fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/sunrpc/gss_api.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/nfs_idmap.h>
  56 #include <linux/sunrpc/bc_xprt.h>
  57 #include <linux/xattr.h>
  58 #include <linux/utsname.h>
  59 #include <linux/freezer.h>
  60
  61 #include "nfs4_fs.h"
  62 #include "delegation.h"
  63 #include "internal.h"
  64 #include "iostat.h"
  65 #include "callback.h"
  66 #include "pnfs.h"
  67
  68 #define NFSDBG_FACILITY         NFSDBG_PROC
  69
  70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  72
  73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  74
  75 static unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  82 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  83 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  84                             struct nfs_fattr *fattr, struct iattr *sattr,
  85                             struct nfs4_state *state);
  86 #ifdef CONFIG_NFS_V4_1
  87 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  88 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  89 #endif
  90 /* Prevent leaks of NFSv4 errors into userland */
  91 static int nfs4_map_errors(int err)
  92 {
  93         if (err >= -1000)
  94                 return err;
  95         switch (err) {
  96         case -NFS4ERR_RESOURCE:
  97                 return -EREMOTEIO;
  98         case -NFS4ERR_WRONGSEC:
  99                 return -EPERM;
 100         case -NFS4ERR_BADOWNER:
 101         case -NFS4ERR_BADNAME:
 102                 return -EINVAL;
 103         default:
 104                 dprintk("%s could not handle NFSv4 error %d\n",
 105                                 __func__, -err);
 106                 break;
 107         }
 108         return -EIO;
 109 }
 110
 111 /*
 112  * This is our standard bitmap for GETATTR requests.
 113  */
 114 const u32 nfs4_fattr_bitmap[2] = {
 115         FATTR4_WORD0_TYPE
 116         | FATTR4_WORD0_CHANGE
 117         | FATTR4_WORD0_SIZE
 118         | FATTR4_WORD0_FSID
 119         | FATTR4_WORD0_FILEID,
 120         FATTR4_WORD1_MODE
 121         | FATTR4_WORD1_NUMLINKS
 122         | FATTR4_WORD1_OWNER
 123         | FATTR4_WORD1_OWNER_GROUP
 124         | FATTR4_WORD1_RAWDEV
 125         | FATTR4_WORD1_SPACE_USED
 126         | FATTR4_WORD1_TIME_ACCESS
 127         | FATTR4_WORD1_TIME_METADATA
 128         | FATTR4_WORD1_TIME_MODIFY
 129 };
 130
 131 const u32 nfs4_statfs_bitmap[2] = {
 132         FATTR4_WORD0_FILES_AVAIL
 133         | FATTR4_WORD0_FILES_FREE
 134         | FATTR4_WORD0_FILES_TOTAL,
 135         FATTR4_WORD1_SPACE_AVAIL
 136         | FATTR4_WORD1_SPACE_FREE
 137         | FATTR4_WORD1_SPACE_TOTAL
 138 };
 139
 140 const u32 nfs4_pathconf_bitmap[2] = {
 141         FATTR4_WORD0_MAXLINK
 142         | FATTR4_WORD0_MAXNAME,
 143         0
 144 };
 145
 146 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 147                         | FATTR4_WORD0_MAXREAD
 148                         | FATTR4_WORD0_MAXWRITE
 149                         | FATTR4_WORD0_LEASE_TIME,
 150                         FATTR4_WORD1_TIME_DELTA
 151                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 152                         FATTR4_WORD2_LAYOUT_BLKSIZE
 153 };
 154
 155 const u32 nfs4_fs_locations_bitmap[2] = {
 156         FATTR4_WORD0_TYPE
 157         | FATTR4_WORD0_CHANGE
 158         | FATTR4_WORD0_SIZE
 159         | FATTR4_WORD0_FSID
 160         | FATTR4_WORD0_FILEID
 161         | FATTR4_WORD0_FS_LOCATIONS,
 162         FATTR4_WORD1_MODE
 163         | FATTR4_WORD1_NUMLINKS
 164         | FATTR4_WORD1_OWNER
 165         | FATTR4_WORD1_OWNER_GROUP
 166         | FATTR4_WORD1_RAWDEV
 167         | FATTR4_WORD1_SPACE_USED
 168         | FATTR4_WORD1_TIME_ACCESS
 169         | FATTR4_WORD1_TIME_METADATA
 170         | FATTR4_WORD1_TIME_MODIFY
 171         | FATTR4_WORD1_MOUNTED_ON_FILEID
 172 };
 173
 174 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 175                 struct nfs4_readdir_arg *readdir)
 176 {
 177         __be32 *start, *p;
 178
 179         BUG_ON(readdir->count < 80);
 180         if (cookie > 2) {
 181                 readdir->cookie = cookie;
 182                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 183                 return;
 184         }
 185
 186         readdir->cookie = 0;
 187         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 188         if (cookie == 2)
 189                 return;
 190
 191         /*
 192          * NFSv4 servers do not return entries for '.' and '..'
 193          * Therefore, we fake these entries here.  We let '.'
 194          * have cookie 0 and '..' have cookie 1.  Note that
 195          * when talking to the server, we always send cookie 0
 196          * instead of 1 or 2.
 197          */
 198         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 199
 200         if (cookie == 0) {
 201                 *p++ = xdr_one;                                  /* next */
 202                 *p++ = xdr_zero;                   /* cookie, first word */
 203                 *p++ = xdr_one;                   /* cookie, second word */
 204                 *p++ = xdr_one;                             /* entry len */
 205                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 206                 p++;
 207                 *p++ = xdr_one;                         /* bitmap length */
 208                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 209                 *p++ = htonl(8);              /* attribute buffer length */
 210                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 211         }
 212
 213         *p++ = xdr_one;                                  /* next */
 214         *p++ = xdr_zero;                   /* cookie, first word */
 215         *p++ = xdr_two;                   /* cookie, second word */
 216         *p++ = xdr_two;                             /* entry len */
 217         memcpy(p, "..\0\0", 4);                         /* entry */
 218         p++;
 219         *p++ = xdr_one;                         /* bitmap length */
 220         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 221         *p++ = htonl(8);              /* attribute buffer length */
 222         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 223
 224         readdir->pgbase = (char *)p - (char *)start;
 225         readdir->count -= readdir->pgbase;
 226         kunmap_atomic(start, KM_USER0);
 227 }
 228
 229 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 230 {
 231         int res;
 232
 233         might_sleep();
 234
 235         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 236                         nfs_wait_bit_killable, TASK_KILLABLE);
 237         return res;
 238 }
 239
 240 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 241 {
 242         int res = 0;
 243
 244         might_sleep();
 245
 246         if (*timeout <= 0)
 247                 *timeout = NFS4_POLL_RETRY_MIN;
 248         if (*timeout > NFS4_POLL_RETRY_MAX)
 249                 *timeout = NFS4_POLL_RETRY_MAX;
 250         freezable_schedule_timeout_killable(*timeout);
 251         if (fatal_signal_pending(current))
 252                 res = -ERESTARTSYS;
 253         *timeout <<= 1;
 254         return res;
 255 }
 256
 257 /* This is the error handling routine for processes that are allowed
 258  * to sleep.
 259  */
 260 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 261 {
 262         struct nfs_client *clp = server->nfs_client;
 263         struct nfs4_state *state = exception->state;
 264         int ret = errorcode;
 265
 266         exception->retry = 0;
 267         switch(errorcode) {
 268                 case 0:
 269                         return 0;
 270                 case -NFS4ERR_ADMIN_REVOKED:
 271                 case -NFS4ERR_BAD_STATEID:
 272                 case -NFS4ERR_OPENMODE:
 273                         if (state == NULL)
 274                                 break;
 275                         nfs4_schedule_stateid_recovery(server, state);
 276                         goto wait_on_recovery;
 277                 case -NFS4ERR_EXPIRED:
 278                         if (state != NULL)
 279                                 nfs4_schedule_stateid_recovery(server, state);
 280                 case -NFS4ERR_STALE_STATEID:
 281                 case -NFS4ERR_STALE_CLIENTID:
 282                         nfs4_schedule_lease_recovery(clp);
 283                         goto wait_on_recovery;
 284 #if defined(CONFIG_NFS_V4_1)
 285                 case -NFS4ERR_BADSESSION:
 286                 case -NFS4ERR_BADSLOT:
 287                 case -NFS4ERR_BAD_HIGH_SLOT:
 288                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 289                 case -NFS4ERR_DEADSESSION:
 290                 case -NFS4ERR_SEQ_FALSE_RETRY:
 291                 case -NFS4ERR_SEQ_MISORDERED:
 292                         dprintk("%s ERROR: %d Reset session\n", __func__,
 293                                 errorcode);
 294                         nfs4_schedule_session_recovery(clp->cl_session);
 295                         exception->retry = 1;
 296                         break;
 297 #endif /* defined(CONFIG_NFS_V4_1) */
 298                 case -NFS4ERR_FILE_OPEN:
 299                         if (exception->timeout > HZ) {
 300                                 /* We have retried a decent amount, time to
 301                                  * fail
 302                                  */
 303                                 ret = -EBUSY;
 304                                 break;
 305                         }
 306                 case -NFS4ERR_GRACE:
 307                 case -NFS4ERR_DELAY:
 308                 case -EKEYEXPIRED:
 309                         ret = nfs4_delay(server->client, &exception->timeout);
 310                         if (ret != 0)
 311                                 break;
 312                 case -NFS4ERR_RETRY_UNCACHED_REP:
 313                 case -NFS4ERR_OLD_STATEID:
 314                         exception->retry = 1;
 315                         break;
 316                 case -NFS4ERR_BADOWNER:
 317                         /* The following works around a Linux server bug! */
 318                 case -NFS4ERR_BADNAME:
 319                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 320                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 321                                 exception->retry = 1;
 322                                 printk(KERN_WARNING "NFS: v4 server %s "
 323                                                 "does not accept raw "
 324                                                 "uid/gids. "
 325                                                 "Reenabling the idmapper.\n",
 326                                                 server->nfs_client->cl_hostname);
 327                         }
 328         }
 329         /* We failed to handle the error */
 330         return nfs4_map_errors(ret);
 331 wait_on_recovery:
 332         ret = nfs4_wait_clnt_recover(clp);
 333         if (ret == 0)
 334                 exception->retry = 1;
 335         return ret;
 336 }
 337
 338
 339 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 340 {
 341         spin_lock(&clp->cl_lock);
 342         if (time_before(clp->cl_last_renewal,timestamp))
 343                 clp->cl_last_renewal = timestamp;
 344         spin_unlock(&clp->cl_lock);
 345 }
 346
 347 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 348 {
 349         do_renew_lease(server->nfs_client, timestamp);
 350 }
 351
 352 #if defined(CONFIG_NFS_V4_1)
 353
 354 /*
 355  * nfs4_free_slot - free a slot and efficiently update slot table.
 356  *
 357  * freeing a slot is trivially done by clearing its respective bit
 358  * in the bitmap.
 359  * If the freed slotid equals highest_used_slotid we want to update it
 360  * so that the server would be able to size down the slot table if needed,
 361  * otherwise we know that the highest_used_slotid is still in use.
 362  * When updating highest_used_slotid there may be "holes" in the bitmap
 363  * so we need to scan down from highest_used_slotid to 0 looking for the now
 364  * highest slotid in use.
 365  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 366  *
 367  * Must be called while holding tbl->slot_tbl_lock
 368  */
 369 static void
 370 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 371 {
 372         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 373         /* clear used bit in bitmap */
 374         __clear_bit(slotid, tbl->used_slots);
 375
 376         /* update highest_used_slotid when it is freed */
 377         if (slotid == tbl->highest_used_slotid) {
 378                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 379                 if (slotid < tbl->max_slots)
 380                         tbl->highest_used_slotid = slotid;
 381                 else
 382                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 383         }
 384         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 385                 slotid, tbl->highest_used_slotid);
 386 }
 387
 388 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 389 {
 390         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 391         return true;
 392 }
 393
 394 /*
 395  * Signal state manager thread if session fore channel is drained
 396  */
 397 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 398 {
 399         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 400                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 401                                 nfs4_set_task_privileged, NULL);
 402                 return;
 403         }
 404
 405         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 406                 return;
 407
 408         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 409         complete(&ses->fc_slot_table.complete);
 410 }
 411
 412 /*
 413  * Signal state manager thread if session back channel is drained
 414  */
 415 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 416 {
 417         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 418             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 419                 return;
 420         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 421         complete(&ses->bc_slot_table.complete);
 422 }
 423
 424 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 425 {
 426         struct nfs4_slot_table *tbl;
 427
 428         tbl = &res->sr_session->fc_slot_table;
 429         if (!res->sr_slot) {
 430                 /* just wake up the next guy waiting since
 431                  * we may have not consumed a slot after all */
 432                 dprintk("%s: No slot\n", __func__);
 433                 return;
 434         }
 435
 436         spin_lock(&tbl->slot_tbl_lock);
 437         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 438         nfs4_check_drain_fc_complete(res->sr_session);
 439         spin_unlock(&tbl->slot_tbl_lock);
 440         res->sr_slot = NULL;
 441 }
 442
 443 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 444 {
 445         unsigned long timestamp;
 446         struct nfs_client *clp;
 447
 448         /*
 449          * sr_status remains 1 if an RPC level error occurred. The server
 450          * may or may not have processed the sequence operation..
 451          * Proceed as if the server received and processed the sequence
 452          * operation.
 453          */
 454         if (res->sr_status == 1)
 455                 res->sr_status = NFS_OK;
 456
 457         /* don't increment the sequence number if the task wasn't sent */
 458         if (!RPC_WAS_SENT(task))
 459                 goto out;
 460
 461         /* Check the SEQUENCE operation status */
 462         switch (res->sr_status) {
 463         case 0:
 464                 /* Update the slot's sequence and clientid lease timer */
 465                 ++res->sr_slot->seq_nr;
 466                 timestamp = res->sr_renewal_time;
 467                 clp = res->sr_session->clp;
 468                 do_renew_lease(clp, timestamp);
 469                 /* Check sequence flags */
 470                 if (res->sr_status_flags != 0)
 471                         nfs4_schedule_lease_recovery(clp);
 472                 break;
 473         case -NFS4ERR_DELAY:
 474                 /* The server detected a resend of the RPC call and
 475                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 476                  * of RFC5661.
 477                  */
 478                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 479                         __func__,
 480                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 481                         res->sr_slot->seq_nr);
 482                 goto out_retry;
 483         default:
 484                 /* Just update the slot sequence no. */
 485                 ++res->sr_slot->seq_nr;
 486         }
 487 out:
 488         /* The session may be reset by one of the error handlers. */
 489         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 490         nfs41_sequence_free_slot(res);
 491         return 1;
 492 out_retry:
 493         if (!rpc_restart_call(task))
 494                 goto out;
 495         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 496         return 0;
 497 }
 498
 499 static int nfs4_sequence_done(struct rpc_task *task,
 500                                struct nfs4_sequence_res *res)
 501 {
 502         if (res->sr_session == NULL)
 503                 return 1;
 504         return nfs41_sequence_done(task, res);
 505 }
 506
 507 /*
 508  * nfs4_find_slot - efficiently look for a free slot
 509  *
 510  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 511  * If found, we mark the slot as used, update the highest_used_slotid,
 512  * and respectively set up the sequence operation args.
 513  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 514  *
 515  * Note: must be called with under the slot_tbl_lock.
 516  */
 517 static u32
 518 nfs4_find_slot(struct nfs4_slot_table *tbl)
 519 {
 520         u32 slotid;
 521         u32 ret_id = NFS4_NO_SLOT;
 522
 523         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 524                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 525                 tbl->max_slots);
 526         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 527         if (slotid >= tbl->max_slots)
 528                 goto out;
 529         __set_bit(slotid, tbl->used_slots);
 530         if (slotid > tbl->highest_used_slotid ||
 531                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 532                 tbl->highest_used_slotid = slotid;
 533         ret_id = slotid;
 534 out:
 535         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 536                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 537         return ret_id;
 538 }
 539
 540 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 541                 struct nfs4_sequence_res *res, int cache_reply)
 542 {
 543         args->sa_session = NULL;
 544         args->sa_cache_this = 0;
 545         if (cache_reply)
 546                 args->sa_cache_this = 1;
 547         res->sr_session = NULL;
 548         res->sr_slot = NULL;
 549 }
 550
 551 int nfs41_setup_sequence(struct nfs4_session *session,
 552                                 struct nfs4_sequence_args *args,
 553                                 struct nfs4_sequence_res *res,
 554                                 struct rpc_task *task)
 555 {
 556         struct nfs4_slot *slot;
 557         struct nfs4_slot_table *tbl;
 558         u32 slotid;
 559
 560         dprintk("--> %s\n", __func__);
 561         /* slot already allocated? */
 562         if (res->sr_slot != NULL)
 563                 return 0;
 564
 565         tbl = &session->fc_slot_table;
 566
 567         spin_lock(&tbl->slot_tbl_lock);
 568         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 569             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 570                 /* The state manager will wait until the slot table is empty */
 571                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 572                 spin_unlock(&tbl->slot_tbl_lock);
 573                 dprintk("%s session is draining\n", __func__);
 574                 return -EAGAIN;
 575         }
 576
 577         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 578             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 579                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 580                 spin_unlock(&tbl->slot_tbl_lock);
 581                 dprintk("%s enforce FIFO order\n", __func__);
 582                 return -EAGAIN;
 583         }
 584
 585         slotid = nfs4_find_slot(tbl);
 586         if (slotid == NFS4_NO_SLOT) {
 587                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 588                 spin_unlock(&tbl->slot_tbl_lock);
 589                 dprintk("<-- %s: no free slots\n", __func__);
 590                 return -EAGAIN;
 591         }
 592         spin_unlock(&tbl->slot_tbl_lock);
 593
 594         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 595         slot = tbl->slots + slotid;
 596         args->sa_session = session;
 597         args->sa_slotid = slotid;
 598
 599         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 600
 601         res->sr_session = session;
 602         res->sr_slot = slot;
 603         res->sr_renewal_time = jiffies;
 604         res->sr_status_flags = 0;
 605         /*
 606          * sr_status is only set in decode_sequence, and so will remain
 607          * set to 1 if an rpc level failure occurs.
 608          */
 609         res->sr_status = 1;
 610         return 0;
 611 }
 612 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 613
 614 int nfs4_setup_sequence(const struct nfs_server *server,
 615                         struct nfs4_sequence_args *args,
 616                         struct nfs4_sequence_res *res,
 617                         struct rpc_task *task)
 618 {
 619         struct nfs4_session *session = nfs4_get_session(server);
 620         int ret = 0;
 621
 622         if (session == NULL)
 623                 goto out;
 624
 625         dprintk("--> %s clp %p session %p sr_slot %td\n",
 626                 __func__, session->clp, session, res->sr_slot ?
