Pileus Git - ~andy/linux/blob - 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         dprintk("%s: begin!\n", __func__);
1987         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1988                 return;
1989         /* hmm. we are done with the inode, and in the process of freeing
1990          * the state_owner. we keep this around to process errors
1991          */
1992         switch (task->tk_status) {
1993                 case 0:
1994                         if (calldata->roc)
1995                                 pnfs_roc_set_barrier(state->inode,
1996                                                      calldata->roc_barrier);
1997                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1998                         renew_lease(server, calldata->timestamp);
1999                         nfs4_close_clear_stateid_flags(state,
2000                                         calldata->arg.fmode);
2001                         break;
2002                 case -NFS4ERR_STALE_STATEID:
2003                 case -NFS4ERR_OLD_STATEID:
2004                 case -NFS4ERR_BAD_STATEID:
2005                 case -NFS4ERR_EXPIRED:
2006                         if (calldata->arg.fmode == 0)
2007                                 break;
2008                 default:
2009                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2010                                 rpc_restart_call_prepare(task);
2011         }
2012         nfs_release_seqid(calldata->arg.seqid);
2013         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2014         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2015 }
2016
2017 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2018 {
2019         struct nfs4_closedata *calldata = data;
2020         struct nfs4_state *state = calldata->state;
2021         int call_close = 0;
2022
2023         dprintk("%s: begin!\n", __func__);
2024         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2025                 return;
2026
2027         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2028         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2029         spin_lock(&state->owner->so_lock);
2030         /* Calculate the change in open mode */
2031         if (state->n_rdwr == 0) {
2032                 if (state->n_rdonly == 0) {
2033                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2034                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2035                         calldata->arg.fmode &= ~FMODE_READ;
2036                 }
2037                 if (state->n_wronly == 0) {
2038                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2039                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2040                         calldata->arg.fmode &= ~FMODE_WRITE;
2041                 }
2042         }
2043         spin_unlock(&state->owner->so_lock);
2044
2045         if (!call_close) {
2046                 /* Note: exit _without_ calling nfs4_close_done */
2047                 task->tk_action = NULL;
2048                 goto out;
2049         }
2050
2051         if (calldata->arg.fmode == 0) {
2052                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2053                 if (calldata->roc &&
2054                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2055                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2056                                      task, NULL);
2057                         goto out;
2058                 }
2059         }
2060
2061         nfs_fattr_init(calldata->res.fattr);
2062         calldata->timestamp = jiffies;
2063         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2064                                 &calldata->arg.seq_args,
2065                                 &calldata->res.seq_res,
2066                                 task))
2067                 goto out;
2068         rpc_call_start(task);
2069 out:
2070         dprintk("%s: done!\n", __func__);
2071 }
2072
2073 static const struct rpc_call_ops nfs4_close_ops = {
2074         .rpc_call_prepare = nfs4_close_prepare,
2075         .rpc_call_done = nfs4_close_done,
2076         .rpc_release = nfs4_free_closedata,
2077 };
2078
2079 /*
2080  * It is possible for data to be read/written from a mem-mapped file
2081  * after the sys_close call (which hits the vfs layer as a flush).
2082  * This means that we can't safely call nfsv4 close on a file until
2083  * the inode is cleared. This in turn means that we are not good
2084  * NFSv4 citizens - we do not indicate to the server to update the file's
2085  * share state even when we are done with one of the three share
2086  * stateid's in the inode.
2087  *
2088  * NOTE: Caller must be holding the sp->so_owner semaphore!
2089  */
2090 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2091 {
2092         struct nfs_server *server = NFS_SERVER(state->inode);
2093         struct nfs4_closedata *calldata;
2094         struct nfs4_state_owner *sp = state->owner;
2095         struct rpc_task *task;
2096         struct rpc_message msg = {
2097                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2098                 .rpc_cred = state->owner->so_cred,
2099         };
2100         struct rpc_task_setup task_setup_data = {
2101                 .rpc_client = server->client,
2102                 .rpc_message = &msg,
2103                 .callback_ops = &nfs4_close_ops,
2104                 .workqueue = nfsiod_workqueue,
2105                 .flags = RPC_TASK_ASYNC,
2106         };
2107         int status = -ENOMEM;
2108
2109         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2110         if (calldata == NULL)
2111                 goto out;
2112         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2113         calldata->inode = state->inode;
2114         calldata->state = state;
2115         calldata->arg.fh = NFS_FH(state->inode);
2116         calldata->arg.stateid = &state->open_stateid;
2117         /* Serialization for the sequence id */
2118         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2119         if (calldata->arg.seqid == NULL)
2120                 goto out_free_calldata;
2121         calldata->arg.fmode = 0;
2122         calldata->arg.bitmask = server->cache_consistency_bitmask;
2123         calldata->res.fattr = &calldata->fattr;
2124         calldata->res.seqid = calldata->arg.seqid;
2125         calldata->res.server = server;
2126         calldata->roc = roc;
2127         nfs_sb_active(calldata->inode->i_sb);
2128
2129         msg.rpc_argp = &calldata->arg;
2130         msg.rpc_resp = &calldata->res;
2131         task_setup_data.callback_data = calldata;
2132         task = rpc_run_task(&task_setup_data);
2133         if (IS_ERR(task))
2134                 return PTR_ERR(task);
2135         status = 0;
2136         if (wait)
2137                 status = rpc_wait_for_completion_task(task);
2138         rpc_put_task(task);
2139         return status;
2140 out_free_calldata:
2141         kfree(calldata);
2142 out:
2143         if (roc)
2144                 pnfs_roc_release(state->inode);
2145         nfs4_put_open_state(state);
2146         nfs4_put_state_owner(sp);
2147         return status;
2148 }
2149
2150 static struct inode *
2151 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2152 {
2153         struct nfs4_state *state;
2154
2155         /* Protect against concurrent sillydeletes */
2156         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2157         if (IS_ERR(state))
2158                 return ERR_CAST(state);
2159         ctx->state = state;
2160         return igrab(state->inode);
2161 }
2162
2163 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2164 {
2165         if (ctx->state == NULL)
2166                 return;
2167         if (is_sync)
2168                 nfs4_close_sync(ctx->state, ctx->mode);
2169         else
2170                 nfs4_close_state(ctx->state, ctx->mode);
2171 }
2172
2173 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2174 {
2175         struct nfs4_server_caps_arg args = {
2176                 .fhandle = fhandle,
2177         };
2178         struct nfs4_server_caps_res res = {};
2179         struct rpc_message msg = {
2180                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2181                 .rpc_argp = &args,
2182                 .rpc_resp = &res,
2183         };
2184         int status;
2185
2186         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2187         if (status == 0) {
2188                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2189                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2190                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2191                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2192                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2193                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2194                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2195                         server->caps |= NFS_CAP_ACLS;
2196                 if (res.has_links != 0)
2197                         server->caps |= NFS_CAP_HARDLINKS;
2198                 if (res.has_symlinks != 0)
2199                         server->caps |= NFS_CAP_SYMLINKS;
2200                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2201                         server->caps |= NFS_CAP_FILEID;
2202                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2203                         server->caps |= NFS_CAP_MODE;
2204                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2205                         server->caps |= NFS_CAP_NLINK;
2206                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2207                         server->caps |= NFS_CAP_OWNER;
2208                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2209                         server->caps |= NFS_CAP_OWNER_GROUP;
2210                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2211                         server->caps |= NFS_CAP_ATIME;
2212                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2213                         server->caps |= NFS_CAP_CTIME;
2214                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2215                         server->caps |= NFS_CAP_MTIME;
2216
2217                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2218                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2219                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2220                 server->acl_bitmask = res.acl_bitmask;
2221         }
2222
2223         return status;
2224 }
2225
2226 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2227 {
2228         struct nfs4_exception exception = { };
2229         int err;
2230         do {
2231                 err = nfs4_handle_exception(server,
2232                                 _nfs4_server_capabilities(server, fhandle),
2233                                 &exception);
2234         } while (exception.retry);
2235         return err;
2236 }
2237
2238 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2239                 struct nfs_fsinfo *info)
2240 {
2241         struct nfs4_lookup_root_arg args = {
2242                 .bitmask = nfs4_fattr_bitmap,
2243         };
2244         struct nfs4_lookup_res res = {
2245                 .server = server,
2246                 .fattr = info->fattr,
2247                 .fh = fhandle,
2248         };
2249         struct rpc_message msg = {
2250                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2251                 .rpc_argp = &args,
2252                 .rpc_resp = &res,
2253         };
2254
2255         nfs_fattr_init(info->fattr);
2256         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2257 }
2258
2259 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2260                 struct nfs_fsinfo *info)
2261 {
2262         struct nfs4_exception exception = { };
2263         int err;
2264         do {
2265                 err = _nfs4_lookup_root(server, fhandle, info);
2266                 switch (err) {
2267                 case 0:
2268                 case -NFS4ERR_WRONGSEC:
2269                         break;
2270                 default:
2271                         err = nfs4_handle_exception(server, err, &exception);
2272                 }
2273         } while (exception.retry);
2274         return err;
2275 }
2276
2277 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2278                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2279 {
2280         struct rpc_auth *auth;
2281         int ret;
2282
2283         auth = rpcauth_create(flavor, server->client);
2284         if (!auth) {
2285                 ret = -EIO;
2286                 goto out;
2287         }
2288         ret = nfs4_lookup_root(server, fhandle, info);
2289 out:
2290         return ret;
2291 }
2292
2293 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2294                               struct nfs_fsinfo *info)
2295 {
2296         int i, len, status = 0;
2297         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2298
2299         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2300         flav_array[len] = RPC_AUTH_NULL;
2301         len += 1;
2302
2303         for (i = 0; i < len; i++) {
2304                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2305                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2306                         continue;
2307                 break;
2308         }
2309         /*
2310          * -EACCESS could mean that the user doesn't have correct permissions
2311          * to access the mount.  It could also mean that we tried to mount
2312          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2313          * existing mount programs don't handle -EACCES very well so it should
2314          * be mapped to -EPERM instead.
2315          */
2316         if (status == -EACCES)
2317                 status = -EPERM;
2318         return status;
2319 }
2320
2321 /*
2322  * get the file handle for the "/" directory on the server
2323  */
2324 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2325                               struct nfs_fsinfo *info)
2326 {
2327         int minor_version = server->nfs_client->cl_minorversion;
2328         int status = nfs4_lookup_root(server, fhandle, info);
2329         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2330                 /*
2331                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2332                  * by nfs4_map_errors() as this function exits.
2333                  */
2334                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2335         if (status == 0)
2336                 status = nfs4_server_capabilities(server, fhandle);
2337         if (status == 0)
2338                 status = nfs4_do_fsinfo(server, fhandle, info);
2339         return nfs4_map_errors(status);
2340 }
2341
2342 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
2343 /*
2344  * Get locations and (maybe) other attributes of a referral.
2345  * Note that we'll actually follow the referral later when
2346  * we detect fsid mismatch in inode revalidation
2347  */
2348 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2349                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2350 {
2351         int status = -ENOMEM;
2352         struct page *page = NULL;
2353         struct nfs4_fs_locations *locations = NULL;
2354
2355         page = alloc_page(GFP_KERNEL);
2356         if (page == NULL)
2357                 goto out;
2358         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2359         if (locations == NULL)
2360                 goto out;
2361
2362         status = nfs4_proc_fs_locations(dir, name, locations, page);
2363         if (status != 0)
2364                 goto out;
2365         /* Make sure server returned a different fsid for the referral */
2366         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2367                 dprintk("%s: server did not return a different fsid for"
2368                         " a referral at %s\n", __func__, name->name);
2369                 status = -EIO;
2370                 goto out;
2371         }
2372         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2373         nfs_fixup_referral_attributes(&locations->fattr);
2374
2375         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2376         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2377         memset(fhandle, 0, sizeof(struct nfs_fh));
2378 out:
2379         if (page)
2380                 __free_page(page);
2381         kfree(locations);
2382         return status;
2383 }
2384
2385 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2386 {
2387         struct nfs4_getattr_arg args = {
2388                 .fh = fhandle,
2389                 .bitmask = server->attr_bitmask,
2390         };
2391         struct nfs4_getattr_res res = {
2392                 .fattr = fattr,
2393                 .server = server,
2394         };
2395         struct rpc_message msg = {
2396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2397                 .rpc_argp = &args,
2398                 .rpc_resp = &res,
2399         };
2400
2401         nfs_fattr_init(fattr);
2402         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2403 }
2404
2405 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2406 {
2407         struct nfs4_exception exception = { };
2408         int err;
2409         do {
2410                 err = nfs4_handle_exception(server,
2411                                 _nfs4_proc_getattr(server, fhandle, fattr),
2412                                 &exception);
2413         } while (exception.retry);
2414         return err;
2415 }
2416
2417 /*
2418  * The file is not closed if it is opened due to the a request to change
2419  * the size of the file. The open call will not be needed once the
2420  * VFS layer lookup-intents are implemented.
2421  *
2422  * Close is called when the inode is destroyed.
2423  * If we haven't opened the file for O_WRONLY, we
2424  * need to in the size_change case to obtain a stateid.
2425  *
2426  * Got race?
2427  * Because OPEN is always done by name in nfsv4, it is
2428  * possible that we opened a different file by the same
2429  * name.  We can recognize this race condition, but we
2430  * can't do anything about it besides returning an error.
2431  *
2432  * This will be fixed with VFS changes (lookup-intent).
