i_lock);
3378 }
3379
3380 static void nfs4_zap_acl_attr(struct inode *inode)
3381 {
3382         nfs4_set_cached_acl(inode, NULL);
3383 }
3384
3385 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3386 {
3387         struct nfs_inode *nfsi = NFS_I(inode);
3388         struct nfs4_cached_acl *acl;
3389         int ret = -ENOENT;
3390
3391         spin_lock(&inode->i_lock);
3392         acl = nfsi->nfs4_acl;
3393         if (acl == NULL)
3394                 goto out;
3395         if (buf == NULL) /* user is just asking for length */
3396                 goto out_len;
3397         if (acl->cached == 0)
3398                 goto out;
3399         ret = -ERANGE; /* see getxattr(2) man page */
3400         if (acl->len > buflen)
3401                 goto out;
3402         memcpy(buf, acl->data, acl->len);
3403 out_len:
3404         ret = acl->len;
3405 out:
3406         spin_unlock(&inode->i_lock);
3407         return ret;
3408 }
3409
3410 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3411 {
3412         struct nfs4_cached_acl *acl;
3413
3414         if (buf && acl_len <= PAGE_SIZE) {
3415                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3416                 if (acl == NULL)
3417                         goto out;
3418                 acl->cached = 1;
3419                 memcpy(acl->data, buf, acl_len);
3420         } else {
3421                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3422                 if (acl == NULL)
3423                         goto out;
3424                 acl->cached = 0;
3425         }
3426         acl->len = acl_len;
3427 out:
3428         nfs4_set_cached_acl(inode, acl);
3429 }
3430
3431 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3432 {
3433         struct page *pages[NFS4ACL_MAXPAGES];
3434         struct nfs_getaclargs args = {
3435                 .fh = NFS_FH(inode),
3436                 .acl_pages = pages,
3437                 .acl_len = buflen,
3438         };
3439         struct nfs_getaclres res = {
3440                 .acl_len = buflen,
3441         };
3442         void *resp_buf;
3443         struct rpc_message msg = {
3444                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3445                 .rpc_argp = &args,
3446                 .rpc_resp = &res,
3447         };
3448         struct page *localpage = NULL;
3449         int ret;
3450
3451         if (buflen < PAGE_SIZE) {
3452                 /* As long as we're doing a round trip to the server anyway,
3453                  * let's be prepared for a page of acl data. */
3454                 localpage = alloc_page(GFP_KERNEL);
3455                 resp_buf = page_address(localpage);
3456                 if (localpage == NULL)
3457                         return -ENOMEM;
3458                 args.acl_pages[0] = localpage;
3459                 args.acl_pgbase = 0;
3460                 args.acl_len = PAGE_SIZE;
3461         } else {
3462                 resp_buf = buf;
3463                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3464         }
3465         ret = nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
3466         if (ret)
3467                 goto out_free;
3468         if (res.acl_len > args.acl_len)
3469                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3470         else
3471                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3472         if (buf) {
3473                 ret = -ERANGE;
3474                 if (res.acl_len > buflen)
3475                         goto out_free;
3476                 if (localpage)
3477                         memcpy(buf, resp_buf, res.acl_len);
3478         }
3479         ret = res.acl_len;
3480 out_free:
3481         if (localpage)
3482                 __free_page(localpage);
3483         return ret;
3484 }
3485
3486 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3487 {
3488         struct nfs4_exception exception = { };
3489         ssize_t ret;
3490         do {
3491                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3492                 if (ret >= 0)
3493                         break;
3494                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3495         } while (exception.retry);
3496         return ret;
3497 }
3498
3499 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3500 {
3501         struct nfs_server *server = NFS_SERVER(inode);
3502         int ret;
3503
3504         if (!nfs4_server_supports_acls(server))
3505                 return -EOPNOTSUPP;
3506         ret = nfs_revalidate_inode(server, inode);
3507         if (ret < 0)
3508                 return ret;
3509         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3510                 nfs_zap_acl_cache(inode);
3511         ret = nfs4_read_cached_acl(inode, buf, buflen);
3512         if (ret != -ENOENT)
3513                 return ret;
3514         return nfs4_get_acl_uncached(inode, buf, buflen);
3515 }
3516
3517 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3518 {
3519         struct nfs_server *server = NFS_SERVER(inode);
3520         struct page *pages[NFS4ACL_MAXPAGES];
3521         struct nfs_setaclargs arg = {
3522                 .fh             = NFS_FH(inode),
3523                 .acl_pages      = pages,
3524                 .acl_len        = buflen,
3525         };
3526         struct nfs_setaclres res;
3527         struct rpc_message msg = {
3528                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3529                 .rpc_argp       = &arg,
3530                 .rpc_resp       = &res,
3531         };
3532         int ret, i;
3533
3534         if (!nfs4_server_supports_acls(server))
3535                 return -EOPNOTSUPP;
3536         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3537         if (i < 0)
3538                 return i;
3539         nfs_inode_return_delegation(inode);
3540         ret = nfs4_call_sync(server, &msg, &arg, &res, 1);
3541
3542         /*
3543          * Free each page after tx, so the only ref left is
3544          * held by the network stack
3545          */
3546         for (; i > 0; i--)
3547                 put_page(pages[i-1]);
3548
3549         /*
3550          * Acl update can result in inode attribute update.
3551          * so mark the attribute cache invalid.
3552          */
3553         spin_lock(&inode->i_lock);
3554         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3555         spin_unlock(&inode->i_lock);
3556         nfs_access_zap_cache(inode);
3557         nfs_zap_acl_cache(inode);
3558         return ret;
3559 }
3560
3561 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3562 {
3563         struct nfs4_exception exception = { };
3564         int err;
3565         do {
3566                 err = nfs4_handle_exception(NFS_SERVER(inode),
3567                                 __nfs4_proc_set_acl(inode, buf, buflen),
3568                                 &exception);
3569         } while (exception.retry);
3570         return err;
3571 }
3572
3573 static int
3574 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3575 {
3576         struct nfs_client *clp = server->nfs_client;
3577
3578         if (task->tk_status >= 0)
3579                 return 0;
3580         switch(task->tk_status) {
3581                 case -NFS4ERR_ADMIN_REVOKED:
3582                 case -NFS4ERR_BAD_STATEID:
3583                 case -NFS4ERR_OPENMODE:
3584                         if (state == NULL)
3585                                 break;
3586                         nfs4_schedule_stateid_recovery(server, state);
3587                         goto wait_on_recovery;
3588                 case -NFS4ERR_STALE_STATEID:
3589                 case -NFS4ERR_STALE_CLIENTID:
3590                 case -NFS4ERR_EXPIRED:
3591                         nfs4_schedule_lease_recovery(clp);
3592                         goto wait_on_recovery;
3593 #if defined(CONFIG_NFS_V4_1)
3594                 case -NFS4ERR_BADSESSION:
3595                 case -NFS4ERR_BADSLOT:
3596                 case -NFS4ERR_BAD_HIGH_SLOT:
3597                 case -NFS4ERR_DEADSESSION:
3598                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3599                 case -NFS4ERR_SEQ_FALSE_RETRY:
3600                 case -NFS4ERR_SEQ_MISORDERED:
3601                         dprintk("%s ERROR %d, Reset session\n", __func__,
3602                                 task->tk_status);
3603                         nfs4_schedule_session_recovery(clp->cl_session);
3604                         task->tk_status = 0;
3605                         return -EAGAIN;
3606 #endif /* CONFIG_NFS_V4_1 */
3607                 case -NFS4ERR_DELAY:
3608                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3609                 case -NFS4ERR_GRACE:
3610                 case -EKEYEXPIRED:
3611                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3612                         task->tk_status = 0;
3613                         return -EAGAIN;
3614                 case -NFS4ERR_OLD_STATEID:
3615                         task->tk_status = 0;
3616                         return -EAGAIN;
3617         }
3618         task->tk_status = nfs4_map_errors(task->tk_status);
3619         return 0;
3620 wait_on_recovery:
3621         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3622         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3623                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3624         task->tk_status = 0;
3625         return -EAGAIN;
3626 }
3627
3628 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3629                 unsigned short port, struct rpc_cred *cred,
3630                 struct nfs4_setclientid_res *res)
3631 {
3632         nfs4_verifier sc_verifier;
3633         struct nfs4_setclientid setclientid = {
3634                 .sc_verifier = &sc_verifier,
3635                 .sc_prog = program,
3636                 .sc_cb_ident = clp->cl_cb_ident,
3637         };
3638         struct rpc_message msg = {
3639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3640                 .rpc_argp = &setclientid,
3641                 .rpc_resp = res,
3642                 .rpc_cred = cred,
3643         };
3644         __be32 *p;
3645         int loop = 0;
3646         int status;
3647
3648         p = (__be32*)sc_verifier.data;
3649         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3650         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3651
3652         for(;;) {
3653                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3654                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3655                                 clp->cl_ipaddr,
3656                                 rpc_peeraddr2str(clp->cl_rpcclient,
3657                                                         RPC_DISPLAY_ADDR),
3658                                 rpc_peeraddr2str(clp->cl_rpcclient,
3659                                                         RPC_DISPLAY_PROTO),
3660                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3661                                 clp->cl_id_uniquifier);
3662                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3663                                 sizeof(setclientid.sc_netid),
3664                                 rpc_peeraddr2str(clp->cl_rpcclient,
3665                                                         RPC_DISPLAY_NETID));
3666                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3667                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3668                                 clp->cl_ipaddr, port >> 8, port & 255);
3669
3670                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3671                 if (status != -NFS4ERR_CLID_INUSE)
3672                         break;
3673                 if (signalled())
3674                         break;
3675                 if (loop++ & 1)
3676                         ssleep(clp->cl_lease_time / HZ + 1);
3677                 else
3678                         if (++clp->cl_id_uniquifier == 0)
3679                                 break;
3680         }
3681         return status;
3682 }
3683
3684 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3685                 struct nfs4_setclientid_res *arg,
3686                 struct rpc_cred *cred)
3687 {
3688         struct nfs_fsinfo fsinfo;
3689         struct rpc_message msg = {
3690                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3691                 .rpc_argp = arg,
3692                 .rpc_resp = &fsinfo,
3693                 .rpc_cred = cred,
3694         };
3695         unsigned long now;
3696         int status;
3697
3698         now = jiffies;
3699         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3700         if (status == 0) {
3701                 spin_lock(&clp->cl_lock);
3702                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3703                 clp->cl_last_renewal = now;
3704                 spin_unlock(&clp->cl_lock);
3705         }
3706         return status;
3707 }
3708