 627                         res->sr_slot - session->fc_slot_table.slots : -1);
 628
 629         ret = nfs41_setup_sequence(session, args, res, task);
 630 out:
 631         dprintk("<-- %s status=%d\n", __func__, ret);
 632         return ret;
 633 }
 634
 635 struct nfs41_call_sync_data {
 636         const struct nfs_server *seq_server;
 637         struct nfs4_sequence_args *seq_args;
 638         struct nfs4_sequence_res *seq_res;
 639 };
 640
 641 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 642 {
 643         struct nfs41_call_sync_data *data = calldata;
 644
 645         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 646
 647         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 648                                 data->seq_res, task))
 649                 return;
 650         rpc_call_start(task);
 651 }
 652
 653 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 654 {
 655         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 656         nfs41_call_sync_prepare(task, calldata);
 657 }
 658
 659 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 660 {
 661         struct nfs41_call_sync_data *data = calldata;
 662
 663         nfs41_sequence_done(task, data->seq_res);
 664 }
 665
 666 struct rpc_call_ops nfs41_call_sync_ops = {
 667         .rpc_call_prepare = nfs41_call_sync_prepare,
 668         .rpc_call_done = nfs41_call_sync_done,
 669 };
 670
 671 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 672         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 673         .rpc_call_done = nfs41_call_sync_done,
 674 };
 675
 676 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 677                                    struct nfs_server *server,
 678                                    struct rpc_message *msg,
 679                                    struct nfs4_sequence_args *args,
 680                                    struct nfs4_sequence_res *res,
 681                                    int privileged)
 682 {
 683         int ret;
 684         struct rpc_task *task;
 685         struct nfs41_call_sync_data data = {
 686                 .seq_server = server,
 687                 .seq_args = args,
 688                 .seq_res = res,
 689         };
 690         struct rpc_task_setup task_setup = {
 691                 .rpc_client = clnt,
 692                 .rpc_message = msg,
 693                 .callback_ops = &nfs41_call_sync_ops,
 694                 .callback_data = &data
 695         };
 696
 697         if (privileged)
 698                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 699         task = rpc_run_task(&task_setup);
 700         if (IS_ERR(task))
 701                 ret = PTR_ERR(task);
 702         else {
 703                 ret = task->tk_status;
 704                 rpc_put_task(task);
 705         }
 706         return ret;
 707 }
 708
 709 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 710                             struct nfs_server *server,
 711                             struct rpc_message *msg,
 712                             struct nfs4_sequence_args *args,
 713                             struct nfs4_sequence_res *res,
 714                             int cache_reply)
 715 {
 716         nfs41_init_sequence(args, res, cache_reply);
 717         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 718 }
 719
 720 #else
 721 static inline
 722 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 723                 struct nfs4_sequence_res *res, int cache_reply)
 724 {
 725 }
 726
 727 static int nfs4_sequence_done(struct rpc_task *task,
 728                                struct nfs4_sequence_res *res)
 729 {
 730         return 1;
 731 }
 732 #endif /* CONFIG_NFS_V4_1 */
 733
 734 int _nfs4_call_sync(struct rpc_clnt *clnt,
 735                     struct nfs_server *server,
 736                     struct rpc_message *msg,
 737                     struct nfs4_sequence_args *args,
 738                     struct nfs4_sequence_res *res,
 739                     int cache_reply)
 740 {
 741         nfs41_init_sequence(args, res, cache_reply);
 742         return rpc_call_sync(clnt, msg, 0);
 743 }
 744
 745 static inline
 746 int nfs4_call_sync(struct rpc_clnt *clnt,
 747                    struct nfs_server *server,
 748                    struct rpc_message *msg,
 749                    struct nfs4_sequence_args *args,
 750                    struct nfs4_sequence_res *res,
 751                    int cache_reply)
 752 {
 753         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 754                                                 args, res, cache_reply);
 755 }
 756
 757 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 758 {
 759         struct nfs_inode *nfsi = NFS_I(dir);
 760
 761         spin_lock(&dir->i_lock);
 762         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 763         if (!cinfo->atomic || cinfo->before != dir->i_version)
 764                 nfs_force_lookup_revalidate(dir);
 765         dir->i_version = cinfo->after;
 766         spin_unlock(&dir->i_lock);
 767 }
 768
 769 struct nfs4_opendata {
 770         struct kref kref;
 771         struct nfs_openargs o_arg;
 772         struct nfs_openres o_res;
 773         struct nfs_open_confirmargs c_arg;
 774         struct nfs_open_confirmres c_res;
 775         struct nfs4_string owner_name;
 776         struct nfs4_string group_name;
 777         struct nfs_fattr f_attr;
 778         struct nfs_fattr dir_attr;
 779         struct dentry *dir;
 780         struct dentry *dentry;
 781         struct nfs4_state_owner *owner;
 782         struct nfs4_state *state;
 783         struct iattr attrs;
 784         unsigned long timestamp;
 785         unsigned int rpc_done : 1;
 786         int rpc_status;
 787         int cancelled;
 788 };
 789
 790
 791 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 792 {
 793         p->o_res.f_attr = &p->f_attr;
 794         p->o_res.dir_attr = &p->dir_attr;
 795         p->o_res.seqid = p->o_arg.seqid;
 796         p->c_res.seqid = p->c_arg.seqid;
 797         p->o_res.server = p->o_arg.server;
 798         nfs_fattr_init(&p->f_attr);
 799         nfs_fattr_init(&p->dir_attr);
 800         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 801 }
 802
 803 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 804                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 805                 const struct iattr *attrs,
 806                 gfp_t gfp_mask)
 807 {
 808         struct dentry *parent = dget_parent(dentry);
 809         struct inode *dir = parent->d_inode;
 810         struct nfs_server *server = NFS_SERVER(dir);
 811         struct nfs4_opendata *p;
 812
 813         p = kzalloc(sizeof(*p), gfp_mask);
 814         if (p == NULL)
 815                 goto err;
 816         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 817         if (p->o_arg.seqid == NULL)
 818                 goto err_free;
 819         nfs_sb_active(dentry->d_sb);
 820         p->dentry = dget(dentry);
 821         p->dir = parent;
 822         p->owner = sp;
 823         atomic_inc(&sp->so_count);
 824         p->o_arg.fh = NFS_FH(dir);
 825         p->o_arg.open_flags = flags;
 826         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 827         p->o_arg.clientid = server->nfs_client->cl_clientid;
 828         p->o_arg.id = sp->so_seqid.owner_id;
 829         p->o_arg.name = &dentry->d_name;
 830         p->o_arg.server = server;
 831         p->o_arg.bitmask = server->attr_bitmask;
 832         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 833         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 834         if (attrs != NULL && attrs->ia_valid != 0) {
 835                 u32 *s;
 836
 837                 p->o_arg.u.attrs = &p->attrs;
 838                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 839                 s = (u32 *) p->o_arg.u.verifier.data;
 840                 s[0] = jiffies;
 841                 s[1] = current->pid;
 842         }
 843         p->c_arg.fh = &p->o_res.fh;
 844         p->c_arg.stateid = &p->o_res.stateid;
 845         p->c_arg.seqid = p->o_arg.seqid;
 846         nfs4_init_opendata_res(p);
 847         kref_init(&p->kref);
 848         return p;
 849 err_free:
 850         kfree(p);
 851 err:
 852         dput(parent);
 853         return NULL;
 854 }
 855
 856 static void nfs4_opendata_free(struct kref *kref)
 857 {
 858         struct nfs4_opendata *p = container_of(kref,
 859                         struct nfs4_opendata, kref);
 860         struct super_block *sb = p->dentry->d_sb;
 861
 862         nfs_free_seqid(p->o_arg.seqid);
 863         if (p->state != NULL)
 864                 nfs4_put_open_state(p->state);
 865         nfs4_put_state_owner(p->owner);
 866         dput(p->dir);
 867         dput(p->dentry);
 868         nfs_sb_deactive(sb);
 869         nfs_fattr_free_names(&p->f_attr);
 870         kfree(p);
 871 }
 872
 873 static void nfs4_opendata_put(struct nfs4_opendata *p)
 874 {
 875         if (p != NULL)
 876                 kref_put(&p->kref, nfs4_opendata_free);
 877 }
 878
 879 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 880 {
 881         int ret;
 882
 883         ret = rpc_wait_for_completion_task(task);
 884         return ret;
 885 }
 886
 887 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 888 {
 889         int ret = 0;
 890
 891         if (open_mode & (O_EXCL|O_TRUNC))
 892                 goto out;
 893         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 894                 case FMODE_READ:
 895                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 896                                 && state->n_rdonly != 0;
 897                         break;
 898                 case FMODE_WRITE:
 899                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 900                                 && state->n_wronly != 0;
 901                         break;
 902                 case FMODE_READ|FMODE_WRITE:
 903                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 904                                 && state->n_rdwr != 0;
 905         }
 906 out:
 907         return ret;
 908 }
 909
 910 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 911 {
 912         if (delegation == NULL)
 913                 return 0;
 914         if ((delegation->type & fmode) != fmode)
 915                 return 0;
 916         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 917                 return 0;
 918         nfs_mark_delegation_referenced(delegation);
 919         return 1;
 920 }
 921
 922 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 923 {
 924         switch (fmode) {
 925                 case FMODE_WRITE:
 926                         state->n_wronly++;
 927                         break;
 928                 case FMODE_READ:
 929                         state->n_rdonly++;
 930                         break;
 931                 case FMODE_READ|FMODE_WRITE:
 932                         state->n_rdwr++;
 933         }
 934         nfs4_state_set_mode_locked(state, state->state | fmode);
 935 }
 936
 937 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 938 {
 939         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 940                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 941         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 942         switch (fmode) {
 943                 case FMODE_READ:
 944                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 945                         break;
 946                 case FMODE_WRITE:
 947                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 948                         break;
 949                 case FMODE_READ|FMODE_WRITE:
 950                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 951         }
 952 }
 953
 954 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 955 {
 956         write_seqlock(&state->seqlock);
 957         nfs_set_open_stateid_locked(state, stateid, fmode);
 958         write_sequnlock(&state->seqlock);
 959 }
 960
 961 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 962 {
 963         /*
 964          * Protect the call to nfs4_state_set_mode_locked and
 965          * serialise the stateid update
 966          */
 967         write_seqlock(&state->seqlock);
 968         if (deleg_stateid != NULL) {
 969                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 970                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 971         }
 972         if (open_stateid != NULL)
 973                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 974         write_sequnlock(&state->seqlock);
 975         spin_lock(&state->owner->so_lock);
 976         update_open_stateflags(state, fmode);
 977         spin_unlock(&state->owner->so_lock);
 978 }
 979
 980 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 981 {
 982         struct nfs_inode *nfsi = NFS_I(state->inode);
 983         struct nfs_delegation *deleg_cur;
 984         int ret = 0;
 985
 986         fmode &= (FMODE_READ|FMODE_WRITE);
 987
 988         rcu_read_lock();
 989         deleg_cur = rcu_dereference(nfsi->delegation);
 990         if (deleg_cur == NULL)
 991                 goto no_delegation;
 992
 993         spin_lock(&deleg_cur->lock);
 994         if (nfsi->delegation != deleg_cur ||
 995             (deleg_cur->type & fmode) != fmode)
 996                 goto no_delegation_unlock;
 997
 998         if (delegation == NULL)
 999                 delegation = &deleg_cur->stateid;
1000         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
1001                 goto no_delegation_unlock;
1002
1003         nfs_mark_delegation_referenced(deleg_cur);
1004         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1005         ret = 1;
1006 no_delegation_unlock:
1007         spin_unlock(&deleg_cur->lock);
1008 no_delegation:
1009         rcu_read_unlock();
1010
1011         if (!ret && open_stateid != NULL) {
1012                 __update_open_stateid(state, open_stateid, NULL, fmode);
1013                 ret = 1;
1014         }
1015
1016         return ret;
1017 }
1018
1019
1020 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1021 {
1022         struct nfs_delegation *delegation;
1023
1024         rcu_read_lock();
1025         delegation = rcu_dereference(NFS_I(inode)->delegation);
1026         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1027                 rcu_read_unlock();
1028                 return;
1029         }
1030         rcu_read_unlock();
1031         nfs_inode_return_delegation(inode);
1032 }
1033
1034 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1035 {
1036         struct nfs4_state *state = opendata->state;
1037         struct nfs_inode *nfsi = NFS_I(state->inode);
1038         struct nfs_delegation *delegation;
1039         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1040         fmode_t fmode = opendata->o_arg.fmode;
1041         nfs4_stateid stateid;
1042         int ret = -EAGAIN;
1043
1044         for (;;) {
1045                 if (can_open_cached(state, fmode, open_mode)) {
1046                         spin_lock(&state->owner->so_lock);
1047                         if (can_open_cached(state, fmode, open_mode)) {
1048                                 update_open_stateflags(state, fmode);
1049                                 spin_unlock(&state->owner->so_lock);
1050                                 goto out_return_state;
1051                         }
1052                         spin_unlock(&state->owner->so_lock);
1053                 }
1054                 rcu_read_lock();
1055                 delegation = rcu_dereference(nfsi->delegation);
1056                 if (!can_open_delegated(delegation, fmode)) {
1057                         rcu_read_unlock();
1058                         break;
1059                 }
1060                 /* Save the delegation */
1061                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1062                 rcu_read_unlock();
1063                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1064                 if (ret != 0)
1065                         goto out;
1066                 ret = -EAGAIN;
1067
1068                 /* Try to update the stateid using the delegation */
1069                 if (update_open_stateid(state, NULL, &stateid, fmode))
1070                         goto out_return_state;
1071         }
1072 out:
1073         return ERR_PTR(ret);
1074 out_return_state:
1075         atomic_inc(&state->count);
1076         return state;
1077 }
1078
1079 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1080 {
1081         struct inode *inode;
1082         struct nfs4_state *state = NULL;
1083         struct nfs_delegation *delegation;
1084         int ret;
1085
1086         if (!data->rpc_done) {
1087                 state = nfs4_try_open_cached(data);
1088                 goto out;
1089         }
1090
1091         ret = -EAGAIN;
1092         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1093                 goto err;
1094         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1095         ret = PTR_ERR(inode);
1096         if (IS_ERR(inode))
1097                 goto err;
1098         ret = -ENOMEM;
1099         state = nfs4_get_open_state(inode, data->owner);
1100         if (state == NULL)
1101                 goto err_put_inode;
1102         if (data->o_res.delegation_type != 0) {
1103                 int delegation_flags = 0;
1104
1105                 rcu_read_lock();
1106                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1107                 if (delegation)
1108                         delegation_flags = delegation->flags;
1109                 rcu_read_unlock();
1110                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1111                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1112                                         "returning a delegation for "
1113                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1114                                         NFS_CLIENT(inode)->cl_server);
1115                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1116                         nfs_inode_set_delegation(state->inode,
1117                                         data->owner->so_cred,
1118                                         &data->o_res);
1119                 else
1120                         nfs_inode_reclaim_delegation(state->inode,
1121                                         data->owner->so_cred,
1122                                         &data->o_res);
1123         }
1124
1125         update_open_stateid(state, &data->o_res.stateid, NULL,
1126                         data->o_arg.fmode);
1127         iput(inode);
1128 out:
1129         return state;
1130 err_put_inode:
1131         iput(inode);
1132 err:
1133         return ERR_PTR(ret);
1134 }
1135
1136 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1137 {
1138         struct nfs_inode *nfsi = NFS_I(state->inode);
1139         struct nfs_open_context *ctx;
1140
1141         spin_lock(&state->inode->i_lock);
1142         list_for_each_entry(ctx, &nfsi->open_files, list) {
1143                 if (ctx->state != state)
1144                         continue;
1145                 get_nfs_open_context(ctx);
1146                 spin_unlock(&state->inode->i_lock);
1147                 return ctx;
1148         }
1149         spin_unlock(&state->inode->i_lock);
1150         return ERR_PTR(-ENOENT);
1151 }
1152
1153 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1154 {
1155         struct nfs4_opendata *opendata;
1156
1157         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1158         if (opendata == NULL)
1159                 return ERR_PTR(-ENOMEM);
1160         opendata->state = state;
1161         atomic_inc(&state->count);
1162         return opendata;
1163 }
1164
1165 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1166 {
1167         struct nfs4_state *newstate;
1168         int ret;
1169
1170         opendata->o_arg.open_flags = 0;
1171         opendata->o_arg.fmode = fmode;
1172         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1173         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1174         nfs4_init_opendata_res(opendata);
1175         ret = _nfs4_recover_proc_open(opendata);
1176         if (ret != 0)
1177                 return ret;
1178         newstate = nfs4_opendata_to_nfs4_state(opendata);
1179         if (IS_ERR(newstate))
1180                 return PTR_ERR(newstate);
1181         nfs4_close_state(newstate, fmode);
1182         *res = newstate;
1183         return 0;
1184 }
1185
1186 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1187 {
1188         struct nfs4_state *newstate;
1189         int ret;
1190
1191         /* memory barrier prior to reading state->n_* */
1192         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1193         smp_rmb();
1194         if (state->n_rdwr != 0) {
1195                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1196                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1197                 if (ret != 0)
1198                         return ret;
1199                 if (newstate != state)
1200                         return -ESTALE;
1201         }
1202         if (state->n_wronly != 0) {
1203                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1204                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1205                 if (ret != 0)
1206                         return ret;
1207                 if (newstate != state)
1208                         return -ESTALE;
1209         }
1210         if (state->n_rdonly != 0) {
1211                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1212                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1213                 if (ret != 0)
1214                         return ret;
1215                 if (newstate != state)
1216                         return -ESTALE;
1217         }
1218         /*
1219          * We may have performed cached opens for all three recoveries.
1220          * Check if we need to update the current stateid.
1221          */
1222         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1223             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1224                 write_seqlock(&state->seqlock);
1225                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1226                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1227                 write_sequnlock(&state->seqlock);
1228         }
1229         return 0;
1230 }
1231
1232 /*
1233  * OPEN_RECLAIM:
1234  *      reclaim state on the server after a reboot.
1235  */
1236 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1237 {
1238         struct nfs_delegation *delegation;
1239         struct nfs4_opendata *opendata;
1240         fmode_t delegation_type = 0;
1241         int status;
1242
1243         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1244         if (IS_ERR(opendata))
1245                 return PTR_ERR(opendata);
1246         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1247         opendata->o_arg.fh = NFS_FH(state->inode);
1248         rcu_read_lock();
1249         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1250         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1251                 delegation_type = delegation->type;
1252         rcu_read_unlock();
1253         opendata->o_arg.u.delegation_type = delegation_type;
1254         status = nfs4_open_recover(opendata, state);
1255         nfs4_opendata_put(opendata);
1256         return status;
1257 }
1258
1259 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1260 {
1261         struct nfs_server *server = NFS_SERVER(state->inode);
1262         struct nfs4_exception exception = { };
1263         int err;
1264         do {
1265                 err = _nfs4_do_open_reclaim(ctx, state);
1266                 if (err != -NFS4ERR_DELAY)
1267                         break;
1268                 nfs4_handle_exception(server, err, &exception);
1269         } while (exception.retry);
1270         return err;
1271 }
1272
1273 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1274 {
1275         struct nfs_open_context *ctx;
1276         int ret;
1277
1278         ctx = nfs4_state_find_open_context(state);
1279         if (IS_ERR(ctx))
1280                 return PTR_ERR(ctx);
1281         ret = nfs4_do_open_reclaim(ctx, state);
1282         put_nfs_open_context(ctx);
1283         return ret;
1284 }
1285
1286 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1287 {
1288         struct nfs4_opendata *opendata;
1289         int ret;
1290
1291         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1292         if (IS_ERR(opendata))
1293                 return PTR_ERR(opendata);
1294         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1295         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1296                         sizeof(opendata->o_arg.u.delegation.data));
1297         ret = nfs4_open_recover(opendata, state);
1298         nfs4_opendata_put(opendata);
1299         return ret;
1300 }
1301
1302 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1303 {
1304         struct nfs4_exception exception = { };
1305         struct nfs_server *server = NFS_SERVER(state->inode);
1306         int err;
1307         do {
1308                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1309                 switch (err) {
1310                         case 0:
1311                         case -ENOENT:
1312                         case -ESTALE:
1313                                 goto out;
1314                         case -NFS4ERR_BADSESSION:
1315                         case -NFS4ERR_BADSLOT:
1316                         case -NFS4ERR_BAD_HIGH_SLOT:
1317                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1318                         case -NFS4ERR_DEADSESSION:
1319                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1320                                 goto out;
1321                         case -NFS4ERR_STALE_CLIENTID:
1322                         case -NFS4ERR_STALE_STATEID:
1323                         case -NFS4ERR_EXPIRED:
1324                                 /* Don't recall a delegation if it was lost */
1325                                 nfs4_schedule_lease_recovery(server->nfs_client);
1326                                 goto out;
1327                         case -ERESTARTSYS:
1328                                 /*
1329                                  * The show must go on: exit, but mark the
1330                                  * stateid as needing recovery.
1331                                  */
1332                         case -NFS4ERR_ADMIN_REVOKED:
1333                         case -NFS4ERR_BAD_STATEID:
1334                                 nfs4_schedule_stateid_recovery(server, state);
1335                         case -EKEYEXPIRED:
1336                                 /*
1337                                  * User RPCSEC_GSS context has expired.
1338                                  * We cannot recover this stateid now, so
1339                                  * skip it and allow recovery thread to
1340                                  * proceed.
1341                                  */
1342                         case -ENOMEM:
1343                                 err = 0;
1344                                 goto out;
1345                 }
1346                 err = nfs4_handle_exception(server, err, &exception);
1347         } while (exception.retry);
1348 out:
1349         return err;
1350 }
1351
1352 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1353 {
1354         struct nfs4_opendata *data = calldata;
1355
1356         data->rpc_status = task->tk_status;
1357         if (data->rpc_status == 0) {
1358                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1359                                 sizeof(data->o_res.stateid.data));
1360                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1361                 renew_lease(data->o_res.server, data->timestamp);
1362                 data->rpc_done = 1;
1363         }
1364 }
1365
1366 static void nfs4_open_confirm_release(void *calldata)
1367 {
1368         struct nfs4_opendata *data = calldata;
1369         struct nfs4_state *state = NULL;
1370
1371         /* If this request hasn't been cancelled, do nothing */
1372         if (data->cancelled == 0)
1373                 goto out_free;
1374         /* In case of error, no cleanup! */
1375         if (!data->rpc_done)
1376                 goto out_free;
1377         state = nfs4_opendata_to_nfs4_state(data);
1378         if (!IS_ERR(state))
1379                 nfs4_close_state(state, data->o_arg.fmode);
1380 out_free:
1381         nfs4_opendata_put(data);
1382 }
1383
1384 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1385         .rpc_call_done = nfs4_open_confirm_done,
1386         .rpc_release = nfs4_open_confirm_release,
1387 };
1388
1389 /*
1390  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1391  */
1392 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1393 {
1394         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1395         struct rpc_task *task;
1396         struct  rpc_message msg = {
1397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1398                 .rpc_argp = &data->c_arg,
1399                 .rpc_resp = &data->c_res,
1400                 .rpc_cred = data->owner->so_cred,
1401         };
1402         struct rpc_task_setup task_setup_data = {
1403                 .rpc_client = server->client,
1404                 .rpc_message = &msg,
1405                 .callback_ops = &nfs4_open_confirm_ops,
1406                 .callback_data = data,
1407                 .workqueue = nfsiod_workqueue,
1408                 .flags = RPC_TASK_ASYNC,
1409         };
1410         int status;
1411
1412         kref_get(&data->kref);
1413         data->rpc_done = 0;
1414         data->rpc_status = 0;
1415         data->timestamp = jiffies;
1416         task = rpc_run_task(&task_setup_data);
1417         if (IS_ERR(task))
1418                 return PTR_ERR(task);
1419         status = nfs4_wait_for_completion_rpc_task(task);
1420         if (status != 0) {
1421                 data->cancelled = 1;
1422                 smp_wmb();
1423         } else
1424                 status = data->rpc_status;
1425         rpc_put_task(task);
1426         return status;
1427 }
1428
1429 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1430 {
1431         struct nfs4_opendata *data = calldata;
1432         struct nfs4_state_owner *sp = data->owner;
1433
1434         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1435                 return;
1436         /*
1437          * Check if we still need to send an OPEN call, or if we can use
1438          * a delegation instead.
1439          */
1440         if (data->state != NULL) {
1441                 struct nfs_delegation *delegation;
1442
1443                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1444                         goto out_no_action;
1445                 rcu_read_lock();
1446                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1447                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1448                     can_open_delegated(delegation, data->o_arg.fmode))
1449                         goto unlock_no_action;
1450                 rcu_read_unlock();
1451         }
1452         /* Update sequence id. */
1453         data->o_arg.id = sp->so_seqid.owner_id;
1454         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1455         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1456                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1457                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1458         }
1459         data->timestamp = jiffies;
1460         if (nfs4_setup_sequence(data->o_arg.server,
1461                                 &data->o_arg.seq_args,
1462                                 &data->o_res.seq_res, task))
1463                 return;
1464         rpc_call_start(task);
1465         return;
1466 unlock_no_action:
1467         rcu_read_unlock();
1468 out_no_action:
1469         task->tk_action = NULL;
1470
1471 }
1472
1473 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1474 {
1475         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1476         nfs4_open_prepare(task, calldata);
1477 }
1478
1479 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1480 {
1481         struct nfs4_opendata *data = calldata;
1482
1483         data->rpc_status = task->tk_status;
1484
1485         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1486                 return;
1487
1488         if (task->tk_status == 0) {
1489                 switch (data->o_res.f_attr->mode & S_IFMT) {
1490                         case S_IFREG:
1491                                 break;
1492                         case S_IFLNK:
1493                                 data->rpc_status = -ELOOP;
1494                                 break;
1495                         case S_IFDIR:
1496                                 data->rpc_status = -EISDIR;
1497                                 break;
1498                         default:
1499                                 data->rpc_status = -ENOTDIR;
1500                 }
1501                 renew_lease(data->o_res.server, data->timestamp);
1502                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1503                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1504         }
1505         data->rpc_done = 1;
1506 }
1507
1508 static void nfs4_open_release(void *calldata)
1509 {
1510         struct nfs4_opendata *data = calldata;
1511         struct nfs4_state *state = NULL;
1512
1513         /* If this request hasn't been cancelled, do nothing */
1514         if (data->cancelled == 0)
1515                 goto out_free;
1516         /* In case of error, no cleanup! */
1517         if (data->rpc_status != 0 || !data->rpc_done)
1518                 goto out_free;
1519         /* In case we need an open_confirm, no cleanup! */
1520         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1521                 goto out_free;
1522         state = nfs4_opendata_to_nfs4_state(data);
1523         if (!IS_ERR(state))
1524                 nfs4_close_state(state, data->o_arg.fmode);
1525 out_free:
1526         nfs4_opendata_put(data);
1527 }
1528
1529 static const struct rpc_call_ops nfs4_open_ops = {
1530         .rpc_call_prepare = nfs4_open_prepare,
1531         .rpc_call_done = nfs4_open_done,
1532         .rpc_release = nfs4_open_release,
1533 };
1534
1535 static const struct rpc_call_ops nfs4_recover_open_ops = {
1536         .rpc_call_prepare = nfs4_recover_open_prepare,
1537         .rpc_call_done = nfs4_open_done,
1538         .rpc_release = nfs4_open_release,
1539 };
1540
1541 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1542 {
1543         struct inode *dir = data->dir->d_inode;
1544         struct nfs_server *server = NFS_SERVER(dir);
1545         struct nfs_openargs *o_arg = &data->o_arg;
1546         struct nfs_openres *o_res = &data->o_res;
1547         struct rpc_task *task;
1548         struct rpc_message msg = {
1549                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1550                 .rpc_argp = o_arg,
1551                 .rpc_resp = o_res,
1552                 .rpc_cred = data->owner->so_cred,
1553         };
1554         struct rpc_task_setup task_setup_data = {
1555                 .rpc_client = server->client,
1556                 .rpc_message = &msg,
1557                 .callback_ops = &nfs4_open_ops,
1558                 .callback_data = data,
1559                 .workqueue = nfsiod_workqueue,
1560                 .flags = RPC_TASK_ASYNC,
1561         };
1562         int status;
1563
1564         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1565         kref_get(&data->kref);
1566         data->rpc_done = 0;
1567         data->rpc_status = 0;
1568         data->cancelled = 0;
1569         if (isrecover)
1570                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1571         task = rpc_run_task(&task_setup_data);
1572         if (IS_ERR(task))
1573                 return PTR_ERR(task);
1574         status = nfs4_wait_for_completion_rpc_task(task);
1575         if (status != 0) {
1576                 data->cancelled = 1;
1577                 smp_wmb();
1578         } else
1579                 status = data->rpc_status;
1580         rpc_put_task(task);
1581
1582         return status;
1583 }
1584
1585 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1586 {
1587         struct inode *dir = data->dir->d_inode;
1588         struct nfs_openres *o_res = &data->o_res;
1589         int status;
1590
1591         status = nfs4_run_open_task(data, 1);
1592         if (status != 0 || !data->rpc_done)
1593                 return status;
1594
1595         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1596
1597         nfs_refresh_inode(dir, o_res->dir_attr);
1598
1599         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1600                 status = _nfs4_proc_open_confirm(data);
1601                 if (status != 0)
1602                         return status;
1603         }
1604
1605         return status;
1606 }
1607
1608 /*
1609  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1610  */
1611 static int _nfs4_proc_open(struct nfs4_opendata *data)
1612 {
1613         struct inode *dir = data->dir->d_inode;
1614         struct nfs_server *server = NFS_SERVER(dir);
1615         struct nfs_openargs *o_arg = &data->o_arg;
1616         struct nfs_openres *o_res = &data->o_res;
1617         int status;
1618
1619         status = nfs4_run_open_task(data, 0);
1620         if (!data->rpc_done)
1621                 return status;
1622         if (status != 0) {
1623                 if (status == -NFS4ERR_BADNAME &&
1624                                 !(o_arg->open_flags & O_CREAT))
1625                         return -ENOENT;
1626                 return status;
1627         }
1628
1629         nfs_fattr_map_and_free_names(server, &data->f_attr);
1630
1631         if (o_arg->open_flags & O_CREAT) {
1632                 update_changeattr(dir, &o_res->cinfo);
1633                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1634         } else
1635                 nfs_refresh_inode(dir, o_res->dir_attr);
1636         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1637                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1638         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1639                 status = _nfs4_proc_open_confirm(data);
1640                 if (status != 0)
1641                         return status;
1642         }
1643         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1644                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1645         return 0;
1646 }
1647
1648 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1649 {
1650         unsigned int loop;
1651         int ret;
1652
1653         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1654                 ret = nfs4_wait_clnt_recover(clp);
1655                 if (ret != 0)
1656                         break;
1657                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1658                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1659                         break;
1660                 nfs4_schedule_state_manager(clp);
1661                 ret = -EIO;
1662         }
1663         return ret;
1664 }
1665
1666 static int nfs4_recover_expired_lease(struct nfs_server *server)
1667 {
1668         return nfs4_client_recover_expired_lease(server->nfs_client);
1669 }
1670
1671 /*
1672  * OPEN_EXPIRED:
1673  *      reclaim state on the server after a network partition.