2433  */
2434 static int
2435 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2436                   struct iattr *sattr)
2437 {
2438         struct inode *inode = dentry->d_inode;
2439         struct rpc_cred *cred = NULL;
2440         struct nfs4_state *state = NULL;
2441         int status;
2442
2443         if (pnfs_ld_layoutret_on_setattr(inode))
2444                 pnfs_return_layout(inode);
2445
2446         nfs_fattr_init(fattr);
2447
2448         /* Search for an existing open(O_WRITE) file */
2449         if (sattr->ia_valid & ATTR_FILE) {
2450                 struct nfs_open_context *ctx;
2451
2452                 ctx = nfs_file_open_context(sattr->ia_file);
2453                 if (ctx) {
2454                         cred = ctx->cred;
2455                         state = ctx->state;
2456                 }
2457         }
2458
2459         /* Deal with open(O_TRUNC) */
2460         if (sattr->ia_valid & ATTR_OPEN)
2461                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2462
2463         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2464         if (status == 0)
2465                 nfs_setattr_update_inode(inode, sattr);
2466         return status;
2467 }
2468
2469 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2470                 const struct qstr *name, struct nfs_fh *fhandle,
2471                 struct nfs_fattr *fattr)
2472 {
2473         struct nfs_server *server = NFS_SERVER(dir);
2474         int                    status;
2475         struct nfs4_lookup_arg args = {
2476                 .bitmask = server->attr_bitmask,
2477                 .dir_fh = NFS_FH(dir),
2478                 .name = name,
2479         };
2480         struct nfs4_lookup_res res = {
2481                 .server = server,
2482                 .fattr = fattr,
2483                 .fh = fhandle,
2484         };
2485         struct rpc_message msg = {
2486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2487                 .rpc_argp = &args,
2488                 .rpc_resp = &res,
2489         };
2490
2491         nfs_fattr_init(fattr);
2492
2493         dprintk("NFS call  lookup %s\n", name->name);
2494         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2495         dprintk("NFS reply lookup: %d\n", status);
2496         return status;
2497 }
2498
2499 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2500 {
2501         memset(fh, 0, sizeof(struct nfs_fh));
2502         fattr->fsid.major = 1;
2503         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2504                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2505         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2506         fattr->nlink = 2;
2507 }
2508
2509 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2510                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2511 {
2512         struct nfs4_exception exception = { };
2513         int err;
2514         do {
2515                 int status;
2516
2517                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2518                 switch (status) {
2519                 case -NFS4ERR_BADNAME:
2520                         return -ENOENT;
2521                 case -NFS4ERR_MOVED:
2522                         return nfs4_get_referral(dir, name, fattr, fhandle);
2523                 case -NFS4ERR_WRONGSEC:
2524                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2525                 }
2526                 err = nfs4_handle_exception(NFS_SERVER(dir),
2527                                 status, &exception);
2528         } while (exception.retry);
2529         return err;
2530 }
2531
2532 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2533 {
2534         struct nfs_server *server = NFS_SERVER(inode);
2535         struct nfs4_accessargs args = {
2536                 .fh = NFS_FH(inode),
2537                 .bitmask = server->cache_consistency_bitmask,
2538         };
2539         struct nfs4_accessres res = {
2540                 .server = server,
2541         };
2542         struct rpc_message msg = {
2543                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2544                 .rpc_argp = &args,
2545                 .rpc_resp = &res,
2546                 .rpc_cred = entry->cred,
2547         };
2548         int mode = entry->mask;
2549         int status;
2550
2551         /*
2552          * Determine which access bits we want to ask for...
2553          */
2554         if (mode & MAY_READ)
2555                 args.access |= NFS4_ACCESS_READ;
2556         if (S_ISDIR(inode->i_mode)) {
2557                 if (mode & MAY_WRITE)
2558                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2559                 if (mode & MAY_EXEC)
2560                         args.access |= NFS4_ACCESS_LOOKUP;
2561         } else {
2562                 if (mode & MAY_WRITE)
2563                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2564                 if (mode & MAY_EXEC)
2565                         args.access |= NFS4_ACCESS_EXECUTE;
2566         }
2567
2568         res.fattr = nfs_alloc_fattr();
2569         if (res.fattr == NULL)
2570                 return -ENOMEM;
2571
2572         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2573         if (!status) {
2574                 entry->mask = 0;
2575                 if (res.access & NFS4_ACCESS_READ)
2576                         entry->mask |= MAY_READ;
2577                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2578                         entry->mask |= MAY_WRITE;
2579                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2580                         entry->mask |= MAY_EXEC;
2581                 nfs_refresh_inode(inode, res.fattr);
2582         }
2583         nfs_free_fattr(res.fattr);
2584         return status;
2585 }
2586
2587 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2588 {
2589         struct nfs4_exception exception = { };
2590         int err;
2591         do {
2592                 err = nfs4_handle_exception(NFS_SERVER(inode),
2593                                 _nfs4_proc_access(inode, entry),
2594                                 &exception);
2595         } while (exception.retry);
2596         return err;
2597 }
2598
2599 /*
2600  * TODO: For the time being, we don't try to get any attributes
2601  * along with any of the zero-copy operations READ, READDIR,
2602  * READLINK, WRITE.
2603  *
2604  * In the case of the first three, we want to put the GETATTR
2605  * after the read-type operation -- this is because it is hard
2606  * to predict the length of a GETATTR response in v4, and thus
2607  * align the READ data correctly.  This means that the GETATTR
2608  * may end up partially falling into the page cache, and we should
2609  * shift it into the 'tail' of the xdr_buf before processing.
2610  * To do this efficiently, we need to know the total length
2611  * of data received, which doesn't seem to be available outside
2612  * of the RPC layer.
2613  *
2614  * In the case of WRITE, we also want to put the GETATTR after
2615  * the operation -- in this case because we want to make sure
2616  * we get the post-operation mtime and size.  This means that
2617  * we can't use xdr_encode_pages() as written: we need a variant
2618  * of it which would leave room in the 'tail' iovec.
2619  *
2620  * Both of these changes to the XDR layer would in fact be quite
2621  * minor, but I decided to leave them for a subsequent patch.
2622  */
2623 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2624                 unsigned int pgbase, unsigned int pglen)
2625 {
2626         struct nfs4_readlink args = {
2627                 .fh       = NFS_FH(inode),
2628                 .pgbase   = pgbase,
2629                 .pglen    = pglen,
2630                 .pages    = &page,
2631         };
2632         struct nfs4_readlink_res res;
2633         struct rpc_message msg = {
2634                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2635                 .rpc_argp = &args,
2636                 .rpc_resp = &res,
2637         };
2638
2639         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2640 }
2641
2642 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2643                 unsigned int pgbase, unsigned int pglen)
2644 {
2645         struct nfs4_exception exception = { };
2646         int err;
2647         do {
2648                 err = nfs4_handle_exception(NFS_SERVER(inode),
2649                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2650                                 &exception);
2651         } while (exception.retry);
2652         return err;
2653 }
2654
2655 /*
2656  * Got race?
2657  * We will need to arrange for the VFS layer to provide an atomic open.
2658  * Until then, this create/open method is prone to inefficiency and race
2659  * conditions due to the lookup, create, and open VFS calls from sys_open()
2660  * placed on the wire.
2661  *
2662  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2663  * The file will be opened again in the subsequent VFS open call
2664  * (nfs4_proc_file_open).
2665  *
2666  * The open for read will just hang around to be used by any process that
2667  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2668  */
2669
2670 static int
2671 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2672                  int flags, struct nfs_open_context *ctx)
2673 {
2674         struct dentry *de = dentry;
2675         struct nfs4_state *state;
2676         struct rpc_cred *cred = NULL;
2677         fmode_t fmode = 0;
2678         int status = 0;
2679
2680         if (ctx != NULL) {
2681                 cred = ctx->cred;
2682                 de = ctx->dentry;
2683                 fmode = ctx->mode;
2684         }
2685         sattr->ia_mode &= ~current_umask();
2686         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2687         d_drop(dentry);
2688         if (IS_ERR(state)) {
2689                 status = PTR_ERR(state);
2690                 goto out;
2691         }
2692         d_add(dentry, igrab(state->inode));
2693         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2694         if (ctx != NULL)
2695                 ctx->state = state;
2696         else
2697                 nfs4_close_sync(state, fmode);
2698 out:
2699         return status;
2700 }
2701
2702 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2703 {
2704         struct nfs_server *server = NFS_SERVER(dir);
2705         struct nfs_removeargs args = {
2706                 .fh = NFS_FH(dir),
2707                 .name.len = name->len,
2708                 .name.name = name->name,
2709                 .bitmask = server->attr_bitmask,
2710         };
2711         struct nfs_removeres res = {
2712                 .server = server,
2713         };
2714         struct rpc_message msg = {
2715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2716                 .rpc_argp = &args,
2717                 .rpc_resp = &res,
2718         };
2719         int status = -ENOMEM;
2720
2721         res.dir_attr = nfs_alloc_fattr();
2722         if (res.dir_attr == NULL)
2723                 goto out;
2724
2725         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2726         if (status == 0) {
2727                 update_changeattr(dir, &res.cinfo);
2728                 nfs_post_op_update_inode(dir, res.dir_attr);
2729         }
2730         nfs_free_fattr(res.dir_attr);
2731 out:
2732         return status;
2733 }
2734
2735 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2736 {
2737         struct nfs4_exception exception = { };
2738         int err;
2739         do {
2740                 err = nfs4_handle_exception(NFS_SERVER(dir),
2741                                 _nfs4_proc_remove(dir, name),
2742                                 &exception);
2743         } while (exception.retry);
2744         return err;
2745 }
2746
2747 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2748 {
2749         struct nfs_server *server = NFS_SERVER(dir);
2750         struct nfs_removeargs *args = msg->rpc_argp;
2751         struct nfs_removeres *res = msg->rpc_resp;
2752
2753         args->bitmask = server->cache_consistency_bitmask;
2754         res->server = server;
2755         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2756         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2757 }
2758
2759 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2760 {
2761         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2762
2763         if (!nfs4_sequence_done(task, &res->seq_res))
2764                 return 0;
2765         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2766                 return 0;
2767         update_changeattr(dir, &res->cinfo);
2768         nfs_post_op_update_inode(dir, res->dir_attr);
2769         return 1;
2770 }
2771
2772 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2773 {
2774         struct nfs_server *server = NFS_SERVER(dir);
2775         struct nfs_renameargs *arg = msg->rpc_argp;
2776         struct nfs_renameres *res = msg->rpc_resp;
2777
2778         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2779         arg->bitmask = server->attr_bitmask;
2780         res->server = server;
2781         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2782 }
2783
2784 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2785                                  struct inode *new_dir)
2786 {
2787         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2788
2789         if (!nfs4_sequence_done(task, &res->seq_res))
2790                 return 0;
2791         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2792                 return 0;
2793
2794         update_changeattr(old_dir, &res->old_cinfo);
2795         nfs_post_op_update_inode(old_dir, res->old_fattr);
2796         update_changeattr(new_dir, &res->new_cinfo);
2797         nfs_post_op_update_inode(new_dir, res->new_fattr);
2798         return 1;
2799 }
2800
2801 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2802                 struct inode *new_dir, struct qstr *new_name)
2803 {
2804         struct nfs_server *server = NFS_SERVER(old_dir);
2805         struct nfs_renameargs arg = {
2806                 .old_dir = NFS_FH(old_dir),
2807                 .new_dir = NFS_FH(new_dir),
2808                 .old_name = old_name,
2809                 .new_name = new_name,
2810                 .bitmask = server->attr_bitmask,
2811         };
2812         struct nfs_renameres res = {
2813                 .server = server,
2814         };
2815         struct rpc_message msg = {
2816                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2817                 .rpc_argp = &arg,
2818                 .rpc_resp = &res,
2819         };
2820         int status = -ENOMEM;
2821
2822         res.old_fattr = nfs_alloc_fattr();
2823         res.new_fattr = nfs_alloc_fattr();
2824         if (res.old_fattr == NULL || res.new_fattr == NULL)
2825                 goto out;
2826
2827         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2828         if (!status) {
2829                 update_changeattr(old_dir, &res.old_cinfo);
2830                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2831                 update_changeattr(new_dir, &res.new_cinfo);
2832                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2833         }
2834 out:
2835         nfs_free_fattr(res.new_fattr);
2836         nfs_free_fattr(res.old_fattr);
2837         return status;
2838 }
2839
2840 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2841                 struct inode *new_dir, struct qstr *new_name)
2842 {
2843         struct nfs4_exception exception = { };
2844         int err;
2845         do {
2846                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2847                                 _nfs4_proc_rename(old_dir, old_name,
2848                                         new_dir, new_name),
2849                                 &exception);
2850         } while (exception.retry);
2851         return err;
2852 }
2853
2854 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2855 {
2856         struct nfs_server *server = NFS_SERVER(inode);
2857         struct nfs4_link_arg arg = {
2858                 .fh     = NFS_FH(inode),
2859                 .dir_fh = NFS_FH(dir),
2860                 .name   = name,
2861                 .bitmask = server->attr_bitmask,
2862         };
2863         struct nfs4_link_res res = {
2864                 .server = server,
2865         };
2866         struct rpc_message msg = {
2867                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2868                 .rpc_argp = &arg,
2869                 .rpc_resp = &res,
2870         };
2871         int status = -ENOMEM;
2872
2873         res.fattr = nfs_alloc_fattr();
2874         res.dir_attr = nfs_alloc_fattr();
2875         if (res.fattr == NULL || res.dir_attr == NULL)
2876                 goto out;
2877
2878         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2879         if (!status) {
2880                 update_changeattr(dir, &res.cinfo);
2881                 nfs_post_op_update_inode(dir, res.dir_attr);
2882                 nfs_post_op_update_inode(inode, res.fattr);
2883         }
2884 out:
2885         nfs_free_fattr(res.dir_attr);
2886         nfs_free_fattr(res.fattr);
2887         return status;
2888 }
2889
2890 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2891 {
2892         struct nfs4_exception exception = { };
2893         int err;
2894         do {
2895                 err = nfs4_handle_exception(NFS_SERVER(inode),
2896                                 _nfs4_proc_link(inode, dir, name),
2897                                 &exception);
2898         } while (exception.retry);
2899         return err;
2900 }
2901
2902 struct nfs4_createdata {