3709 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3710                 struct nfs4_setclientid_res *arg,
3711                 struct rpc_cred *cred)
3712 {
3713         long timeout = 0;
3714         int err;
3715         do {
3716                 err = _nfs4_proc_setclientid_confirm(clp, arg, cred);
3717                 switch (err) {
3718                         case 0:
3719                                 return err;
3720                         case -NFS4ERR_RESOURCE:
3721                                 /* The IBM lawyers misread another document! */
3722                         case -NFS4ERR_DELAY:
3723                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3724                 }
3725         } while (err == 0);
3726         return err;
3727 }
3728
3729 struct nfs4_delegreturndata {
3730         struct nfs4_delegreturnargs args;
3731         struct nfs4_delegreturnres res;
3732         struct nfs_fh fh;
3733         nfs4_stateid stateid;
3734         unsigned long timestamp;
3735         struct nfs_fattr fattr;
3736         int rpc_status;
3737 };
3738
3739 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3740 {
3741         struct nfs4_delegreturndata *data = calldata;
3742
3743         if (!nfs4_sequence_done(task, &data->res.seq_res))
3744                 return;
3745
3746         switch (task->tk_status) {
3747         case -NFS4ERR_STALE_STATEID:
3748         case -NFS4ERR_EXPIRED:
3749         case 0:
3750                 renew_lease(data->res.server, data->timestamp);
3751                 break;
3752         default:
3753                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3754                                 -EAGAIN) {
3755                         nfs_restart_rpc(task, data->res.server->nfs_client);
3756                         return;
3757                 }
3758         }
3759         data->rpc_status = task->tk_status;
3760 }
3761
3762 static void nfs4_delegreturn_release(void *calldata)
3763 {
3764         kfree(calldata);
3765 }
3766
3767 #if defined(CONFIG_NFS_V4_1)
3768 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3769 {
3770         struct nfs4_delegreturndata *d_data;
3771
3772         d_data = (struct nfs4_delegreturndata *)data;
3773
3774         if (nfs4_setup_sequence(d_data->res.server,
3775                                 &d_data->args.seq_args,
3776                                 &d_data->res.seq_res, 1, task))
3777                 return;
3778         rpc_call_start(task);
3779 }
3780 #endif /* CONFIG_NFS_V4_1 */
3781
3782 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3783 #if defined(CONFIG_NFS_V4_1)
3784         .rpc_call_prepare = nfs4_delegreturn_prepare,
3785 #endif /* CONFIG_NFS_V4_1 */
3786         .rpc_call_done = nfs4_delegreturn_done,
3787         .rpc_release = nfs4_delegreturn_release,
3788 };
3789
3790 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3791 {
3792         struct nfs4_delegreturndata *data;
3793         struct nfs_server *server = NFS_SERVER(inode);
3794         struct rpc_task *task;
3795         struct rpc_message msg = {
3796                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3797                 .rpc_cred = cred,
3798         };
3799         struct rpc_task_setup task_setup_data = {
3800                 .rpc_client = server->client,
3801                 .rpc_message = &msg,
3802                 .callback_ops = &nfs4_delegreturn_ops,
3803                 .flags = RPC_TASK_ASYNC,
3804         };
3805         int status = 0;
3806
3807         data = kzalloc(sizeof(*data), GFP_NOFS);
3808         if (data == NULL)
3809                 return -ENOMEM;
3810         data->args.fhandle = &data->fh;
3811         data->args.stateid = &data->stateid;
3812         data->args.bitmask = server->attr_bitmask;
3813         nfs_copy_fh(&data->fh, NFS_FH(inode));
3814         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3815         data->res.fattr = &data->fattr;
3816         data->res.server = server;
3817         nfs_fattr_init(data->res.fattr);
3818         data->timestamp = jiffies;
3819         data->rpc_status = 0;
3820
3821         task_setup_data.callback_data = data;
3822         msg.rpc_argp = &data->args;
3823         msg.rpc_resp = &data->res;
3824         task = rpc_run_task(&task_setup_data);
3825         if (IS_ERR(task))
3826                 return PTR_ERR(task);
3827         if (!issync)
3828                 goto out;
3829         status = nfs4_wait_for_completion_rpc_task(task);
3830         if (status != 0)
3831                 goto out;
3832         status = data->rpc_status;
3833         if (status != 0)
3834                 goto out;
3835         nfs_refresh_inode(inode, &data->fattr);
3836 out:
3837         rpc_put_task(task);
3838         return status;
3839 }
3840
3841 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3842 {
3843         struct nfs_server *server = NFS_SERVER(inode);
3844         struct nfs4_exception exception = { };
3845         int err;
3846         do {
3847                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3848                 switch (err) {
3849                         case -NFS4ERR_STALE_STATEID:
3850                         case -NFS4ERR_EXPIRED:
3851                         case 0:
3852                                 return 0;
3853                 }
3854                 err = nfs4_handle_exception(server, err, &exception);
3855         } while (exception.retry);
3856         return err;
3857 }
3858
3859 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3860 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3861
3862 /*
3863  * sleep, with exponential backoff, and retry the LOCK operation.
3864  */
3865 static unsigned long
3866 nfs4_set_lock_task_retry(unsigned long timeout)
3867 {
3868         schedule_timeout_killable(timeout);
3869         timeout <<= 1;
3870         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3871                 return NFS4_LOCK_MAXTIMEOUT;
3872         return timeout;
3873 }
3874
3875 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3876 {
3877         struct inode *inode = state->inode;
3878         struct nfs_server *server = NFS_SERVER(inode);
3879         struct nfs_client *clp = server->nfs_client;
3880         struct nfs_lockt_args arg = {
3881                 .fh = NFS_FH(inode),
3882                 .fl = request,
3883         };
3884         struct nfs_lockt_res res = {
3885                 .denied = request,
3886         };
3887         struct rpc_message msg = {
3888                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3889                 .rpc_argp       = &arg,
3890                 .rpc_resp       = &res,
3891                 .rpc_cred       = state->owner->so_cred,
3892         };
3893         struct nfs4_lock_state *lsp;
3894         int status;
3895
3896         arg.lock_owner.clientid = clp->cl_clientid;
3897         status = nfs4_set_lock_state(state, request);
3898         if (status != 0)
3899                 goto out;
3900         lsp = request->fl_u.nfs4_fl.owner;
3901         arg.lock_owner.id = lsp->ls_id.id;
3902         arg.lock_owner.s_dev = server->s_dev;
3903         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
3904         switch (status) {
3905                 case 0:
3906                         request->fl_type = F_UNLCK;
3907                         break;
3908                 case -NFS4ERR_DENIED:
3909                         status = 0;
3910         }
3911         request->fl_ops->fl_release_private(request);
3912 out:
3913         return status;
3914 }
3915
3916 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3917 {
3918         struct nfs4_exception exception = { };
3919         int err;
3920
3921         do {
3922                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3923                                 _nfs4_proc_getlk(state, cmd, request),
3924                                 &exception);
3925         } while (exception.retry);
3926         return err;
3927 }
3928
3929 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3930 {
3931         int res = 0;
3932         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3933                 case FL_POSIX:
3934                         res = posix_lock_file_wait(file, fl);
3935                         break;
3936                 case FL_FLOCK:
3937                         res = flock_lock_file_wait(file, fl);
3938                         break;
3939                 default:
3940                         BUG();
3941         }
3942         return res;
3943 }
3944
3945 struct nfs4_unlockdata {
3946         struct nfs_locku_args arg;
3947         struct nfs_locku_res res;
3948         struct nfs4_lock_state *lsp;
3949         struct nfs_open_context *ctx;
3950         struct file_lock fl;
3951         const struct nfs_server *server;
3952         unsigned long timestamp;
3953 };
3954
3955 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3956                 struct nfs_open_context *ctx,
3957                 struct nfs4_lock_state *lsp,
3958                 struct nfs_seqid *seqid)
3959 {
3960         struct nfs4_unlockdata *p;
3961         struct inode *inode = lsp->ls_state->inode;
3962
3963         p = kzalloc(sizeof(*p), GFP_NOFS);
3964         if (p == NULL)
3965                 return NULL;
3966         p->arg.fh = NFS_FH(inode);
3967         p->arg.fl = &p->fl;
3968         p->arg.seqid = seqid;
3969         p->res.seqid = seqid;
3970         p->arg.stateid = &lsp->ls_stateid;
3971         p->lsp = lsp;
3972         atomic_inc(&lsp->ls_count);
3973         /* Ensure we don't close file until we're done freeing locks! */
3974         p->ctx = get_nfs_open_context(ctx);
3975         memcpy(&p->fl, fl, sizeof(p->fl));
3976         p->server = NFS_SERVER(inode);
3977         return p;
3978 }
3979
3980 static void nfs4_locku_release_calldata(void *data)
3981 {
3982         struct nfs4_unlockdata *calldata = data;
3983         nfs_free_seqid(calldata->arg.seqid);
3984         nfs4_put_lock_state(calldata->lsp);
3985         put_nfs_open_context(calldata->ctx);
3986         kfree(calldata);
3987 }
3988
3989 static void nfs4_locku_done(struct rpc_task *task, void *data)
3990 {
3991         struct nfs4_unlockdata *calldata = data;
3992
3993         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3994                 return;
3995         switch (task->tk_status) {
3996                 case 0:
3997                         memcpy(calldata->lsp->ls_stateid.data,
3998                                         calldata->res.stateid.data,
3999                                         sizeof(calldata->lsp->ls_stateid.data));
4000                         renew_lease(calldata->server, calldata->timestamp);
4001                         break;
4002                 case -NFS4ERR_BAD_STATEID:
4003                 case -NFS4ERR_OLD_STATEID:
4004                 case -NFS4ERR_STALE_STATEID:
4005                 case -NFS4ERR_EXPIRED:
4006                         break;
4007                 default:
4008                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4009                                 nfs_restart_rpc(task,
4010                                                  calldata->server->nfs_client);
4011         }
4012 }
4013
4014 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4015 {
4016         struct nfs4_unlockdata *calldata = data;
4017
4018         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4019                 return;
4020         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4021                 /* Note: exit _without_ running nfs4_locku_done */
4022                 task->tk_action = NULL;
4023                 return;
4024         }
4025         calldata->timestamp = jiffies;
4026         if (nfs4_setup_sequence(calldata->server,
4027                                 &calldata->arg.seq_args,
4028                                 &calldata->res.seq_res, 1, task))
4029                 return;
4030         rpc_call_start(task);
4031 }
4032
4033 static const struct rpc_call_ops nfs4_locku_ops = {
4034         .rpc_call_prepare = nfs4_locku_prepare,
4035         .rpc_call_done = nfs4_locku_done,
4036         .rpc_release = nfs4_locku_release_calldata,
4037 };
4038
4039 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4040                 struct nfs_open_context *ctx,
4041                 struct nfs4_lock_state *lsp,
4042                 struct nfs_seqid *seqid)
4043 {
4044         struct nfs4_unlockdata *data;
4045         struct rpc_message msg = {
4046                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4047                 .rpc_cred = ctx->cred,
4048         };
4049         struct rpc_task_setup task_setup_data = {
4050                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4051                 .rpc_message = &msg,
4052                 .callback_ops = &nfs4_locku_ops,
4053                 .workqueue = nfsiod_workqueue,
4054                 .flags = RPC_TASK_ASYNC,
4055         };
4056
4057         /* Ensure this is an unlock - when canceling a lock, the
4058          * canceled lock is passed in, and it won't be an unlock.