1674  *      Assumes caller holds the appropriate lock
1675  */
1676 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1677 {
1678         struct nfs4_opendata *opendata;
1679         int ret;
1680
1681         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1682         if (IS_ERR(opendata))
1683                 return PTR_ERR(opendata);
1684         ret = nfs4_open_recover(opendata, state);
1685         if (ret == -ESTALE)
1686                 d_drop(ctx->dentry);
1687         nfs4_opendata_put(opendata);
1688         return ret;
1689 }
1690
1691 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1692 {
1693         struct nfs_server *server = NFS_SERVER(state->inode);
1694         struct nfs4_exception exception = { };
1695         int err;
1696
1697         do {
1698                 err = _nfs4_open_expired(ctx, state);
1699                 switch (err) {
1700                 default:
1701                         goto out;
1702                 case -NFS4ERR_GRACE:
1703                 case -NFS4ERR_DELAY:
1704                         nfs4_handle_exception(server, err, &exception);
1705                         err = 0;
1706                 }
1707         } while (exception.retry);
1708 out:
1709         return err;
1710 }
1711
1712 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1713 {
1714         struct nfs_open_context *ctx;
1715         int ret;
1716
1717         ctx = nfs4_state_find_open_context(state);
1718         if (IS_ERR(ctx))
1719                 return PTR_ERR(ctx);
1720         ret = nfs4_do_open_expired(ctx, state);
1721         put_nfs_open_context(ctx);
1722         return ret;
1723 }
1724
1725 #if defined(CONFIG_NFS_V4_1)
1726 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1727 {
1728         int status = NFS_OK;
1729         struct nfs_server *server = NFS_SERVER(state->inode);
1730
1731         if (state->flags & flags) {
1732                 status = nfs41_test_stateid(server, stateid);
1733                 if (status != NFS_OK) {
1734                         nfs41_free_stateid(server, stateid);
1735                         state->flags &= ~flags;
1736                 }
1737         }
1738         return status;
1739 }
1740
1741 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1742 {
1743         int deleg_status, open_status;
1744         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1745         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1746
1747         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1748         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1749
1750         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1751                 return NFS_OK;
1752         return nfs4_open_expired(sp, state);
1753 }
1754 #endif
1755
1756 /*
1757  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1758  * fields corresponding to attributes that were used to store the verifier.
1759  * Make sure we clobber those fields in the later setattr call
1760  */
1761 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1762 {
1763         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1764             !(sattr->ia_valid & ATTR_ATIME_SET))
1765                 sattr->ia_valid |= ATTR_ATIME;
1766
1767         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1768             !(sattr->ia_valid & ATTR_MTIME_SET))
1769                 sattr->ia_valid |= ATTR_MTIME;
1770 }
1771
1772 /*
1773  * Returns a referenced nfs4_state
1774  */
1775 static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1776 {
1777         struct nfs4_state_owner  *sp;
1778         struct nfs4_state     *state = NULL;
1779         struct nfs_server       *server = NFS_SERVER(dir);
1780         struct nfs4_opendata *opendata;
1781         int status;
1782
1783         /* Protect against reboot recovery conflicts */
1784         status = -ENOMEM;
1785         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1786         if (sp == NULL) {
1787                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1788                 goto out_err;
1789         }
1790         status = nfs4_recover_expired_lease(server);
1791         if (status != 0)
1792                 goto err_put_state_owner;
1793         if (dentry->d_inode != NULL)
1794                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1795         status = -ENOMEM;
1796         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1797         if (opendata == NULL)
1798                 goto err_put_state_owner;
1799
1800         if (dentry->d_inode != NULL)
1801                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1802
1803         status = _nfs4_proc_open(opendata);
1804         if (status != 0)
1805                 goto err_opendata_put;
1806
1807         state = nfs4_opendata_to_nfs4_state(opendata);
1808         status = PTR_ERR(state);
1809         if (IS_ERR(state))
1810                 goto err_opendata_put;
1811         if (server->caps & NFS_CAP_POSIX_LOCK)
1812                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1813
1814         if (opendata->o_arg.open_flags & O_EXCL) {
1815                 nfs4_exclusive_attrset(opendata, sattr);
1816
1817                 nfs_fattr_init(opendata->o_res.f_attr);
1818                 status = nfs4_do_setattr(state->inode, cred,
1819                                 opendata->o_res.f_attr, sattr,
1820                                 state);
1821                 if (status == 0)
1822                         nfs_setattr_update_inode(state->inode, sattr);
1823                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1824         }
1825         nfs4_opendata_put(opendata);
1826         nfs4_put_state_owner(sp);
1827         *res = state;
1828         return 0;
1829 err_opendata_put:
1830         nfs4_opendata_put(opendata);
1831 err_put_state_owner:
1832         nfs4_put_state_owner(sp);
1833 out_err:
1834         *res = NULL;
1835         return status;
1836 }
1837
1838
1839 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1840 {
1841         struct nfs4_exception exception = { };
1842         struct nfs4_state *res;
1843         int status;
1844
1845         do {
1846                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1847                 if (status == 0)
1848                         break;
1849                 /* NOTE: BAD_SEQID means the server and client disagree about the
1850                  * book-keeping w.r.t. state-changing operations
1851                  * (OPEN/CLOSE/LOCK/LOCKU...)
1852                  * It is actually a sign of a bug on the client or on the server.
1853                  *
1854                  * If we receive a BAD_SEQID error in the particular case of
1855                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1856                  * have unhashed the old state_owner for us, and that we can
1857                  * therefore safely retry using a new one. We should still warn
1858                  * the user though...
1859                  */
1860                 if (status == -NFS4ERR_BAD_SEQID) {
1861                         printk(KERN_WARNING "NFS: v4 server %s "
1862                                         " returned a bad sequence-id error!\n",
1863                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1864                         exception.retry = 1;
1865                         continue;
1866                 }
1867                 /*
1868                  * BAD_STATEID on OPEN means that the server cancelled our
1869                  * state before it received the OPEN_CONFIRM.
1870                  * Recover by retrying the request as per the discussion
1871                  * on Page 181 of RFC3530.
1872                  */
1873                 if (status == -NFS4ERR_BAD_STATEID) {
1874                         exception.retry = 1;
1875                         continue;
1876                 }
1877                 if (status == -EAGAIN) {
1878                         /* We must have found a delegation */
1879                         exception.retry = 1;
1880                         continue;
1881                 }
1882                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1883                                         status, &exception));
1884         } while (exception.retry);
1885         return res;
1886 }
1887
1888 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1889                             struct nfs_fattr *fattr, struct iattr *sattr,
1890                             struct nfs4_state *state)
1891 {
1892         struct nfs_server *server = NFS_SERVER(inode);
1893         struct nfs_setattrargs  arg = {
1894                 .fh             = NFS_FH(inode),
1895                 .iap            = sattr,
1896                 .server         = server,
1897                 .bitmask = server->attr_bitmask,
1898         };
1899         struct nfs_setattrres  res = {
1900                 .fattr          = fattr,
1901                 .server         = server,
1902         };
1903         struct rpc_message msg = {
1904                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1905                 .rpc_argp       = &arg,
1906                 .rpc_resp       = &res,
1907                 .rpc_cred       = cred,
1908         };
1909         unsigned long timestamp = jiffies;
1910         int status;
1911
1912         nfs_fattr_init(fattr);
1913
1914         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1915                 /* Use that stateid */
1916         } else if (state != NULL) {
1917                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1918         } else
1919                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1920
1921         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1922         if (status == 0 && state != NULL)
1923                 renew_lease(server, timestamp);
1924         return status;
1925 }
1926
1927 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1928                            struct nfs_fattr *fattr, struct iattr *sattr,
1929                            struct nfs4_state *state)
1930 {
1931         struct nfs_server *server = NFS_SERVER(inode);
1932         struct nfs4_exception exception = { };
1933         int err;
1934         do {
1935                 err = nfs4_handle_exception(server,
1936                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1937                                 &exception);
1938         } while (exception.retry);
1939         return err;
1940 }
1941
1942 struct nfs4_closedata {
1943         struct inode *inode;
1944         struct nfs4_state *state;
1945         struct nfs_closeargs arg;
1946         struct nfs_closeres res;
1947         struct nfs_fattr fattr;
1948         unsigned long timestamp;
1949         bool roc;
1950         u32 roc_barrier;
1951 };
1952
1953 static void nfs4_free_closedata(void *data)
1954 {
1955         struct nfs4_closedata *calldata = data;
1956         struct nfs4_state_owner *sp = calldata->state->owner;
1957         struct super_block *sb = calldata->state->inode->i_sb;
1958
1959         if (calldata->roc)
1960                 pnfs_roc_release(calldata->state->inode);
1961         nfs4_put_open_state(calldata->state);
1962         nfs_free_seqid(calldata->arg.seqid);
1963         nfs4_put_state_owner(sp);
1964         nfs_sb_deactive(sb);
1965         kfree(calldata);
1966 }
1967
1968 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1969                 fmode_t fmode)
1970 {
1971         spin_lock(&state->owner->so_lock);
1972         if (!(fmode & FMODE_READ))
1973                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1974         if (!(fmode & FMODE_WRITE))
1975                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1976         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1977         spin_unlock(&state->owner->so_lock);
1978 }
1979
1980 static void nfs4_close_done(struct rpc_task *task, void *data)
1981 {
1982         struct nfs4_closedata *calldata = data;
1983         struct nfs4_state *state = calldata->state;
1984         struct nfs_server *server = NFS_SERVER(calldata->inode);
1985
1986         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1987                 return;
1988         /* hmm. we are done with the inode, and in the process of freeing
1989          * the state_owner. we keep this around to process errors
1990          */
1991         switch (task->tk_status) {
1992                 case 0:
1993                         if (calldata->roc)
1994                                 pnfs_roc_set_barrier(state->inode,
1995                                                      calldata->roc_barrier);
1996                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1997                         renew_lease(server, calldata->timestamp);
1998                         nfs4_close_clear_stateid_flags(state,
1999                                         calldata->arg.fmode);
2000                         break;
2001                 case -NFS4ERR_STALE_STATEID:
2002                 case -NFS4ERR_OLD_STATEID:
2003                 case -NFS4ERR_BAD_STATEID:
2004                 case -NFS4ERR_EXPIRED:
2005                         if (calldata->arg.fmode == 0)
2006                                 break;
2007                 default:
2008                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2009                                 rpc_restart_call_prepare(task);
2010         }
2011         nfs_release_seqid(calldata->arg.seqid);
2012         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2013 }
2014
2015 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2016 {
2017         struct nfs4_closedata *calldata = data;
2018         struct nfs4_state *state = calldata->state;
2019         int call_close = 0;
2020
2021         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2022                 return;
2023
2024         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2025         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2026         spin_lock(&state->owner->so_lock);
2027         /* Calculate the change in open mode */
2028         if (state->n_rdwr == 0) {
2029                 if (state->n_rdonly == 0) {
2030                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2031                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2032                         calldata->arg.fmode &= ~FMODE_READ;
2033                 }
2034                 if (state->n_wronly == 0) {
2035                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2036                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2037                         calldata->arg.fmode &= ~FMODE_WRITE;
2038                 }
2039         }
2040         spin_unlock(&state->owner->so_lock);
2041
2042         if (!call_close) {
2043                 /* Note: exit _without_ calling nfs4_close_done */
2044                 task->tk_action = NULL;
2045                 return;
2046         }
2047
2048         if (calldata->arg.fmode == 0) {
2049                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2050                 if (calldata->roc &&
2051                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2052                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2053                                      task, NULL);
2054                         return;
2055                 }
2056         }
2057
2058         nfs_fattr_init(calldata->res.fattr);
2059         calldata->timestamp = jiffies;
2060         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2061                                 &calldata->arg.seq_args,
2062                                 &calldata->res.seq_res,
2063                                 task))
2064                 return;
2065         rpc_call_start(task);
2066 }
2067
2068 static const struct rpc_call_ops nfs4_close_ops = {
2069         .rpc_call_prepare = nfs4_close_prepare,
2070         .rpc_call_done = nfs4_close_done,
2071         .rpc_release = nfs4_free_closedata,
2072 };
2073
2074 /*
2075  * It is possible for data to be read/written from a mem-mapped file
2076  * after the sys_close call (which hits the vfs layer as a flush).
2077  * This means that we can't safely call nfsv4 close on a file until
2078  * the inode is cleared. This in turn means that we are not good
2079  * NFSv4 citizens - we do not indicate to the server to update the file's
2080  * share state even when we are done with one of the three share
2081  * stateid's in the inode.
2082  *
2083  * NOTE: Caller must be holding the sp->so_owner semaphore!
2084  */
2085 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2086 {
2087         struct nfs_server *server = NFS_SERVER(state->inode);
2088         struct nfs4_closedata *calldata;
2089         struct nfs4_state_owner *sp = state->owner;
2090         struct rpc_task *task;
2091         struct rpc_message msg = {
2092                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2093                 .rpc_cred = state->owner->so_cred,
2094         };
2095         struct rpc_task_setup task_setup_data = {
2096                 .rpc_client = server->client,
2097                 .rpc_message = &msg,
2098                 .callback_ops = &nfs4_close_ops,
2099                 .workqueue = nfsiod_workqueue,
2100                 .flags = RPC_TASK_ASYNC,
2101         };
2102         int status = -ENOMEM;
2103
2104         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2105         if (calldata == NULL)
2106                 goto out;
2107         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2108         calldata->inode = state->inode;
2109         calldata->state = state;
2110         calldata->arg.fh = NFS_FH(state->inode);
2111         calldata->arg.stateid = &state->open_stateid;
2112         /* Serialization for the sequence id */
2113         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2114         if (calldata->arg.seqid == NULL)
2115                 goto out_free_calldata;
2116         calldata->arg.fmode = 0;
2117         calldata->arg.bitmask = server->cache_consistency_bitmask;
2118         calldata->res.fattr = &calldata->fattr;
2119         calldata->res.seqid = calldata->arg.seqid;
2120         calldata->res.server = server;
2121         calldata->roc = roc;
2122         nfs_sb_active(calldata->inode->i_sb);
2123
2124         msg.rpc_argp = &calldata->arg;
2125         msg.rpc_resp = &calldata->res;
2126         task_setup_data.callback_data = calldata;
2127         task = rpc_run_task(&task_setup_data);
2128         if (IS_ERR(task))
2129                 return PTR_ERR(task);
2130         status = 0;
2131         if (wait)
2132                 status = rpc_wait_for_completion_task(task);
2133         rpc_put_task(task);
2134         return status;
2135 out_free_calldata:
2136         kfree(calldata);
2137 out:
2138         if (roc)
2139                 pnfs_roc_release(state->inode);
2140         nfs4_put_open_state(state);
2141         nfs4_put_state_owner(sp);
2142         return status;
2143 }
2144
2145 static struct inode *
2146 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2147 {
2148         struct nfs4_state *state;
2149
2150         /* Protect against concurrent sillydeletes */
2151         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2152         if (IS_ERR(state))
2153                 return ERR_CAST(state);
2154         ctx->state = state;
2155         return igrab(state->inode);
2156 }
2157
2158 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2159 {
2160         if (ctx->state == NULL)
2161                 return;
2162         if (is_sync)
2163                 nfs4_close_sync(ctx->state, ctx->mode);
2164         else
2165                 nfs4_close_state(ctx->state, ctx->mode);
2166 }
2167
2168 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2169 {
2170         struct nfs4_server_caps_arg args = {
2171                 .fhandle = fhandle,
2172         };
2173         struct nfs4_server_caps_res res = {};
2174         struct rpc_message msg = {
2175                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2176                 .rpc_argp = &args,
2177                 .rpc_resp = &res,
2178         };
2179         int status;
2180
2181         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2182         if (status == 0) {
2183                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2184                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2185                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2186                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2187                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2188                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2189                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2190                         server->caps |= NFS_CAP_ACLS;
2191                 if (res.has_links != 0)
2192                         server->caps |= NFS_CAP_HARDLINKS;
2193                 if (res.has_symlinks != 0)
2194                         server->caps |= NFS_CAP_SYMLINKS;
2195                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2196                         server->caps |= NFS_CAP_FILEID;
2197                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2198                         server->caps |= NFS_CAP_MODE;
2199                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2200                         server->caps |= NFS_CAP_NLINK;
2201                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2202                         server->caps |= NFS_CAP_OWNER;
2203                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2204                         server->caps |= NFS_CAP_OWNER_GROUP;
2205                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2206                         server->caps |= NFS_CAP_ATIME;
2207                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2208                         server->caps |= NFS_CAP_CTIME;
2209                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2210                         server->caps |= NFS_CAP_MTIME;
2211
2212                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2213                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2214                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2215                 server->acl_bitmask = res.acl_bitmask;
2216         }
2217
2218         return status;
2219 }
2220
2221 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2222 {
2223         struct nfs4_exception exception = { };
2224         int err;
2225         do {
2226                 err = nfs4_handle_exception(server,
2227                                 _nfs4_server_capabilities(server, fhandle),
2228                                 &exception);
2229         } while (exception.retry);
2230         return err;
2231 }
2232
2233 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2234                 struct nfs_fsinfo *info)
2235 {
2236         struct nfs4_lookup_root_arg args = {
2237                 .bitmask = nfs4_fattr_bitmap,
2238         };
2239         struct nfs4_lookup_res res = {
2240                 .server = server,
2241                 .fattr = info->fattr,
2242                 .fh = fhandle,
2243         };
2244         struct rpc_message msg = {
2245                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2246                 .rpc_argp = &args,
2247                 .rpc_resp = &res,
2248         };
2249
2250         nfs_fattr_init(info->fattr);
2251         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2252 }
2253
2254 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2255                 struct nfs_fsinfo *info)
2256 {
2257         struct nfs4_exception exception = { };
2258         int err;
2259         do {
2260                 err = _nfs4_lookup_root(server, fhandle, info);
2261                 switch (err) {
2262                 case 0:
2263                 case -NFS4ERR_WRONGSEC:
2264                         break;
2265                 default:
2266                         err = nfs4_handle_exception(server, err, &exception);
2267                 }
2268         } while (exception.retry);
2269         return err;
2270 }
2271
2272 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2273                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2274 {
2275         struct rpc_auth *auth;
2276         int ret;
2277
2278         auth = rpcauth_create(flavor, server->client);
2279         if (!auth) {
2280                 ret = -EIO;
2281                 goto out;
2282         }
2283         ret = nfs4_lookup_root(server, fhandle, info);
2284 out:
2285         return ret;
2286 }
2287
2288 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2289                               struct nfs_fsinfo *info)
2290 {
2291         int i, len, status = 0;
2292         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2293
2294         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2295         flav_array[len] = RPC_AUTH_NULL;
2296         len += 1;
2297
2298         for (i = 0; i < len; i++) {
2299                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2300                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2301                         continue;
2302                 break;
2303         }
2304         /*
2305          * -EACCESS could mean that the user doesn't have correct permissions
2306          * to access the mount.  It could also mean that we tried to mount
2307          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2308          * existing mount programs don't handle -EACCES very well so it should
2309          * be mapped to -EPERM instead.
2310          */
2311         if (status == -EACCES)
2312                 status = -EPERM;
2313         return status;
2314 }
2315
2316 /*
2317  * get the file handle for the "/" directory on the server
2318  */
2319 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2320                               struct nfs_fsinfo *info)
2321 {
2322         int minor_version = server->nfs_client->cl_minorversion;
2323         int status = nfs4_lookup_root(server, fhandle, info);
2324         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2325                 /*
2326                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2327                  * by nfs4_map_errors() as this function exits.
2328                  */
2329                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2330         if (status == 0)
2331                 status = nfs4_server_capabilities(server, fhandle);
2332         if (status == 0)
2333                 status = nfs4_do_fsinfo(server, fhandle, info);
2334         return nfs4_map_errors(status);
2335 }
2336
2337 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
2338 /*
2339  * Get locations and (maybe) other attributes of a referral.
2340  * Note that we'll actually follow the referral later when
2341  * we detect fsid mismatch in inode revalidation
2342  */
2343 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2344                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2345 {
2346         int status = -ENOMEM;
2347         struct page *page = NULL;
2348         struct nfs4_fs_locations *locations = NULL;
2349
2350         page = alloc_page(GFP_KERNEL);
2351         if (page == NULL)
2352                 goto out;
2353         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2354         if (locations == NULL)
2355                 goto out;
2356
2357         status = nfs4_proc_fs_locations(dir, name, locations, page);
2358         if (status != 0)
2359                 goto out;
2360         /* Make sure server returned a different fsid for the referral */
2361         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2362                 dprintk("%s: server did not return a different fsid for"
2363                         " a referral at %s\n", __func__, name->name);
2364                 status = -EIO;
2365                 goto out;
2366         }
2367         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2368         nfs_fixup_referral_attributes(&locations->fattr);
2369
2370         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2371         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2372         memset(fhandle, 0, sizeof(struct nfs_fh));
2373 out:
2374         if (page)
2375                 __free_page(page);
2376         kfree(locations);
2377         return status;
2378 }
2379
2380 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2381 {
2382         struct nfs4_getattr_arg args = {
2383                 .fh = fhandle,
2384                 .bitmask = server->attr_bitmask,
2385         };
2386         struct nfs4_getattr_res res = {
2387                 .fattr = fattr,
2388                 .server = server,
2389         };
2390         struct rpc_message msg = {
2391                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2392                 .rpc_argp = &args,
2393                 .rpc_resp = &res,
2394         };
2395
2396         nfs_fattr_init(fattr);
2397         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2398 }
2399
2400 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2401 {
2402         struct nfs4_exception exception = { };
2403         int err;
2404         do {
2405                 err = nfs4_handle_exception(server,
2406                                 _nfs4_proc_getattr(server, fhandle, fattr),
2407                                 &exception);
2408         } while (exception.retry);
2409         return err;
2410 }
2411
2412 /*
2413  * The file is not closed if it is opened due to the a request to change
2414  * the size of the file. The open call will not be needed once the
2415  * VFS layer lookup-intents are implemented.