2903         struct rpc_message msg;
2904         struct nfs4_create_arg arg;
2905         struct nfs4_create_res res;
2906         struct nfs_fh fh;
2907         struct nfs_fattr fattr;
2908         struct nfs_fattr dir_fattr;
2909 };
2910
2911 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2912                 struct qstr *name, struct iattr *sattr, u32 ftype)
2913 {
2914         struct nfs4_createdata *data;
2915
2916         data = kzalloc(sizeof(*data), GFP_KERNEL);
2917         if (data != NULL) {
2918                 struct nfs_server *server = NFS_SERVER(dir);
2919
2920                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2921                 data->msg.rpc_argp = &data->arg;
2922                 data->msg.rpc_resp = &data->res;
2923                 data->arg.dir_fh = NFS_FH(dir);
2924                 data->arg.server = server;
2925                 data->arg.name = name;
2926                 data->arg.attrs = sattr;
2927                 data->arg.ftype = ftype;
2928                 data->arg.bitmask = server->attr_bitmask;
2929                 data->res.server = server;
2930                 data->res.fh = &data->fh;
2931                 data->res.fattr = &data->fattr;
2932                 data->res.dir_fattr = &data->dir_fattr;
2933                 nfs_fattr_init(data->res.fattr);
2934                 nfs_fattr_init(data->res.dir_fattr);
2935         }
2936         return data;
2937 }
2938
2939 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2940 {
2941         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2942                                     &data->arg.seq_args, &data->res.seq_res, 1);
2943         if (status == 0) {
2944                 update_changeattr(dir, &data->res.dir_cinfo);
2945                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2946                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2947         }
2948         return status;
2949 }
2950
2951 static void nfs4_free_createdata(struct nfs4_createdata *data)
2952 {
2953         kfree(data);
2954 }
2955
2956 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2957                 struct page *page, unsigned int len, struct iattr *sattr)
2958 {
2959         struct nfs4_createdata *data;
2960         int status = -ENAMETOOLONG;
2961
2962         if (len > NFS4_MAXPATHLEN)
2963                 goto out;
2964
2965         status = -ENOMEM;
2966         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2967         if (data == NULL)
2968                 goto out;
2969
2970         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2971         data->arg.u.symlink.pages = &page;
2972         data->arg.u.symlink.len = len;
2973
2974         status = nfs4_do_create(dir, dentry, data);
2975
2976         nfs4_free_createdata(data);
2977 out:
2978         return status;
2979 }
2980
2981 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2982                 struct page *page, unsigned int len, struct iattr *sattr)
2983 {
2984         struct nfs4_exception exception = { };
2985         int err;
2986         do {
2987                 err = nfs4_handle_exception(NFS_SERVER(dir),
2988                                 _nfs4_proc_symlink(dir, dentry, page,
2989                                                         len, sattr),
2990                                 &exception);
2991         } while (exception.retry);
2992         return err;
2993 }
2994
2995 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2996                 struct iattr *sattr)
2997 {
2998         struct nfs4_createdata *data;
2999         int status = -ENOMEM;
3000
3001         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3002         if (data == NULL)
3003                 goto out;
3004
3005         status = nfs4_do_create(dir, dentry, data);
3006
3007         nfs4_free_createdata(data);
3008 out:
3009         return status;
3010 }
3011
3012 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3013                 struct iattr *sattr)
3014 {
3015         struct nfs4_exception exception = { };
3016         int err;
3017
3018         sattr->ia_mode &= ~current_umask();
3019         do {
3020                 err = nfs4_handle_exception(NFS_SERVER(dir),
3021                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3022                                 &exception);
3023         } while (exception.retry);
3024         return err;
3025 }
3026
3027 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3028                 u64 cookie, struct page **pages, unsigned int count, int plus)
3029 {
3030         struct inode            *dir = dentry->d_inode;
3031         struct nfs4_readdir_arg args = {
3032                 .fh = NFS_FH(dir),
3033                 .pages = pages,
3034                 .pgbase = 0,
3035                 .count = count,
3036                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3037                 .plus = plus,
3038         };
3039         struct nfs4_readdir_res res;
3040         struct rpc_message msg = {
3041                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3042                 .rpc_argp = &args,
3043                 .rpc_resp = &res,
3044                 .rpc_cred = cred,
3045         };
3046         int                     status;
3047
3048         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3049                         dentry->d_parent->d_name.name,
3050                         dentry->d_name.name,
3051                         (unsigned long long)cookie);
3052         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3053         res.pgbase = args.pgbase;
3054         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3055         if (status >= 0) {
3056                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3057                 status += args.pgbase;
3058         }
3059
3060         nfs_invalidate_atime(dir);
3061
3062         dprintk("%s: returns %d\n", __func__, status);
3063         return status;
3064 }
3065
3066 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3067                 u64 cookie, struct page **pages, unsigned int count, int plus)
3068 {
3069         struct nfs4_exception exception = { };
3070         int err;
3071         do {
3072                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3073                                 _nfs4_proc_readdir(dentry, cred, cookie,
3074                                         pages, count, plus),
3075                                 &exception);
3076         } while (exception.retry);
3077         return err;
3078 }
3079
3080 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3081                 struct iattr *sattr, dev_t rdev)
3082 {
3083         struct nfs4_createdata *data;
3084         int mode = sattr->ia_mode;
3085         int status = -ENOMEM;
3086
3087         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3088         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3089
3090         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3091         if (data == NULL)
3092                 goto out;
3093
3094         if (S_ISFIFO(mode))
3095                 data->arg.ftype = NF4FIFO;
3096         else if (S_ISBLK(mode)) {
3097                 data->arg.ftype = NF4BLK;
3098                 data->arg.u.device.specdata1 = MAJOR(rdev);
3099                 data->arg.u.device.specdata2 = MINOR(rdev);
3100         }
3101         else if (S_ISCHR(mode)) {
3102                 data->arg.ftype = NF4CHR;
3103                 data->arg.u.device.specdata1 = MAJOR(rdev);
3104                 data->arg.u.device.specdata2 = MINOR(rdev);
3105         }
3106
3107         status = nfs4_do_create(dir, dentry, data);
3108
3109         nfs4_free_createdata(data);
3110 out:
3111         return status;
3112 }
3113
3114 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3115                 struct iattr *sattr, dev_t rdev)
3116 {
3117         struct nfs4_exception exception = { };
3118         int err;
3119
3120         sattr->ia_mode &= ~current_umask();
3121         do {
3122                 err = nfs4_handle_exception(NFS_SERVER(dir),
3123                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3124                                 &exception);
3125         } while (exception.retry);
3126         return err;
3127 }
3128
3129 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3130                  struct nfs_fsstat *fsstat)
3131 {
3132         struct nfs4_statfs_arg args = {
3133                 .fh = fhandle,
3134                 .bitmask = server->attr_bitmask,
3135         };
3136         struct nfs4_statfs_res res = {
3137                 .fsstat = fsstat,
3138         };
3139         struct rpc_message msg = {
3140                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3141                 .rpc_argp = &args,
3142                 .rpc_resp = &res,
3143         };
3144
3145         nfs_fattr_init(fsstat->fattr);
3146         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3147 }
3148
3149 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3150 {
3151         struct nfs4_exception exception = { };
3152         int err;
3153         do {
3154                 err = nfs4_handle_exception(server,
3155                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3156                                 &exception);
3157         } while (exception.retry);
3158         return err;
3159 }
3160
3161 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3162                 struct nfs_fsinfo *fsinfo)
3163 {
3164         struct nfs4_fsinfo_arg args = {
3165                 .fh = fhandle,
3166                 .bitmask = server->attr_bitmask,
3167         };
3168         struct nfs4_fsinfo_res res = {
3169                 .fsinfo = fsinfo,
3170         };
3171         struct rpc_message msg = {
3172                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3173                 .rpc_argp = &args,
3174                 .rpc_resp = &res,
3175         };
3176
3177         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3178 }
3179
3180 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3181 {
3182         struct nfs4_exception exception = { };
3183         int err;
3184
3185         do {
3186                 err = nfs4_handle_exception(server,
3187                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3188                                 &exception);
3189         } while (exception.retry);
3190         return err;
3191 }
3192
3193 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3194 {
3195         nfs_fattr_init(fsinfo->fattr);
3196         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3197 }
3198
3199 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3200                 struct nfs_pathconf *pathconf)
3201 {
3202         struct nfs4_pathconf_arg args = {
3203                 .fh = fhandle,
3204                 .bitmask = server->attr_bitmask,
3205         };
3206         struct nfs4_pathconf_res res = {
3207                 .pathconf = pathconf,
3208         };
3209         struct rpc_message msg = {
3210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3211                 .rpc_argp = &args,
3212                 .rpc_resp = &res,
3213         };
3214
3215         /* None of the pathconf attributes are mandatory to implement */
3216         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3217                 memset(pathconf, 0, sizeof(*pathconf));
3218                 return 0;
3219         }
3220
3221         nfs_fattr_init(pathconf->fattr);
3222         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3223 }
3224
3225 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3226                 struct nfs_pathconf *pathconf)
3227 {
3228         struct nfs4_exception exception = { };
3229         int err;
3230
3231         do {
3232                 err = nfs4_handle_exception(server,
3233                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3234                                 &exception);
3235         } while (exception.retry);
3236         return err;
3237 }
3238
3239 void __nfs4_read_done_cb(struct nfs_read_data *data)
3240 {
3241         nfs_invalidate_atime(data->inode);
3242 }
3243
3244 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3245 {
3246         struct nfs_server *server = NFS_SERVER(data->inode);
3247
3248         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3249                 rpc_restart_call_prepare(task);
3250                 return -EAGAIN;
3251         }
3252
3253         __nfs4_read_done_cb(data);
3254         if (task->tk_status > 0)
3255                 renew_lease(server, data->timestamp);
3256         return 0;
3257 }
3258
3259 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3260 {
3261
3262         dprintk("--> %s\n", __func__);
3263
3264         if (!nfs4_sequence_done(task, &data->res.seq_res))
3265                 return -EAGAIN;
3266
3267         return data->read_done_cb ? data->read_done_cb(task, data) :
3268                                     nfs4_read_done_cb(task, data);
3269 }
3270
3271 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3272 {
3273         data->timestamp   = jiffies;
3274         data->read_done_cb = nfs4_read_done_cb;
3275         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3276         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3277 }
3278
3279 /* Reset the the nfs_read_data to send the read to the MDS. */
3280 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3281 {
3282         dprintk("%s Reset task for i/o through\n", __func__);
3283         put_lseg(data->lseg);
3284         data->lseg = NULL;
3285         /* offsets will differ in the dense stripe case */
3286         data->args.offset = data->mds_offset;
3287         data->ds_clp = NULL;
3288         data->args.fh     = NFS_FH(data->inode);
3289         data->read_done_cb = nfs4_read_done_cb;
3290         task->tk_ops = data->mds_ops;
3291         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3292 }
3293 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3294
3295 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3296 {
3297         struct inode *inode = data->inode;
3298
3299         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3300                 rpc_restart_call_prepare(task);
3301                 return -EAGAIN;
3302         }
3303         if (task->tk_status >= 0) {
3304                 renew_lease(NFS_SERVER(inode), data->timestamp);
3305                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3306         }
3307         return 0;
3308 }
3309
3310 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3311 {
3312         if (!nfs4_sequence_done(task, &data->res.seq_res))
3313                 return -EAGAIN;
3314         return data->write_done_cb ? data->write_done_cb(task, data) :
3315                 nfs4_write_done_cb(task, data);
3316 }
3317
3318 /* Reset the the nfs_write_data to send the write to the MDS. */
3319 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3320 {
3321         dprintk("%s Reset task for i/o through\n", __func__);
3322         put_lseg(data->lseg);
3323         data->lseg          = NULL;
3324         data->ds_clp        = NULL;
3325         data->write_done_cb = nfs4_write_done_cb;
3326         data->args.fh       = NFS_FH(data->inode);
3327         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3328         data->args.offset   = data->mds_offset;
3329         data->res.fattr     = &data->fattr;
3330         task->tk_ops        = data->mds_ops;
3331         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3332 }
3333 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3334
3335 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3336 {
3337         struct nfs_server *server = NFS_SERVER(data->inode);
3338
3339         if (data->lseg) {
3340                 data->args.bitmask = NULL;
3341                 data->res.fattr = NULL;
3342         } else
3343                 data->args.bitmask = server->cache_consistency_bitmask;
3344         if (!data->write_done_cb)
3345                 data->write_done_cb = nfs4_write_done_cb;
3346         data->res.server = server;
3347         data->timestamp   = jiffies;
3348
3349         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3350         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3351 }
3352
3353 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3354 {
3355         struct inode *inode = data->inode;
3356
3357         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3358                 rpc_restart_call_prepare(task);
3359                 return -EAGAIN;
3360         }
3361         nfs_refresh_inode(inode, data->res.fattr);
3362         return 0;
3363 }
3364
3365 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3366 {
3367         if (!nfs4_sequence_done(task, &data->res.seq_res))
3368                 return -EAGAIN;
3369         return data->write_done_cb(task, data);
3370 }
3371
3372 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3373 {
3374         struct nfs_server *server = NFS_SERVER(data->inode);
3375
3376         if (data->lseg) {
3377                 data->args.bitmask = NULL;
3378                 data->res.fattr = NULL;
3379         } else
3380                 data->args.bitmask = server->cache_consistency_bitmask;
3381         if (!data->write_done_cb)
3382                 data->write_done_cb = nfs4_commit_done_cb;
3383         data->res.server = server;
3384         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3385         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3386 }
3387
3388 struct nfs4_renewdata {
3389         struct nfs_client       *client;
3390         unsigned long           timestamp;
3391 };
3392
3393 /*
3394  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3395  * standalone procedure for queueing an asynchronous RENEW.