4059          */
4060         fl->fl_type = F_UNLCK;
4061
4062         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4063         if (data == NULL) {
4064                 nfs_free_seqid(seqid);
4065                 return ERR_PTR(-ENOMEM);
4066         }
4067
4068         msg.rpc_argp = &data->arg;
4069         msg.rpc_resp = &data->res;
4070         task_setup_data.callback_data = data;
4071         return rpc_run_task(&task_setup_data);
4072 }
4073
4074 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4075 {
4076         struct nfs_inode *nfsi = NFS_I(state->inode);
4077         struct nfs_seqid *seqid;
4078         struct nfs4_lock_state *lsp;
4079         struct rpc_task *task;
4080         int status = 0;
4081         unsigned char fl_flags = request->fl_flags;
4082
4083         status = nfs4_set_lock_state(state, request);
4084         /* Unlock _before_ we do the RPC call */
4085         request->fl_flags |= FL_EXISTS;
4086         down_read(&nfsi->rwsem);
4087         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4088                 up_read(&nfsi->rwsem);
4089                 goto out;
4090         }
4091         up_read(&nfsi->rwsem);
4092         if (status != 0)
4093                 goto out;
4094         /* Is this a delegated lock? */
4095         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4096                 goto out;
4097         lsp = request->fl_u.nfs4_fl.owner;
4098         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4099         status = -ENOMEM;
4100         if (seqid == NULL)
4101                 goto out;
4102         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4103         status = PTR_ERR(task);
4104         if (IS_ERR(task))
4105                 goto out;
4106         status = nfs4_wait_for_completion_rpc_task(task);
4107         rpc_put_task(task);
4108 out:
4109         request->fl_flags = fl_flags;
4110         return status;
4111 }
4112
4113 struct nfs4_lockdata {
4114         struct nfs_lock_args arg;
4115         struct nfs_lock_res res;
4116         struct nfs4_lock_state *lsp;
4117         struct nfs_open_context *ctx;
4118         struct file_lock fl;
4119         unsigned long timestamp;
4120         int rpc_status;
4121         int cancelled;
4122         struct nfs_server *server;
4123 };
4124
4125 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4126                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4127                 gfp_t gfp_mask)
4128 {
4129         struct nfs4_lockdata *p;
4130         struct inode *inode = lsp->ls_state->inode;
4131         struct nfs_server *server = NFS_SERVER(inode);
4132
4133         p = kzalloc(sizeof(*p), gfp_mask);
4134         if (p == NULL)
4135                 return NULL;
4136
4137         p->arg.fh = NFS_FH(inode);
4138         p->arg.fl = &p->fl;
4139         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4140         if (p->arg.open_seqid == NULL)
4141                 goto out_free;
4142         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4143         if (p->arg.lock_seqid == NULL)
4144                 goto out_free_seqid;
4145         p->arg.lock_stateid = &lsp->ls_stateid;
4146         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4147         p->arg.lock_owner.id = lsp->ls_id.id;
4148         p->arg.lock_owner.s_dev = server->s_dev;
4149         p->res.lock_seqid = p->arg.lock_seqid;
4150         p->lsp = lsp;
4151         p->server = server;
4152         atomic_inc(&lsp->ls_count);
4153         p->ctx = get_nfs_open_context(ctx);
4154         memcpy(&p->fl, fl, sizeof(p->fl));
4155         return p;
4156 out_free_seqid:
4157         nfs_free_seqid(p->arg.open_seqid);
4158 out_free:
4159         kfree(p);
4160         return NULL;
4161 }
4162
4163 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4164 {
4165         struct nfs4_lockdata *data = calldata;
4166         struct nfs4_state *state = data->lsp->ls_state;
4167
4168         dprintk("%s: begin!\n", __func__);
4169         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4170                 return;
4171         /* Do we need to do an open_to_lock_owner? */
4172         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4173                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4174                         return;
4175                 data->arg.open_stateid = &state->stateid;
4176                 data->arg.new_lock_owner = 1;
4177                 data->res.open_seqid = data->arg.open_seqid;
4178         } else
4179                 data->arg.new_lock_owner = 0;
4180         data->timestamp = jiffies;
4181         if (nfs4_setup_sequence(data->server,
4182                                 &data->arg.seq_args,
4183                                 &data->res.seq_res, 1, task))
4184                 return;
4185         rpc_call_start(task);
4186         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4187 }
4188
4189 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4190 {
4191         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4192         nfs4_lock_prepare(task, calldata);
4193 }
4194
4195 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4196 {
4197         struct nfs4_lockdata *data = calldata;
4198
4199         dprintk("%s: begin!\n", __func__);
4200
4201         if (!nfs4_sequence_done(task, &data->res.seq_res))
4202                 return;
4203
4204         data->rpc_status = task->tk_status;
4205         if (data->arg.new_lock_owner != 0) {
4206                 if (data->rpc_status == 0)
4207                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4208                 else
4209                         goto out;
4210         }
4211         if (data->rpc_status == 0) {
4212                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4213                                         sizeof(data->lsp->ls_stateid.data));
4214                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4215                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4216         }
4217 out:
4218         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4219 }
4220
4221 static void nfs4_lock_release(void *calldata)
4222 {
4223         struct nfs4_lockdata *data = calldata;
4224
4225         dprintk("%s: begin!\n", __func__);
4226         nfs_free_seqid(data->arg.open_seqid);
4227         if (data->cancelled != 0) {
4228                 struct rpc_task *task;
4229                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4230                                 data->arg.lock_seqid);
4231                 if (!IS_ERR(task))
4232                         rpc_put_task_async(task);
4233                 dprintk("%s: cancelling lock!\n", __func__);
4234         } else
4235                 nfs_free_seqid(data->arg.lock_seqid);
4236         nfs4_put_lock_state(data->lsp);
4237         put_nfs_open_context(data->ctx);
4238         kfree(data);
4239         dprintk("%s: done!\n", __func__);
4240 }
4241
4242 static const struct rpc_call_ops nfs4_lock_ops = {
4243         .rpc_call_prepare = nfs4_lock_prepare,
4244         .rpc_call_done = nfs4_lock_done,
4245         .rpc_release = nfs4_lock_release,
4246 };
4247
4248 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4249         .rpc_call_prepare = nfs4_recover_lock_prepare,
4250         .rpc_call_done = nfs4_lock_done,
4251         .rpc_release = nfs4_lock_release,
4252 };
4253
4254 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4255 {
4256         switch (error) {
4257         case -NFS4ERR_ADMIN_REVOKED:
4258         case -NFS4ERR_BAD_STATEID:
4259                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4260                 if (new_lock_owner != 0 ||
4261                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4262                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4263                 break;
4264         case -NFS4ERR_STALE_STATEID:
4265                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4266         case -NFS4ERR_EXPIRED:
4267                 nfs4_schedule_lease_recovery(server->nfs_client);
4268         };
4269 }
4270
4271 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4272 {
4273         struct nfs4_lockdata *data;
4274         struct rpc_task *task;
4275         struct rpc_message msg = {
4276                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4277                 .rpc_cred = state->owner->so_cred,
4278         };
4279         struct rpc_task_setup task_setup_data = {
4280                 .rpc_client = NFS_CLIENT(state->inode),
4281                 .rpc_message = &msg,
4282                 .callback_ops = &nfs4_lock_ops,
4283                 .workqueue = nfsiod_workqueue,
4284                 .flags = RPC_TASK_ASYNC,
4285         };
4286         int ret;
4287
4288         dprintk("%s: begin!\n", __func__);
4289         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4290                         fl->fl_u.nfs4_fl.owner,
4291                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4292         if (data == NULL)
4293                 return -ENOMEM;
4294         if (IS_SETLKW(cmd))
4295                 data->arg.block = 1;
4296         if (recovery_type > NFS_LOCK_NEW) {
4297                 if (recovery_type == NFS_LOCK_RECLAIM)
4298                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4299                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4300         }
4301         msg.rpc_argp = &data->arg;
4302         msg.rpc_resp = &data->res;
4303         task_setup_data.callback_data = data;
4304         task = rpc_run_task(&task_setup_data);
4305         if (IS_ERR(task))
4306                 return PTR_ERR(task);
4307         ret = nfs4_wait_for_completion_rpc_task(task);
4308         if (ret == 0) {
4309                 ret = data->rpc_status;
4310                 if (ret)
4311                         nfs4_handle_setlk_error(data->server, data->lsp,
4312                                         data->arg.new_lock_owner, ret);
4313         } else
4314                 data->cancelled = 1;
4315         rpc_put_task(task);
4316         dprintk("%s: done, ret = %d!\n", __func__, ret);
4317         return ret;
4318 }
4319
4320 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4321 {
4322         struct nfs_server *server = NFS_SERVER(state->inode);
4323         struct nfs4_exception exception = { };
4324         int err;
4325
4326         do {
4327                 /* Cache the lock if possible... */
4328                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4329                         return 0;
4330                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4331                 if (err != -NFS4ERR_DELAY)
4332                         break;
4333                 nfs4_handle_exception(server, err, &exception);
4334         } while (exception.retry);
4335         return err;
4336 }
4337
4338 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4339 {
4340         struct nfs_server *server = NFS_SERVER(state->inode);
4341         struct nfs4_exception exception = { };
4342         int err;
4343
4344         err = nfs4_set_lock_state(state, request);
4345         if (err != 0)
4346                 return err;
4347         do {
4348                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4349                         return 0;
4350                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4351                 switch (err) {
4352                 default:
4353                         goto out;