2416  *
2417  * Close is called when the inode is destroyed.
2418  * If we haven't opened the file for O_WRONLY, we
2419  * need to in the size_change case to obtain a stateid.
2420  *
2421  * Got race?
2422  * Because OPEN is always done by name in nfsv4, it is
2423  * possible that we opened a different file by the same
2424  * name.  We can recognize this race condition, but we
2425  * can't do anything about it besides returning an error.
2426  *
2427  * This will be fixed with VFS changes (lookup-intent).
2428  */
2429 static int
2430 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2431                   struct iattr *sattr)
2432 {
2433         struct inode *inode = dentry->d_inode;
2434         struct rpc_cred *cred = NULL;
2435         struct nfs4_state *state = NULL;
2436         int status;
2437
2438         if (pnfs_ld_layoutret_on_setattr(inode))
2439                 pnfs_return_layout(inode);
2440
2441         nfs_fattr_init(fattr);
2442
2443         /* Search for an existing open(O_WRITE) file */
2444         if (sattr->ia_valid & ATTR_FILE) {
2445                 struct nfs_open_context *ctx;
2446
2447                 ctx = nfs_file_open_context(sattr->ia_file);
2448                 if (ctx) {
2449                         cred = ctx->cred;
2450                         state = ctx->state;
2451                 }
2452         }
2453
2454         /* Deal with open(O_TRUNC) */
2455         if (sattr->ia_valid & ATTR_OPEN)
2456                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2457
2458         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2459         if (status == 0)
2460                 nfs_setattr_update_inode(inode, sattr);
2461         return status;
2462 }
2463
2464 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2465                 const struct qstr *name, struct nfs_fh *fhandle,
2466                 struct nfs_fattr *fattr)
2467 {
2468         struct nfs_server *server = NFS_SERVER(dir);
2469         int                    status;
2470         struct nfs4_lookup_arg args = {
2471                 .bitmask = server->attr_bitmask,
2472                 .dir_fh = NFS_FH(dir),
2473                 .name = name,
2474         };
2475         struct nfs4_lookup_res res = {
2476                 .server = server,
2477                 .fattr = fattr,
2478                 .fh = fhandle,
2479         };
2480         struct rpc_message msg = {
2481                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2482                 .rpc_argp = &args,
2483                 .rpc_resp = &res,
2484         };
2485
2486         nfs_fattr_init(fattr);
2487
2488         dprintk("NFS call  lookup %s\n", name->name);
2489         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2490         dprintk("NFS reply lookup: %d\n", status);
2491         return status;
2492 }
2493
2494 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2495 {
2496         memset(fh, 0, sizeof(struct nfs_fh));
2497         fattr->fsid.major = 1;
2498         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2499                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2500         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2501         fattr->nlink = 2;
2502 }
2503
2504 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2505                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2506 {
2507         struct nfs4_exception exception = { };
2508         int err;
2509         do {
2510                 int status;
2511
2512                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2513                 switch (status) {
2514                 case -NFS4ERR_BADNAME:
2515                         return -ENOENT;
2516                 case -NFS4ERR_MOVED:
2517                         return nfs4_get_referral(dir, name, fattr, fhandle);
2518                 case -NFS4ERR_WRONGSEC:
2519                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2520                 }
2521                 err = nfs4_handle_exception(NFS_SERVER(dir),
2522                                 status, &exception);
2523         } while (exception.retry);
2524         return err;
2525 }
2526
2527 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2528 {
2529         struct nfs_server *server = NFS_SERVER(inode);
2530         struct nfs4_accessargs args = {
2531                 .fh = NFS_FH(inode),
2532                 .bitmask = server->cache_consistency_bitmask,
2533         };
2534         struct nfs4_accessres res = {
2535                 .server = server,
2536         };
2537         struct rpc_message msg = {
2538                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2539                 .rpc_argp = &args,
2540                 .rpc_resp = &res,
2541                 .rpc_cred = entry->cred,
2542         };
2543         int mode = entry->mask;
2544         int status;
2545
2546         /*
2547          * Determine which access bits we want to ask for...
2548          */
2549         if (mode & MAY_READ)
2550                 args.access |= NFS4_ACCESS_READ;
2551         if (S_ISDIR(inode->i_mode)) {
2552                 if (mode & MAY_WRITE)
2553                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2554                 if (mode & MAY_EXEC)
2555                         args.access |= NFS4_ACCESS_LOOKUP;
2556         } else {
2557                 if (mode & MAY_WRITE)
2558                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2559                 if (mode & MAY_EXEC)
2560                         args.access |= NFS4_ACCESS_EXECUTE;
2561         }
2562
2563         res.fattr = nfs_alloc_fattr();
2564         if (res.fattr == NULL)
2565                 return -ENOMEM;
2566
2567         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2568         if (!status) {
2569                 entry->mask = 0;
2570                 if (res.access & NFS4_ACCESS_READ)
2571                         entry->mask |= MAY_READ;
2572                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2573                         entry->mask |= MAY_WRITE;
2574                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2575                         entry->mask |= MAY_EXEC;
2576                 nfs_refresh_inode(inode, res.fattr);
2577         }
2578         nfs_free_fattr(res.fattr);
2579         return status;
2580 }
2581
2582 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2583 {
2584         struct nfs4_exception exception = { };
2585         int err;
2586         do {
2587                 err = nfs4_handle_exception(NFS_SERVER(inode),
2588                                 _nfs4_proc_access(inode, entry),
2589                                 &exception);
2590         } while (exception.retry);
2591         return err;
2592 }
2593
2594 /*
2595  * TODO: For the time being, we don't try to get any attributes
2596  * along with any of the zero-copy operations READ, READDIR,
2597  * READLINK, WRITE.
2598  *
2599  * In the case of the first three, we want to put the GETATTR
2600  * after the read-type operation -- this is because it is hard
2601  * to predict the length of a GETATTR response in v4, and thus
2602  * align the READ data correctly.  This means that the GETATTR
2603  * may end up partially falling into the page cache, and we should
2604  * shift it into the 'tail' of the xdr_buf before processing.
2605  * To do this efficiently, we need to know the total length
2606  * of data received, which doesn't seem to be available outside
2607  * of the RPC layer.
2608  *
2609  * In the case of WRITE, we also want to put the GETATTR after
2610  * the operation -- in this case because we want to make sure
2611  * we get the post-operation mtime and size.  This means that
2612  * we can't use xdr_encode_pages() as written: we need a variant
2613  * of it which would leave room in the 'tail' iovec.
2614  *
2615  * Both of these changes to the XDR layer would in fact be quite
2616  * minor, but I decided to leave them for a subsequent patch.
2617  */
2618 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2619                 unsigned int pgbase, unsigned int pglen)
2620 {
2621         struct nfs4_readlink args = {
2622                 .fh       = NFS_FH(inode),
2623                 .pgbase   = pgbase,
2624                 .pglen    = pglen,
2625                 .pages    = &page,
2626         };
2627         struct nfs4_readlink_res res;
2628         struct rpc_message msg = {
2629                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2630                 .rpc_argp = &args,
2631                 .rpc_resp = &res,
2632         };
2633
2634         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2635 }
2636
2637 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2638                 unsigned int pgbase, unsigned int pglen)
2639 {
2640         struct nfs4_exception exception = { };
2641         int err;
2642         do {
2643                 err = nfs4_handle_exception(NFS_SERVER(inode),
2644                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2645                                 &exception);
2646         } while (exception.retry);
2647         return err;
2648 }
2649
2650 /*
2651  * Got race?
2652  * We will need to arrange for the VFS layer to provide an atomic open.
2653  * Until then, this create/open method is prone to inefficiency and race
2654  * conditions due to the lookup, create, and open VFS calls from sys_open()
2655  * placed on the wire.
2656  *
2657  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2658  * The file will be opened again in the subsequent VFS open call
2659  * (nfs4_proc_file_open).
2660  *
2661  * The open for read will just hang around to be used by any process that
2662  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2663  */
2664
2665 static int
2666 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2667                  int flags, struct nfs_open_context *ctx)
2668 {
2669         struct dentry *de = dentry;
2670         struct nfs4_state *state;
2671         struct rpc_cred *cred = NULL;
2672         fmode_t fmode = 0;
2673         int status = 0;
2674
2675         if (ctx != NULL) {
2676                 cred = ctx->cred;
2677                 de = ctx->dentry;
2678                 fmode = ctx->mode;
2679         }
2680         sattr->ia_mode &= ~current_umask();
2681         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2682         d_drop(dentry);
2683         if (IS_ERR(state)) {
2684                 status = PTR_ERR(state);
2685                 goto out;
2686         }
2687         d_add(dentry, igrab(state->inode));
2688         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2689         if (ctx != NULL)
2690                 ctx->state = state;
2691         else
2692                 nfs4_close_sync(state, fmode);
2693 out:
2694         return status;
2695 }
2696
2697 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2698 {
2699         struct nfs_server *server = NFS_SERVER(dir);
2700         struct nfs_removeargs args = {
2701                 .fh = NFS_FH(dir),
2702                 .name.len = name->len,
2703                 .name.name = name->name,
2704                 .bitmask = server->attr_bitmask,
2705         };
2706         struct nfs_removeres res = {
2707                 .server = server,
2708         };
2709         struct rpc_message msg = {
2710                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2711                 .rpc_argp = &args,
2712                 .rpc_resp = &res,
2713         };
2714         int status = -ENOMEM;
2715
2716         res.dir_attr = nfs_alloc_fattr();
2717         if (res.dir_attr == NULL)
2718                 goto out;
2719
2720         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2721         if (status == 0) {
2722                 update_changeattr(dir, &res.cinfo);
2723                 nfs_post_op_update_inode(dir, res.dir_attr);
2724         }
2725         nfs_free_fattr(res.dir_attr);
2726 out:
2727         return status;
2728 }
2729
2730 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2731 {
2732         struct nfs4_exception exception = { };
2733         int err;
2734         do {
2735                 err = nfs4_handle_exception(NFS_SERVER(dir),
2736                                 _nfs4_proc_remove(dir, name),
2737                                 &exception);
2738         } while (exception.retry);
2739         return err;
2740 }
2741
2742 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2743 {
2744         struct nfs_server *server = NFS_SERVER(dir);
2745         struct nfs_removeargs *args = msg->rpc_argp;
2746         struct nfs_removeres *res = msg->rpc_resp;
2747
2748         args->bitmask = server->cache_consistency_bitmask;
2749         res->server = server;
2750         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2751         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2752 }
2753
2754 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2755 {
2756         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2757
2758         if (!nfs4_sequence_done(task, &res->seq_res))
2759                 return 0;
2760         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2761                 return 0;
2762         update_changeattr(dir, &res->cinfo);
2763         nfs_post_op_update_inode(dir, res->dir_attr);
2764         return 1;
2765 }
2766
2767 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2768 {
2769         struct nfs_server *server = NFS_SERVER(dir);
2770         struct nfs_renameargs *arg = msg->rpc_argp;
2771         struct nfs_renameres *res = msg->rpc_resp;
2772
2773         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2774         arg->bitmask = server->attr_bitmask;
2775         res->server = server;
2776         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2777 }
2778
2779 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2780                                  struct inode *new_dir)
2781 {
2782         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2783
2784         if (!nfs4_sequence_done(task, &res->seq_res))
2785                 return 0;
2786         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2787                 return 0;
2788
2789         update_changeattr(old_dir, &res->old_cinfo);
2790         nfs_post_op_update_inode(old_dir, res->old_fattr);
2791         update_changeattr(new_dir, &res->new_cinfo);
2792         nfs_post_op_update_inode(new_dir, res->new_fattr);
2793         return 1;
2794 }
2795
2796 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2797                 struct inode *new_dir, struct qstr *new_name)
2798 {
2799         struct nfs_server *server = NFS_SERVER(old_dir);
2800         struct nfs_renameargs arg = {
2801                 .old_dir = NFS_FH(old_dir),
2802                 .new_dir = NFS_FH(new_dir),
2803                 .old_name = old_name,
2804                 .new_name = new_name,
2805                 .bitmask = server->attr_bitmask,
2806         };
2807         struct nfs_renameres res = {
2808                 .server = server,
2809         };
2810         struct rpc_message msg = {
2811                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2812                 .rpc_argp = &arg,
2813                 .rpc_resp = &res,
2814         };
2815         int status = -ENOMEM;
2816
2817         res.old_fattr = nfs_alloc_fattr();
2818         res.new_fattr = nfs_alloc_fattr();
2819         if (res.old_fattr == NULL || res.new_fattr == NULL)
2820                 goto out;
2821
2822         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2823         if (!status) {
2824                 update_changeattr(old_dir, &res.old_cinfo);
2825                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2826                 update_changeattr(new_dir, &res.new_cinfo);
2827                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2828         }
2829 out:
2830         nfs_free_fattr(res.new_fattr);
2831         nfs_free_fattr(res.old_fattr);
2832         return status;
2833 }
2834
2835 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2836                 struct inode *new_dir, struct qstr *new_name)
2837 {
2838         struct nfs4_exception exception = { };
2839         int err;
2840         do {
2841                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2842                                 _nfs4_proc_rename(old_dir, old_name,
2843                                         new_dir, new_name),
2844                                 &exception);
2845         } while (exception.retry);
2846         return err;
2847 }
2848
2849 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2850 {
2851         struct nfs_server *server = NFS_SERVER(inode);
2852         struct nfs4_link_arg arg = {
2853                 .fh     = NFS_FH(inode),
2854                 .dir_fh = NFS_FH(dir),
2855                 .name   = name,
2856                 .bitmask = server->attr_bitmask,
2857         };
2858         struct nfs4_link_res res = {
2859                 .server = server,
2860         };
2861         struct rpc_message msg = {
2862                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2863                 .rpc_argp = &arg,
2864                 .rpc_resp = &res,
2865         };
2866         int status = -ENOMEM;
2867
2868         res.fattr = nfs_alloc_fattr();
2869         res.dir_attr = nfs_alloc_fattr();
2870         if (res.fattr == NULL || res.dir_attr == NULL)
2871                 goto out;
2872
2873         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2874         if (!status) {
2875                 update_changeattr(dir, &res.cinfo);
2876                 nfs_post_op_update_inode(dir, res.dir_attr);
2877                 nfs_post_op_update_inode(inode, res.fattr);
2878         }
2879 out:
2880         nfs_free_fattr(res.dir_attr);
2881         nfs_free_fattr(res.fattr);
2882         return status;
2883 }
2884
2885 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2886 {
2887         struct nfs4_exception exception = { };
2888         int err;
2889         do {
2890                 err = nfs4_handle_exception(NFS_SERVER(inode),
2891                                 _nfs4_proc_link(inode, dir, name),
2892                                 &exception);
2893         } while (exception.retry);
2894         return err;
2895 }
2896
2897 struct nfs4_createdata {
2898         struct rpc_message msg;
2899         struct nfs4_create_arg arg;
2900         struct nfs4_create_res res;
2901         struct nfs_fh fh;
2902         struct nfs_fattr fattr;
2903         struct nfs_fattr dir_fattr;
2904 };
2905
2906 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2907                 struct qstr *name, struct iattr *sattr, u32 ftype)
2908 {
2909         struct nfs4_createdata *data;
2910
2911         data = kzalloc(sizeof(*data), GFP_KERNEL);
2912         if (data != NULL) {
2913                 struct nfs_server *server = NFS_SERVER(dir);
2914
2915                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2916                 data->msg.rpc_argp = &data->arg;
2917                 data->msg.rpc_resp = &data->res;
2918                 data->arg.dir_fh = NFS_FH(dir);
2919                 data->arg.server = server;
2920                 data->arg.name = name;
2921                 data->arg.attrs = sattr;
2922                 data->arg.ftype = ftype;
2923                 data->arg.bitmask = server->attr_bitmask;
2924                 data->res.server = server;
2925                 data->res.fh = &data->fh;
2926                 data->res.fattr = &data->fattr;
2927                 data->res.dir_fattr = &data->dir_fattr;
2928                 nfs_fattr_init(data->res.fattr);
2929                 nfs_fattr_init(data->res.dir_fattr);
2930         }
2931         return data;
2932 }
2933
2934 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2935 {
2936         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2937                                     &data->arg.seq_args, &data->res.seq_res, 1);
2938         if (status == 0) {
2939                 update_changeattr(dir, &data->res.dir_cinfo);
2940                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2941                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2942         }
2943         return status;
2944 }
2945
2946 static void nfs4_free_createdata(struct nfs4_createdata *data)
2947 {
2948         kfree(data);
2949 }
2950
2951 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2952                 struct page *page, unsigned int len, struct iattr *sattr)
2953 {
2954         struct nfs4_createdata *data;
2955         int status = -ENAMETOOLONG;
2956
2957         if (len > NFS4_MAXPATHLEN)
2958                 goto out;
2959
2960         status = -ENOMEM;
2961         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2962         if (data == NULL)
2963                 goto out;
2964
2965         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2966         data->arg.u.symlink.pages = &page;
2967         data->arg.u.symlink.len = len;
2968
2969         status = nfs4_do_create(dir, dentry, data);
2970
2971         nfs4_free_createdata(data);
2972 out:
2973         return status;
2974 }
2975
2976 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2977                 struct page *page, unsigned int len, struct iattr *sattr)
2978 {
2979         struct nfs4_exception exception = { };
2980         int err;
2981         do {
2982                 err = nfs4_handle_exception(NFS_SERVER(dir),
2983                                 _nfs4_proc_symlink(dir, dentry, page,
2984                                                         len, sattr),
2985                                 &exception);
2986         } while (exception.retry);
2987         return err;
2988 }
2989
2990 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2991                 struct iattr *sattr)
2992 {
2993         struct nfs4_createdata *data;
2994         int status = -ENOMEM;
2995
2996         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2997         if (data == NULL)
2998                 goto out;
2999
3000         status = nfs4_do_create(dir, dentry, data);
3001
3002         nfs4_free_createdata(data);
3003 out:
3004         return status;
3005 }
3006
3007 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3008                 struct iattr *sattr)
3009 {
3010         struct nfs4_exception exception = { };
3011         int err;
3012
3013         sattr->ia_mode &= ~current_umask();
3014         do {
3015                 err = nfs4_handle_exception(NFS_SERVER(dir),
3016                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3017                                 &exception);
3018         } while (exception.retry);
3019         return err;
3020 }
3021
3022 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3023                 u64 cookie, struct page **pages, unsigned int count, int plus)
3024 {
3025         struct inode            *dir = dentry->d_inode;
3026         struct nfs4_readdir_arg args = {
3027                 .fh = NFS_FH(dir),
3028                 .pages = pages,
3029                 .pgbase = 0,
3030                 .count = count,
3031                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3032                 .plus = plus,
3033         };
3034         struct nfs4_readdir_res res;
3035         struct rpc_message msg = {
3036                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3037                 .rpc_argp = &args,
3038                 .rpc_resp = &res,
3039                 .rpc_cred = cred,
3040         };
3041         int                     status;
3042
3043         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3044                         dentry->d_parent->d_name.name,
3045                         dentry->d_name.name,
3046                         (unsigned long long)cookie);
3047         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3048         res.pgbase = args.pgbase;
3049         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3050         if (status >= 0) {
3051                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3052                 status += args.pgbase;
3053         }
3054
3055         nfs_invalidate_atime(dir);
3056
3057         dprintk("%s: returns %d\n", __func__, status);
3058         return status;
3059 }
3060
3061 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3062                 u64 cookie, struct page **pages, unsigned int count, int plus)
3063 {
3064         struct nfs4_exception exception = { };
3065         int err;
3066         do {
3067                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3068                                 _nfs4_proc_readdir(dentry, cred, cookie,
3069                                         pages, count, plus),
3070                                 &exception);
3071         } while (exception.retry);
3072         return err;
3073 }
3074
3075 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3076                 struct iattr *sattr, dev_t rdev)
3077 {
3078         struct nfs4_createdata *data;
3079         int mode = sattr->ia_mode;
3080         int status = -ENOMEM;
3081
3082         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3083         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3084
3085         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3086         if (data == NULL)
3087                 goto out;
3088
3089         if (S_ISFIFO(mode))
3090                 data->arg.ftype = NF4FIFO;
3091         else if (S_ISBLK(mode)) {
3092                 data->arg.ftype = NF4BLK;
3093                 data->arg.u.device.specdata1 = MAJOR(rdev);
3094                 data->arg.u.device.specdata2 = MINOR(rdev);
3095         }
3096         else if (S_ISCHR(mode)) {
3097                 data->arg.ftype = NF4CHR;
3098                 data->arg.u.device.specdata1 = MAJOR(rdev);
3099                 data->arg.u.device.specdata2 = MINOR(rdev);
3100         }
3101
3102         status = nfs4_do_create(dir, dentry, data);
3103
3104         nfs4_free_createdata(data);
3105 out:
3106         return status;
3107 }
3108
3109 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3110                 struct iattr *sattr, dev_t rdev)
3111 {
3112         struct nfs4_exception exception = { };
3113         int err;
3114
3115         sattr->ia_mode &= ~current_umask();
3116         do {
3117                 err = nfs4_handle_exception(NFS_SERVER(dir),
3118                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3119                                 &exception);
3120         } while (exception.retry);
3121         return err;
3122 }
3123
3124 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3125                  struct nfs_fsstat *fsstat)
3126 {
3127         struct nfs4_statfs_arg args = {
3128                 .fh = fhandle,
3129                 .bitmask = server->attr_bitmask,
3130         };
3131         struct nfs4_statfs_res res = {
3132                 .fsstat = fsstat,
3133         };
3134         struct rpc_message msg = {
3135                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3136                 .rpc_argp = &args,
3137                 .rpc_resp = &res,
3138         };
3139
3140         nfs_fattr_init(fsstat->fattr);
3141         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3142 }
3143
3144 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3145 {
3146         struct nfs4_exception exception = { };
3147         int err;
3148         do {
3149                 err = nfs4_handle_exception(server,
3150                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3151                                 &exception);
3152         } while (exception.retry);
3153         return err;
3154 }
3155
3156 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3157                 struct nfs_fsinfo *fsinfo)
3158 {
3159         struct nfs4_fsinfo_arg args = {
3160                 .fh = fhandle,
3161                 .bitmask = server->attr_bitmask,
3162         };
3163         struct nfs4_fsinfo_res res = {
3164                 .fsinfo = fsinfo,
3165         };
3166         struct rpc_message msg = {
3167                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3168                 .rpc_argp = &args,
3169                 .rpc_resp = &res,
3170         };
3171
3172         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3173 }
3174
3175 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3176 {
3177         struct nfs4_exception exception = { };
3178         int err;
3179
3180         do {
3181                 err = nfs4_handle_exception(server,
3182                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3183                                 &exception);
3184         } while (exception.retry);
3185         return err;
3186 }
3187
3188 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3189 {
3190         nfs_fattr_init(fsinfo->fattr);
3191         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3192 }
3193
3194 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3195                 struct nfs_pathconf *pathconf)
3196 {
3197         struct nfs4_pathconf_arg args = {
3198                 .fh = fhandle,
3199                 .bitmask = server->attr_bitmask,
3200         };
3201         struct nfs4_pathconf_res res = {
3202                 .pathconf = pathconf,
3203         };
3204         struct rpc_message msg = {
3205                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3206                 .rpc_argp = &args,
3207                 .rpc_resp = &res,
3208         };
3209
3210         /* None of the pathconf attributes are mandatory to implement */
3211         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3212                 memset(pathconf, 0, sizeof(*pathconf));
3213                 return 0;
3214         }
3215
3216         nfs_fattr_init(pathconf->fattr);
3217         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3218 }
3219
3220 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3221                 struct nfs_pathconf *pathconf)
3222 {
3223         struct nfs4_exception exception = { };
3224         int err;
3225
3226         do {
3227                 err = nfs4_handle_exception(server,
3228                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3229                                 &exception);
3230         } while (exception.retry);
3231         return err;
3232 }
3233
3234 void __nfs4_read_done_cb(struct nfs_read_data *data)
3235 {
3236         nfs_invalidate_atime(data->inode);
3237 }
3238
3239 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3240 {
3241         struct nfs_server *server = NFS_SERVER(data->inode);
3242
3243         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3244                 rpc_restart_call_prepare(task);
3245                 return -EAGAIN;
3246         }
3247
3248         __nfs4_read_done_cb(data);
3249         if (task->tk_status > 0)
3250                 renew_lease(server, data->timestamp);
3251         return 0;
3252 }
3253
3254 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3255 {
3256
3257         dprintk("--> %s\n", __func__);
3258
3259         if (!nfs4_sequence_done(task, &data->res.seq_res))
3260                 return -EAGAIN;
3261
3262         return data->read_done_cb ? data->read_done_cb(task, data) :
3263                                     nfs4_read_done_cb(task, data);
3264 }
3265
3266 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3267 {
3268         data->timestamp   = jiffies;
3269         data->read_done_cb = nfs4_read_done_cb;
3270         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3271         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3272 }
3273
3274 /* Reset the the nfs_read_data to send the read to the MDS. */
3275 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3276 {
3277         dprintk("%s Reset task for i/o through\n", __func__);
3278         put_lseg(data->lseg);
3279         data->lseg = NULL;
3280         /* offsets will differ in the dense stripe case */
3281         data->args.offset = data->mds_offset;
3282         data->ds_clp = NULL;
3283         data->args.fh     = NFS_FH(data->inode);
3284         data->read_done_cb = nfs4_read_done_cb;
3285         task->tk_ops = data->mds_ops;
3286         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3287 }
3288 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3289
3290 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3291 {
3292         struct inode *inode = data->inode;
3293
3294         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3295                 rpc_restart_call_prepare(task);
3296                 return -EAGAIN;
3297         }
3298         if (task->tk_status >= 0) {
3299                 renew_lease(NFS_SERVER(inode), data->timestamp);
3300                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3301         }
3302         return 0;
3303 }
3304
3305 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3306 {
3307         if (!nfs4_sequence_done(task, &data->res.seq_res))
3308                 return -EAGAIN;
3309         return data->write_done_cb ? data->write_done_cb(task, data) :
3310                 nfs4_write_done_cb(task, data);
3311 }
3312
3313 /* Reset the the nfs_write_data to send the write to the MDS. */
3314 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3315 {
3316         dprintk("%s Reset task for i/o through\n", __func__);
3317         put_lseg(data->lseg);
3318         data->lseg          = NULL;
3319         data->ds_clp        = NULL;
3320         data->write_done_cb = nfs4_write_done_cb;
3321         data->args.fh       = NFS_FH(data->inode);
3322         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3323         data->args.offset   = data->mds_offset;
3324         data->res.fattr     = &data->fattr;
3325         task->tk_ops        = data->mds_ops;
3326         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3327 }
3328 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3329
3330 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3331 {
3332         struct nfs_server *server = NFS_SERVER(data->inode);
3333
3334         if (data->lseg) {
3335                 data->args.bitmask = NULL;
3336                 data->res.fattr = NULL;
3337         } else
3338                 data->args.bitmask = server->cache_consistency_bitmask;
3339         if (!data->write_done_cb)
3340                 data->write_done_cb = nfs4_write_done_cb;
3341         data->res.server = server;
3342         data->timestamp   = jiffies;
3343
3344         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3345         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3346 }
3347
3348 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3349 {
3350         struct inode *inode = data->inode;
3351
3352         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3353                 rpc_restart_call_prepare(task);
3354                 return -EAGAIN;
3355         }
3356         nfs_refresh_inode(inode, data->res.fattr);
3357         return 0;
3358 }
3359
3360 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3361 {
3362         if (!nfs4_sequence_done(task, &data->res.seq_res))
3363                 return -EAGAIN;
3364         return data->write_done_cb(task, data);
3365 }
3366
3367 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3368 {
3369         struct nfs_server *server = NFS_SERVER(data->inode);
3370
3371         if (data->lseg) {
3372                 data->args.bitmask = NULL;
3373                 data->res.fattr = NULL;
3374         } else
3375                 data->args.bitmask = server->cache_consistency_bitmask;
3376         if (!data->write_done_cb)
3377                 data->write_done_cb = nfs4_commit_done_cb;
3378         data->res.server = server;
3379         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3380         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3381 }
3382
3383 struct nfs4_renewdata {
3384         struct nfs_client       *client;
3385         unsigned long           timestamp;
3386 };
3387
3388 /*
3389  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3390  * standalone procedure for queueing an asynchronous RENEW.