3396  */
3397 static void nfs4_renew_release(void *calldata)
3398 {
3399         struct nfs4_renewdata *data = calldata;
3400         struct nfs_client *clp = data->client;
3401
3402         if (atomic_read(&clp->cl_count) > 1)
3403                 nfs4_schedule_state_renewal(clp);
3404         nfs_put_client(clp);
3405         kfree(data);
3406 }
3407
3408 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3409 {
3410         struct nfs4_renewdata *data = calldata;
3411         struct nfs_client *clp = data->client;
3412         unsigned long timestamp = data->timestamp;
3413
3414         if (task->tk_status < 0) {
3415                 /* Unless we're shutting down, schedule state recovery! */
3416                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3417                         return;
3418                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3419                         nfs4_schedule_lease_recovery(clp);
3420                         return;
3421                 }
3422                 nfs4_schedule_path_down_recovery(clp);
3423         }
3424         do_renew_lease(clp, timestamp);
3425 }
3426
3427 static const struct rpc_call_ops nfs4_renew_ops = {
3428         .rpc_call_done = nfs4_renew_done,
3429         .rpc_release = nfs4_renew_release,
3430 };
3431
3432 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3433 {
3434         struct rpc_message msg = {
3435                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3436                 .rpc_argp       = clp,
3437                 .rpc_cred       = cred,
3438         };
3439         struct nfs4_renewdata *data;
3440
3441         if (renew_flags == 0)
3442                 return 0;
3443         if (!atomic_inc_not_zero(&clp->cl_count))
3444                 return -EIO;
3445         data = kmalloc(sizeof(*data), GFP_NOFS);
3446         if (data == NULL)
3447                 return -ENOMEM;
3448         data->client = clp;
3449         data->timestamp = jiffies;
3450         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3451                         &nfs4_renew_ops, data);
3452 }
3453
3454 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3455 {
3456         struct rpc_message msg = {
3457                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3458                 .rpc_argp       = clp,
3459                 .rpc_cred       = cred,
3460         };
3461         unsigned long now = jiffies;
3462         int status;
3463
3464         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3465         if (status < 0)
3466                 return status;
3467         do_renew_lease(clp, now);
3468         return 0;
3469 }
3470
3471 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3472 {
3473         return (server->caps & NFS_CAP_ACLS)
3474                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3475                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3476 }
3477
3478 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3479  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3480  * the stack.
3481  */
3482 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3483
3484 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3485                 struct page **pages, unsigned int *pgbase)
3486 {
3487         struct page *newpage, **spages;
3488         int rc = 0;
3489         size_t len;
3490         spages = pages;
3491
3492         do {
3493                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3494                 newpage = alloc_page(GFP_KERNEL);
3495
3496                 if (newpage == NULL)
3497                         goto unwind;
3498                 memcpy(page_address(newpage), buf, len);
3499                 buf += len;
3500                 buflen -= len;
3501                 *pages++ = newpage;
3502                 rc++;
3503         } while (buflen != 0);
3504
3505         return rc;
3506
3507 unwind:
3508         for(; rc > 0; rc--)
3509                 __free_page(spages[rc-1]);
3510         return -ENOMEM;
3511 }
3512
3513 struct nfs4_cached_acl {
3514         int cached;
3515         size_t len;
3516         char data[0];
3517 };
3518
3519 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3520 {
3521         struct nfs_inode *nfsi = NFS_I(inode);
3522
3523         spin_lock(&inode->i_lock);
3524         kfree(nfsi->nfs4_acl);
3525         nfsi->nfs4_acl = acl;
3526         spin_unlock(&inode->i_lock);
3527 }
3528
3529 static void nfs4_zap_acl_attr(struct inode *inode)
3530 {
3531         nfs4_set_cached_acl(inode, NULL);
3532 }
3533
3534 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3535 {
3536         struct nfs_inode *nfsi = NFS_I(inode);
3537         struct nfs4_cached_acl *acl;
3538         int ret = -ENOENT;
3539
3540         spin_lock(&inode->i_lock);
3541         acl = nfsi->nfs4_acl;
3542         if (acl == NULL)
3543                 goto out;
3544         if (buf == NULL) /* user is just asking for length */
3545                 goto out_len;
3546         if (acl->cached == 0)
3547                 goto out;
3548         ret = -ERANGE; /* see getxattr(2) man page */
3549         if (acl->len > buflen)
3550                 goto out;
3551         memcpy(buf, acl->data, acl->len);
3552 out_len:
3553         ret = acl->len;
3554 out:
3555         spin_unlock(&inode->i_lock);
3556         return ret;
3557 }
3558
3559 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3560 {
3561         struct nfs4_cached_acl *acl;
3562
3563         if (buf && acl_len <= PAGE_SIZE) {
3564                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3565                 if (acl == NULL)
3566                         goto out;
3567                 acl->cached = 1;
3568                 memcpy(acl->data, buf, acl_len);
3569         } else {
3570                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3571                 if (acl == NULL)
3572                         goto out;
3573                 acl->cached = 0;
3574         }
3575         acl->len = acl_len;
3576 out:
3577         nfs4_set_cached_acl(inode, acl);
3578 }
3579
3580 /*
3581  * The getxattr API returns the required buffer length when called with a
3582  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3583  * the required buf.  On a NULL buf, we send a page of data to the server
3584  * guessing that the ACL request can be serviced by a page. If so, we cache
3585  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3586  * the cache. If not so, we throw away the page, and cache the required
3587  * length. The next getxattr call will then produce another round trip to
3588  * the server, this time with the input buf of the required size.
3589  */
3590 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3591 {
3592         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3593         struct nfs_getaclargs args = {
3594                 .fh = NFS_FH(inode),
3595                 .acl_pages = pages,
3596                 .acl_len = buflen,
3597         };
3598         struct nfs_getaclres res = {
3599                 .acl_len = buflen,
3600         };
3601         void *resp_buf;
3602         struct rpc_message msg = {
3603                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3604                 .rpc_argp = &args,
3605                 .rpc_resp = &res,
3606         };
3607         int ret = -ENOMEM, npages, i, acl_len = 0;
3608
3609         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3610         /* As long as we're doing a round trip to the server anyway,
3611          * let's be prepared for a page of acl data. */
3612         if (npages == 0)
3613                 npages = 1;
3614
3615         for (i = 0; i < npages; i++) {
3616                 pages[i] = alloc_page(GFP_KERNEL);
3617                 if (!pages[i])
3618                         goto out_free;
3619         }
3620         if (npages > 1) {
3621                 /* for decoding across pages */
3622                 args.acl_scratch = alloc_page(GFP_KERNEL);
3623                 if (!args.acl_scratch)
3624                         goto out_free;
3625         }
3626         args.acl_len = npages * PAGE_SIZE;
3627         args.acl_pgbase = 0;
3628         /* Let decode_getfacl know not to fail if the ACL data is larger than
3629          * the page we send as a guess */
3630         if (buf == NULL)
3631                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3632         resp_buf = page_address(pages[0]);
3633
3634         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3635                 __func__, buf, buflen, npages, args.acl_len);
3636         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3637                              &msg, &args.seq_args, &res.seq_res, 0);
3638         if (ret)
3639                 goto out_free;
3640
3641         acl_len = res.acl_len - res.acl_data_offset;
3642         if (acl_len > args.acl_len)
3643                 nfs4_write_cached_acl(inode, NULL, acl_len);
3644         else
3645                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3646                                       acl_len);
3647         if (buf) {
3648                 ret = -ERANGE;
3649                 if (acl_len > buflen)
3650                         goto out_free;
3651                 _copy_from_pages(buf, pages, res.acl_data_offset,
3652                                 res.acl_len);
3653         }
3654         ret = acl_len;
3655 out_free:
3656         for (i = 0; i < npages; i++)
3657                 if (pages[i])
3658                         __free_page(pages[i]);
3659         if (args.acl_scratch)
3660                 __free_page(args.acl_scratch);
3661         return ret;
3662 }
3663
3664 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3665 {
3666         struct nfs4_exception exception = { };
3667         ssize_t ret;
3668         do {
3669                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3670                 if (ret >= 0)
3671                         break;
3672                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3673         } while (exception.retry);
3674         return ret;
3675 }
3676
3677 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3678 {
3679         struct nfs_server *server = NFS_SERVER(inode);
3680         int ret;
3681
3682         if (!nfs4_server_supports_acls(server))
3683                 return -EOPNOTSUPP;
3684         ret = nfs_revalidate_inode(server, inode);
3685         if (ret < 0)
3686                 return ret;
3687         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3688                 nfs_zap_acl_cache(inode);
3689         ret = nfs4_read_cached_acl(inode, buf, buflen);
3690         if (ret != -ENOENT)
3691                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3692                  * but no cached acl data, just the acl length */
3693                 return ret;
3694         return nfs4_get_acl_uncached(inode, buf, buflen);
3695 }
3696
3697 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3698 {
3699         struct nfs_server *server = NFS_SERVER(inode);
3700         struct page *pages[NFS4ACL_MAXPAGES];
3701         struct nfs_setaclargs arg = {
3702                 .fh             = NFS_FH(inode),
3703                 .acl_pages      = pages,
3704                 .acl_len        = buflen,
3705         };
3706         struct nfs_setaclres res;
3707         struct rpc_message msg = {
3708                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3709                 .rpc_argp       = &arg,
3710                 .rpc_resp       = &res,
3711         };
3712         int ret, i;
3713
3714         if (!nfs4_server_supports_acls(server))
3715                 return -EOPNOTSUPP;
3716         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3717         if (i < 0)
3718                 return i;
3719         nfs_inode_return_delegation(inode);
3720         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3721
3722         /*
3723          * Free each page after tx, so the only ref left is
3724          * held by the network stack
3725          */
3726         for (; i > 0; i--)
3727                 put_page(pages[i-1]);
3728
3729         /*
3730          * Acl update can result in inode attribute update.
3731          * so mark the attribute cache invalid.