4354                 case -NFS4ERR_GRACE:
4355                 case -NFS4ERR_DELAY:
4356                         nfs4_handle_exception(server, err, &exception);
4357                         err = 0;
4358                 }
4359         } while (exception.retry);
4360 out:
4361         return err;
4362 }
4363
4364 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4365 {
4366         struct nfs_inode *nfsi = NFS_I(state->inode);
4367         unsigned char fl_flags = request->fl_flags;
4368         int status = -ENOLCK;
4369
4370         if ((fl_flags & FL_POSIX) &&
4371                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4372                 goto out;
4373         /* Is this a delegated open? */
4374         status = nfs4_set_lock_state(state, request);
4375         if (status != 0)
4376                 goto out;
4377         request->fl_flags |= FL_ACCESS;
4378         status = do_vfs_lock(request->fl_file, request);
4379         if (status < 0)
4380                 goto out;
4381         down_read(&nfsi->rwsem);
4382         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4383                 /* Yes: cache locks! */
4384                 /* ...but avoid races with delegation recall... */
4385                 request->fl_flags = fl_flags & ~FL_SLEEP;
4386                 status = do_vfs_lock(request->fl_file, request);
4387                 goto out_unlock;
4388         }
4389         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4390         if (status != 0)
4391                 goto out_unlock;
4392         /* Note: we always want to sleep here! */
4393         request->fl_flags = fl_flags | FL_SLEEP;
4394         if (do_vfs_lock(request->fl_file, request) < 0)
4395                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4396 out_unlock:
4397         up_read(&nfsi->rwsem);
4398 out:
4399         request->fl_flags = fl_flags;
4400         return status;
4401 }
4402
4403 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4404 {
4405         struct nfs4_exception exception = { };
4406         int err;
4407
4408         do {
4409                 err = _nfs4_proc_setlk(state, cmd, request);
4410                 if (err == -NFS4ERR_DENIED)
4411                         err = -EAGAIN;
4412                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4413                                 err, &exception);
4414         } while (exception.retry);
4415         return err;
4416 }
4417
4418 static int
4419 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4420 {
4421         struct nfs_open_context *ctx;
4422         struct nfs4_state *state;
4423         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4424         int status;
4425
4426         /* verify open state */
4427         ctx = nfs_file_open_context(filp);
4428         state = ctx->state;
4429
4430         if (request->fl_start < 0 || request->fl_end < 0)
4431                 return -EINVAL;
4432
4433         if (IS_GETLK(cmd)) {
4434                 if (state != NULL)
4435                         return nfs4_proc_getlk(state, F_GETLK, request);
4436                 return 0;
4437         }
4438
4439         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4440                 return -EINVAL;
4441
4442         if (request->fl_type == F_UNLCK) {
4443                 if (state != NULL)
4444                         return nfs4_proc_unlck(state, cmd, request);
4445                 return 0;
4446         }
4447
4448         if (state == NULL)
4449                 return -ENOLCK;
4450         do {
4451                 status = nfs4_proc_setlk(state, cmd, request);
4452                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4453                         break;
4454                 timeout = nfs4_set_lock_task_retry(timeout);
4455                 status = -ERESTARTSYS;
4456                 if (signalled())
4457                         break;
4458         } while(status < 0);
4459         return status;
4460 }
4461
4462 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4463 {
4464         struct nfs_server *server = NFS_SERVER(state->inode);
4465         struct nfs4_exception exception = { };
4466         int err;
4467
4468         err = nfs4_set_lock_state(state, fl);
4469         if (err != 0)
4470                 goto out;
4471         do {
4472                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4473                 switch (err) {
4474                         default:
4475                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4476                                                 __func__, err);
4477                         case 0:
4478                         case -ESTALE:
4479                                 goto out;
4480                         case -NFS4ERR_EXPIRED:
4481                         case -NFS4ERR_STALE_CLIENTID:
4482                         case -NFS4ERR_STALE_STATEID:
4483                                 nfs4_schedule_lease_recovery(server->nfs_client);
4484                                 goto out;
4485                         case -NFS4ERR_BADSESSION:
4486                         case -NFS4ERR_BADSLOT:
4487                         case -NFS4ERR_BAD_HIGH_SLOT:
4488                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4489                         case -NFS4ERR_DEADSESSION:
4490                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4491                                 goto out;
4492                         case -ERESTARTSYS:
4493                                 /*
4494                                  * The show must go on: exit, but mark the
4495                                  * stateid as needing recovery.
4496                                  */
4497                         case -NFS4ERR_ADMIN_REVOKED:
4498                         case -NFS4ERR_BAD_STATEID:
4499                         case -NFS4ERR_OPENMODE:
4500                                 nfs4_schedule_stateid_recovery(server, state);
4501                                 err = 0;
4502                                 goto out;
4503                         case -EKEYEXPIRED:
4504                                 /*
4505                                  * User RPCSEC_GSS context has expired.
4506                                  * We cannot recover this stateid now, so
4507                                  * skip it and allow recovery thread to
4508                                  * proceed.
4509                                  */
4510                                 err = 0;
4511                                 goto out;
4512                         case -ENOMEM:
4513                         case -NFS4ERR_DENIED:
4514                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4515                                 err = 0;
4516                                 goto out;
4517                         case -NFS4ERR_DELAY:
4518                                 break;
4519                 }
4520                 err = nfs4_handle_exception(server, err, &exception);
4521         } while (exception.retry);
4522 out:
4523         return err;
4524 }
4525
4526 static void nfs4_release_lockowner_release(void *calldata)
4527 {
4528         kfree(calldata);
4529 }
4530
4531 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4532         .rpc_release = nfs4_release_lockowner_release,
4533 };
4534
4535 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4536 {
4537         struct nfs_server *server = lsp->ls_state->owner->so_server;
4538         struct nfs_release_lockowner_args *args;
4539         struct rpc_message msg = {
4540                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4541         };
4542
4543         if (server->nfs_client->cl_mvops->minor_version != 0)
4544                 return;
4545         args = kmalloc(sizeof(*args), GFP_NOFS);
4546         if (!args)
4547                 return;
4548         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4549         args->lock_owner.id = lsp->ls_id.id;
4550         args->lock_owner.s_dev = server->s_dev;
4551         msg.rpc_argp = args;
4552         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4553 }
4554
4555 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4556
4557 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4558                                    const void *buf, size_t buflen,
4559                                    int flags, int type)
4560 {
4561         if (strcmp(key, "") != 0)
4562                 return -EINVAL;
4563
4564         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4565 }
4566
4567 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4568                                    void *buf, size_t buflen, int type)
4569 {
4570         if (strcmp(key, "") != 0)
4571                 return -EINVAL;
4572
4573         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4574 }
4575
4576 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4577                                        size_t list_len, const char *name,
4578                                        size_t name_len, int type)
4579 {
4580         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4581
4582         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4583                 return 0;
4584
4585         if (list && len <= list_len)
4586                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4587         return len;
4588 }
4589
4590 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4591 {
4592         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4593                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4594                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4595                 return;
4596
4597         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4598                 NFS_ATTR_FATTR_NLINK;
4599         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4600         fattr->nlink = 2;
4601 }
4602
4603 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4604                 struct nfs4_fs_locations *fs_locations, struct page *page)
4605 {
4606         struct nfs_server *server = NFS_SERVER(dir);
4607         u32 bitmask[2] = {
4608                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4609                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4610         };
4611         struct nfs4_fs_locations_arg args = {
4612                 .dir_fh = NFS_FH(dir),
4613                 .name = name,
4614                 .page = page,
4615                 .bitmask = bitmask,
4616         };
4617         struct nfs4_fs_locations_res res = {
4618                 .fs_locations = fs_locations,
4619         };
4620         struct rpc_message msg = {
4621                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4622                 .rpc_argp = &args,
4623                 .rpc_resp = &res,
4624         };
4625         int status;
4626
4627         dprintk("%s: start\n", __func__);
4628         nfs_fattr_init(&fs_locations->fattr);
4629         fs_locations->server = server;
4630         fs_locations->nlocations = 0;
4631         status = nfs4_call_sync(server, &msg, &args, &res, 0);
4632         nfs_fixup_referral_attributes(&fs_locations->fattr);
4633         dprintk("%s: returned status = %d\n", __func__, status);
4634         return status;
4635 }
4636
4637 #ifdef CONFIG_NFS_V4_1
4638 /*
4639  * Check the exchange flags returned by the server for invalid flags, having
4640  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4641  * DS flags set.