3391  */
3392 static void nfs4_renew_release(void *calldata)
3393 {
3394         struct nfs4_renewdata *data = calldata;
3395         struct nfs_client *clp = data->client;
3396
3397         if (atomic_read(&clp->cl_count) > 1)
3398                 nfs4_schedule_state_renewal(clp);
3399         nfs_put_client(clp);
3400         kfree(data);
3401 }
3402
3403 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3404 {
3405         struct nfs4_renewdata *data = calldata;
3406         struct nfs_client *clp = data->client;
3407         unsigned long timestamp = data->timestamp;
3408
3409         if (task->tk_status < 0) {
3410                 /* Unless we're shutting down, schedule state recovery! */
3411                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3412                         return;
3413                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3414                         nfs4_schedule_lease_recovery(clp);
3415                         return;
3416                 }
3417                 nfs4_schedule_path_down_recovery(clp);
3418         }
3419         do_renew_lease(clp, timestamp);
3420 }
3421
3422 static const struct rpc_call_ops nfs4_renew_ops = {
3423         .rpc_call_done = nfs4_renew_done,
3424         .rpc_release = nfs4_renew_release,
3425 };
3426
3427 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3428 {
3429         struct rpc_message msg = {
3430                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3431                 .rpc_argp       = clp,
3432                 .rpc_cred       = cred,
3433         };
3434         struct nfs4_renewdata *data;
3435
3436         if (renew_flags == 0)
3437                 return 0;
3438         if (!atomic_inc_not_zero(&clp->cl_count))
3439                 return -EIO;
3440         data = kmalloc(sizeof(*data), GFP_NOFS);
3441         if (data == NULL)
3442                 return -ENOMEM;
3443         data->client = clp;
3444         data->timestamp = jiffies;
3445         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3446                         &nfs4_renew_ops, data);
3447 }
3448
3449 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3450 {
3451         struct rpc_message msg = {
3452                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3453                 .rpc_argp       = clp,
3454                 .rpc_cred       = cred,
3455         };
3456         unsigned long now = jiffies;
3457         int status;
3458
3459         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3460         if (status < 0)
3461                 return status;
3462         do_renew_lease(clp, now);
3463         return 0;
3464 }
3465
3466 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3467 {
3468         return (server->caps & NFS_CAP_ACLS)
3469                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3470                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3471 }
3472
3473 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3474  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3475  * the stack.
3476  */
3477 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3478
3479 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3480                 struct page **pages, unsigned int *pgbase)
3481 {
3482         struct page *newpage, **spages;
3483         int rc = 0;
3484         size_t len;
3485         spages = pages;
3486
3487         do {
3488                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3489                 newpage = alloc_page(GFP_KERNEL);
3490
3491                 if (newpage == NULL)
3492                         goto unwind;
3493                 memcpy(page_address(newpage), buf, len);
3494                 buf += len;
3495                 buflen -= len;
3496                 *pages++ = newpage;
3497                 rc++;
3498         } while (buflen != 0);
3499
3500         return rc;
3501
3502 unwind:
3503         for(; rc > 0; rc--)
3504                 __free_page(spages[rc-1]);
3505         return -ENOMEM;
3506 }
3507
3508 struct nfs4_cached_acl {
3509         int cached;
3510         size_t len;
3511         char data[0];
3512 };
3513
3514 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3515 {
3516         struct nfs_inode *nfsi = NFS_I(inode);
3517
3518         spin_lock(&inode->i_lock);
3519         kfree(nfsi->nfs4_acl);
3520         nfsi->nfs4_acl = acl;
3521         spin_unlock(&inode->i_lock);
3522 }
3523
3524 static void nfs4_zap_acl_attr(struct inode *inode)
3525 {
3526         nfs4_set_cached_acl(inode, NULL);
3527 }
3528
3529 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3530 {
3531         struct nfs_inode *nfsi = NFS_I(inode);
3532         struct nfs4_cached_acl *acl;
3533         int ret = -ENOENT;
3534
3535         spin_lock(&inode->i_lock);
3536         acl = nfsi->nfs4_acl;
3537         if (acl == NULL)
3538                 goto out;
3539         if (buf == NULL) /* user is just asking for length */
3540                 goto out_len;
3541         if (acl->cached == 0)
3542                 goto out;
3543         ret = -ERANGE; /* see getxattr(2) man page */
3544         if (acl->len > buflen)
3545                 goto out;
3546         memcpy(buf, acl->data, acl->len);
3547 out_len:
3548         ret = acl->len;
3549 out:
3550         spin_unlock(&inode->i_lock);
3551         return ret;
3552 }
3553
3554 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3555 {
3556         struct nfs4_cached_acl *acl;
3557
3558         if (buf && acl_len <= PAGE_SIZE) {
3559                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3560                 if (acl == NULL)
3561                         goto out;
3562                 acl->cached = 1;
3563                 memcpy(acl->data, buf, acl_len);
3564         } else {
3565                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3566                 if (acl == NULL)
3567                         goto out;
3568                 acl->cached = 0;
3569         }
3570         acl->len = acl_len;
3571 out:
3572         nfs4_set_cached_acl(inode, acl);
3573 }
3574
3575 /*
3576  * The getxattr API returns the required buffer length when called with a
3577  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3578  * the required buf.  On a NULL buf, we send a page of data to the server
3579  * guessing that the ACL request can be serviced by a page. If so, we cache
3580  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3581  * the cache. If not so, we throw away the page, and cache the required
3582  * length. The next getxattr call will then produce another round trip to
3583  * the server, this time with the input buf of the required size.
3584  */
3585 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3586 {
3587         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3588         struct nfs_getaclargs args = {
3589                 .fh = NFS_FH(inode),
3590                 .acl_pages = pages,
3591                 .acl_len = buflen,
3592         };
3593         struct nfs_getaclres res = {
3594                 .acl_len = buflen,
3595         };
3596         void *resp_buf;
3597         struct rpc_message msg = {
3598                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3599                 .rpc_argp = &args,
3600                 .rpc_resp = &res,
3601         };
3602         int ret = -ENOMEM, npages, i, acl_len = 0;
3603
3604         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3605         /* As long as we're doing a round trip to the server anyway,
3606          * let's be prepared for a page of acl data. */
3607         if (npages == 0)
3608                 npages = 1;
3609
3610         for (i = 0; i < npages; i++) {
3611                 pages[i] = alloc_page(GFP_KERNEL);
3612                 if (!pages[i])
3613                         goto out_free;
3614         }
3615         if (npages > 1) {
3616                 /* for decoding across pages */
3617                 args.acl_scratch = alloc_page(GFP_KERNEL);
3618                 if (!args.acl_scratch)
3619                         goto out_free;
3620         }
3621         args.acl_len = npages * PAGE_SIZE;
3622         args.acl_pgbase = 0;
3623         /* Let decode_getfacl know not to fail if the ACL data is larger than
3624          * the page we send as a guess */
3625         if (buf == NULL)
3626                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3627         resp_buf = page_address(pages[0]);
3628
3629         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3630                 __func__, buf, buflen, npages, args.acl_len);
3631         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3632                              &msg, &args.seq_args, &res.seq_res, 0);
3633         if (ret)
3634                 goto out_free;
3635
3636         acl_len = res.acl_len - res.acl_data_offset;
3637         if (acl_len > args.acl_len)
3638                 nfs4_write_cached_acl(inode, NULL, acl_len);
3639         else
3640                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3641                                       acl_len);
3642         if (buf) {
3643                 ret = -ERANGE;
3644                 if (acl_len > buflen)
3645                         goto out_free;
3646                 _copy_from_pages(buf, pages, res.acl_data_offset,
3647                                 res.acl_len);
3648         }
3649         ret = acl_len;
3650 out_free:
3651         for (i = 0; i < npages; i++)
3652                 if (pages[i])
3653                         __free_page(pages[i]);
3654         if (args.acl_scratch)
3655                 __free_page(args.acl_scratch);
3656         return ret;
3657 }
3658
3659 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3660 {
3661         struct nfs4_exception exception = { };
3662         ssize_t ret;
3663         do {
3664                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3665                 if (ret >= 0)
3666                         break;
3667                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3668         } while (exception.retry);
3669         return ret;
3670 }
3671
3672 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3673 {
3674         struct nfs_server *server = NFS_SERVER(inode);
3675         int ret;
3676
3677         if (!nfs4_server_supports_acls(server))
3678                 return -EOPNOTSUPP;
3679         ret = nfs_revalidate_inode(server, inode);
3680         if (ret < 0)
3681                 return ret;
3682         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3683                 nfs_zap_acl_cache(inode);
3684         ret = nfs4_read_cached_acl(inode, buf, buflen);
3685         if (ret != -ENOENT)
3686                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3687                  * but no cached acl data, just the acl length */
3688                 return ret;
3689         return nfs4_get_acl_uncached(inode, buf, buflen);
3690 }
3691
3692 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3693 {
3694         struct nfs_server *server = NFS_SERVER(inode);
3695         struct page *pages[NFS4ACL_MAXPAGES];
3696         struct nfs_setaclargs arg = {
3697                 .fh             = NFS_FH(inode),
3698                 .acl_pages      = pages,
3699                 .acl_len        = buflen,
3700         };
3701         struct nfs_setaclres res;
3702         struct rpc_message msg = {
3703                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3704                 .rpc_argp       = &arg,
3705                 .rpc_resp       = &res,
3706         };
3707         int ret, i;
3708
3709         if (!nfs4_server_supports_acls(server))
3710                 return -EOPNOTSUPP;
3711         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3712         if (i < 0)
3713                 return i;
3714         nfs_inode_return_delegation(inode);
3715         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3716
3717         /*
3718          * Free each page after tx, so the only ref left is
3719          * held by the network stack
3720          */
3721         for (; i > 0; i--)
3722                 put_page(pages[i-1]);
3723
3724         /*
3725          * Acl update can result in inode attribute update.
3726          * so mark the attribute cache invalid.
3727          */
3728         spin_lock(&inode->i_lock);
3729         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3730         spin_unlock(&inode->i_lock);
3731         nfs_access_zap_cache(inode);
3732         nfs_zap_acl_cache(inode);
3733         return ret;
3734 }
3735
3736 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3737 {
3738         struct nfs4_exception exception = { };
3739         int err;
3740         do {
3741                 err = nfs4_handle_exception(NFS_SERVER(inode),
3742                                 __nfs4_proc_set_acl(inode, buf, buflen),
3743                                 &exception);
3744         } while (exception.retry);
3745         return err;
3746 }
3747
3748 static int
3749 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3750 {
3751         struct nfs_client *clp = server->nfs_client;
3752
3753         if (task->tk_status >= 0)
3754                 return 0;
3755         switch(task->tk_status) {
3756                 case -NFS4ERR_ADMIN_REVOKED:
3757                 case -NFS4ERR_BAD_STATEID:
3758                 case -NFS4ERR_OPENMODE:
3759                         if (state == NULL)
3760                                 break;
3761                         nfs4_schedule_stateid_recovery(server, state);
3762                         goto wait_on_recovery;
3763                 case -NFS4ERR_EXPIRED:
3764                         if (state != NULL)
3765                                 nfs4_schedule_stateid_recovery(server, state);
3766                 case -NFS4ERR_STALE_STATEID:
3767                 case -NFS4ERR_STALE_CLIENTID:
3768                         nfs4_schedule_lease_recovery(clp);
3769                         goto wait_on_recovery;
3770 #if defined(CONFIG_NFS_V4_1)
3771                 case -NFS4ERR_BADSESSION:
3772                 case -NFS4ERR_BADSLOT:
3773                 case -NFS4ERR_BAD_HIGH_SLOT:
3774                 case -NFS4ERR_DEADSESSION:
3775                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3776                 case -NFS4ERR_SEQ_FALSE_RETRY:
3777                 case -NFS4ERR_SEQ_MISORDERED:
3778                         dprintk("%s ERROR %d, Reset session\n", __func__,
3779                                 task->tk_status);
3780                         nfs4_schedule_session_recovery(clp->cl_session);
3781                         task->tk_status = 0;
3782                         return -EAGAIN;
3783 #endif /* CONFIG_NFS_V4_1 */
3784                 case -NFS4ERR_DELAY:
3785                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3786                 case -NFS4ERR_GRACE:
3787                 case -EKEYEXPIRED:
3788                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3789                         task->tk_status = 0;
3790                         return -EAGAIN;
3791                 case -NFS4ERR_RETRY_UNCACHED_REP:
3792                 case -NFS4ERR_OLD_STATEID:
3793                         task->tk_status = 0;
3794                         return -EAGAIN;
3795         }
3796         task->tk_status = nfs4_map_errors(task->tk_status);
3797         return 0;
3798 wait_on_recovery:
3799         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3800         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3801                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3802         task->tk_status = 0;
3803         return -EAGAIN;
3804 }
3805
3806 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3807                 unsigned short port, struct rpc_cred *cred,
3808                 struct nfs4_setclientid_res *res)
3809 {
3810         nfs4_verifier sc_verifier;
3811         struct nfs4_setclientid setclientid = {
3812                 .sc_verifier = &sc_verifier,
3813                 .sc_prog = program,
3814                 .sc_cb_ident = clp->cl_cb_ident,
3815         };
3816         struct rpc_message msg = {
3817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3818                 .rpc_argp = &setclientid,
3819                 .rpc_resp = res,
3820                 .rpc_cred = cred,
3821         };
3822         __be32 *p;
3823         int loop = 0;
3824         int status;
3825
3826         p = (__be32*)sc_verifier.data;
3827         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3828         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3829
3830         for(;;) {
3831                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3832                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3833                                 clp->cl_ipaddr,
3834                                 rpc_peeraddr2str(clp->cl_rpcclient,
3835                                                         RPC_DISPLAY_ADDR),
3836                                 rpc_peeraddr2str(clp->cl_rpcclient,
3837                                                         RPC_DISPLAY_PROTO),
3838                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3839                                 clp->cl_id_uniquifier);
3840                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3841                                 sizeof(setclientid.sc_netid),
3842                                 rpc_peeraddr2str(clp->cl_rpcclient,
3843                                                         RPC_DISPLAY_NETID));
3844                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3845                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3846                                 clp->cl_ipaddr, port >> 8, port & 255);
3847
3848                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3849                 if (status != -NFS4ERR_CLID_INUSE)
3850                         break;
3851                 if (loop != 0) {
3852                         ++clp->cl_id_uniquifier;
3853                         break;
3854                 }
3855                 ++loop;
3856                 ssleep(clp->cl_lease_time / HZ + 1);
3857         }
3858         return status;
3859 }
3860
3861 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3862                 struct nfs4_setclientid_res *arg,
3863                 struct rpc_cred *cred)
3864 {
3865         struct nfs_fsinfo fsinfo;
3866         struct rpc_message msg = {
3867                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3868                 .rpc_argp = arg,
3869                 .rpc_resp = &fsinfo,
3870                 .rpc_cred = cred,
3871         };
3872         unsigned long now;
3873         int status;
3874
3875         now = jiffies;
3876         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3877         if (status == 0) {
3878                 spin_lock(&clp->cl_lock);
3879                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3880                 clp->cl_last_renewal = now;
3881                 spin_unlock(&clp->cl_lock);
3882         }
3883         return status;
3884 }
3885
3886 struct nfs4_delegreturndata {
3887         struct nfs4_delegreturnargs args;
3888         struct nfs4_delegreturnres res;
3889         struct nfs_fh fh;
3890         nfs4_stateid stateid;
3891         unsigned long timestamp;
3892         struct nfs_fattr fattr;
3893         int rpc_status;
3894 };
3895
3896 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3897 {
3898         struct nfs4_delegreturndata *data = calldata;
3899
3900         if (!nfs4_sequence_done(task, &data->res.seq_res))
3901                 return;
3902
3903         switch (task->tk_status) {
3904         case -NFS4ERR_STALE_STATEID:
3905         case -NFS4ERR_EXPIRED:
3906         case 0:
3907                 renew_lease(data->res.server, data->timestamp);
3908                 break;
3909         default:
3910                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3911                                 -EAGAIN) {
3912                         rpc_restart_call_prepare(task);
3913                         return;
3914                 }
3915         }
3916         data->rpc_status = task->tk_status;
3917 }
3918
3919 static void nfs4_delegreturn_release(void *calldata)
3920 {
3921         kfree(calldata);
3922 }
3923
3924 #if defined(CONFIG_NFS_V4_1)
3925 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3926 {
3927         struct nfs4_delegreturndata *d_data;
3928
3929         d_data = (struct nfs4_delegreturndata *)data;
3930
3931         if (nfs4_setup_sequence(d_data->res.server,
3932                                 &d_data->args.seq_args,
3933                                 &d_data->res.seq_res, task))
3934                 return;
3935         rpc_call_start(task);
3936 }
3937 #endif /* CONFIG_NFS_V4_1 */
3938
3939 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3940 #if defined(CONFIG_NFS_V4_1)
3941         .rpc_call_prepare = nfs4_delegreturn_prepare,
3942 #endif /* CONFIG_NFS_V4_1 */
3943         .rpc_call_done = nfs4_delegreturn_done,
3944         .rpc_release = nfs4_delegreturn_release,
3945 };
3946
3947 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3948 {
3949         struct nfs4_delegreturndata *data;
3950         struct nfs_server *server = NFS_SERVER(inode);
3951         struct rpc_task *task;
3952         struct rpc_message msg = {
3953                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3954                 .rpc_cred = cred,
3955         };
3956         struct rpc_task_setup task_setup_data = {
3957                 .rpc_client = server->client,
3958                 .rpc_message = &msg,
3959                 .callback_ops = &nfs4_delegreturn_ops,
3960                 .flags = RPC_TASK_ASYNC,
3961         };
3962         int status = 0;
3963
3964         data = kzalloc(sizeof(*data), GFP_NOFS);
3965         if (data == NULL)
3966                 return -ENOMEM;
3967         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3968         data->args.fhandle = &data->fh;
3969         data->args.stateid = &data->stateid;
3970         data->args.bitmask = server->attr_bitmask;
3971         nfs_copy_fh(&data->fh, NFS_FH(inode));
3972         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3973         data->res.fattr = &data->fattr;
3974         data->res.server = server;
3975         nfs_fattr_init(data->res.fattr);
3976         data->timestamp = jiffies;
3977         data->rpc_status = 0;
3978
3979         task_setup_data.callback_data = data;
3980         msg.rpc_argp = &data->args;
3981         msg.rpc_resp = &data->res;
3982         task = rpc_run_task(&task_setup_data);
3983         if (IS_ERR(task))
3984                 return PTR_ERR(task);
3985         if (!issync)
3986                 goto out;
3987         status = nfs4_wait_for_completion_rpc_task(task);
3988         if (status != 0)
3989                 goto out;
3990         status = data->rpc_status;
3991         if (status != 0)
3992                 goto out;
3993         nfs_refresh_inode(inode, &data->fattr);
3994 out:
3995         rpc_put_task(task);
3996         return status;
3997 }
3998
3999 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4000 {
4001         struct nfs_server *server = NFS_SERVER(inode);
4002         struct nfs4_exception exception = { };
4003         int err;
4004         do {
4005                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4006                 switch (err) {
4007                         case -NFS4ERR_STALE_STATEID:
4008                         case -NFS4ERR_EXPIRED:
4009                         case 0:
4010                                 return 0;
4011                 }
4012                 err = nfs4_handle_exception(server, err, &exception);
4013         } while (exception.retry);
4014         return err;
4015 }
4016
4017 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4018 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4019
4020 /*
4021  * sleep, with exponential backoff, and retry the LOCK operation.