3732          */
3733         spin_lock(&inode->i_lock);
3734         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3735         spin_unlock(&inode->i_lock);
3736         nfs_access_zap_cache(inode);
3737         nfs_zap_acl_cache(inode);
3738         return ret;
3739 }
3740
3741 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3742 {
3743         struct nfs4_exception exception = { };
3744         int err;
3745         do {
3746                 err = nfs4_handle_exception(NFS_SERVER(inode),
3747                                 __nfs4_proc_set_acl(inode, buf, buflen),
3748                                 &exception);
3749         } while (exception.retry);
3750         return err;
3751 }
3752
3753 static int
3754 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3755 {
3756         struct nfs_client *clp = server->nfs_client;
3757
3758         if (task->tk_status >= 0)
3759                 return 0;
3760         switch(task->tk_status) {
3761                 case -NFS4ERR_ADMIN_REVOKED:
3762                 case -NFS4ERR_BAD_STATEID:
3763                 case -NFS4ERR_OPENMODE:
3764                         if (state == NULL)
3765                                 break;
3766                         nfs4_schedule_stateid_recovery(server, state);
3767                         goto wait_on_recovery;
3768                 case -NFS4ERR_EXPIRED:
3769                         if (state != NULL)
3770                                 nfs4_schedule_stateid_recovery(server, state);
3771                 case -NFS4ERR_STALE_STATEID:
3772                 case -NFS4ERR_STALE_CLIENTID:
3773                         nfs4_schedule_lease_recovery(clp);
3774                         goto wait_on_recovery;
3775 #if defined(CONFIG_NFS_V4_1)
3776                 case -NFS4ERR_BADSESSION:
3777                 case -NFS4ERR_BADSLOT:
3778                 case -NFS4ERR_BAD_HIGH_SLOT:
3779                 case -NFS4ERR_DEADSESSION:
3780                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3781                 case -NFS4ERR_SEQ_FALSE_RETRY:
3782                 case -NFS4ERR_SEQ_MISORDERED:
3783                         dprintk("%s ERROR %d, Reset session\n", __func__,
3784                                 task->tk_status);
3785                         nfs4_schedule_session_recovery(clp->cl_session);
3786                         task->tk_status = 0;
3787                         return -EAGAIN;
3788 #endif /* CONFIG_NFS_V4_1 */
3789                 case -NFS4ERR_DELAY:
3790                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3791                 case -NFS4ERR_GRACE:
3792                 case -EKEYEXPIRED:
3793                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3794                         task->tk_status = 0;
3795                         return -EAGAIN;
3796                 case -NFS4ERR_RETRY_UNCACHED_REP:
3797                 case -NFS4ERR_OLD_STATEID:
3798                         task->tk_status = 0;
3799                         return -EAGAIN;
3800         }
3801         task->tk_status = nfs4_map_errors(task->tk_status);
3802         return 0;
3803 wait_on_recovery:
3804         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3805         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3806                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3807         task->tk_status = 0;
3808         return -EAGAIN;
3809 }
3810
3811 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3812                 unsigned short port, struct rpc_cred *cred,
3813                 struct nfs4_setclientid_res *res)
3814 {
3815         nfs4_verifier sc_verifier;
3816         struct nfs4_setclientid setclientid = {
3817                 .sc_verifier = &sc_verifier,
3818                 .sc_prog = program,
3819                 .sc_cb_ident = clp->cl_cb_ident,
3820         };
3821         struct rpc_message msg = {
3822                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3823                 .rpc_argp = &setclientid,
3824                 .rpc_resp = res,
3825                 .rpc_cred = cred,
3826         };
3827         __be32 *p;
3828         int loop = 0;
3829         int status;
3830
3831         p = (__be32*)sc_verifier.data;
3832         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3833         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3834
3835         for(;;) {
3836                 rcu_read_lock();
3837                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3838                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3839                                 clp->cl_ipaddr,
3840                                 rpc_peeraddr2str(clp->cl_rpcclient,
3841                                                         RPC_DISPLAY_ADDR),
3842                                 rpc_peeraddr2str(clp->cl_rpcclient,
3843                                                         RPC_DISPLAY_PROTO),
3844                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3845                                 clp->cl_id_uniquifier);
3846                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3847                                 sizeof(setclientid.sc_netid),
3848                                 rpc_peeraddr2str(clp->cl_rpcclient,
3849                                                         RPC_DISPLAY_NETID));
3850                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3851                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3852                                 clp->cl_ipaddr, port >> 8, port & 255);
3853                 rcu_read_unlock();
3854
3855                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3856                 if (status != -NFS4ERR_CLID_INUSE)
3857                         break;
3858                 if (loop != 0) {
3859                         ++clp->cl_id_uniquifier;
3860                         break;
3861                 }
3862                 ++loop;
3863                 ssleep(clp->cl_lease_time / HZ + 1);
3864         }
3865         return status;
3866 }
3867
3868 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3869                 struct nfs4_setclientid_res *arg,
3870                 struct rpc_cred *cred)
3871 {
3872         struct nfs_fsinfo fsinfo;
3873         struct rpc_message msg = {
3874                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3875                 .rpc_argp = arg,
3876                 .rpc_resp = &fsinfo,
3877                 .rpc_cred = cred,
3878         };
3879         unsigned long now;
3880         int status;
3881
3882         now = jiffies;
3883         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3884         if (status == 0) {
3885                 spin_lock(&clp->cl_lock);
3886                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3887                 clp->cl_last_renewal = now;
3888                 spin_unlock(&clp->cl_lock);
3889         }
3890         return status;
3891 }
3892
3893 struct nfs4_delegreturndata {
3894         struct nfs4_delegreturnargs args;
3895         struct nfs4_delegreturnres res;
3896         struct nfs_fh fh;
3897         nfs4_stateid stateid;
3898         unsigned long timestamp;
3899         struct nfs_fattr fattr;
3900         int rpc_status;
3901 };
3902
3903 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3904 {
3905         struct nfs4_delegreturndata *data = calldata;
3906
3907         if (!nfs4_sequence_done(task, &data->res.seq_res))
3908                 return;
3909
3910         switch (task->tk_status) {
3911         case -NFS4ERR_STALE_STATEID:
3912         case -NFS4ERR_EXPIRED:
3913         case 0:
3914                 renew_lease(data->res.server, data->timestamp);
3915                 break;
3916         default:
3917                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3918                                 -EAGAIN) {
3919                         rpc_restart_call_prepare(task);
3920                         return;
3921                 }
3922         }
3923         data->rpc_status = task->tk_status;
3924 }
3925
3926 static void nfs4_delegreturn_release(void *calldata)
3927 {
3928         kfree(calldata);
3929 }
3930
3931 #if defined(CONFIG_NFS_V4_1)
3932 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3933 {
3934         struct nfs4_delegreturndata *d_data;
3935
3936         d_data = (struct nfs4_delegreturndata *)data;
3937
3938         if (nfs4_setup_sequence(d_data->res.server,
3939                                 &d_data->args.seq_args,
3940                                 &d_data->res.seq_res, task))
3941                 return;
3942         rpc_call_start(task);
3943 }
3944 #endif /* CONFIG_NFS_V4_1 */
3945
3946 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3947 #if defined(CONFIG_NFS_V4_1)
3948         .rpc_call_prepare = nfs4_delegreturn_prepare,
3949 #endif /* CONFIG_NFS_V4_1 */
3950         .rpc_call_done = nfs4_delegreturn_done,
3951         .rpc_release = nfs4_delegreturn_release,
3952 };
3953
3954 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3955 {
3956         struct nfs4_delegreturndata *data;
3957         struct nfs_server *server = NFS_SERVER(inode);
3958         struct rpc_task *task;
3959         struct rpc_message msg = {
3960                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3961                 .rpc_cred = cred,
3962         };
3963         struct rpc_task_setup task_setup_data = {
3964                 .rpc_client = server->client,
3965                 .rpc_message = &msg,
3966                 .callback_ops = &nfs4_delegreturn_ops,
3967                 .flags = RPC_TASK_ASYNC,
3968         };
3969         int status = 0;
3970
3971         data = kzalloc(sizeof(*data), GFP_NOFS);
3972         if (data == NULL)
3973                 return -ENOMEM;
3974         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3975         data->args.fhandle = &data->fh;
3976         data->args.stateid = &data->stateid;
3977         data->args.bitmask = server->attr_bitmask;
3978         nfs_copy_fh(&data->fh, NFS_FH(inode));
3979         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3980         data->res.fattr = &data->fattr;
3981         data->res.server = server;
3982         nfs_fattr_init(data->res.fattr);
3983         data->timestamp = jiffies;
3984         data->rpc_status = 0;
3985
3986         task_setup_data.callback_data = data;
3987         msg.rpc_argp = &data->args;
3988         msg.rpc_resp = &data->res;
3989         task = rpc_run_task(&task_setup_data);
3990         if (IS_ERR(task))
3991                 return PTR_ERR(task);
3992         if (!issync)
3993                 goto out;
3994         status = nfs4_wait_for_completion_rpc_task(task);
3995         if (status != 0)
3996                 goto out;
3997         status = data->rpc_status;
3998         if (status != 0)
3999                 goto out;
4000         nfs_refresh_inode(inode, &data->fattr);
4001 out:
4002         rpc_put_task(task);
4003         return status;
4004 }
4005
4006 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4007 {
4008         struct nfs_server *server = NFS_SERVER(inode);
4009         struct nfs4_exception exception = { };
4010         int err;
4011         do {
4012                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4013                 switch (err) {
4014                         case -NFS4ERR_STALE_STATEID:
4015                         case -NFS4ERR_EXPIRED:
4016                         case 0:
4017                                 return 0;
4018                 }
4019                 err = nfs4_handle_exception(server, err, &exception);
4020         } while (exception.retry);
4021         return err;
4022 }
4023
4024 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4025 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4026
4027 /*
4028  * sleep, with exponential backoff, and retry the LOCK operation.
4029  */
4030 static unsigned long
4031 nfs4_set_lock_task_retry(unsigned long timeout)
4032 {
4033         freezable_schedule_timeout_killable(timeout);
4034         timeout <<= 1;
4035         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4036                 return NFS4_LOCK_MAXTIMEOUT;
4037         return timeout;
4038 }
4039
4040 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4041 {
4042         struct inode *inode = state->inode;
4043         struct nfs_server *server = NFS_SERVER(inode);
4044         struct nfs_client *clp = server->nfs_client;
4045         struct nfs_lockt_args arg = {
4046                 .fh = NFS_FH(inode),
4047                 .fl = request,
4048         };
4049         struct nfs_lockt_res res = {
4050                 .denied = request,
4051         };
4052         struct rpc_message msg = {
4053                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4054                 .rpc_argp       = &arg,
4055                 .rpc_resp       = &res,
4056                 .rpc_cred       = state->owner->so_cred,
4057         };
4058         struct nfs4_lock_state *lsp;
4059         int status;
4060
4061         arg.lock_owner.clientid = clp->cl_clientid;
4062         status = nfs4_set_lock_state(state, request);
4063         if (status != 0)
4064                 goto out;
4065         lsp = request->fl_u.nfs4_fl.owner;
4066         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4067         arg.lock_owner.s_dev = server->s_dev;
4068         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4069         switch (status) {
4070                 case 0:
4071                         request->fl_type = F_UNLCK;
4072                         break;
4073                 case -NFS4ERR_DENIED:
4074                         status = 0;
4075         }
4076         request->fl_ops->fl_release_private(request);
4077 out:
4078         return status;
4079 }
4080
4081 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4082 {
4083         struct nfs4_exception exception = { };
4084         int err;
4085
4086         do {
4087                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4088                                 _nfs4_proc_getlk(state, cmd, request),
4089                                 &exception);
4090         } while (exception.retry);
4091         return err;
4092 }
4093
4094 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4095 {
4096         int res = 0;
4097         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4098                 case FL_POSIX:
4099                         res = posix_lock_file_wait(file, fl);
4100                         break;
4101                 case FL_FLOCK:
4102                         res = flock_lock_file_wait(file, fl);
4103                         break;
4104                 default:
4105                         BUG();
4106         }
4107         return res;
4108 }
4109
4110 struct nfs4_unlockdata {
4111         struct nfs_locku_args arg;
4112         struct nfs_locku_res res;
4113         struct nfs4_lock_state *lsp;
4114         struct nfs_open_context *ctx;
4115         struct file_lock fl;
4116         const struct nfs_server *server;
4117         unsigned long timestamp;
4118 };
4119
4120 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4121                 struct nfs_open_context *ctx,
4122                 struct nfs4_lock_state *lsp,
4123                 struct nfs_seqid *seqid)
4124 {
4125         struct nfs4_unlockdata *p;
4126         struct inode *inode = lsp->ls_state->inode;
4127
4128         p = kzalloc(sizeof(*p), GFP_NOFS);
4129         if (p == NULL)
4130                 return NULL;
4131         p->arg.fh = NFS_FH(inode);
4132         p->arg.fl = &p->fl;
4133         p->arg.seqid = seqid;
4134         p->res.seqid = seqid;
4135         p->arg.stateid = &lsp->ls_stateid;
4136         p->lsp = lsp;
4137         atomic_inc(&lsp->ls_count);
4138         /* Ensure we don't close file until we're done freeing locks! */
4139         p->ctx = get_nfs_open_context(ctx);
4140         memcpy(&p->fl, fl, sizeof(p->fl));
4141         p->server = NFS_SERVER(inode);
4142         return p;
4143 }
4144
4145 static void nfs4_locku_release_calldata(void *data)
4146 {
4147         struct nfs4_unlockdata *calldata = data;
4148         nfs_free_seqid(calldata->arg.seqid);
4149         nfs4_put_lock_state(calldata->lsp);
4150         put_nfs_open_context(calldata->ctx);
4151         kfree(calldata);
4152 }
4153
4154 static void nfs4_locku_done(struct rpc_task *task, void *data)
4155 {
4156         struct nfs4_unlockdata *calldata = data;
4157
4158         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4159                 return;
4160         switch (task->tk_status) {
4161                 case 0:
4162                         memcpy(calldata->lsp->ls_stateid.data,
4163                                         calldata->res.stateid.data,
4164                                         sizeof(calldata->lsp->ls_stateid.data));
4165                         renew_lease(calldata->server, calldata->timestamp);
4166                         break;
4167                 case -NFS4ERR_BAD_STATEID:
4168                 case -NFS4ERR_OLD_STATEID:
4169                 case -NFS4ERR_STALE_STATEID:
4170                 case -NFS4ERR_EXPIRED:
4171                         break;
4172                 default:
4173                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4174                                 rpc_restart_call_prepare(task);
4175         }
4176 }
4177
4178 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4179 {
4180         struct nfs4_unlockdata *calldata = data;
4181
4182         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4183                 return;
4184         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4185                 /* Note: exit _without_ running nfs4_locku_done */
4186                 task->tk_action = NULL;
4187                 return;
4188         }
4189         calldata->timestamp = jiffies;
4190         if (nfs4_setup_sequence(calldata->server,
4191                                 &calldata->arg.seq_args,
4192                                 &calldata->res.seq_res, task))
4193                 return;
4194         rpc_call_start(task);
4195 }
4196
4197 static const struct rpc_call_ops nfs4_locku_ops = {
4198         .rpc_call_prepare = nfs4_locku_prepare,
4199         .rpc_call_done = nfs4_locku_done,
4200         .rpc_release = nfs4_locku_release_calldata,
4201 };
4202
4203 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4204                 struct nfs_open_context *ctx,
4205                 struct nfs4_lock_state *lsp,
4206                 struct nfs_seqid *seqid)
4207 {
4208         struct nfs4_unlockdata *data;
4209         struct rpc_message msg = {
4210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4211                 .rpc_cred = ctx->cred,
4212         };
4213         struct rpc_task_setup task_setup_data = {
4214                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4215                 .rpc_message = &msg,
4216                 .callback_ops = &nfs4_locku_ops,
4217                 .workqueue = nfsiod_workqueue,
4218                 .flags = RPC_TASK_ASYNC,
4219         };
4220
4221         /* Ensure this is an unlock - when canceling a lock, the
4222          * canceled lock is passed in, and it won't be an unlock.