4642  */
4643 static int nfs4_check_cl_exchange_flags(u32 flags)
4644 {
4645         if (flags & ~EXCHGID4_FLAG_MASK_R)
4646                 goto out_inval;
4647         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4648             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4649                 goto out_inval;
4650         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4651                 goto out_inval;
4652         return NFS_OK;
4653 out_inval:
4654         return -NFS4ERR_INVAL;
4655 }
4656
4657 /*
4658  * nfs4_proc_exchange_id()
4659  *
4660  * Since the clientid has expired, all compounds using sessions
4661  * associated with the stale clientid will be returning
4662  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4663  * be in some phase of session reset.
4664  */
4665 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4666 {
4667         nfs4_verifier verifier;
4668         struct nfs41_exchange_id_args args = {
4669                 .client = clp,
4670                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4671         };
4672         struct nfs41_exchange_id_res res = {
4673                 .client = clp,
4674         };
4675         int status;
4676         struct rpc_message msg = {
4677                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4678                 .rpc_argp = &args,
4679                 .rpc_resp = &res,
4680                 .rpc_cred = cred,
4681         };
4682         __be32 *p;
4683
4684         dprintk("--> %s\n", __func__);
4685         BUG_ON(clp == NULL);
4686
4687         p = (u32 *)verifier.data;
4688         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4689         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4690         args.verifier = &verifier;
4691
4692         args.id_len = scnprintf(args.id, sizeof(args.id),
4693                                 "%s/%s.%s/%u",
4694                                 clp->cl_ipaddr,
4695                                 init_utsname()->nodename,
4696                                 init_utsname()->domainname,
4697                                 clp->cl_rpcclient->cl_auth->au_flavor);
4698
4699         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4700         if (!status)
4701                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4702         dprintk("<-- %s status= %d\n", __func__, status);
4703         return status;
4704 }
4705
4706 struct nfs4_get_lease_time_data {
4707         struct nfs4_get_lease_time_args *args;
4708         struct nfs4_get_lease_time_res *res;
4709         struct nfs_client *clp;
4710 };
4711
4712 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4713                                         void *calldata)
4714 {
4715         int ret;
4716         struct nfs4_get_lease_time_data *data =
4717                         (struct nfs4_get_lease_time_data *)calldata;
4718
4719         dprintk("--> %s\n", __func__);
4720         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4721         /* just setup sequence, do not trigger session recovery
4722            since we're invoked within one */
4723         ret = nfs41_setup_sequence(data->clp->cl_session,
4724                                    &data->args->la_seq_args,
4725                                    &data->res->lr_seq_res, 0, task);
4726
4727         BUG_ON(ret == -EAGAIN);
4728         rpc_call_start(task);
4729         dprintk("<-- %s\n", __func__);
4730 }
4731
4732 /*
4733  * Called from nfs4_state_manager thread for session setup, so don't recover
4734  * from sequence operation or clientid errors.
4735  */
4736 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4737 {
4738         struct nfs4_get_lease_time_data *data =
4739                         (struct nfs4_get_lease_time_data *)calldata;
4740
4741         dprintk("--> %s\n", __func__);
4742         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4743                 return;
4744         switch (task->tk_status) {
4745         case -NFS4ERR_DELAY:
4746         case -NFS4ERR_GRACE:
4747                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4748                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4749                 task->tk_status = 0;
4750                 nfs_restart_rpc(task, data->clp);
4751                 return;
4752         }
4753         dprintk("<-- %s\n", __func__);
4754 }
4755
4756 struct rpc_call_ops nfs4_get_lease_time_ops = {
4757         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4758         .rpc_call_done = nfs4_get_lease_time_done,
4759 };
4760
4761 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4762 {
4763         struct rpc_task *task;
4764         struct nfs4_get_lease_time_args args;
4765         struct nfs4_get_lease_time_res res = {
4766                 .lr_fsinfo = fsinfo,
4767         };
4768         struct nfs4_get_lease_time_data data = {
4769                 .args = &args,
4770                 .res = &res,
4771                 .clp = clp,
4772         };
4773         struct rpc_message msg = {
4774                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4775                 .rpc_argp = &args,
4776                 .rpc_resp = &res,
4777         };
4778         struct rpc_task_setup task_setup = {
4779                 .rpc_client = clp->cl_rpcclient,
4780                 .rpc_message = &msg,
4781                 .callback_ops = &nfs4_get_lease_time_ops,
4782                 .callback_data = &data
4783         };
4784         int status;
4785
4786         dprintk("--> %s\n", __func__);
4787         task = rpc_run_task(&task_setup);
4788
4789         if (IS_ERR(task))
4790                 status = PTR_ERR(task);
4791         else {
4792                 status = task->tk_status;
4793                 rpc_put_task(task);
4794         }
4795         dprintk("<-- %s return %d\n", __func__, status);
4796
4797         return status;
4798 }
4799
4800 /*
4801  * Reset a slot table
4802  */
4803 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4804                                  int ivalue)
4805 {
4806         struct nfs4_slot *new = NULL;
4807         int i;
4808         int ret = 0;
4809
4810         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4811                 max_reqs, tbl->max_slots);
4812
4813         /* Does the newly negotiated max_reqs match the existing slot table? */
4814         if (max_reqs != tbl->max_slots) {
4815                 ret = -ENOMEM;
4816                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4817                               GFP_NOFS);
4818                 if (!new)
4819                         goto out;
4820                 ret = 0;
4821                 kfree(tbl->slots);
4822         }
4823         spin_lock(&tbl->slot_tbl_lock);
4824         if (new) {
4825                 tbl->slots = new;
4826                 tbl->max_slots = max_reqs;
4827         }
4828         for (i = 0; i < tbl->max_slots; ++i)
4829                 tbl->slots[i].seq_nr = ivalue;
4830         spin_unlock(&tbl->slot_tbl_lock);
4831         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4832                 tbl, tbl->slots, tbl->max_slots);
4833 out:
4834         dprintk("<-- %s: return %d\n", __func__, ret);
4835         return ret;
4836 }
4837
4838 /*
4839  * Reset the forechannel and backchannel slot tables
4840  */
4841 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4842 {
4843         int status;
4844
4845         status = nfs4_reset_slot_table(&session->fc_slot_table,
4846                         session->fc_attrs.max_reqs, 1);
4847         if (status)
4848                 return status;
4849
4850         status = nfs4_reset_slot_table(&session->bc_slot_table,
4851                         session->bc_attrs.max_reqs, 0);
4852         return status;
4853 }
4854
4855 /* Destroy the slot table */
4856 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4857 {
4858         if (session->fc_slot_table.slots != NULL) {
4859                 kfree(session->fc_slot_table.slots);
4860                 session->fc_slot_table.slots = NULL;
4861         }
4862         if (session->bc_slot_table.slots != NULL) {
4863                 kfree(session->bc_slot_table.slots);
4864                 session->bc_slot_table.slots = NULL;
4865         }
4866         return;
4867 }
4868
4869 /*
4870  * Initialize slot table
4871  */
4872 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4873                 int max_slots, int ivalue)
4874 {
4875         struct nfs4_slot *slot;
4876         int ret = -ENOMEM;
4877
4878         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4879
4880         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4881
4882         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4883         if (!slot)
4884                 goto out;
4885         ret = 0;
4886
4887         spin_lock(&tbl->slot_tbl_lock);
4888         tbl->max_slots = max_slots;
4889         tbl->slots = slot;
4890         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4891         spin_unlock(&tbl->slot_tbl_lock);
4892         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4893                 tbl, tbl->slots, tbl->max_slots);
4894 out:
4895         dprintk("<-- %s: return %d\n", __func__, ret);
4896         return ret;
4897 }
4898
4899 /*
4900  * Initialize the forechannel and backchannel tables
4901  */
4902 static int nfs4_init_slot_tables(struct nfs4_session *session)
4903 {
4904         struct nfs4_slot_table *tbl;
4905         int status = 0;
4906
4907         tbl = &session->fc_slot_table;
4908         if (tbl->slots == NULL) {
4909                 status = nfs4_init_slot_table(tbl,
4910                                 session->fc_attrs.max_reqs, 1);
4911                 if (status)
4912                         return status;
4913         }
4914
4915         tbl = &session->bc_slot_table;
4916         if (tbl->slots == NULL) {
4917                 status = nfs4_init_slot_table(tbl,
4918                                 session->bc_attrs.max_reqs, 0);
4919                 if (status)
4920                         nfs4_destroy_slot_tables(session);
4921         }
4922
4923         return status;
4924 }
4925
4926 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4927 {
4928         struct nfs4_session *session;
4929         struct nfs4_slot_table *tbl;
4930
4931         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
4932         if (!session)
4933                 return NULL;
4934
4935         tbl = &session->fc_slot_table;
4936         tbl->highest_used_slotid = -1;
4937         spin_lock_init(&tbl->slot_tbl_lock);
4938         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
4939         init_completion(&tbl->complete);
4940
4941         tbl = &session->bc_slot_table;
4942         tbl->highest_used_slotid = -1;
4943         spin_lock_init(&tbl->slot_tbl_lock);
4944         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
4945         init_completion(&tbl->complete);
4946
4947         session->session_state = 1<<NFS4_SESSION_INITING;
4948
4949         session->clp = clp;
4950         return session;
4951 }
4952
4953 void nfs4_destroy_session(struct nfs4_session *session)
4954 {
4955         nfs4_proc_destroy_session(session);
4956         dprintk("%s Destroy backchannel for xprt %p\n",
4957                 __func__, session->clp->cl_rpcclient->cl_xprt);
4958         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
4959                                 NFS41_BC_MIN_CALLBACKS);
4960         nfs4_destroy_slot_tables(session);
4961         kfree(session);
4962 }
4963
4964 /*
4965  * Initialize the values to be used by the client in CREATE_SESSION
4966  * If nfs4_init_session set the fore channel request and response sizes,
4967  * use them.