4022  */
4023 static unsigned long
4024 nfs4_set_lock_task_retry(unsigned long timeout)
4025 {
4026         freezable_schedule_timeout_killable(timeout);
4027         timeout <<= 1;
4028         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4029                 return NFS4_LOCK_MAXTIMEOUT;
4030         return timeout;
4031 }
4032
4033 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4034 {
4035         struct inode *inode = state->inode;
4036         struct nfs_server *server = NFS_SERVER(inode);
4037         struct nfs_client *clp = server->nfs_client;
4038         struct nfs_lockt_args arg = {
4039                 .fh = NFS_FH(inode),
4040                 .fl = request,
4041         };
4042         struct nfs_lockt_res res = {
4043                 .denied = request,
4044         };
4045         struct rpc_message msg = {
4046                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4047                 .rpc_argp       = &arg,
4048                 .rpc_resp       = &res,
4049                 .rpc_cred       = state->owner->so_cred,
4050         };
4051         struct nfs4_lock_state *lsp;
4052         int status;
4053
4054         arg.lock_owner.clientid = clp->cl_clientid;
4055         status = nfs4_set_lock_state(state, request);
4056         if (status != 0)
4057                 goto out;
4058         lsp = request->fl_u.nfs4_fl.owner;
4059         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4060         arg.lock_owner.s_dev = server->s_dev;
4061         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4062         switch (status) {
4063                 case 0:
4064                         request->fl_type = F_UNLCK;
4065                         break;
4066                 case -NFS4ERR_DENIED:
4067                         status = 0;
4068         }
4069         request->fl_ops->fl_release_private(request);
4070 out:
4071         return status;
4072 }
4073
4074 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4075 {
4076         struct nfs4_exception exception = { };
4077         int err;
4078
4079         do {
4080                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4081                                 _nfs4_proc_getlk(state, cmd, request),
4082                                 &exception);
4083         } while (exception.retry);
4084         return err;
4085 }
4086
4087 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4088 {
4089         int res = 0;
4090         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4091                 case FL_POSIX:
4092                         res = posix_lock_file_wait(file, fl);
4093                         break;
4094                 case FL_FLOCK:
4095                         res = flock_lock_file_wait(file, fl);
4096                         break;
4097                 default:
4098                         BUG();
4099         }
4100         return res;
4101 }
4102
4103 struct nfs4_unlockdata {
4104         struct nfs_locku_args arg;
4105         struct nfs_locku_res res;
4106         struct nfs4_lock_state *lsp;
4107         struct nfs_open_context *ctx;
4108         struct file_lock fl;
4109         const struct nfs_server *server;
4110         unsigned long timestamp;
4111 };
4112
4113 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4114                 struct nfs_open_context *ctx,
4115                 struct nfs4_lock_state *lsp,
4116                 struct nfs_seqid *seqid)
4117 {
4118         struct nfs4_unlockdata *p;
4119         struct inode *inode = lsp->ls_state->inode;
4120
4121         p = kzalloc(sizeof(*p), GFP_NOFS);
4122         if (p == NULL)
4123                 return NULL;
4124         p->arg.fh = NFS_FH(inode);
4125         p->arg.fl = &p->fl;
4126         p->arg.seqid = seqid;
4127         p->res.seqid = seqid;
4128         p->arg.stateid = &lsp->ls_stateid;
4129         p->lsp = lsp;
4130         atomic_inc(&lsp->ls_count);
4131         /* Ensure we don't close file until we're done freeing locks! */
4132         p->ctx = get_nfs_open_context(ctx);
4133         memcpy(&p->fl, fl, sizeof(p->fl));
4134         p->server = NFS_SERVER(inode);
4135         return p;
4136 }
4137
4138 static void nfs4_locku_release_calldata(void *data)
4139 {
4140         struct nfs4_unlockdata *calldata = data;
4141         nfs_free_seqid(calldata->arg.seqid);
4142         nfs4_put_lock_state(calldata->lsp);
4143         put_nfs_open_context(calldata->ctx);
4144         kfree(calldata);
4145 }
4146
4147 static void nfs4_locku_done(struct rpc_task *task, void *data)
4148 {
4149         struct nfs4_unlockdata *calldata = data;
4150
4151         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4152                 return;
4153         switch (task->tk_status) {
4154                 case 0:
4155                         memcpy(calldata->lsp->ls_stateid.data,
4156                                         calldata->res.stateid.data,
4157                                         sizeof(calldata->lsp->ls_stateid.data));
4158                         renew_lease(calldata->server, calldata->timestamp);
4159                         break;
4160                 case -NFS4ERR_BAD_STATEID:
4161                 case -NFS4ERR_OLD_STATEID:
4162                 case -NFS4ERR_STALE_STATEID:
4163                 case -NFS4ERR_EXPIRED:
4164                         break;
4165                 default:
4166                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4167                                 rpc_restart_call_prepare(task);
4168         }
4169 }
4170
4171 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4172 {
4173         struct nfs4_unlockdata *calldata = data;
4174
4175         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4176                 return;
4177         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4178                 /* Note: exit _without_ running nfs4_locku_done */
4179                 task->tk_action = NULL;
4180                 return;
4181         }
4182         calldata->timestamp = jiffies;
4183         if (nfs4_setup_sequence(calldata->server,
4184                                 &calldata->arg.seq_args,
4185                                 &calldata->res.seq_res, task))
4186                 return;
4187         rpc_call_start(task);
4188 }
4189
4190 static const struct rpc_call_ops nfs4_locku_ops = {
4191         .rpc_call_prepare = nfs4_locku_prepare,
4192         .rpc_call_done = nfs4_locku_done,
4193         .rpc_release = nfs4_locku_release_calldata,
4194 };
4195
4196 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4197                 struct nfs_open_context *ctx,
4198                 struct nfs4_lock_state *lsp,
4199                 struct nfs_seqid *seqid)
4200 {
4201         struct nfs4_unlockdata *data;
4202         struct rpc_message msg = {
4203                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4204                 .rpc_cred = ctx->cred,
4205         };
4206         struct rpc_task_setup task_setup_data = {
4207                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4208                 .rpc_message = &msg,
4209                 .callback_ops = &nfs4_locku_ops,
4210                 .workqueue = nfsiod_workqueue,
4211                 .flags = RPC_TASK_ASYNC,
4212         };
4213
4214         /* Ensure this is an unlock - when canceling a lock, the
4215          * canceled lock is passed in, and it won't be an unlock.
4216          */
4217         fl->fl_type = F_UNLCK;
4218
4219         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4220         if (data == NULL) {
4221                 nfs_free_seqid(seqid);
4222                 return ERR_PTR(-ENOMEM);
4223         }
4224
4225         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4226         msg.rpc_argp = &data->arg;
4227         msg.rpc_resp = &data->res;
4228         task_setup_data.callback_data = data;
4229         return rpc_run_task(&task_setup_data);
4230 }
4231
4232 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4233 {
4234         struct nfs_inode *nfsi = NFS_I(state->inode);
4235         struct nfs_seqid *seqid;
4236         struct nfs4_lock_state *lsp;
4237         struct rpc_task *task;
4238         int status = 0;
4239         unsigned char fl_flags = request->fl_flags;
4240
4241         status = nfs4_set_lock_state(state, request);
4242         /* Unlock _before_ we do the RPC call */
4243         request->fl_flags |= FL_EXISTS;
4244         down_read(&nfsi->rwsem);
4245         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4246                 up_read(&nfsi->rwsem);
4247                 goto out;
4248         }
4249         up_read(&nfsi->rwsem);
4250         if (status != 0)
4251                 goto out;
4252         /* Is this a delegated lock? */
4253         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4254                 goto out;
4255         lsp = request->fl_u.nfs4_fl.owner;
4256         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4257         status = -ENOMEM;
4258         if (seqid == NULL)
4259                 goto out;
4260         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4261         status = PTR_ERR(task);
4262         if (IS_ERR(task))
4263                 goto out;
4264         status = nfs4_wait_for_completion_rpc_task(task);
4265         rpc_put_task(task);
4266 out:
4267         request->fl_flags = fl_flags;
4268         return status;
4269 }
4270
4271 struct nfs4_lockdata {
4272         struct nfs_lock_args arg;
4273         struct nfs_lock_res res;
4274         struct nfs4_lock_state *lsp;
4275         struct nfs_open_context *ctx;
4276         struct file_lock fl;
4277         unsigned long timestamp;
4278         int rpc_status;
4279         int cancelled;
4280         struct nfs_server *server;
4281 };
4282
4283 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4284                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4285                 gfp_t gfp_mask)
4286 {
4287         struct nfs4_lockdata *p;
4288         struct inode *inode = lsp->ls_state->inode;
4289         struct nfs_server *server = NFS_SERVER(inode);
4290
4291         p = kzalloc(sizeof(*p), gfp_mask);
4292         if (p == NULL)
4293                 return NULL;
4294
4295         p->arg.fh = NFS_FH(inode);
4296         p->arg.fl = &p->fl;
4297         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4298         if (p->arg.open_seqid == NULL)
4299                 goto out_free;
4300         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4301         if (p->arg.lock_seqid == NULL)
4302                 goto out_free_seqid;
4303         p->arg.lock_stateid = &lsp->ls_stateid;
4304         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4305         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4306         p->arg.lock_owner.s_dev = server->s_dev;
4307         p->res.lock_seqid = p->arg.lock_seqid;
4308         p->lsp = lsp;
4309         p->server = server;
4310         atomic_inc(&lsp->ls_count);
4311         p->ctx = get_nfs_open_context(ctx);
4312         memcpy(&p->fl, fl, sizeof(p->fl));
4313         return p;
4314 out_free_seqid:
4315         nfs_free_seqid(p->arg.open_seqid);
4316 out_free:
4317         kfree(p);
4318         return NULL;
4319 }
4320
4321 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4322 {
4323         struct nfs4_lockdata *data = calldata;
4324         struct nfs4_state *state = data->lsp->ls_state;
4325
4326         dprintk("%s: begin!\n", __func__);
4327         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4328                 return;
4329         /* Do we need to do an open_to_lock_owner? */
4330         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4331                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4332                         return;
4333                 data->arg.open_stateid = &state->stateid;
4334                 data->arg.new_lock_owner = 1;
4335                 data->res.open_seqid = data->arg.open_seqid;
4336         } else
4337                 data->arg.new_lock_owner = 0;
4338         data->timestamp = jiffies;
4339         if (nfs4_setup_sequence(data->server,
4340                                 &data->arg.seq_args,
4341                                 &data->res.seq_res, task))
4342                 return;
4343         rpc_call_start(task);
4344         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4345 }
4346
4347 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4348 {
4349         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4350         nfs4_lock_prepare(task, calldata);
4351 }
4352
4353 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4354 {
4355         struct nfs4_lockdata *data = calldata;
4356
4357         dprintk("%s: begin!\n", __func__);
4358
4359         if (!nfs4_sequence_done(task, &data->res.seq_res))
4360                 return;
4361
4362         data->rpc_status = task->tk_status;
4363         if (data->arg.new_lock_owner != 0) {
4364                 if (data->rpc_status == 0)
4365                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4366                 else
4367                         goto out;
4368         }
4369         if (data->rpc_status == 0) {
4370                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4371                                         sizeof(data->lsp->ls_stateid.data));
4372                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4373                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4374         }
4375 out:
4376         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4377 }
4378
4379 static void nfs4_lock_release(void *calldata)
4380 {
4381         struct nfs4_lockdata *data = calldata;
4382
4383         dprintk("%s: begin!\n", __func__);
4384         nfs_free_seqid(data->arg.open_seqid);
4385         if (data->cancelled != 0) {
4386                 struct rpc_task *task;
4387                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4388                                 data->arg.lock_seqid);
4389                 if (!IS_ERR(task))
4390                         rpc_put_task_async(task);
4391                 dprintk("%s: cancelling lock!\n", __func__);
4392         } else
4393                 nfs_free_seqid(data->arg.lock_seqid);
4394         nfs4_put_lock_state(data->lsp);
4395         put_nfs_open_context(data->ctx);
4396         kfree(data);
4397         dprintk("%s: done!\n", __func__);
4398 }
4399
4400 static const struct rpc_call_ops nfs4_lock_ops = {
4401         .rpc_call_prepare = nfs4_lock_prepare,
4402         .rpc_call_done = nfs4_lock_done,
4403         .rpc_release = nfs4_lock_release,
4404 };
4405
4406 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4407         .rpc_call_prepare = nfs4_recover_lock_prepare,
4408         .rpc_call_done = nfs4_lock_done,
4409         .rpc_release = nfs4_lock_release,
4410 };
4411
4412 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4413 {
4414         switch (error) {
4415         case -NFS4ERR_ADMIN_REVOKED:
4416         case -NFS4ERR_BAD_STATEID:
4417                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4418                 if (new_lock_owner != 0 ||
4419                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4420                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4421                 break;
4422         case -NFS4ERR_STALE_STATEID:
4423                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4424         case -NFS4ERR_EXPIRED:
4425                 nfs4_schedule_lease_recovery(server->nfs_client);
4426         };
4427 }
4428
4429 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4430 {
4431         struct nfs4_lockdata *data;
4432         struct rpc_task *task;
4433         struct rpc_message msg = {
4434                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4435                 .rpc_cred = state->owner->so_cred,
4436         };
4437         struct rpc_task_setup task_setup_data = {
4438                 .rpc_client = NFS_CLIENT(state->inode),
4439                 .rpc_message = &msg,
4440                 .callback_ops = &nfs4_lock_ops,
4441                 .workqueue = nfsiod_workqueue,
4442                 .flags = RPC_TASK_ASYNC,
4443         };
4444         int ret;
4445
4446         dprintk("%s: begin!\n", __func__);
4447         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4448                         fl->fl_u.nfs4_fl.owner,
4449                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4450         if (data == NULL)
4451                 return -ENOMEM;
4452         if (IS_SETLKW(cmd))
4453                 data->arg.block = 1;
4454         if (recovery_type > NFS_LOCK_NEW) {
4455                 if (recovery_type == NFS_LOCK_RECLAIM)
4456                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4457                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4458         }
4459         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4460         msg.rpc_argp = &data->arg;
4461         msg.rpc_resp = &data->res;
4462         task_setup_data.callback_data = data;
4463         task = rpc_run_task(&task_setup_data);
4464         if (IS_ERR(task))
4465                 return PTR_ERR(task);
4466         ret = nfs4_wait_for_completion_rpc_task(task);
4467         if (ret == 0) {
4468                 ret = data->rpc_status;
4469                 if (ret)
4470                         nfs4_handle_setlk_error(data->server, data->lsp,
4471                                         data->arg.new_lock_owner, ret);
4472         } else
4473                 data->cancelled = 1;
4474         rpc_put_task(task);
4475         dprintk("%s: done, ret = %d!\n", __func__, ret);
4476         return ret;
4477 }
4478
4479 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4480 {
4481         struct nfs_server *server = NFS_SERVER(state->inode);
4482         struct nfs4_exception exception = { };
4483         int err;
4484
4485         do {
4486                 /* Cache the lock if possible... */
4487                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4488                         return 0;
4489                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4490                 if (err != -NFS4ERR_DELAY)
4491                         break;
4492                 nfs4_handle_exception(server, err, &exception);
4493         } while (exception.retry);
4494         return err;
4495 }
4496
4497 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4498 {
4499         struct nfs_server *server = NFS_SERVER(state->inode);
4500         struct nfs4_exception exception = { };
4501         int err;
4502
4503         err = nfs4_set_lock_state(state, request);
4504         if (err != 0)
4505                 return err;
4506         do {
4507                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4508                         return 0;
4509                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4510                 switch (err) {
4511                 default:
4512                         goto out;
4513                 case -NFS4ERR_GRACE:
4514                 case -NFS4ERR_DELAY:
4515                         nfs4_handle_exception(server, err, &exception);
4516                         err = 0;
4517                 }
4518         } while (exception.retry);
4519 out:
4520         return err;
4521 }
4522
4523 #if defined(CONFIG_NFS_V4_1)
4524 static int nfs41_check_expired_locks(struct nfs4_state *state)
4525 {
4526         int status, ret = NFS_OK;
4527         struct nfs4_lock_state *lsp;
4528         struct nfs_server *server = NFS_SERVER(state->inode);
4529
4530         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4531                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4532                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4533                         if (status != NFS_OK) {
4534                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4535                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4536                                 ret = status;
4537                         }
4538                 }
4539         };
4540
4541         return ret;
4542 }
4543
4544 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4545 {
4546         int status = NFS_OK;
4547
4548         if (test_bit(LK_STATE_IN_USE, &state->flags))
4549                 status = nfs41_check_expired_locks(state);
4550         if (status == NFS_OK)
4551                 return status;
4552         return nfs4_lock_expired(state, request);
4553 }
4554 #endif
4555
4556 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4557 {
4558         struct nfs_inode *nfsi = NFS_I(state->inode);
4559         unsigned char fl_flags = request->fl_flags;
4560         int status = -ENOLCK;
4561
4562         if ((fl_flags & FL_POSIX) &&
4563                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4564                 goto out;
4565         /* Is this a delegated open? */
4566         status = nfs4_set_lock_state(state, request);
4567         if (status != 0)
4568                 goto out;
4569         request->fl_flags |= FL_ACCESS;
4570         status = do_vfs_lock(request->fl_file, request);
4571         if (status < 0)
4572                 goto out;
4573         down_read(&nfsi->rwsem);
4574         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4575                 /* Yes: cache locks! */
4576                 /* ...but avoid races with delegation recall... */
4577                 request->fl_flags = fl_flags & ~FL_SLEEP;
4578                 status = do_vfs_lock(request->fl_file, request);
4579                 goto out_unlock;
4580         }
4581         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4582         if (status != 0)
4583                 goto out_unlock;
4584         /* Note: we always want to sleep here! */
4585         request->fl_flags = fl_flags | FL_SLEEP;
4586         if (do_vfs_lock(request->fl_file, request) < 0)
4587                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4588                         "manager!\n", __func__);
4589 out_unlock:
4590         up_read(&nfsi->rwsem);
4591 out:
4592         request->fl_flags = fl_flags;
4593         return status;
4594 }
4595
4596 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4597 {
4598         struct nfs4_exception exception = { };
4599         int err;
4600
4601         do {
4602                 err = _nfs4_proc_setlk(state, cmd, request);
4603                 if (err == -NFS4ERR_DENIED)
4604                         err = -EAGAIN;
4605                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4606                                 err, &exception);
4607         } while (exception.retry);
4608         return err;
4609 }
4610
4611 static int
4612 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4613 {
4614         struct nfs_open_context *ctx;
4615         struct nfs4_state *state;
4616         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4617         int status;
4618
4619         /* verify open state */
4620         ctx = nfs_file_open_context(filp);
4621         state = ctx->state;
4622
4623         if (request->fl_start < 0 || request->fl_end < 0)
4624                 return -EINVAL;
4625
4626         if (IS_GETLK(cmd)) {
4627                 if (state != NULL)
4628                         return nfs4_proc_getlk(state, F_GETLK, request);
4629                 return 0;
4630         }
4631
4632         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4633                 return -EINVAL;
4634
4635         if (request->fl_type == F_UNLCK) {
4636                 if (state != NULL)
4637                         return nfs4_proc_unlck(state, cmd, request);
4638                 return 0;
4639         }
4640
4641         if (state == NULL)
4642                 return -ENOLCK;
4643         do {
4644                 status = nfs4_proc_setlk(state, cmd, request);
4645                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4646                         break;
4647                 timeout = nfs4_set_lock_task_retry(timeout);
4648                 status = -ERESTARTSYS;
4649                 if (signalled())
4650                         break;
4651         } while(status < 0);
4652         return status;
4653 }
4654
4655 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4656 {
4657         struct nfs_server *server = NFS_SERVER(state->inode);
4658         struct nfs4_exception exception = { };
4659         int err;
4660
4661         err = nfs4_set_lock_state(state, fl);
4662         if (err != 0)
4663                 goto out;
4664         do {
4665                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4666                 switch (err) {
4667                         default:
4668                                 printk(KERN_ERR "NFS: %s: unhandled error "
4669                                         "%d.\n", __func__, err);
4670                         case 0:
4671                         case -ESTALE:
4672                                 goto out;
4673                         case -NFS4ERR_EXPIRED:
4674                                 nfs4_schedule_stateid_recovery(server, state);
4675                         case -NFS4ERR_STALE_CLIENTID:
4676                         case -NFS4ERR_STALE_STATEID:
4677                                 nfs4_schedule_lease_recovery(server->nfs_client);
4678                                 goto out;
4679                         case -NFS4ERR_BADSESSION:
4680                         case -NFS4ERR_BADSLOT:
4681                         case -NFS4ERR_BAD_HIGH_SLOT:
4682                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4683                         case -NFS4ERR_DEADSESSION:
4684                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4685                                 goto out;
4686                         case -ERESTARTSYS:
4687                                 /*
4688                                  * The show must go on: exit, but mark the
4689                                  * stateid as needing recovery.
4690                                  */
4691                         case -NFS4ERR_ADMIN_REVOKED:
4692                         case -NFS4ERR_BAD_STATEID:
4693                         case -NFS4ERR_OPENMODE:
4694                                 nfs4_schedule_stateid_recovery(server, state);
4695                                 err = 0;
4696                                 goto out;
4697                         case -EKEYEXPIRED:
4698                                 /*
4699                                  * User RPCSEC_GSS context has expired.
4700                                  * We cannot recover this stateid now, so
4701                                  * skip it and allow recovery thread to
4702                                  * proceed.