4223          */
4224         fl->fl_type = F_UNLCK;
4225
4226         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4227         if (data == NULL) {
4228                 nfs_free_seqid(seqid);
4229                 return ERR_PTR(-ENOMEM);
4230         }
4231
4232         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4233         msg.rpc_argp = &data->arg;
4234         msg.rpc_resp = &data->res;
4235         task_setup_data.callback_data = data;
4236         return rpc_run_task(&task_setup_data);
4237 }
4238
4239 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4240 {
4241         struct nfs_inode *nfsi = NFS_I(state->inode);
4242         struct nfs_seqid *seqid;
4243         struct nfs4_lock_state *lsp;
4244         struct rpc_task *task;
4245         int status = 0;
4246         unsigned char fl_flags = request->fl_flags;
4247
4248         status = nfs4_set_lock_state(state, request);
4249         /* Unlock _before_ we do the RPC call */
4250         request->fl_flags |= FL_EXISTS;
4251         down_read(&nfsi->rwsem);
4252         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4253                 up_read(&nfsi->rwsem);
4254                 goto out;
4255         }
4256         up_read(&nfsi->rwsem);
4257         if (status != 0)
4258                 goto out;
4259         /* Is this a delegated lock? */
4260         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4261                 goto out;
4262         lsp = request->fl_u.nfs4_fl.owner;
4263         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4264         status = -ENOMEM;
4265         if (seqid == NULL)
4266                 goto out;
4267         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4268         status = PTR_ERR(task);
4269         if (IS_ERR(task))
4270                 goto out;
4271         status = nfs4_wait_for_completion_rpc_task(task);
4272         rpc_put_task(task);
4273 out:
4274         request->fl_flags = fl_flags;
4275         return status;
4276 }
4277
4278 struct nfs4_lockdata {
4279         struct nfs_lock_args arg;
4280         struct nfs_lock_res res;
4281         struct nfs4_lock_state *lsp;
4282         struct nfs_open_context *ctx;
4283         struct file_lock fl;
4284         unsigned long timestamp;
4285         int rpc_status;
4286         int cancelled;
4287         struct nfs_server *server;
4288 };
4289
4290 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4291                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4292                 gfp_t gfp_mask)
4293 {
4294         struct nfs4_lockdata *p;
4295         struct inode *inode = lsp->ls_state->inode;
4296         struct nfs_server *server = NFS_SERVER(inode);
4297
4298         p = kzalloc(sizeof(*p), gfp_mask);
4299         if (p == NULL)
4300                 return NULL;
4301
4302         p->arg.fh = NFS_FH(inode);
4303         p->arg.fl = &p->fl;
4304         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4305         if (p->arg.open_seqid == NULL)
4306                 goto out_free;
4307         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4308         if (p->arg.lock_seqid == NULL)
4309                 goto out_free_seqid;
4310         p->arg.lock_stateid = &lsp->ls_stateid;
4311         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4312         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4313         p->arg.lock_owner.s_dev = server->s_dev;
4314         p->res.lock_seqid = p->arg.lock_seqid;
4315         p->lsp = lsp;
4316         p->server = server;
4317         atomic_inc(&lsp->ls_count);
4318         p->ctx = get_nfs_open_context(ctx);
4319         memcpy(&p->fl, fl, sizeof(p->fl));
4320         return p;
4321 out_free_seqid:
4322         nfs_free_seqid(p->arg.open_seqid);
4323 out_free:
4324         kfree(p);
4325         return NULL;
4326 }
4327
4328 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4329 {
4330         struct nfs4_lockdata *data = calldata;
4331         struct nfs4_state *state = data->lsp->ls_state;
4332
4333         dprintk("%s: begin!\n", __func__);
4334         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4335                 return;
4336         /* Do we need to do an open_to_lock_owner? */
4337         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4338                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4339                         return;
4340                 data->arg.open_stateid = &state->stateid;
4341                 data->arg.new_lock_owner = 1;
4342                 data->res.open_seqid = data->arg.open_seqid;
4343         } else
4344                 data->arg.new_lock_owner = 0;
4345         data->timestamp = jiffies;
4346         if (nfs4_setup_sequence(data->server,
4347                                 &data->arg.seq_args,
4348                                 &data->res.seq_res, task))
4349                 return;
4350         rpc_call_start(task);
4351         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4352 }
4353
4354 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4355 {
4356         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4357         nfs4_lock_prepare(task, calldata);
4358 }
4359
4360 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4361 {
4362         struct nfs4_lockdata *data = calldata;
4363
4364         dprintk("%s: begin!\n", __func__);
4365
4366         if (!nfs4_sequence_done(task, &data->res.seq_res))
4367                 return;
4368
4369         data->rpc_status = task->tk_status;
4370         if (data->arg.new_lock_owner != 0) {
4371                 if (data->rpc_status == 0)
4372                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4373                 else
4374                         goto out;
4375         }
4376         if (data->rpc_status == 0) {
4377                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4378                                         sizeof(data->lsp->ls_stateid.data));
4379                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4380                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4381         }
4382 out:
4383         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4384 }
4385
4386 static void nfs4_lock_release(void *calldata)
4387 {
4388         struct nfs4_lockdata *data = calldata;
4389
4390         dprintk("%s: begin!\n", __func__);
4391         nfs_free_seqid(data->arg.open_seqid);
4392         if (data->cancelled != 0) {
4393                 struct rpc_task *task;
4394                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4395                                 data->arg.lock_seqid);
4396                 if (!IS_ERR(task))
4397                         rpc_put_task_async(task);
4398                 dprintk("%s: cancelling lock!\n", __func__);
4399         } else
4400                 nfs_free_seqid(data->arg.lock_seqid);
4401         nfs4_put_lock_state(data->lsp);
4402         put_nfs_open_context(data->ctx);
4403         kfree(data);
4404         dprintk("%s: done!\n", __func__);
4405 }
4406
4407 static const struct rpc_call_ops nfs4_lock_ops = {
4408         .rpc_call_prepare = nfs4_lock_prepare,
4409         .rpc_call_done = nfs4_lock_done,
4410         .rpc_release = nfs4_lock_release,
4411 };
4412
4413 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4414         .rpc_call_prepare = nfs4_recover_lock_prepare,
4415         .rpc_call_done = nfs4_lock_done,
4416         .rpc_release = nfs4_lock_release,
4417 };
4418
4419 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4420 {
4421         switch (error) {
4422         case -NFS4ERR_ADMIN_REVOKED:
4423         case -NFS4ERR_BAD_STATEID:
4424                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4425                 if (new_lock_owner != 0 ||
4426                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4427                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4428                 break;
4429         case -NFS4ERR_STALE_STATEID:
4430                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4431         case -NFS4ERR_EXPIRED:
4432                 nfs4_schedule_lease_recovery(server->nfs_client);
4433         };
4434 }
4435
4436 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4437 {
4438         struct nfs4_lockdata *data;
4439         struct rpc_task *task;
4440         struct rpc_message msg = {
4441                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4442                 .rpc_cred = state->owner->so_cred,
4443         };
4444         struct rpc_task_setup task_setup_data = {
4445                 .rpc_client = NFS_CLIENT(state->inode),
4446                 .rpc_message = &msg,
4447                 .callback_ops = &nfs4_lock_ops,
4448                 .workqueue = nfsiod_workqueue,
4449                 .flags = RPC_TASK_ASYNC,
4450         };
4451         int ret;
4452
4453         dprintk("%s: begin!\n", __func__);
4454         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4455                         fl->fl_u.nfs4_fl.owner,
4456                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4457         if (data == NULL)
4458                 return -ENOMEM;
4459         if (IS_SETLKW(cmd))
4460                 data->arg.block = 1;
4461         if (recovery_type > NFS_LOCK_NEW) {
4462                 if (recovery_type == NFS_LOCK_RECLAIM)
4463                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4464                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4465         }
4466         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4467         msg.rpc_argp = &data->arg;
4468         msg.rpc_resp = &data->res;
4469         task_setup_data.callback_data = data;
4470         task = rpc_run_task(&task_setup_data);
4471         if (IS_ERR(task))
4472                 return PTR_ERR(task);
4473         ret = nfs4_wait_for_completion_rpc_task(task);
4474         if (ret == 0) {
4475                 ret = data->rpc_status;
4476                 if (ret)
4477                         nfs4_handle_setlk_error(data->server, data->lsp,
4478                                         data->arg.new_lock_owner, ret);
4479         } else
4480                 data->cancelled = 1;
4481         rpc_put_task(task);
4482         dprintk("%s: done, ret = %d!\n", __func__, ret);
4483         return ret;
4484 }
4485
4486 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4487 {
4488         struct nfs_server *server = NFS_SERVER(state->inode);
4489         struct nfs4_exception exception = { };
4490         int err;
4491
4492         do {
4493                 /* Cache the lock if possible... */
4494                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4495                         return 0;
4496                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4497                 if (err != -NFS4ERR_DELAY)
4498                         break;
4499                 nfs4_handle_exception(server, err, &exception);
4500         } while (exception.retry);
4501         return err;
4502 }
4503
4504 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4505 {
4506         struct nfs_server *server = NFS_SERVER(state->inode);
4507         struct nfs4_exception exception = { };
4508         int err;
4509
4510         err = nfs4_set_lock_state(state, request);
4511         if (err != 0)
4512                 return err;
4513         do {
4514                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4515                         return 0;
4516                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4517                 switch (err) {
4518                 default:
4519                         goto out;
4520                 case -NFS4ERR_GRACE:
4521                 case -NFS4ERR_DELAY:
4522                         nfs4_handle_exception(server, err, &exception);
4523                         err = 0;
4524                 }
4525         } while (exception.retry);
4526 out:
4527         return err;
4528 }
4529
4530 #if defined(CONFIG_NFS_V4_1)
4531 static int nfs41_check_expired_locks(struct nfs4_state *state)
4532 {
4533         int status, ret = NFS_OK;
4534         struct nfs4_lock_state *lsp;
4535         struct nfs_server *server = NFS_SERVER(state->inode);
4536
4537         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4538                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4539                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4540                         if (status != NFS_OK) {
4541                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4542                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4543                                 ret = status;
4544                         }
4545                 }
4546         };
4547
4548         return ret;
4549 }
4550
4551 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4552 {
4553         int status = NFS_OK;
4554
4555         if (test_bit(LK_STATE_IN_USE, &state->flags))
4556                 status = nfs41_check_expired_locks(state);
4557         if (status == NFS_OK)
4558                 return status;
4559         return nfs4_lock_expired(state, request);
4560 }
4561 #endif
4562
4563 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4564 {
4565         struct nfs_inode *nfsi = NFS_I(state->inode);
4566         unsigned char fl_flags = request->fl_flags;
4567         int status = -ENOLCK;
4568
4569         if ((fl_flags & FL_POSIX) &&
4570                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4571                 goto out;
4572         /* Is this a delegated open? */
4573         status = nfs4_set_lock_state(state, request);
4574         if (status != 0)
4575                 goto out;
4576         request->fl_flags |= FL_ACCESS;
4577         status = do_vfs_lock(request->fl_file, request);
4578         if (status < 0)
4579                 goto out;
4580         down_read(&nfsi->rwsem);
4581         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4582                 /* Yes: cache locks! */
4583                 /* ...but avoid races with delegation recall... */
4584                 request->fl_flags = fl_flags & ~FL_SLEEP;
4585                 status = do_vfs_lock(request->fl_file, request);
4586                 goto out_unlock;
4587         }
4588         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4589         if (status != 0)
4590                 goto out_unlock;
4591         /* Note: we always want to sleep here! */
4592         request->fl_flags = fl_flags | FL_SLEEP;
4593         if (do_vfs_lock(request->fl_file, request) < 0)
4594                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4595                         "manager!\n", __func__);
4596 out_unlock:
4597         up_read(&nfsi->rwsem);
4598 out:
4599         request->fl_flags = fl_flags;
4600         return status;
4601 }
4602
4603 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4604 {
4605         struct nfs4_exception exception = { };
4606         int err;
4607
4608         do {
4609                 err = _nfs4_proc_setlk(state, cmd, request);
4610                 if (err == -NFS4ERR_DENIED)
4611                         err = -EAGAIN;
4612                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4613                                 err, &exception);
4614         } while (exception.retry);
4615         return err;
4616 }
4617
4618 static int
4619 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4620 {
4621         struct nfs_open_context *ctx;
4622         struct nfs4_state *state;
4623         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4624         int status;
4625
4626         /* verify open state */
4627         ctx = nfs_file_open_context(filp);
4628         state = ctx->state;
4629
4630         if (request->fl_start < 0 || request->fl_end < 0)
4631                 return -EINVAL;
4632
4633         if (IS_GETLK(cmd)) {
4634                 if (state != NULL)
4635                         return nfs4_proc_getlk(state, F_GETLK, request);
4636                 return 0;
4637         }
4638
4639         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4640                 return -EINVAL;
4641
4642         if (request->fl_type == F_UNLCK) {
4643                 if (state != NULL)
4644                         return nfs4_proc_unlck(state, cmd, request);
4645                 return 0;
4646         }
4647
4648         if (state == NULL)
4649                 return -ENOLCK;
4650         do {
4651                 status = nfs4_proc_setlk(state, cmd, request);
4652                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4653                         break;
4654                 timeout = nfs4_set_lock_task_retry(timeout);
4655                 status = -ERESTARTSYS;
4656                 if (signalled())
4657                         break;
4658         } while(status < 0);
4659         return status;
4660 }
4661
4662 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4663 {
4664         struct nfs_server *server = NFS_SERVER(state->inode);
4665         struct nfs4_exception exception = { };
4666         int err;
4667
4668         err = nfs4_set_lock_state(state, fl);
4669         if (err != 0)
4670                 goto out;
4671         do {
4672                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4673                 switch (err) {
4674                         default:
4675                                 printk(KERN_ERR "NFS: %s: unhandled error "
4676                                         "%d.\n", __func__, err);
4677                         case 0:
4678                         case -ESTALE:
4679                                 goto out;
4680                         case -NFS4ERR_EXPIRED:
4681                                 nfs4_schedule_stateid_recovery(server, state);
4682                         case -NFS4ERR_STALE_CLIENTID:
4683                         case -NFS4ERR_STALE_STATEID:
4684                                 nfs4_schedule_lease_recovery(server->nfs_client);
4685                                 goto out;
4686                         case -NFS4ERR_BADSESSION:
4687                         case -NFS4ERR_BADSLOT:
4688                         case -NFS4ERR_BAD_HIGH_SLOT:
4689                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4690                         case -NFS4ERR_DEADSESSION:
4691                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4692                                 goto out;
4693                         case -ERESTARTSYS:
4694                                 /*
4695                                  * The show must go on: exit, but mark the
4696                                  * stateid as needing recovery.
4697                                  */
4698                         case -NFS4ERR_ADMIN_REVOKED:
4699                         case -NFS4ERR_BAD_STATEID:
4700                         case -NFS4ERR_OPENMODE:
4701                                 nfs4_schedule_stateid_recovery(server, state);
4702                                 err = 0;
4703                                 goto out;
4704                         case -EKEYEXPIRED:
4705                                 /*
4706                                  * User RPCSEC_GSS context has expired.
4707                                  * We cannot recover this stateid now, so
4708                                  * skip it and allow recovery thread to
4709                                  * proceed.