4968  *
4969  * Set the back channel max_resp_sz_cached to zero to force the client to
4970  * always set csa_cachethis to FALSE because the current implementation
4971  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
4972  */
4973 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
4974 {
4975         struct nfs4_session *session = args->client->cl_session;
4976         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
4977                      mxresp_sz = session->fc_attrs.max_resp_sz;
4978
4979         if (mxrqst_sz == 0)
4980                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
4981         if (mxresp_sz == 0)
4982                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
4983         /* Fore channel attributes */
4984         args->fc_attrs.headerpadsz = 0;
4985         args->fc_attrs.max_rqst_sz = mxrqst_sz;
4986         args->fc_attrs.max_resp_sz = mxresp_sz;
4987         args->fc_attrs.max_ops = NFS4_MAX_OPS;
4988         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
4989
4990         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
4991                 "max_ops=%u max_reqs=%u\n",
4992                 __func__,
4993                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
4994                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
4995
4996         /* Back channel attributes */
4997         args->bc_attrs.headerpadsz = 0;
4998         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
4999         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5000         args->bc_attrs.max_resp_sz_cached = 0;
5001         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5002         args->bc_attrs.max_reqs = 1;
5003
5004         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5005                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5006                 __func__,
5007                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5008                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5009                 args->bc_attrs.max_reqs);
5010 }
5011
5012 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5013 {
5014         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5015         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5016
5017         if (rcvd->headerpadsz > sent->headerpadsz)
5018                 return -EINVAL;
5019         if (rcvd->max_resp_sz > sent->max_resp_sz)
5020                 return -EINVAL;
5021         /*
5022          * Our requested max_ops is the minimum we need; we're not
5023          * prepared to break up compounds into smaller pieces than that.
5024          * So, no point even trying to continue if the server won't
5025          * cooperate:
5026          */
5027         if (rcvd->max_ops < sent->max_ops)
5028                 return -EINVAL;
5029         if (rcvd->max_reqs == 0)
5030                 return -EINVAL;
5031         return 0;
5032 }
5033
5034 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5035 {
5036         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5037         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5038
5039         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5040                 return -EINVAL;
5041         if (rcvd->max_resp_sz < sent->max_resp_sz)
5042                 return -EINVAL;
5043         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5044                 return -EINVAL;
5045         /* These would render the backchannel useless: */
5046         if (rcvd->max_ops  == 0)
5047                 return -EINVAL;
5048         if (rcvd->max_reqs == 0)
5049                 return -EINVAL;
5050         return 0;
5051 }
5052
5053 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5054                                      struct nfs4_session *session)
5055 {
5056         int ret;
5057
5058         ret = nfs4_verify_fore_channel_attrs(args, session);
5059         if (ret)
5060                 return ret;
5061         return nfs4_verify_back_channel_attrs(args, session);
5062 }
5063
5064 static int _nfs4_proc_create_session(struct nfs_client *clp)
5065 {
5066         struct nfs4_session *session = clp->cl_session;
5067         struct nfs41_create_session_args args = {
5068                 .client = clp,
5069                 .cb_program = NFS4_CALLBACK,
5070         };
5071         struct nfs41_create_session_res res = {
5072                 .client = clp,
5073         };
5074         struct rpc_message msg = {
5075                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5076                 .rpc_argp = &args,
5077                 .rpc_resp = &res,
5078         };
5079         int status;
5080
5081         nfs4_init_channel_attrs(&args);
5082         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5083
5084         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5085
5086         if (!status)
5087                 /* Verify the session's negotiated channel_attrs values */
5088                 status = nfs4_verify_channel_attrs(&args, session);
5089         if (!status) {
5090                 /* Increment the clientid slot sequence id */
5091                 clp->cl_seqid++;
5092         }
5093
5094         return status;
5095 }
5096
5097 /*
5098  * Issues a CREATE_SESSION operation to the server.
5099  * It is the responsibility of the caller to verify the session is
5100  * expired before calling this routine.
5101  */
5102 int nfs4_proc_create_session(struct nfs_client *clp)
5103 {
5104         int status;
5105         unsigned *ptr;
5106         struct nfs4_session *session = clp->cl_session;
5107         long timeout = 0;
5108         int err;
5109
5110         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5111
5112         do {
5113                 status = _nfs4_proc_create_session(clp);
5114                 if (status == -NFS4ERR_DELAY) {
5115                         err = nfs4_delay(clp->cl_rpcclient, &timeout);
5116                         if (err)
5117                                 status = err;
5118                 }
5119         } while (status == -NFS4ERR_DELAY);
5120
5121         if (status)
5122                 goto out;
5123
5124         /* Init and reset the fore channel */
5125         status = nfs4_init_slot_tables(session);
5126         dprintk("slot table initialization returned %d\n", status);
5127         if (status)
5128                 goto out;
5129         status = nfs4_reset_slot_tables(session);
5130         dprintk("slot table reset returned %d\n", status);
5131         if (status)
5132                 goto out;
5133
5134         ptr = (unsigned *)&session->sess_id.data[0];
5135         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5136                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5137 out:
5138         dprintk("<-- %s\n", __func__);
5139         return status;
5140 }
5141
5142 /*
5143  * Issue the over-the-wire RPC DESTROY_SESSION.
5144  * The caller must serialize access to this routine.
5145  */
5146 int nfs4_proc_destroy_session(struct nfs4_session *session)
5147 {
5148         int status = 0;
5149         struct rpc_message msg;
5150
5151         dprintk("--> nfs4_proc_destroy_session\n");
5152
5153         /* session is still being setup */
5154         if (session->clp->cl_cons_state != NFS_CS_READY)
5155                 return status;
5156
5157         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5158         msg.rpc_argp = session;
5159         msg.rpc_resp = NULL;
5160         msg.rpc_cred = NULL;
5161         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
5162
5163         if (status)
5164                 printk(KERN_WARNING
5165                         "Got error %d from the server on DESTROY_SESSION. "
5166                         "Session has been destroyed regardless...\n", status);
5167
5168         dprintk("<-- nfs4_proc_destroy_session\n");
5169         return status;
5170 }
5171
5172 int nfs4_init_session(struct nfs_server *server)
5173 {
5174         struct nfs_client *clp = server->nfs_client;
5175         struct nfs4_session *session;
5176         unsigned int rsize, wsize;
5177         int ret;
5178
5179         if (!nfs4_has_session(clp))
5180                 return 0;
5181
5182         session = clp->cl_session;
5183         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5184                 return 0;
5185
5186         rsize = server->rsize;
5187         if (rsize == 0)
5188                 rsize = NFS_MAX_FILE_IO_SIZE;
5189         wsize = server->wsize;
5190         if (wsize == 0)
5191                 wsize = NFS_MAX_FILE_IO_SIZE;
5192
5193         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5194         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5195
5196         ret = nfs4_recover_expired_lease(server);
5197         if (!ret)
5198                 ret = nfs4_check_client_ready(clp);
5199         return ret;
5200 }
5201
5202 int nfs4_init_ds_session(struct nfs_client *clp)
5203 {
5204         struct nfs4_session *session = clp->cl_session;
5205         int ret;
5206
5207         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5208                 return 0;
5209
5210         ret = nfs4_client_recover_expired_lease(clp);
5211         if (!ret)
5212                 /* Test for the DS role */
5213                 if (!is_ds_client(clp))
5214                         ret = -ENODEV;
5215         if (!ret)
5216                 ret = nfs4_check_client_ready(clp);
5217         return ret;
5218
5219 }
5220 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5221
5222
5223 /*
5224  * Renew the cl_session lease.