4703                                  */
4704                                 err = 0;
4705                                 goto out;
4706                         case -ENOMEM:
4707                         case -NFS4ERR_DENIED:
4708                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4709                                 err = 0;
4710                                 goto out;
4711                         case -NFS4ERR_DELAY:
4712                                 break;
4713                 }
4714                 err = nfs4_handle_exception(server, err, &exception);
4715         } while (exception.retry);
4716 out:
4717         return err;
4718 }
4719
4720 static void nfs4_release_lockowner_release(void *calldata)
4721 {
4722         kfree(calldata);
4723 }
4724
4725 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4726         .rpc_release = nfs4_release_lockowner_release,
4727 };
4728
4729 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4730 {
4731         struct nfs_server *server = lsp->ls_state->owner->so_server;
4732         struct nfs_release_lockowner_args *args;
4733         struct rpc_message msg = {
4734                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4735         };
4736
4737         if (server->nfs_client->cl_mvops->minor_version != 0)
4738                 return;
4739         args = kmalloc(sizeof(*args), GFP_NOFS);
4740         if (!args)
4741                 return;
4742         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4743         args->lock_owner.id = lsp->ls_seqid.owner_id;
4744         args->lock_owner.s_dev = server->s_dev;
4745         msg.rpc_argp = args;
4746         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4747 }
4748
4749 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4750
4751 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4752                                    const void *buf, size_t buflen,
4753                                    int flags, int type)
4754 {
4755         if (strcmp(key, "") != 0)
4756                 return -EINVAL;
4757
4758         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4759 }
4760
4761 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4762                                    void *buf, size_t buflen, int type)
4763 {
4764         if (strcmp(key, "") != 0)
4765                 return -EINVAL;
4766
4767         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4768 }
4769
4770 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4771                                        size_t list_len, const char *name,
4772                                        size_t name_len, int type)
4773 {
4774         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4775
4776         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4777                 return 0;
4778
4779         if (list && len <= list_len)
4780                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4781         return len;
4782 }
4783
4784 /*
4785  * nfs_fhget will use either the mounted_on_fileid or the fileid
4786  */
4787 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4788 {
4789         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4790                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4791               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4792               (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4793                 return;
4794
4795         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4796                 NFS_ATTR_FATTR_NLINK;
4797         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4798         fattr->nlink = 2;
4799 }
4800
4801 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4802                 struct nfs4_fs_locations *fs_locations, struct page *page)
4803 {
4804         struct nfs_server *server = NFS_SERVER(dir);
4805         u32 bitmask[2] = {
4806                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4807         };
4808         struct nfs4_fs_locations_arg args = {
4809                 .dir_fh = NFS_FH(dir),
4810                 .name = name,
4811                 .page = page,
4812                 .bitmask = bitmask,
4813         };
4814         struct nfs4_fs_locations_res res = {
4815                 .fs_locations = fs_locations,
4816         };
4817         struct rpc_message msg = {
4818                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4819                 .rpc_argp = &args,
4820                 .rpc_resp = &res,
4821         };
4822         int status;
4823
4824         dprintk("%s: start\n", __func__);
4825
4826         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4827          * is not supported */
4828         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4829                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4830         else
4831                 bitmask[0] |= FATTR4_WORD0_FILEID;
4832
4833         nfs_fattr_init(&fs_locations->fattr);
4834         fs_locations->server = server;
4835         fs_locations->nlocations = 0;
4836         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4837         dprintk("%s: returned status = %d\n", __func__, status);
4838         return status;
4839 }
4840
4841 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4842 {
4843         int status;
4844         struct nfs4_secinfo_arg args = {
4845                 .dir_fh = NFS_FH(dir),
4846                 .name   = name,
4847         };
4848         struct nfs4_secinfo_res res = {
4849                 .flavors     = flavors,
4850         };
4851         struct rpc_message msg = {
4852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4853                 .rpc_argp = &args,
4854                 .rpc_resp = &res,
4855         };
4856
4857         dprintk("NFS call  secinfo %s\n", name->name);
4858         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4859         dprintk("NFS reply  secinfo: %d\n", status);
4860         return status;
4861 }
4862
4863 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4864 {
4865         struct nfs4_exception exception = { };
4866         int err;
4867         do {
4868                 err = nfs4_handle_exception(NFS_SERVER(dir),
4869                                 _nfs4_proc_secinfo(dir, name, flavors),
4870                                 &exception);
4871         } while (exception.retry);
4872         return err;
4873 }
4874
4875 #ifdef CONFIG_NFS_V4_1
4876 /*
4877  * Check the exchange flags returned by the server for invalid flags, having
4878  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4879  * DS flags set.
4880  */
4881 static int nfs4_check_cl_exchange_flags(u32 flags)
4882 {
4883         if (flags & ~EXCHGID4_FLAG_MASK_R)
4884                 goto out_inval;
4885         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4886             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4887                 goto out_inval;
4888         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4889                 goto out_inval;
4890         return NFS_OK;
4891 out_inval:
4892         return -NFS4ERR_INVAL;
4893 }
4894
4895 static bool
4896 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4897 {
4898         if (a->server_scope_sz == b->server_scope_sz &&
4899             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4900                 return true;
4901
4902         return false;
4903 }
4904
4905 /*
4906  * nfs4_proc_exchange_id()
4907  *
4908  * Since the clientid has expired, all compounds using sessions
4909  * associated with the stale clientid will be returning
4910  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4911  * be in some phase of session reset.
4912  */
4913 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4914 {
4915         nfs4_verifier verifier;
4916         struct nfs41_exchange_id_args args = {
4917                 .client = clp,
4918                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4919         };
4920         struct nfs41_exchange_id_res res = {
4921                 .client = clp,
4922         };
4923         int status;
4924         struct rpc_message msg = {
4925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4926                 .rpc_argp = &args,
4927                 .rpc_resp = &res,
4928                 .rpc_cred = cred,
4929         };
4930         __be32 *p;
4931
4932         dprintk("--> %s\n", __func__);
4933         BUG_ON(clp == NULL);
4934
4935         p = (u32 *)verifier.data;
4936         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4937         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4938         args.verifier = &verifier;
4939
4940         args.id_len = scnprintf(args.id, sizeof(args.id),
4941                                 "%s/%s.%s/%u",
4942                                 clp->cl_ipaddr,
4943                                 init_utsname()->nodename,
4944                                 init_utsname()->domainname,
4945                                 clp->cl_rpcclient->cl_auth->au_flavor);
4946
4947         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4948         if (unlikely(!res.server_scope))
4949                 return -ENOMEM;
4950
4951         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4952         if (!status)
4953                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4954
4955         if (!status) {
4956                 if (clp->server_scope &&
4957                     !nfs41_same_server_scope(clp->server_scope,
4958                                              res.server_scope)) {
4959                         dprintk("%s: server_scope mismatch detected\n",
4960                                 __func__);
4961                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4962                         kfree(clp->server_scope);
4963                         clp->server_scope = NULL;
4964                 }
4965
4966                 if (!clp->server_scope)
4967                         clp->server_scope = res.server_scope;
4968                 else
4969                         kfree(res.server_scope);
4970         }
4971
4972         dprintk("<-- %s status= %d\n", __func__, status);
4973         return status;
4974 }
4975
4976 struct nfs4_get_lease_time_data {
4977         struct nfs4_get_lease_time_args *args;
4978         struct nfs4_get_lease_time_res *res;
4979         struct nfs_client *clp;
4980 };
4981
4982 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4983                                         void *calldata)
4984 {
4985         int ret;
4986         struct nfs4_get_lease_time_data *data =
4987                         (struct nfs4_get_lease_time_data *)calldata;
4988
4989         dprintk("--> %s\n", __func__);
4990         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4991         /* just setup sequence, do not trigger session recovery
4992            since we're invoked within one */
4993         ret = nfs41_setup_sequence(data->clp->cl_session,
4994                                    &data->args->la_seq_args,
4995                                    &data->res->lr_seq_res, task);
4996
4997         BUG_ON(ret == -EAGAIN);
4998         rpc_call_start(task);
4999         dprintk("<-- %s\n", __func__);
5000 }
5001
5002 /*
5003  * Called from nfs4_state_manager thread for session setup, so don't recover
5004  * from sequence operation or clientid errors.
5005  */
5006 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5007 {
5008         struct nfs4_get_lease_time_data *data =
5009                         (struct nfs4_get_lease_time_data *)calldata;
5010
5011         dprintk("--> %s\n", __func__);
5012         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5013                 return;
5014         switch (task->tk_status) {
5015         case -NFS4ERR_DELAY:
5016         case -NFS4ERR_GRACE:
5017                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5018                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5019                 task->tk_status = 0;
5020                 /* fall through */
5021         case -NFS4ERR_RETRY_UNCACHED_REP:
5022                 rpc_restart_call_prepare(task);
5023                 return;
5024         }
5025         dprintk("<-- %s\n", __func__);
5026 }
5027
5028 struct rpc_call_ops nfs4_get_lease_time_ops = {
5029         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5030         .rpc_call_done = nfs4_get_lease_time_done,
5031 };
5032
5033 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5034 {
5035         struct rpc_task *task;
5036         struct nfs4_get_lease_time_args args;
5037         struct nfs4_get_lease_time_res res = {
5038                 .lr_fsinfo = fsinfo,
5039         };
5040         struct nfs4_get_lease_time_data data = {
5041                 .args = &args,
5042                 .res = &res,
5043                 .clp = clp,
5044         };
5045         struct rpc_message msg = {
5046                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5047                 .rpc_argp = &args,
5048                 .rpc_resp = &res,
5049         };
5050         struct rpc_task_setup task_setup = {
5051                 .rpc_client = clp->cl_rpcclient,
5052                 .rpc_message = &msg,
5053                 .callback_ops = &nfs4_get_lease_time_ops,
5054                 .callback_data = &data,
5055                 .flags = RPC_TASK_TIMEOUT,
5056         };
5057         int status;
5058
5059         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5060         dprintk("--> %s\n", __func__);
5061         task = rpc_run_task(&task_setup);
5062
5063         if (IS_ERR(task))
5064                 status = PTR_ERR(task);
5065         else {
5066                 status = task->tk_status;
5067                 rpc_put_task(task);
5068         }
5069         dprintk("<-- %s return %d\n", __func__, status);
5070
5071         return status;
5072 }
5073
5074 /*
5075  * Reset a slot table
5076  */
5077 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5078                                  int ivalue)
5079 {
5080         struct nfs4_slot *new = NULL;
5081         int i;
5082         int ret = 0;
5083
5084         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5085                 max_reqs, tbl->max_slots);
5086
5087         /* Does the newly negotiated max_reqs match the existing slot table? */
5088         if (max_reqs != tbl->max_slots) {
5089                 ret = -ENOMEM;
5090                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
5091                               GFP_NOFS);
5092                 if (!new)
5093                         goto out;
5094                 ret = 0;
5095                 kfree(tbl->slots);
5096         }
5097         spin_lock(&tbl->slot_tbl_lock);
5098         if (new) {
5099                 tbl->slots = new;
5100                 tbl->max_slots = max_reqs;
5101         }
5102         for (i = 0; i < tbl->max_slots; ++i)
5103                 tbl->slots[i].seq_nr = ivalue;
5104         spin_unlock(&tbl->slot_tbl_lock);
5105         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5106                 tbl, tbl->slots, tbl->max_slots);
5107 out:
5108         dprintk("<-- %s: return %d\n", __func__, ret);
5109         return ret;
5110 }
5111
5112 /* Destroy the slot table */
5113 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5114 {
5115         if (session->fc_slot_table.slots != NULL) {
5116                 kfree(session->fc_slot_table.slots);
5117                 session->fc_slot_table.slots = NULL;
5118         }
5119         if (session->bc_slot_table.slots != NULL) {
5120                 kfree(session->bc_slot_table.slots);
5121                 session->bc_slot_table.slots = NULL;
5122         }
5123         return;
5124 }
5125
5126 /*
5127  * Initialize slot table
5128  */
5129 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
5130                 int max_slots, int ivalue)
5131 {
5132         struct nfs4_slot *slot;
5133         int ret = -ENOMEM;
5134
5135         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
5136
5137         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
5138
5139         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
5140         if (!slot)
5141                 goto out;
5142         ret = 0;
5143
5144         spin_lock(&tbl->slot_tbl_lock);
5145         tbl->max_slots = max_slots;
5146         tbl->slots = slot;
5147         tbl->highest_used_slotid = NFS4_NO_SLOT;  /* no slot is currently used */
5148         spin_unlock(&tbl->slot_tbl_lock);
5149         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5150                 tbl, tbl->slots, tbl->max_slots);
5151 out:
5152         dprintk("<-- %s: return %d\n", __func__, ret);
5153         return ret;
5154 }
5155
5156 /*
5157  * Initialize or reset the forechannel and backchannel tables
5158  */
5159 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5160 {
5161         struct nfs4_slot_table *tbl;
5162         int status;
5163
5164         dprintk("--> %s\n", __func__);
5165         /* Fore channel */
5166         tbl = &ses->fc_slot_table;
5167         if (tbl->slots == NULL) {
5168                 status = nfs4_init_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5169                 if (status) /* -ENOMEM */
5170                         return status;
5171         } else {
5172                 status = nfs4_reset_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5173                 if (status)
5174                         return status;
5175         }
5176         /* Back channel */
5177         tbl = &ses->bc_slot_table;
5178         if (tbl->slots == NULL) {
5179                 status = nfs4_init_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5180                 if (status)
5181                         /* Fore and back channel share a connection so get
5182                          * both slot tables or neither */
5183                         nfs4_destroy_slot_tables(ses);
5184         } else
5185                 status = nfs4_reset_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5186         return status;
5187 }
5188
5189 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5190 {
5191         struct nfs4_session *session;
5192         struct nfs4_slot_table *tbl;
5193
5194         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5195         if (!session)
5196                 return NULL;
5197
5198         tbl = &session->fc_slot_table;
5199         tbl->highest_used_slotid = NFS4_NO_SLOT;
5200         spin_lock_init(&tbl->slot_tbl_lock);
5201         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5202         init_completion(&tbl->complete);
5203
5204         tbl = &session->bc_slot_table;
5205         tbl->highest_used_slotid = NFS4_NO_SLOT;
5206         spin_lock_init(&tbl->slot_tbl_lock);
5207         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5208         init_completion(&tbl->complete);
5209
5210         session->session_state = 1<<NFS4_SESSION_INITING;
5211
5212         session->clp = clp;
5213         return session;
5214 }
5215
5216 void nfs4_destroy_session(struct nfs4_session *session)
5217 {
5218         nfs4_proc_destroy_session(session);
5219         dprintk("%s Destroy backchannel for xprt %p\n",
5220                 __func__, session->clp->cl_rpcclient->cl_xprt);
5221         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5222                                 NFS41_BC_MIN_CALLBACKS);
5223         nfs4_destroy_slot_tables(session);
5224         kfree(session);
5225 }
5226
5227 /*
5228  * Initialize the values to be used by the client in CREATE_SESSION
5229  * If nfs4_init_session set the fore channel request and response sizes,
5230  * use them.
5231  *
5232  * Set the back channel max_resp_sz_cached to zero to force the client to
5233  * always set csa_cachethis to FALSE because the current implementation
5234  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5235  */
5236 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5237 {
5238         struct nfs4_session *session = args->client->cl_session;
5239         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5240                      mxresp_sz = session->fc_attrs.max_resp_sz;
5241
5242         if (mxrqst_sz == 0)
5243                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5244         if (mxresp_sz == 0)
5245                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5246         /* Fore channel attributes */
5247         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5248         args->fc_attrs.max_resp_sz = mxresp_sz;
5249         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5250         args->fc_attrs.max_reqs = max_session_slots;
5251
5252         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5253                 "max_ops=%u max_reqs=%u\n",
5254                 __func__,
5255                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5256                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5257
5258         /* Back channel attributes */
5259         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5260         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5261         args->bc_attrs.max_resp_sz_cached = 0;
5262         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5263         args->bc_attrs.max_reqs = 1;
5264
5265         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5266                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5267                 __func__,
5268                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5269                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5270                 args->bc_attrs.max_reqs);
5271 }
5272
5273 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5274 {
5275         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5276         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5277
5278         if (rcvd->max_resp_sz > sent->max_resp_sz)
5279                 return -EINVAL;
5280         /*
5281          * Our requested max_ops is the minimum we need; we're not
5282          * prepared to break up compounds into smaller pieces than that.
5283          * So, no point even trying to continue if the server won't
5284          * cooperate:
5285          */
5286         if (rcvd->max_ops < sent->max_ops)
5287                 return -EINVAL;
5288         if (rcvd->max_reqs == 0)
5289                 return -EINVAL;
5290         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5291                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5292         return 0;
5293 }
5294
5295 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5296 {
5297         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5298         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5299
5300         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5301                 return -EINVAL;
5302         if (rcvd->max_resp_sz < sent->max_resp_sz)
5303                 return -EINVAL;
5304         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5305                 return -EINVAL;
5306         /* These would render the backchannel useless: */
5307         if (rcvd->max_ops != sent->max_ops)
5308                 return -EINVAL;
5309         if (rcvd->max_reqs != sent->max_reqs)
5310                 return -EINVAL;
5311         return 0;
5312 }
5313
5314 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5315                                      struct nfs4_session *session)
5316 {
5317         int ret;
5318
5319         ret = nfs4_verify_fore_channel_attrs(args, session);
5320         if (ret)
5321                 return ret;
5322         return nfs4_verify_back_channel_attrs(args, session);
5323 }
5324
5325 static int _nfs4_proc_create_session(struct nfs_client *clp)
5326 {
5327         struct nfs4_session *session = clp->cl_session;
5328         struct nfs41_create_session_args args = {
5329                 .client = clp,
5330                 .cb_program = NFS4_CALLBACK,
5331         };
5332         struct nfs41_create_session_res res = {
5333                 .client = clp,
5334         };
5335         struct rpc_message msg = {
5336                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5337                 .rpc_argp = &args,
5338                 .rpc_resp = &res,
5339         };
5340         int status;
5341
5342         nfs4_init_channel_attrs(&args);
5343         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5344
5345         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5346
5347         if (!status)
5348                 /* Verify the session's negotiated channel_attrs values */
5349                 status = nfs4_verify_channel_attrs(&args, session);
5350         if (!status) {
5351                 /* Increment the clientid slot sequence id */
5352                 clp->cl_seqid++;
5353         }
5354
5355         return status;
5356 }
5357
5358 /*
5359  * Issues a CREATE_SESSION operation to the server.
5360  * It is the responsibility of the caller to verify the session is
5361  * expired before calling this routine.
5362  */
5363 int nfs4_proc_create_session(struct nfs_client *clp)
5364 {
5365         int status;
5366         unsigned *ptr;
5367         struct nfs4_session *session = clp->cl_session;
5368
5369         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5370
5371         status = _nfs4_proc_create_session(clp);
5372         if (status)
5373                 goto out;
5374
5375         /* Init or reset the session slot tables */
5376         status = nfs4_setup_session_slot_tables(session);
5377         dprintk("slot table setup returned %d\n", status);
5378         if (status)
5379                 goto out;
5380
5381         ptr = (unsigned *)&session->sess_id.data[0];
5382         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5383                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5384 out:
5385         dprintk("<-- %s\n", __func__);
5386         return status;
5387 }
5388
5389 /*
5390  * Issue the over-the-wire RPC DESTROY_SESSION.
5391  * The caller must serialize access to this routine.
5392  */
5393 int nfs4_proc_destroy_session(struct nfs4_session *session)
5394 {
5395         int status = 0;
5396         struct rpc_message msg;
5397
5398         dprintk("--> nfs4_proc_destroy_session\n");
5399
5400         /* session is still being setup */
5401         if (session->clp->cl_cons_state != NFS_CS_READY)
5402                 return status;
5403
5404         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5405         msg.rpc_argp = session;
5406         msg.rpc_resp = NULL;
5407         msg.rpc_cred = NULL;
5408         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5409
5410         if (status)
5411                 printk(KERN_WARNING
5412                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5413                         "Session has been destroyed regardless...\n", status);
5414
5415         dprintk("<-- nfs4_proc_destroy_session\n");
5416         return status;
5417 }
5418
5419 int nfs4_init_session(struct nfs_server *server)
5420 {
5421         struct nfs_client *clp = server->nfs_client;
5422         struct nfs4_session *session;
5423         unsigned int rsize, wsize;
5424         int ret;
5425
5426         if (!nfs4_has_session(clp))
5427                 return 0;
5428
5429         session = clp->cl_session;
5430         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5431                 return 0;
5432
5433         rsize = server->rsize;
5434         if (rsize == 0)
5435                 rsize = NFS_MAX_FILE_IO_SIZE;
5436         wsize = server->wsize;
5437         if (wsize == 0)
5438                 wsize = NFS_MAX_FILE_IO_SIZE;
5439
5440         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5441         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5442
5443         ret = nfs4_recover_expired_lease(server);
5444         if (!ret)
5445                 ret = nfs4_check_client_ready(clp);
5446         return ret;
5447 }
5448
5449 int nfs4_init_ds_session(struct nfs_client *clp)
5450 {
5451         struct nfs4_session *session = clp->cl_session;
5452         int ret;
5453
5454         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5455                 return 0;
5456
5457         ret = nfs4_client_recover_expired_lease(clp);
5458         if (!ret)
5459                 /* Test for the DS role */
5460                 if (!is_ds_client(clp))
5461                         ret = -ENODEV;
5462         if (!ret)
5463                 ret = nfs4_check_client_ready(clp);
5464         return ret;
5465
5466 }
5467 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5468
5469
5470 /*
5471  * Renew the cl_session lease.
5472  */
5473 struct nfs4_sequence_data {
5474         struct nfs_client *clp;
5475         struct nfs4_sequence_args args;
5476         struct nfs4_sequence_res res;
5477 };
5478
5479 static void nfs41_sequence_release(void *data)
5480 {
5481         struct nfs4_sequence_data *calldata = data;
5482         struct nfs_client *clp = calldata->clp;
5483
5484         if (atomic_read(&clp->cl_count) > 1)
5485                 nfs4_schedule_state_renewal(clp);
5486         nfs_put_client(clp);
5487         kfree(calldata);
5488 }
5489
5490 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5491 {
5492         switch(task->tk_status) {
5493         case -NFS4ERR_DELAY:
5494                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5495                 return -EAGAIN;
5496         default:
5497                 nfs4_schedule_lease_recovery(clp);
5498         }
5499         return 0;
5500 }
5501
5502 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5503 {
5504         struct nfs4_sequence_data *calldata = data;
5505         struct nfs_client *clp = calldata->clp;
5506
5507         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5508                 return;
5509
5510         if (task->tk_status < 0) {
5511                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5512                 if (atomic_read(&clp->cl_count) == 1)
5513                         goto out;
5514
5515                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5516                         rpc_restart_call_prepare(task);
5517                         return;
5518                 }
5519         }
5520         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5521 out:
5522         dprintk("<-- %s\n", __func__);
5523 }
5524
5525 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5526 {
5527         struct nfs4_sequence_data *calldata = data;
5528         struct nfs_client *clp = calldata->clp;
5529         struct nfs4_sequence_args *args;
5530         struct nfs4_sequence_res *res;
5531
5532         args = task->tk_msg.rpc_argp;
5533         res = task->tk_msg.rpc_resp;
5534
5535         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5536                 return;
5537         rpc_call_start(task);
5538 }
5539
5540 static const struct rpc_call_ops nfs41_sequence_ops = {
5541         .rpc_call_done = nfs41_sequence_call_done,
5542         .rpc_call_prepare = nfs41_sequence_prepare,
5543         .rpc_release = nfs41_sequence_release,
5544 };
5545
5546 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5547 {
5548         struct nfs4_sequence_data *calldata;
5549         struct rpc_message msg = {
5550                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5551                 .rpc_cred = cred,
5552         };
5553         struct rpc_task_setup task_setup_data = {
5554                 .rpc_client = clp->cl_rpcclient,
5555                 .rpc_message = &msg,
5556                 .callback_ops = &nfs41_sequence_ops,
5557                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5558         };
5559
5560         if (!atomic_inc_not_zero(&clp->cl_count))
5561                 return ERR_PTR(-EIO);
5562         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5563         if (calldata == NULL) {
5564                 nfs_put_client(clp);
5565                 return ERR_PTR(-ENOMEM);
5566         }
5567         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5568         msg.rpc_argp = &calldata->args;
5569         msg.rpc_resp = &calldata->res;
5570         calldata->clp = clp;
5571         task_setup_data.callback_data = calldata;
5572
5573         return rpc_run_task(&task_setup_data);
5574 }
5575
5576 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5577 {
5578         struct rpc_task *task;
5579         int ret = 0;
5580
5581         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5582                 return 0;
5583         task = _nfs41_proc_sequence(clp, cred);
5584         if (IS_ERR(task))
5585                 ret = PTR_ERR(task);
5586         else
5587                 rpc_put_task_async(task);
5588         dprintk("<-- %s status=%d\n", __func__, ret);
5589         return ret;
5590 }
5591
5592 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5593 {
5594         struct rpc_task *task;
5595         int ret;
5596
5597         task = _nfs41_proc_sequence(clp, cred);
5598         if (IS_ERR(task)) {
5599                 ret = PTR_ERR(task);
5600                 goto out;
5601         }
5602         ret = rpc_wait_for_completion_task(task);
5603         if (!ret) {
5604                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5605
5606                 if (task->tk_status == 0)
5607                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5608                 ret = task->tk_status;
5609         }
5610         rpc_put_task(task);
5611 out:
5612         dprintk("<-- %s status=%d\n", __func__, ret);
5613         return ret;
5614 }
5615
5616 struct nfs4_reclaim_complete_data {
5617         struct nfs_client *clp;
5618         struct nfs41_reclaim_complete_args arg;
5619         struct nfs41_reclaim_complete_res res;
5620 };
5621
5622 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5623 {
5624         struct nfs4_reclaim_complete_data *calldata = data;
5625
5626         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5627         if (nfs41_setup_sequence(calldata->clp->cl_session,
5628                                 &calldata->arg.seq_args,
5629                                 &calldata->res.seq_res, task))
5630                 return;
5631
5632         rpc_call_start(task);
5633 }
5634
5635 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5636 {
5637         switch(task->tk_status) {
5638         case 0:
5639         case -NFS4ERR_COMPLETE_ALREADY:
5640         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5641                 break;
5642         case -NFS4ERR_DELAY:
5643                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5644                 /* fall through */
5645         case -NFS4ERR_RETRY_UNCACHED_REP:
5646                 return -EAGAIN;
5647         default:
5648                 nfs4_schedule_lease_recovery(clp);
5649         }
5650         return 0;
5651 }
5652
5653 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5654 {
5655         struct nfs4_reclaim_complete_data *calldata = data;
5656         struct nfs_client *clp = calldata->clp;
5657         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5658
5659         dprintk("--> %s\n", __func__);
5660         if (!nfs41_sequence_done(task, res))
5661                 return;
5662
5663         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5664                 rpc_restart_call_prepare(task);
5665                 return;
5666         }
5667         dprintk("<-- %s\n", __func__);
5668 }
5669
5670 static void nfs4_free_reclaim_complete_data(void *data)
5671 {
5672         struct nfs4_reclaim_complete_data *calldata = data;
5673
5674         kfree(calldata);
5675 }
5676
5677 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5678         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5679         .rpc_call_done = nfs4_reclaim_complete_done,
5680         .rpc_release = nfs4_free_reclaim_complete_data,
5681 };
5682
5683 /*
5684  * Issue a global reclaim complete.