4710                                  */
4711                                 err = 0;
4712                                 goto out;
4713                         case -ENOMEM:
4714                         case -NFS4ERR_DENIED:
4715                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4716                                 err = 0;
4717                                 goto out;
4718                         case -NFS4ERR_DELAY:
4719                                 break;
4720                 }
4721                 err = nfs4_handle_exception(server, err, &exception);
4722         } while (exception.retry);
4723 out:
4724         return err;
4725 }
4726
4727 static void nfs4_release_lockowner_release(void *calldata)
4728 {
4729         kfree(calldata);
4730 }
4731
4732 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4733         .rpc_release = nfs4_release_lockowner_release,
4734 };
4735
4736 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4737 {
4738         struct nfs_server *server = lsp->ls_state->owner->so_server;
4739         struct nfs_release_lockowner_args *args;
4740         struct rpc_message msg = {
4741                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4742         };
4743
4744         if (server->nfs_client->cl_mvops->minor_version != 0)
4745                 return;
4746         args = kmalloc(sizeof(*args), GFP_NOFS);
4747         if (!args)
4748                 return;
4749         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4750         args->lock_owner.id = lsp->ls_seqid.owner_id;
4751         args->lock_owner.s_dev = server->s_dev;
4752         msg.rpc_argp = args;
4753         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4754 }
4755
4756 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4757
4758 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4759                                    const void *buf, size_t buflen,
4760                                    int flags, int type)
4761 {
4762         if (strcmp(key, "") != 0)
4763                 return -EINVAL;
4764
4765         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4766 }
4767
4768 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4769                                    void *buf, size_t buflen, int type)
4770 {
4771         if (strcmp(key, "") != 0)
4772                 return -EINVAL;
4773
4774         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4775 }
4776
4777 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4778                                        size_t list_len, const char *name,
4779                                        size_t name_len, int type)
4780 {
4781         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4782
4783         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4784                 return 0;
4785
4786         if (list && len <= list_len)
4787                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4788         return len;
4789 }
4790
4791 /*
4792  * nfs_fhget will use either the mounted_on_fileid or the fileid
4793  */
4794 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4795 {
4796         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4797                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4798               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4799               (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4800                 return;
4801
4802         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4803                 NFS_ATTR_FATTR_NLINK;
4804         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4805         fattr->nlink = 2;
4806 }
4807
4808 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4809                 struct nfs4_fs_locations *fs_locations, struct page *page)
4810 {
4811         struct nfs_server *server = NFS_SERVER(dir);
4812         u32 bitmask[2] = {
4813                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4814         };
4815         struct nfs4_fs_locations_arg args = {
4816                 .dir_fh = NFS_FH(dir),
4817                 .name = name,
4818                 .page = page,
4819                 .bitmask = bitmask,
4820         };
4821         struct nfs4_fs_locations_res res = {
4822                 .fs_locations = fs_locations,
4823         };
4824         struct rpc_message msg = {
4825                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4826                 .rpc_argp = &args,
4827                 .rpc_resp = &res,
4828         };
4829         int status;
4830
4831         dprintk("%s: start\n", __func__);
4832
4833         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4834          * is not supported */
4835         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4836                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4837         else
4838                 bitmask[0] |= FATTR4_WORD0_FILEID;
4839
4840         nfs_fattr_init(&fs_locations->fattr);
4841         fs_locations->server = server;
4842         fs_locations->nlocations = 0;
4843         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4844         dprintk("%s: returned status = %d\n", __func__, status);
4845         return status;
4846 }
4847
4848 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4849 {
4850         int status;
4851         struct nfs4_secinfo_arg args = {
4852                 .dir_fh = NFS_FH(dir),
4853                 .name   = name,
4854         };
4855         struct nfs4_secinfo_res res = {
4856                 .flavors     = flavors,
4857         };
4858         struct rpc_message msg = {
4859                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4860                 .rpc_argp = &args,
4861                 .rpc_resp = &res,
4862         };
4863
4864         dprintk("NFS call  secinfo %s\n", name->name);
4865         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4866         dprintk("NFS reply  secinfo: %d\n", status);
4867         return status;
4868 }
4869
4870 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4871 {
4872         struct nfs4_exception exception = { };
4873         int err;
4874         do {
4875                 err = nfs4_handle_exception(NFS_SERVER(dir),
4876                                 _nfs4_proc_secinfo(dir, name, flavors),
4877                                 &exception);
4878         } while (exception.retry);
4879         return err;
4880 }
4881
4882 #ifdef CONFIG_NFS_V4_1
4883 /*
4884  * Check the exchange flags returned by the server for invalid flags, having
4885  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4886  * DS flags set.
4887  */
4888 static int nfs4_check_cl_exchange_flags(u32 flags)
4889 {
4890         if (flags & ~EXCHGID4_FLAG_MASK_R)
4891                 goto out_inval;
4892         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4893             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4894                 goto out_inval;
4895         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4896                 goto out_inval;
4897         return NFS_OK;
4898 out_inval:
4899         return -NFS4ERR_INVAL;
4900 }
4901
4902 static bool
4903 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4904 {
4905         if (a->server_scope_sz == b->server_scope_sz &&
4906             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4907                 return true;
4908
4909         return false;
4910 }
4911
4912 /*
4913  * nfs4_proc_exchange_id()
4914  *
4915  * Since the clientid has expired, all compounds using sessions
4916  * associated with the stale clientid will be returning
4917  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4918  * be in some phase of session reset.
4919  */
4920 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4921 {
4922         nfs4_verifier verifier;
4923         struct nfs41_exchange_id_args args = {
4924                 .client = clp,
4925                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4926         };
4927         struct nfs41_exchange_id_res res = {
4928                 .client = clp,
4929         };
4930         int status;
4931         struct rpc_message msg = {
4932                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4933                 .rpc_argp = &args,
4934                 .rpc_resp = &res,
4935                 .rpc_cred = cred,
4936         };
4937         __be32 *p;
4938
4939         dprintk("--> %s\n", __func__);
4940         BUG_ON(clp == NULL);
4941
4942         p = (u32 *)verifier.data;
4943         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4944         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4945         args.verifier = &verifier;
4946
4947         args.id_len = scnprintf(args.id, sizeof(args.id),
4948                                 "%s/%s.%s/%u",
4949                                 clp->cl_ipaddr,
4950                                 init_utsname()->nodename,
4951                                 init_utsname()->domainname,
4952                                 clp->cl_rpcclient->cl_auth->au_flavor);
4953
4954         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4955         if (unlikely(!res.server_scope)) {
4956                 status = -ENOMEM;
4957                 goto out;
4958         }
4959
4960         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
4961         if (unlikely(!res.impl_id)) {
4962                 status = -ENOMEM;
4963                 goto out_server_scope;
4964         }
4965
4966         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4967         if (!status)
4968                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4969
4970         if (!status) {
4971                 /* use the most recent implementation id */
4972                 kfree(clp->impl_id);
4973                 clp->impl_id = res.impl_id;
4974         } else
4975                 kfree(res.impl_id);
4976
4977         if (!status) {
4978                 if (clp->server_scope &&
4979                     !nfs41_same_server_scope(clp->server_scope,
4980                                              res.server_scope)) {
4981                         dprintk("%s: server_scope mismatch detected\n",
4982                                 __func__);
4983                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4984                         kfree(clp->server_scope);
4985                         clp->server_scope = NULL;
4986                 }
4987
4988                 if (!clp->server_scope) {
4989                         clp->server_scope = res.server_scope;
4990                         goto out;
4991                 }
4992         }
4993
4994 out_server_scope:
4995         kfree(res.server_scope);
4996 out:
4997         if (clp->impl_id)
4998                 dprintk("%s: Server Implementation ID: "
4999                         "domain: %s, name: %s, date: %llu,%u\n",
5000                         __func__, clp->impl_id->domain, clp->impl_id->name,
5001                         clp->impl_id->date.seconds,
5002                         clp->impl_id->date.nseconds);
5003         dprintk("<-- %s status= %d\n", __func__, status);
5004         return status;
5005 }
5006
5007 struct nfs4_get_lease_time_data {
5008         struct nfs4_get_lease_time_args *args;
5009         struct nfs4_get_lease_time_res *res;
5010         struct nfs_client *clp;
5011 };
5012
5013 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5014                                         void *calldata)
5015 {
5016         int ret;
5017         struct nfs4_get_lease_time_data *data =
5018                         (struct nfs4_get_lease_time_data *)calldata;
5019
5020         dprintk("--> %s\n", __func__);
5021         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5022         /* just setup sequence, do not trigger session recovery
5023            since we're invoked within one */
5024         ret = nfs41_setup_sequence(data->clp->cl_session,
5025                                    &data->args->la_seq_args,
5026                                    &data->res->lr_seq_res, task);
5027
5028         BUG_ON(ret == -EAGAIN);
5029         rpc_call_start(task);
5030         dprintk("<-- %s\n", __func__);
5031 }
5032
5033 /*
5034  * Called from nfs4_state_manager thread for session setup, so don't recover
5035  * from sequence operation or clientid errors.
5036  */
5037 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5038 {
5039         struct nfs4_get_lease_time_data *data =
5040                         (struct nfs4_get_lease_time_data *)calldata;
5041
5042         dprintk("--> %s\n", __func__);
5043         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5044                 return;
5045         switch (task->tk_status) {
5046         case -NFS4ERR_DELAY:
5047         case -NFS4ERR_GRACE:
5048                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5049                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5050                 task->tk_status = 0;
5051                 /* fall through */
5052         case -NFS4ERR_RETRY_UNCACHED_REP:
5053                 rpc_restart_call_prepare(task);
5054                 return;
5055         }
5056         dprintk("<-- %s\n", __func__);
5057 }
5058
5059 struct rpc_call_ops nfs4_get_lease_time_ops = {
5060         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5061         .rpc_call_done = nfs4_get_lease_time_done,
5062 };
5063
5064 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5065 {
5066         struct rpc_task *task;
5067         struct nfs4_get_lease_time_args args;
5068         struct nfs4_get_lease_time_res res = {
5069                 .lr_fsinfo = fsinfo,
5070         };
5071         struct nfs4_get_lease_time_data data = {
5072                 .args = &args,
5073                 .res = &res,
5074                 .clp = clp,
5075         };
5076         struct rpc_message msg = {
5077                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5078                 .rpc_argp = &args,
5079                 .rpc_resp = &res,
5080         };
5081         struct rpc_task_setup task_setup = {
5082                 .rpc_client = clp->cl_rpcclient,
5083                 .rpc_message = &msg,
5084                 .callback_ops = &nfs4_get_lease_time_ops,
5085                 .callback_data = &data,
5086                 .flags = RPC_TASK_TIMEOUT,
5087         };
5088         int status;
5089
5090         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5091         dprintk("--> %s\n", __func__);
5092         task = rpc_run_task(&task_setup);
5093
5094         if (IS_ERR(task))
5095                 status = PTR_ERR(task);
5096         else {
5097                 status = task->tk_status;
5098                 rpc_put_task(task);
5099         }
5100         dprintk("<-- %s return %d\n", __func__, status);
5101
5102         return status;
5103 }
5104
5105 /*
5106  * Reset a slot table
5107  */
5108 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5109                                  int ivalue)
5110 {
5111         struct nfs4_slot *new = NULL;
5112         int i;
5113         int ret = 0;
5114
5115         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5116                 max_reqs, tbl->max_slots);
5117
5118         /* Does the newly negotiated max_reqs match the existing slot table? */
5119         if (max_reqs != tbl->max_slots) {
5120                 ret = -ENOMEM;
5121                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
5122                               GFP_NOFS);
5123                 if (!new)
5124                         goto out;
5125                 ret = 0;
5126                 kfree(tbl->slots);
5127         }
5128         spin_lock(&tbl->slot_tbl_lock);
5129         if (new) {
5130                 tbl->slots = new;
5131                 tbl->max_slots = max_reqs;
5132         }
5133         for (i = 0; i < tbl->max_slots; ++i)
5134                 tbl->slots[i].seq_nr = ivalue;
5135         spin_unlock(&tbl->slot_tbl_lock);
5136         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5137                 tbl, tbl->slots, tbl->max_slots);
5138 out:
5139         dprintk("<-- %s: return %d\n", __func__, ret);
5140         return ret;
5141 }
5142
5143 /* Destroy the slot table */
5144 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5145 {
5146         if (session->fc_slot_table.slots != NULL) {
5147                 kfree(session->fc_slot_table.slots);
5148                 session->fc_slot_table.slots = NULL;
5149         }
5150         if (session->bc_slot_table.slots != NULL) {
5151                 kfree(session->bc_slot_table.slots);
5152                 session->bc_slot_table.slots = NULL;
5153         }
5154         return;
5155 }
5156
5157 /*
5158  * Initialize slot table
5159  */
5160 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
5161                 int max_slots, int ivalue)
5162 {
5163         struct nfs4_slot *slot;
5164         int ret = -ENOMEM;
5165
5166         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
5167
5168         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
5169
5170         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
5171         if (!slot)
5172                 goto out;
5173         ret = 0;
5174
5175         spin_lock(&tbl->slot_tbl_lock);
5176         tbl->max_slots = max_slots;
5177         tbl->slots = slot;
5178         tbl->highest_used_slotid = NFS4_NO_SLOT;  /* no slot is currently used */
5179         spin_unlock(&tbl->slot_tbl_lock);
5180         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5181                 tbl, tbl->slots, tbl->max_slots);
5182 out:
5183         dprintk("<-- %s: return %d\n", __func__, ret);
5184         return ret;
5185 }
5186
5187 /*
5188  * Initialize or reset the forechannel and backchannel tables
5189  */
5190 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5191 {
5192         struct nfs4_slot_table *tbl;
5193         int status;
5194
5195         dprintk("--> %s\n", __func__);
5196         /* Fore channel */
5197         tbl = &ses->fc_slot_table;
5198         if (tbl->slots == NULL) {
5199                 status = nfs4_init_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5200                 if (status) /* -ENOMEM */
5201                         return status;
5202         } else {
5203                 status = nfs4_reset_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5204                 if (status)
5205                         return status;
5206         }
5207         /* Back channel */
5208         tbl = &ses->bc_slot_table;
5209         if (tbl->slots == NULL) {
5210                 status = nfs4_init_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5211                 if (status)
5212                         /* Fore and back channel share a connection so get
5213                          * both slot tables or neither */
5214                         nfs4_destroy_slot_tables(ses);
5215         } else
5216                 status = nfs4_reset_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5217         return status;
5218 }
5219
5220 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5221 {
5222         struct nfs4_session *session;
5223         struct nfs4_slot_table *tbl;
5224
5225         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5226         if (!session)
5227                 return NULL;
5228
5229         tbl = &session->fc_slot_table;
5230         tbl->highest_used_slotid = NFS4_NO_SLOT;
5231         spin_lock_init(&tbl->slot_tbl_lock);
5232         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5233         init_completion(&tbl->complete);
5234
5235         tbl = &session->bc_slot_table;
5236         tbl->highest_used_slotid = NFS4_NO_SLOT;
5237         spin_lock_init(&tbl->slot_tbl_lock);
5238         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5239         init_completion(&tbl->complete);
5240
5241         session->session_state = 1<<NFS4_SESSION_INITING;
5242
5243         session->clp = clp;
5244         return session;
5245 }
5246
5247 void nfs4_destroy_session(struct nfs4_session *session)
5248 {
5249         struct rpc_xprt *xprt;
5250
5251         nfs4_proc_destroy_session(session);
5252
5253         rcu_read_lock();
5254         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5255         rcu_read_unlock();
5256         dprintk("%s Destroy backchannel for xprt %p\n",
5257                 __func__, xprt);
5258         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5259         nfs4_destroy_slot_tables(session);
5260         kfree(session);
5261 }
5262
5263 /*
5264  * Initialize the values to be used by the client in CREATE_SESSION
5265  * If nfs4_init_session set the fore channel request and response sizes,
5266  * use them.