5225  */
5226 struct nfs4_sequence_data {
5227         struct nfs_client *clp;
5228         struct nfs4_sequence_args args;
5229         struct nfs4_sequence_res res;
5230 };
5231
5232 static void nfs41_sequence_release(void *data)
5233 {
5234         struct nfs4_sequence_data *calldata = data;
5235         struct nfs_client *clp = calldata->clp;
5236
5237         if (atomic_read(&clp->cl_count) > 1)
5238                 nfs4_schedule_state_renewal(clp);
5239         nfs_put_client(clp);
5240         kfree(calldata);
5241 }
5242
5243 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5244 {
5245         switch(task->tk_status) {
5246         case -NFS4ERR_DELAY:
5247                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5248                 return -EAGAIN;
5249         default:
5250                 nfs4_schedule_lease_recovery(clp);
5251         }
5252         return 0;
5253 }
5254
5255 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5256 {
5257         struct nfs4_sequence_data *calldata = data;
5258         struct nfs_client *clp = calldata->clp;
5259
5260         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5261                 return;
5262
5263         if (task->tk_status < 0) {
5264                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5265                 if (atomic_read(&clp->cl_count) == 1)
5266                         goto out;
5267
5268                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5269                         rpc_restart_call_prepare(task);
5270                         return;
5271                 }
5272         }
5273         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5274 out:
5275         dprintk("<-- %s\n", __func__);
5276 }
5277
5278 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5279 {
5280         struct nfs4_sequence_data *calldata = data;
5281         struct nfs_client *clp = calldata->clp;
5282         struct nfs4_sequence_args *args;
5283         struct nfs4_sequence_res *res;
5284
5285         args = task->tk_msg.rpc_argp;
5286         res = task->tk_msg.rpc_resp;
5287
5288         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5289                 return;
5290         rpc_call_start(task);
5291 }
5292
5293 static const struct rpc_call_ops nfs41_sequence_ops = {
5294         .rpc_call_done = nfs41_sequence_call_done,
5295         .rpc_call_prepare = nfs41_sequence_prepare,
5296         .rpc_release = nfs41_sequence_release,
5297 };
5298
5299 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5300 {
5301         struct nfs4_sequence_data *calldata;
5302         struct rpc_message msg = {
5303                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5304                 .rpc_cred = cred,
5305         };
5306         struct rpc_task_setup task_setup_data = {
5307                 .rpc_client = clp->cl_rpcclient,
5308                 .rpc_message = &msg,
5309                 .callback_ops = &nfs41_sequence_ops,
5310                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5311         };
5312
5313         if (!atomic_inc_not_zero(&clp->cl_count))
5314                 return ERR_PTR(-EIO);
5315         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5316         if (calldata == NULL) {
5317                 nfs_put_client(clp);
5318                 return ERR_PTR(-ENOMEM);
5319         }
5320         msg.rpc_argp = &calldata->args;
5321         msg.rpc_resp = &calldata->res;
5322         calldata->clp = clp;
5323         task_setup_data.callback_data = calldata;
5324
5325         return rpc_run_task(&task_setup_data);
5326 }
5327
5328 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5329 {
5330         struct rpc_task *task;
5331         int ret = 0;
5332
5333         task = _nfs41_proc_sequence(clp, cred);
5334         if (IS_ERR(task))
5335                 ret = PTR_ERR(task);
5336         else
5337                 rpc_put_task_async(task);
5338         dprintk("<-- %s status=%d\n", __func__, ret);
5339         return ret;
5340 }
5341
5342 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5343 {
5344         struct rpc_task *task;
5345         int ret;
5346
5347         task = _nfs41_proc_sequence(clp, cred);
5348         if (IS_ERR(task)) {
5349                 ret = PTR_ERR(task);
5350                 goto out;
5351         }
5352         ret = rpc_wait_for_completion_task(task);
5353         if (!ret) {
5354                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5355
5356                 if (task->tk_status == 0)
5357                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5358                 ret = task->tk_status;
5359         }
5360         rpc_put_task(task);
5361 out:
5362         dprintk("<-- %s status=%d\n", __func__, ret);
5363         return ret;
5364 }
5365
5366 struct nfs4_reclaim_complete_data {
5367         struct nfs_client *clp;
5368         struct nfs41_reclaim_complete_args arg;
5369         struct nfs41_reclaim_complete_res res;
5370 };
5371
5372 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5373 {
5374         struct nfs4_reclaim_complete_data *calldata = data;
5375
5376         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5377         if (nfs41_setup_sequence(calldata->clp->cl_session,
5378                                 &calldata->arg.seq_args,
5379                                 &calldata->res.seq_res, 0, task))
5380                 return;
5381
5382         rpc_call_start(task);
5383 }
5384
5385 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5386 {
5387         switch(task->tk_status) {
5388         case 0:
5389         case -NFS4ERR_COMPLETE_ALREADY:
5390         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5391                 break;
5392         case -NFS4ERR_DELAY:
5393                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5394                 return -EAGAIN;
5395         default:
5396                 nfs4_schedule_lease_recovery(clp);
5397         }
5398         return 0;
5399 }
5400
5401 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5402 {
5403         struct nfs4_reclaim_complete_data *calldata = data;
5404         struct nfs_client *clp = calldata->clp;
5405         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5406
5407         dprintk("--> %s\n", __func__);
5408         if (!nfs41_sequence_done(task, res))
5409                 return;
5410
5411         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5412                 rpc_restart_call_prepare(task);
5413                 return;
5414         }
5415         dprintk("<-- %s\n", __func__);
5416 }
5417
5418 static void nfs4_free_reclaim_complete_data(void *data)
5419 {
5420         struct nfs4_reclaim_complete_data *calldata = data;
5421
5422         kfree(calldata);
5423 }
5424
5425 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5426         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5427         .rpc_call_done = nfs4_reclaim_complete_done,
5428         .rpc_release = nfs4_free_reclaim_complete_data,
5429 };
5430
5431 /*
5432  * Issue a global reclaim complete.
5433  */
5434 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5435 {
5436         struct nfs4_reclaim_complete_data *calldata;
5437         struct rpc_task *task;
5438         struct rpc_message msg = {
5439                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5440         };
5441         struct rpc_task_setup task_setup_data = {
5442                 .rpc_client = clp->cl_rpcclient,
5443                 .rpc_message = &msg,
5444                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5445                 .flags = RPC_TASK_ASYNC,
5446         };
5447         int status = -ENOMEM;
5448
5449         dprintk("--> %s\n", __func__);
5450         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5451         if (calldata == NULL)
5452                 goto out;
5453         calldata->clp = clp;
5454         calldata->arg.one_fs = 0;
5455
5456         msg.rpc_argp = &calldata->arg;
5457         msg.rpc_resp = &calldata->res;
5458         task_setup_data.callback_data = calldata;
5459         task = rpc_run_task(&task_setup_data);
5460         if (IS_ERR(task)) {
5461                 status = PTR_ERR(task);
5462                 goto out;
5463         }
5464         status = nfs4_wait_for_completion_rpc_task(task);
5465         if (status == 0)
5466                 status = task->tk_status;
5467         rpc_put_task(task);
5468         return 0;
5469 out:
5470         dprintk("<-- %s status=%d\n", __func__, status);
5471         return status;
5472 }
5473
5474 static void
5475 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5476 {
5477         struct nfs4_layoutget *lgp = calldata;
5478         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5479
5480         dprintk("--> %s\n", __func__);
5481         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5482          * right now covering the LAYOUTGET we are about to send.
5483          * However, that is not so catastrophic, and there seems
5484          * to be no way to prevent it completely.
5485          */
5486         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5487                                 &lgp->res.seq_res, 0, task))
5488                 return;
5489         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5490                                           NFS_I(lgp->args.inode)->layout,
5491                                           lgp->args.ctx->state)) {
5492                 rpc_exit(task, NFS4_OK);
5493                 return;
5494         }
5495         rpc_call_start(task);
5496 }
5497
5498 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5499 {
5500         struct nfs4_layoutget *lgp = calldata;
5501         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5502
5503         dprintk("--> %s\n", __func__);
5504
5505         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5506                 return;
5507
5508         switch (task->tk_status) {
5509         case 0:
5510                 break;
5511         case -NFS4ERR_LAYOUTTRYLATER:
5512         case -NFS4ERR_RECALLCONFLICT:
5513                 task->tk_status = -NFS4ERR_DELAY;
5514                 /* Fall through */
5515         default:
5516                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5517                         rpc_restart_call_prepare(task);
5518                         return;
5519                 }
5520         }
5521         dprintk("<-- %s\n", __func__);
5522 }
5523
5524 static void nfs4_layoutget_release(void *calldata)
5525 {
5526         struct nfs4_layoutget *lgp = calldata;
5527
5528         dprintk("--> %s\n", __func__);
5529         put_nfs_open_context(lgp->args.ctx);
5530         kfree(calldata);
5531         dprintk("<-- %s\n", __func__);
5532 }
5533
5534 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5535         .rpc_call_prepare = nfs4_layoutget_prepare,
5536         .rpc_call_done = nfs4_layoutget_done,
5537         .rpc_release = nfs4_layoutget_release,
5538 };
5539
5540 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5541 {
5542         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5543         struct rpc_task *task;
5544         struct rpc_message msg = {
5545                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5546                 .rpc_argp = &lgp->args,
5547                 .rpc_resp = &lgp->res,
5548         };
5549         struct rpc_task_setup task_setup_data = {
5550                 .rpc_client = server->client,
5551                 .rpc_message = &msg,
5552                 .callback_ops = &nfs4_layoutget_call_ops,
5553                 .callback_data = lgp,
5554                 .flags = RPC_TASK_ASYNC,
5555         };
5556         int status = 0;
5557
5558         dprintk("--> %s\n", __func__);
5559
5560         lgp->res.layoutp = &lgp->args.layout;