5685  */
5686 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5687 {
5688         struct nfs4_reclaim_complete_data *calldata;
5689         struct rpc_task *task;
5690         struct rpc_message msg = {
5691                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5692         };
5693         struct rpc_task_setup task_setup_data = {
5694                 .rpc_client = clp->cl_rpcclient,
5695                 .rpc_message = &msg,
5696                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5697                 .flags = RPC_TASK_ASYNC,
5698         };
5699         int status = -ENOMEM;
5700
5701         dprintk("--> %s\n", __func__);
5702         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5703         if (calldata == NULL)
5704                 goto out;
5705         calldata->clp = clp;
5706         calldata->arg.one_fs = 0;
5707
5708         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5709         msg.rpc_argp = &calldata->arg;
5710         msg.rpc_resp = &calldata->res;
5711         task_setup_data.callback_data = calldata;
5712         task = rpc_run_task(&task_setup_data);
5713         if (IS_ERR(task)) {
5714                 status = PTR_ERR(task);
5715                 goto out;
5716         }
5717         status = nfs4_wait_for_completion_rpc_task(task);
5718         if (status == 0)
5719                 status = task->tk_status;
5720         rpc_put_task(task);
5721         return 0;
5722 out:
5723         dprintk("<-- %s status=%d\n", __func__, status);
5724         return status;
5725 }
5726
5727 static void
5728 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5729 {
5730         struct nfs4_layoutget *lgp = calldata;
5731         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5732
5733         dprintk("--> %s\n", __func__);
5734         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5735          * right now covering the LAYOUTGET we are about to send.
5736          * However, that is not so catastrophic, and there seems
5737          * to be no way to prevent it completely.
5738          */
5739         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5740                                 &lgp->res.seq_res, task))
5741                 return;
5742         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5743                                           NFS_I(lgp->args.inode)->layout,
5744                                           lgp->args.ctx->state)) {
5745                 rpc_exit(task, NFS4_OK);
5746                 return;
5747         }
5748         rpc_call_start(task);
5749 }
5750
5751 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5752 {
5753         struct nfs4_layoutget *lgp = calldata;
5754         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5755
5756         dprintk("--> %s\n", __func__);
5757
5758         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5759                 return;
5760
5761         switch (task->tk_status) {
5762         case 0:
5763                 break;
5764         case -NFS4ERR_LAYOUTTRYLATER:
5765         case -NFS4ERR_RECALLCONFLICT:
5766                 task->tk_status = -NFS4ERR_DELAY;
5767                 /* Fall through */
5768         default:
5769                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5770                         rpc_restart_call_prepare(task);
5771                         return;
5772                 }
5773         }
5774         dprintk("<-- %s\n", __func__);
5775 }
5776
5777 static void nfs4_layoutget_release(void *calldata)
5778 {
5779         struct nfs4_layoutget *lgp = calldata;
5780
5781         dprintk("--> %s\n", __func__);
5782         put_nfs_open_context(lgp->args.ctx);
5783         kfree(calldata);
5784         dprintk("<-- %s\n", __func__);
5785 }
5786
5787 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5788         .rpc_call_prepare = nfs4_layoutget_prepare,
5789         .rpc_call_done = nfs4_layoutget_done,
5790         .rpc_release = nfs4_layoutget_release,
5791 };
5792
5793 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5794 {
5795         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5796         struct rpc_task *task;
5797         struct rpc_message msg = {
5798                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5799                 .rpc_argp = &lgp->args,
5800                 .rpc_resp = &lgp->res,
5801         };
5802         struct rpc_task_setup task_setup_data = {
5803                 .rpc_client = server->client,
5804                 .rpc_message = &msg,
5805                 .callback_ops = &nfs4_layoutget_call_ops,
5806                 .callback_data = lgp,
5807                 .flags = RPC_TASK_ASYNC,
5808         };
5809         int status = 0;
5810
5811         dprintk("--> %s\n", __func__);
5812
5813         lgp->res.layoutp = &lgp->args.layout;
5814         lgp->res.seq_res.sr_slot = NULL;
5815         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5816         task = rpc_run_task(&task_setup_data);
5817         if (IS_ERR(task))
5818                 return PTR_ERR(task);
5819         status = nfs4_wait_for_completion_rpc_task(task);
5820         if (status == 0)
5821                 status = task->tk_status;
5822         if (status == 0)
5823                 status = pnfs_layout_process(lgp);
5824         rpc_put_task(task);
5825         dprintk("<-- %s status=%d\n", __func__, status);
5826         return status;
5827 }
5828
5829 static void
5830 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5831 {
5832         struct nfs4_layoutreturn *lrp = calldata;
5833
5834         dprintk("--> %s\n", __func__);
5835         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5836                                 &lrp->res.seq_res, task))
5837                 return;
5838         rpc_call_start(task);
5839 }
5840
5841 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5842 {
5843         struct nfs4_layoutreturn *lrp = calldata;
5844         struct nfs_server *server;
5845         struct pnfs_layout_hdr *lo = lrp->args.layout;
5846
5847         dprintk("--> %s\n", __func__);
5848
5849         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5850                 return;
5851
5852         server = NFS_SERVER(lrp->args.inode);
5853         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5854                 rpc_restart_call_prepare(task);
5855                 return;
5856         }
5857         spin_lock(&lo->plh_inode->i_lock);
5858         if (task->tk_status == 0) {
5859                 if (lrp->res.lrs_present) {
5860                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5861                 } else
5862                         BUG_ON(!list_empty(&lo->plh_segs));
5863         }
5864         lo->plh_block_lgets--;
5865         spin_unlock(&lo->plh_inode->i_lock);
5866         dprintk("<-- %s\n", __func__);
5867 }
5868
5869 static void nfs4_layoutreturn_release(void *calldata)
5870 {
5871         struct nfs4_layoutreturn *lrp = calldata;
5872
5873         dprintk("--> %s\n", __func__);
5874         put_layout_hdr(lrp->args.layout);
5875         kfree(calldata);
5876         dprintk("<-- %s\n", __func__);
5877 }
5878
5879 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5880         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5881         .rpc_call_done = nfs4_layoutreturn_done,
5882         .rpc_release = nfs4_layoutreturn_release,
5883 };
5884
5885 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5886 {
5887         struct rpc_task *task;
5888         struct rpc_message msg = {
5889                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5890                 .rpc_argp = &lrp->args,
5891                 .rpc_resp = &lrp->res,
5892         };
5893         struct rpc_task_setup task_setup_data = {
5894                 .rpc_client = lrp->clp->cl_rpcclient,
5895                 .rpc_message = &msg,
5896                 .callback_ops = &nfs4_layoutreturn_call_ops,
5897                 .callback_data = lrp,
5898         };
5899         int status;
5900
5901         dprintk("--> %s\n", __func__);
5902         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5903         task = rpc_run_task(&task_setup_data);
5904         if (IS_ERR(task))
5905                 return PTR_ERR(task);
5906         status = task->tk_status;
5907         dprintk("<-- %s status=%d\n", __func__, status);
5908         rpc_put_task(task);
5909         return status;
5910 }
5911
5912 /*
5913  * Retrieve the list of Data Server devices from the MDS.
5914  */
5915 static int _nfs4_getdevicelist(struct nfs_server *server,
5916                                     const struct nfs_fh *fh,
5917                                     struct pnfs_devicelist *devlist)
5918 {
5919         struct nfs4_getdevicelist_args args = {
5920                 .fh = fh,
5921                 .layoutclass = server->pnfs_curr_ld->id,
5922         };
5923         struct nfs4_getdevicelist_res res = {
5924                 .devlist = devlist,
5925         };
5926         struct rpc_message msg = {
5927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5928                 .rpc_argp = &args,
5929                 .rpc_resp = &res,
5930         };
5931         int status;
5932
5933         dprintk("--> %s\n", __func__);
5934         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5935                                 &res.seq_res, 0);
5936         dprintk("<-- %s status=%d\n", __func__, status);
5937         return status;
5938 }
5939
5940 int nfs4_proc_getdevicelist(struct nfs_server *server,
5941                             const struct nfs_fh *fh,
5942                             struct pnfs_devicelist *devlist)
5943 {
5944         struct nfs4_exception exception = { };
5945         int err;
5946
5947         do {
5948                 err = nfs4_handle_exception(server,
5949                                 _nfs4_getdevicelist(server, fh, devlist),
5950                                 &exception);
5951         } while (exception.retry);
5952
5953         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5954                 err, devlist->num_devs);
5955
5956         return err;
5957 }
5958 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5959
5960 static int
5961 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5962 {
5963         struct nfs4_getdeviceinfo_args args = {
5964                 .pdev = pdev,
5965         };
5966         struct nfs4_getdeviceinfo_res res = {
5967                 .pdev = pdev,
5968         };
5969         struct rpc_message msg = {
5970                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5971                 .rpc_argp = &args,
5972                 .rpc_resp = &res,
5973         };
5974         int status;
5975
5976         dprintk("--> %s\n", __func__);
5977         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5978         dprintk("<-- %s status=%d\n", __func__, status);
5979
5980         return status;
5981 }
5982
5983 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5984 {
5985         struct nfs4_exception exception = { };
5986         int err;
5987
5988         do {
5989                 err = nfs4_handle_exception(server,
5990                                         _nfs4_proc_getdeviceinfo(server, pdev),
5991                                         &exception);
5992         } while (exception.retry);
5993         return err;
5994 }
5995 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5996
5997 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5998 {
5999         struct nfs4_layoutcommit_data *data = calldata;
6000         struct nfs_server *server = NFS_SERVER(data->args.inode);
6001
6002         if (nfs4_setup_sequence(server, &data->args.seq_args,
6003                                 &data->res.seq_res, task))
6004                 return;
6005         rpc_call_start(task);
6006 }
6007
6008 static void
6009 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6010 {
6011         struct nfs4_layoutcommit_data *data = calldata;
6012         struct nfs_server *server = NFS_SERVER(data->args.inode);
6013
6014         if (!nfs4_sequence_done(task, &data->res.seq_res))
6015                 return;
6016
6017         switch (task->tk_status) { /* Just ignore these failures */
6018         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6019         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6020         case NFS4ERR_BADLAYOUT:     /* no layout */
6021         case NFS4ERR_GRACE:         /* loca_recalim always false */
6022                 task->tk_status = 0;
6023         }
6024
6025         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6026                 rpc_restart_call_prepare(task);
6027                 return;
6028         }
6029
6030         if (task->tk_status == 0)
6031                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6032                                                    data->res.fattr);
6033 }
6034
6035 static void nfs4_layoutcommit_release(void *calldata)
6036 {
6037         struct nfs4_layoutcommit_data *data = calldata;
6038         struct pnfs_layout_segment *lseg, *tmp;
6039         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6040
6041         pnfs_cleanup_layoutcommit(data);
6042         /* Matched by references in pnfs_set_layoutcommit */
6043         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6044                 list_del_init(&lseg->pls_lc_list);
6045                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6046                                        &lseg->pls_flags))
6047                         put_lseg(lseg);
6048         }
6049
6050         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6051         smp_mb__after_clear_bit();
6052         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6053
6054         put_rpccred(data->cred);
6055         kfree(data);
6056 }
6057
6058 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6059         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6060         .rpc_call_done = nfs4_layoutcommit_done,
6061         .rpc_release = nfs4_layoutcommit_release,
6062 };
6063
6064 int
6065 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6066 {
6067         struct rpc_message msg = {
6068                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6069                 .rpc_argp = &data->args,
6070                 .rpc_resp = &data->res,
6071                 .rpc_cred = data->cred,
6072         };
6073         struct rpc_task_setup task_setup_data = {
6074                 .task = &data->task,
6075                 .rpc_client = NFS_CLIENT(data->args.inode),
6076                 .rpc_message = &msg,
6077                 .callback_ops = &nfs4_layoutcommit_ops,
6078                 .callback_data = data,
6079                 .flags = RPC_TASK_ASYNC,
6080         };
6081         struct rpc_task *task;
6082         int status = 0;
6083
6084         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6085                 "lbw: %llu inode %lu\n",
6086                 data->task.tk_pid, sync,
6087                 data->args.lastbytewritten,
6088                 data->args.inode->i_ino);
6089
6090         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6091         task = rpc_run_task(&task_setup_data);
6092         if (IS_ERR(task))
6093                 return PTR_ERR(task);
6094         if (sync == false)
6095                 goto out;
6096         status = nfs4_wait_for_completion_rpc_task(task);
6097         if (status != 0)
6098                 goto out;
6099         status = task->tk_status;
6100 out:
6101         dprintk("%s: status %d\n", __func__, status);
6102         rpc_put_task(task);
6103         return status;
6104 }
6105
6106 static int
6107 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6108                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6109 {
6110         struct nfs41_secinfo_no_name_args args = {
6111                 .style = SECINFO_STYLE_CURRENT_FH,
6112         };
6113         struct nfs4_secinfo_res res = {
6114                 .flavors = flavors,
6115         };
6116         struct rpc_message msg = {
6117                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6118                 .rpc_argp = &args,
6119                 .rpc_resp = &res,
6120         };
6121         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6122 }
6123
6124 static int
6125 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6126                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6127 {
6128         struct nfs4_exception exception = { };
6129         int err;
6130         do {
6131                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6132                 switch (err) {
6133                 case 0:
6134                 case -NFS4ERR_WRONGSEC:
6135                 case -NFS4ERR_NOTSUPP:
6136                         break;
6137                 default:
6138                         err = nfs4_handle_exception(server, err, &exception);
6139                 }
6140         } while (exception.retry);
6141         return err;
6142 }
6143
6144 static int
6145 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6146                     struct nfs_fsinfo *info)
6147 {
6148         int err;
6149         struct page *page;
6150         rpc_authflavor_t flavor;
6151         struct nfs4_secinfo_flavors *flavors;
6152
6153         page = alloc_page(GFP_KERNEL);
6154         if (!page) {
6155                 err = -ENOMEM;
6156                 goto out;
6157         }
6158
6159         flavors = page_address(page);
6160         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6161
6162         /*
6163          * Fall back on "guess and check" method if
6164          * the server doesn't support SECINFO_NO_NAME
6165          */
6166         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6167                 err = nfs4_find_root_sec(server, fhandle, info);
6168                 goto out_freepage;
6169         }
6170         if (err)
6171                 goto out_freepage;
6172
6173         flavor = nfs_find_best_sec(flavors);
6174         if (err == 0)
6175                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6176
6177 out_freepage:
6178         put_page(page);
6179         if (err == -EACCES)
6180                 return -EPERM;
6181 out:
6182         return err;
6183 }
6184
6185 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6186 {
6187         int status;
6188         struct nfs41_test_stateid_args args = {
6189                 .stateid = stateid,
6190         };
6191         struct nfs41_test_stateid_res res;
6192         struct rpc_message msg = {
6193                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6194                 .rpc_argp = &args,
6195                 .rpc_resp = &res,
6196         };
6197
6198         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6199         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6200
6201         if (status == NFS_OK)
6202                 return res.status;
6203         return status;
6204 }
6205
6206 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6207 {
6208         struct nfs4_exception exception = { };
6209         int err;
6210         do {
6211                 err = nfs4_handle_exception(server,
6212                                 _nfs41_test_stateid(server, stateid),
6213                                 &exception);
6214         } while (exception.retry);
6215         return err;
6216 }
6217
6218 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6219 {
6220         struct nfs41_free_stateid_args args = {
6221                 .stateid = stateid,
6222         };
6223         struct nfs41_free_stateid_res res;
6224         struct rpc_message msg = {
6225                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6226                 .rpc_argp = &args,
6227                 .rpc_resp = &res,
6228         };
6229
6230         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6231         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6232 }
6233
6234 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6235 {
6236         struct nfs4_exception exception = { };
6237         int err;
6238         do {
6239                 err = nfs4_handle_exception(server,
6240                                 _nfs4_free_stateid(server, stateid),
6241                                 &exception);
6242         } while (exception.retry);
6243         return err;
6244 }
6245 #endif /* CONFIG_NFS_V4_1 */
6246
6247 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6248         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6249         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6250         .recover_open   = nfs4_open_reclaim,
6251         .recover_lock   = nfs4_lock_reclaim,
6252         .establish_clid = nfs4_init_clientid,
6253         .get_clid_cred  = nfs4_get_setclientid_cred,
6254 };
6255
6256 #if defined(CONFIG_NFS_V4_1)
6257 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6258         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6259         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6260         .recover_open   = nfs4_open_reclaim,
6261         .recover_lock   = nfs4_lock_reclaim,
6262         .establish_clid = nfs41_init_clientid,
6263         .get_clid_cred  = nfs4_get_exchange_id_cred,
6264         .reclaim_complete = nfs41_proc_reclaim_complete,
6265 };
6266 #endif /* CONFIG_NFS_V4_1 */
6267
6268 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6269         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6270         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6271         .recover_open   = nfs4_open_expired,
6272         .recover_lock   = nfs4_lock_expired,
6273         .establish_clid = nfs4_init_clientid,
6274         .get_clid_cred  = nfs4_get_setclientid_cred,
6275 };
6276
6277 #if defined(CONFIG_NFS_V4_1)
6278 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6279         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6280         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6281         .recover_open   = nfs41_open_expired,
6282         .recover_lock   = nfs41_lock_expired,
6283         .establish_clid = nfs41_init_clientid,
6284         .get_clid_cred  = nfs4_get_exchange_id_cred,
6285 };
6286 #endif /* CONFIG_NFS_V4_1 */
6287
6288 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6289         .sched_state_renewal = nfs4_proc_async_renew,
6290         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6291         .renew_lease = nfs4_proc_renew,
6292 };
6293
6294 #if defined(CONFIG_NFS_V4_1)
6295 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6296         .sched_state_renewal = nfs41_proc_async_sequence,
6297         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6298         .renew_lease = nfs4_proc_sequence,
6299 };
6300 #endif
6301
6302 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6303         .minor_version = 0,
6304         .call_sync = _nfs4_call_sync,
6305         .validate_stateid = nfs4_validate_delegation_stateid,
6306         .find_root_sec = nfs4_find_root_sec,
6307         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6308         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6309         .state_renewal_ops = &nfs40_state_renewal_ops,
6310 };
6311
6312 #if defined(CONFIG_NFS_V4_1)
6313 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6314         .minor_version = 1,
6315         .call_sync = _nfs4_call_sync_session,
6316         .validate_stateid = nfs41_validate_delegation_stateid,
6317         .find_root_sec = nfs41_find_root_sec,
6318         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6319         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6320         .state_renewal_ops = &nfs41_state_renewal_ops,
6321 };
6322 #endif
6323
6324 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6325         [0] = &nfs_v4_0_minor_ops,
6326 #if defined(CONFIG_NFS_V4_1)
6327         [1] = &nfs_v4_1_minor_ops,
6328 #endif
6329 };
6330
6331 static const struct inode_operations nfs4_file_inode_operations = {
6332         .permission     = nfs_permission,
6333         .getattr        = nfs_getattr,
6334         .setattr        = nfs_setattr,
6335         .getxattr       = generic_getxattr,
6336         .setxattr       = generic_setxattr,
6337         .listxattr      = generic_listxattr,
6338         .removexattr    = generic_removexattr,
6339 };
6340
6341 const struct nfs_rpc_ops nfs_v4_clientops = {
6342         .version        = 4,                    /* protocol version */
6343         .dentry_ops     = &nfs4_dentry_operations,
6344         .dir_inode_ops  = &nfs4_dir_inode_operations,
6345         .file_inode_ops = &nfs4_file_inode_operations,
6346         .file_ops       = &nfs4_file_operations,
6347         .getroot        = nfs4_proc_get_root,
6348         .getattr        = nfs4_proc_getattr,
6349         .setattr        = nfs4_proc_setattr,
6350         .lookup         = nfs4_proc_lookup,
6351         .access         = nfs4_proc_access,
6352         .readlink       = nfs4_proc_readlink,
6353         .create         = nfs4_proc_create,
6354         .remove         = nfs4_proc_remove,
6355         .unlink_setup   = nfs4_proc_unlink_setup,
6356         .unlink_done    = nfs4_proc_unlink_done,
6357         .rename         = nfs4_proc_rename,
6358         .rename_setup   = nfs4_proc_rename_setup,
6359         .rename_done    = nfs4_proc_rename_done,
6360         .link           = nfs4_proc_link,
6361         .symlink        = nfs4_proc_symlink,
6362         .mkdir          = nfs4_proc_mkdir,
6363         .rmdir          = nfs4_proc_remove,
6364         .readdir        = nfs4_proc_readdir,
6365         .mknod          = nfs4_proc_mknod,
6366         .statfs         = nfs4_proc_statfs,
6367         .fsinfo         = nfs4_proc_fsinfo,
6368         .pathconf       = nfs4_proc_pathconf,
6369         .set_capabilities = nfs4_server_capabilities,
6370         .decode_dirent  = nfs4_decode_dirent,
6371         .read_setup     = nfs4_proc_read_setup,
6372         .read_done      = nfs4_read_done,
6373         .write_setup    = nfs4_proc_write_setup,
6374         .write_done     = nfs4_write_done,
6375         .commit_setup   = nfs4_proc_commit_setup,
6376         .commit_done    = nfs4_commit_done,
6377         .lock           = nfs4_proc_lock,
6378         .clear_acl_cache = nfs4_zap_acl_attr,
6379         .close_context  = nfs4_close_context,
6380         .open_context   = nfs4_atomic_open,
6381         .init_client    = nfs4_init_client,
6382         .secinfo        = nfs4_proc_secinfo,
6383 };
6384
6385 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6386         .prefix = XATTR_NAME_NFSV4_ACL,
6387         .list   = nfs4_xattr_list_nfs4_acl,
6388         .get    = nfs4_xattr_get_nfs4_acl,
6389         .set    = nfs4_xattr_set_nfs4_acl,
6390 };
6391
6392 const struct xattr_handler *nfs4_xattr_handlers[] = {
6393         &nfs4_xattr_nfs4_acl_handler,
6394         NULL
6395 };
6396
6397 module_param(max_session_slots, ushort, 0644);
6398 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6399                 "requests the client will negotiate");
6400
6401 /*
6402  * Local variables:
6403  *  c-basic-offset: 8
6404  * End:
6405  */