5561         lgp->res.seq_res.sr_slot = NULL;
5562         task = rpc_run_task(&task_setup_data);
5563         if (IS_ERR(task))
5564                 return PTR_ERR(task);
5565         status = nfs4_wait_for_completion_rpc_task(task);
5566         if (status == 0)
5567                 status = task->tk_status;
5568         if (status == 0)
5569                 status = pnfs_layout_process(lgp);
5570         rpc_put_task(task);
5571         dprintk("<-- %s status=%d\n", __func__, status);
5572         return status;
5573 }
5574
5575 static int
5576 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5577 {
5578         struct nfs4_getdeviceinfo_args args = {
5579                 .pdev = pdev,
5580         };
5581         struct nfs4_getdeviceinfo_res res = {
5582                 .pdev = pdev,
5583         };
5584         struct rpc_message msg = {
5585                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5586                 .rpc_argp = &args,
5587                 .rpc_resp = &res,
5588         };
5589         int status;
5590
5591         dprintk("--> %s\n", __func__);
5592         status = nfs4_call_sync(server, &msg, &args, &res, 0);
5593         dprintk("<-- %s status=%d\n", __func__, status);
5594
5595         return status;
5596 }
5597
5598 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5599 {
5600         struct nfs4_exception exception = { };
5601         int err;
5602
5603         do {
5604                 err = nfs4_handle_exception(server,
5605                                         _nfs4_proc_getdeviceinfo(server, pdev),
5606                                         &exception);
5607         } while (exception.retry);
5608         return err;
5609 }
5610 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5611
5612 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5613 {
5614         struct nfs4_layoutcommit_data *data = calldata;
5615         struct nfs_server *server = NFS_SERVER(data->args.inode);
5616
5617         if (nfs4_setup_sequence(server, &data->args.seq_args,
5618                                 &data->res.seq_res, 1, task))
5619                 return;
5620         rpc_call_start(task);
5621 }
5622
5623 static void
5624 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5625 {
5626         struct nfs4_layoutcommit_data *data = calldata;
5627         struct nfs_server *server = NFS_SERVER(data->args.inode);
5628
5629         if (!nfs4_sequence_done(task, &data->res.seq_res))
5630                 return;
5631
5632         switch (task->tk_status) { /* Just ignore these failures */
5633         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5634         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5635         case NFS4ERR_BADLAYOUT:     /* no layout */
5636         case NFS4ERR_GRACE:         /* loca_recalim always false */
5637                 task->tk_status = 0;
5638         }
5639
5640         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5641                 nfs_restart_rpc(task, server->nfs_client);
5642                 return;
5643         }
5644
5645         if (task->tk_status == 0)
5646                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5647                                                    data->res.fattr);
5648 }
5649
5650 static void nfs4_layoutcommit_release(void *calldata)
5651 {
5652         struct nfs4_layoutcommit_data *data = calldata;
5653
5654         /* Matched by references in pnfs_set_layoutcommit */
5655         put_lseg(data->lseg);
5656         put_rpccred(data->cred);
5657         kfree(data);
5658 }
5659
5660 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5661         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5662         .rpc_call_done = nfs4_layoutcommit_done,
5663         .rpc_release = nfs4_layoutcommit_release,
5664 };
5665
5666 int
5667 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5668 {
5669         struct rpc_message msg = {
5670                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5671                 .rpc_argp = &data->args,
5672                 .rpc_resp = &data->res,
5673                 .rpc_cred = data->cred,
5674         };
5675         struct rpc_task_setup task_setup_data = {
5676                 .task = &data->task,
5677                 .rpc_client = NFS_CLIENT(data->args.inode),
5678                 .rpc_message = &msg,
5679                 .callback_ops = &nfs4_layoutcommit_ops,
5680                 .callback_data = data,
5681                 .flags = RPC_TASK_ASYNC,
5682         };
5683         struct rpc_task *task;
5684         int status = 0;
5685
5686         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
5687                 "lbw: %llu inode %lu\n",
5688                 data->task.tk_pid, sync,
5689                 data->args.lastbytewritten,
5690                 data->args.inode->i_ino);
5691
5692         task = rpc_run_task(&task_setup_data);
5693         if (IS_ERR(task))
5694                 return PTR_ERR(task);
5695         if (sync == false)
5696                 goto out;
5697         status = nfs4_wait_for_completion_rpc_task(task);
5698         if (status != 0)
5699                 goto out;
5700         status = task->tk_status;
5701 out:
5702         dprintk("%s: status %d\n", __func__, status);
5703         rpc_put_task(task);
5704         return status;
5705 }
5706 #endif /* CONFIG_NFS_V4_1 */
5707
5708 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5709         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5710         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5711         .recover_open   = nfs4_open_reclaim,
5712         .recover_lock   = nfs4_lock_reclaim,
5713         .establish_clid = nfs4_init_clientid,
5714         .get_clid_cred  = nfs4_get_setclientid_cred,
5715 };
5716
5717 #if defined(CONFIG_NFS_V4_1)
5718 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5719         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5720         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5721         .recover_open   = nfs4_open_reclaim,
5722         .recover_lock   = nfs4_lock_reclaim,
5723         .establish_clid = nfs41_init_clientid,
5724         .get_clid_cred  = nfs4_get_exchange_id_cred,
5725         .reclaim_complete = nfs41_proc_reclaim_complete,
5726 };
5727 #endif /* CONFIG_NFS_V4_1 */
5728
5729 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5730         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5731         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5732         .recover_open   = nfs4_open_expired,
5733         .recover_lock   = nfs4_lock_expired,
5734         .establish_clid = nfs4_init_clientid,
5735         .get_clid_cred  = nfs4_get_setclientid_cred,
5736 };
5737
5738 #if defined(CONFIG_NFS_V4_1)
5739 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5740         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5741         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5742         .recover_open   = nfs4_open_expired,
5743         .recover_lock   = nfs4_lock_expired,
5744         .establish_clid = nfs41_init_clientid,
5745         .get_clid_cred  = nfs4_get_exchange_id_cred,
5746 };
5747 #endif /* CONFIG_NFS_V4_1 */
5748
5749 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5750         .sched_state_renewal = nfs4_proc_async_renew,
5751         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5752         .renew_lease = nfs4_proc_renew,
5753 };
5754
5755 #if defined(CONFIG_NFS_V4_1)
5756 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5757         .sched_state_renewal = nfs41_proc_async_sequence,
5758         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5759         .renew_lease = nfs4_proc_sequence,
5760 };
5761 #endif
5762
5763 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5764         .minor_version = 0,
5765         .call_sync = _nfs4_call_sync,
5766         .validate_stateid = nfs4_validate_delegation_stateid,
5767         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5768         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5769         .state_renewal_ops = &nfs40_state_renewal_ops,
5770 };
5771
5772 #if defined(CONFIG_NFS_V4_1)
5773 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5774         .minor_version = 1,
5775         .call_sync = _nfs4_call_sync_session,
5776         .validate_stateid = nfs41_validate_delegation_stateid,
5777         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5778         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5779         .state_renewal_ops = &nfs41_state_renewal_ops,
5780 };
5781 #endif
5782
5783 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5784         [0] = &nfs_v4_0_minor_ops,
5785 #if defined(CONFIG_NFS_V4_1)
5786         [1] = &nfs_v4_1_minor_ops,
5787 #endif
5788 };
5789
5790 static const struct inode_operations nfs4_file_inode_operations = {
5791         .permission     = nfs_permission,
5792         .getattr        = nfs_getattr,
5793         .setattr        = nfs_setattr,
5794         .getxattr       = generic_getxattr,
5795         .setxattr       = generic_setxattr,
5796         .listxattr      = generic_listxattr,
5797         .removexattr    = generic_removexattr,
5798 };
5799
5800 const struct nfs_rpc_ops nfs_v4_clientops = {
5801         .version        = 4,                    /* protocol version */
5802         .dentry_ops     = &nfs4_dentry_operations,
5803         .dir_inode_ops  = &nfs4_dir_inode_operations,
5804         .file_inode_ops = &nfs4_file_inode_operations,
5805         .getroot        = nfs4_proc_get_root,
5806         .getattr        = nfs4_proc_getattr,
5807         .setattr        = nfs4_proc_setattr,
5808         .lookupfh       = nfs4_proc_lookupfh,
5809         .lookup         = nfs4_proc_lookup,
5810         .access         = nfs4_proc_access,
5811         .readlink       = nfs4_proc_readlink,
5812         .create         = nfs4_proc_create,
5813         .remove         = nfs4_proc_remove,
5814         .unlink_setup   = nfs4_proc_unlink_setup,
5815         .unlink_done    = nfs4_proc_unlink_done,
5816         .rename         = nfs4_proc_rename,
5817         .rename_setup   = nfs4_proc_rename_setup,
5818         .rename_done    = nfs4_proc_rename_done,
5819         .link           = nfs4_proc_link,
5820         .symlink        = nfs4_proc_symlink,
5821         .mkdir          = nfs4_proc_mkdir,
5822         .rmdir          = nfs4_proc_remove,
5823         .readdir        = nfs4_proc_readdir,
5824         .mknod          = nfs4_proc_mknod,
5825         .statfs         = nfs4_proc_statfs,
5826         .fsinfo         = nfs4_proc_fsinfo,
5827         .pathconf       = nfs4_proc_pathconf,
5828         .set_capabilities = nfs4_server_capabilities,
5829         .decode_dirent  = nfs4_decode_dirent,
5830         .read_setup     = nfs4_proc_read_setup,
5831         .read_done      = nfs4_read_done,
5832         .write_setup    = nfs4_proc_write_setup,
5833         .write_done     = nfs4_write_done,
5834         .commit_setup   = nfs4_proc_commit_setup,
5835         .commit_done    = nfs4_commit_done,
5836         .lock           = nfs4_proc_lock,
5837         .clear_acl_cache = nfs4_zap_acl_attr,
5838         .close_context  = nfs4_close_context,
5839         .open_context   = nfs4_atomic_open,
5840         .init_client    = nfs4_init_client,
5841 };
5842
5843 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5844         .prefix = XATTR_NAME_NFSV4_ACL,
5845         .list   = nfs4_xattr_list_nfs4_acl,
5846         .get    = nfs4_xattr_get_nfs4_acl,
5847         .set    = nfs4_xattr_set_nfs4_acl,
5848 };
5849
5850 const struct xattr_handler *nfs4_xattr_handlers[] = {
5851         &nfs4_xattr_nfs4_acl_handler,
5852         NULL
5853 };
5854
5855 /*
5856  * Local variables:
5857  *  c-basic-offset: 8
5858  * End:
5859  */