5267  *
5268  * Set the back channel max_resp_sz_cached to zero to force the client to
5269  * always set csa_cachethis to FALSE because the current implementation
5270  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5271  */
5272 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5273 {
5274         struct nfs4_session *session = args->client->cl_session;
5275         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5276                      mxresp_sz = session->fc_attrs.max_resp_sz;
5277
5278         if (mxrqst_sz == 0)
5279                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5280         if (mxresp_sz == 0)
5281                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5282         /* Fore channel attributes */
5283         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5284         args->fc_attrs.max_resp_sz = mxresp_sz;
5285         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5286         args->fc_attrs.max_reqs = max_session_slots;
5287
5288         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5289                 "max_ops=%u max_reqs=%u\n",
5290                 __func__,
5291                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5292                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5293
5294         /* Back channel attributes */
5295         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5296         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5297         args->bc_attrs.max_resp_sz_cached = 0;
5298         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5299         args->bc_attrs.max_reqs = 1;
5300
5301         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5302                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5303                 __func__,
5304                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5305                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5306                 args->bc_attrs.max_reqs);
5307 }
5308
5309 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5310 {
5311         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5312         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5313
5314         if (rcvd->max_resp_sz > sent->max_resp_sz)
5315                 return -EINVAL;
5316         /*
5317          * Our requested max_ops is the minimum we need; we're not
5318          * prepared to break up compounds into smaller pieces than that.
5319          * So, no point even trying to continue if the server won't
5320          * cooperate:
5321          */
5322         if (rcvd->max_ops < sent->max_ops)
5323                 return -EINVAL;
5324         if (rcvd->max_reqs == 0)
5325                 return -EINVAL;
5326         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5327                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5328         return 0;
5329 }
5330
5331 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5332 {
5333         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5334         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5335
5336         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5337                 return -EINVAL;
5338         if (rcvd->max_resp_sz < sent->max_resp_sz)
5339                 return -EINVAL;
5340         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5341                 return -EINVAL;
5342         /* These would render the backchannel useless: */
5343         if (rcvd->max_ops != sent->max_ops)
5344                 return -EINVAL;
5345         if (rcvd->max_reqs != sent->max_reqs)
5346                 return -EINVAL;
5347         return 0;
5348 }
5349
5350 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5351                                      struct nfs4_session *session)
5352 {
5353         int ret;
5354
5355         ret = nfs4_verify_fore_channel_attrs(args, session);
5356         if (ret)
5357                 return ret;
5358         return nfs4_verify_back_channel_attrs(args, session);
5359 }
5360
5361 static int _nfs4_proc_create_session(struct nfs_client *clp)
5362 {
5363         struct nfs4_session *session = clp->cl_session;
5364         struct nfs41_create_session_args args = {
5365                 .client = clp,
5366                 .cb_program = NFS4_CALLBACK,
5367         };
5368         struct nfs41_create_session_res res = {
5369                 .client = clp,
5370         };
5371         struct rpc_message msg = {
5372                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5373                 .rpc_argp = &args,
5374                 .rpc_resp = &res,
5375         };
5376         int status;
5377
5378         nfs4_init_channel_attrs(&args);
5379         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5380
5381         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5382
5383         if (!status)
5384                 /* Verify the session's negotiated channel_attrs values */
5385                 status = nfs4_verify_channel_attrs(&args, session);
5386         if (!status) {
5387                 /* Increment the clientid slot sequence id */
5388                 clp->cl_seqid++;
5389         }
5390
5391         return status;
5392 }
5393
5394 /*
5395  * Issues a CREATE_SESSION operation to the server.
5396  * It is the responsibility of the caller to verify the session is
5397  * expired before calling this routine.
5398  */
5399 int nfs4_proc_create_session(struct nfs_client *clp)
5400 {
5401         int status;
5402         unsigned *ptr;
5403         struct nfs4_session *session = clp->cl_session;
5404
5405         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5406
5407         status = _nfs4_proc_create_session(clp);
5408         if (status)
5409                 goto out;
5410
5411         /* Init or reset the session slot tables */
5412         status = nfs4_setup_session_slot_tables(session);
5413         dprintk("slot table setup returned %d\n", status);
5414         if (status)
5415                 goto out;
5416
5417         ptr = (unsigned *)&session->sess_id.data[0];
5418         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5419                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5420 out:
5421         dprintk("<-- %s\n", __func__);
5422         return status;
5423 }
5424
5425 /*
5426  * Issue the over-the-wire RPC DESTROY_SESSION.
5427  * The caller must serialize access to this routine.
5428  */
5429 int nfs4_proc_destroy_session(struct nfs4_session *session)
5430 {
5431         int status = 0;
5432         struct rpc_message msg;
5433
5434         dprintk("--> nfs4_proc_destroy_session\n");
5435
5436         /* session is still being setup */
5437         if (session->clp->cl_cons_state != NFS_CS_READY)
5438                 return status;
5439
5440         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5441         msg.rpc_argp = session;
5442         msg.rpc_resp = NULL;
5443         msg.rpc_cred = NULL;
5444         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5445
5446         if (status)
5447                 printk(KERN_WARNING
5448                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5449                         "Session has been destroyed regardless...\n", status);
5450
5451         dprintk("<-- nfs4_proc_destroy_session\n");
5452         return status;
5453 }
5454
5455 int nfs4_init_session(struct nfs_server *server)
5456 {
5457         struct nfs_client *clp = server->nfs_client;
5458         struct nfs4_session *session;
5459         unsigned int rsize, wsize;
5460         int ret;
5461
5462         if (!nfs4_has_session(clp))
5463                 return 0;
5464
5465         session = clp->cl_session;
5466         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5467                 return 0;
5468
5469         rsize = server->rsize;
5470         if (rsize == 0)
5471                 rsize = NFS_MAX_FILE_IO_SIZE;
5472         wsize = server->wsize;
5473         if (wsize == 0)
5474                 wsize = NFS_MAX_FILE_IO_SIZE;
5475
5476         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5477         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5478
5479         ret = nfs4_recover_expired_lease(server);
5480         if (!ret)
5481                 ret = nfs4_check_client_ready(clp);
5482         return ret;
5483 }
5484
5485 int nfs4_init_ds_session(struct nfs_client *clp)
5486 {
5487         struct nfs4_session *session = clp->cl_session;
5488         int ret;
5489
5490         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5491                 return 0;
5492
5493         ret = nfs4_client_recover_expired_lease(clp);
5494         if (!ret)
5495                 /* Test for the DS role */
5496                 if (!is_ds_client(clp))
5497                         ret = -ENODEV;
5498         if (!ret)
5499                 ret = nfs4_check_client_ready(clp);
5500         return ret;
5501
5502 }
5503 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5504
5505
5506 /*
5507  * Renew the cl_session lease.
5508  */
5509 struct nfs4_sequence_data {
5510         struct nfs_client *clp;
5511         struct nfs4_sequence_args args;
5512         struct nfs4_sequence_res res;
5513 };
5514
5515 static void nfs41_sequence_release(void *data)
5516 {
5517         struct nfs4_sequence_data *calldata = data;
5518         struct nfs_client *clp = calldata->clp;
5519
5520         if (atomic_read(&clp->cl_count) > 1)
5521                 nfs4_schedule_state_renewal(clp);
5522         nfs_put_client(clp);
5523         kfree(calldata);
5524 }
5525
5526 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5527 {
5528         switch(task->tk_status) {
5529         case -NFS4ERR_DELAY:
5530                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5531                 return -EAGAIN;
5532         default:
5533                 nfs4_schedule_lease_recovery(clp);
5534         }
5535         return 0;
5536 }
5537
5538 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5539 {
5540         struct nfs4_sequence_data *calldata = data;
5541         struct nfs_client *clp = calldata->clp;
5542
5543         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5544                 return;
5545
5546         if (task->tk_status < 0) {
5547                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5548                 if (atomic_read(&clp->cl_count) == 1)
5549                         goto out;
5550
5551                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5552                         rpc_restart_call_prepare(task);
5553                         return;
5554                 }
5555         }
5556         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5557 out:
5558         dprintk("<-- %s\n", __func__);
5559 }
5560
5561 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5562 {
5563         struct nfs4_sequence_data *calldata = data;
5564         struct nfs_client *clp = calldata->clp;
5565         struct nfs4_sequence_args *args;
5566         struct nfs4_sequence_res *res;
5567
5568         args = task->tk_msg.rpc_argp;
5569         res = task->tk_msg.rpc_resp;
5570
5571         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5572                 return;
5573         rpc_call_start(task);
5574 }
5575
5576 static const struct rpc_call_ops nfs41_sequence_ops = {
5577         .rpc_call_done = nfs41_sequence_call_done,
5578         .rpc_call_prepare = nfs41_sequence_prepare,
5579         .rpc_release = nfs41_sequence_release,
5580 };
5581
5582 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5583 {
5584         struct nfs4_sequence_data *calldata;
5585         struct rpc_message msg = {
5586                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5587                 .rpc_cred = cred,
5588         };
5589         struct rpc_task_setup task_setup_data = {
5590                 .rpc_client = clp->cl_rpcclient,
5591                 .rpc_message = &msg,
5592                 .callback_ops = &nfs41_sequence_ops,
5593                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5594         };
5595
5596         if (!atomic_inc_not_zero(&clp->cl_count))
5597                 return ERR_PTR(-EIO);
5598         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5599         if (calldata == NULL) {
5600                 nfs_put_client(clp);
5601                 return ERR_PTR(-ENOMEM);
5602         }
5603         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5604         msg.rpc_argp = &calldata->args;
5605         msg.rpc_resp = &calldata->res;
5606         calldata->clp = clp;
5607         task_setup_data.callback_data = calldata;
5608
5609         return rpc_run_task(&task_setup_data);
5610 }
5611
5612 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5613 {
5614         struct rpc_task *task;
5615         int ret = 0;
5616
5617         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5618                 return 0;
5619         task = _nfs41_proc_sequence(clp, cred);
5620         if (IS_ERR(task))
5621                 ret = PTR_ERR(task);
5622         else
5623                 rpc_put_task_async(task);
5624         dprintk("<-- %s status=%d\n", __func__, ret);
5625         return ret;
5626 }
5627
5628 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5629 {
5630         struct rpc_task *task;
5631         int ret;
5632
5633         task = _nfs41_proc_sequence(clp, cred);
5634         if (IS_ERR(task)) {
5635                 ret = PTR_ERR(task);
5636                 goto out;
5637         }
5638         ret = rpc_wait_for_completion_task(task);
5639         if (!ret) {
5640                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5641
5642                 if (task->tk_status == 0)
5643                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5644                 ret = task->tk_status;
5645         }
5646         rpc_put_task(task);
5647 out:
5648         dprintk("<-- %s status=%d\n", __func__, ret);
5649         return ret;
5650 }
5651
5652 struct nfs4_reclaim_complete_data {
5653         struct nfs_client *clp;
5654         struct nfs41_reclaim_complete_args arg;
5655         struct nfs41_reclaim_complete_res res;
5656 };
5657
5658 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5659 {
5660         struct nfs4_reclaim_complete_data *calldata = data;
5661
5662         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5663         if (nfs41_setup_sequence(calldata->clp->cl_session,
5664                                 &calldata->arg.seq_args,
5665                                 &calldata->res.seq_res, task))
5666                 return;
5667
5668         rpc_call_start(task);
5669 }
5670
5671 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5672 {
5673         switch(task->tk_status) {
5674         case 0:
5675         case -NFS4ERR_COMPLETE_ALREADY:
5676         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5677                 break;
5678         case -NFS4ERR_DELAY:
5679                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5680                 /* fall through */
5681         case -NFS4ERR_RETRY_UNCACHED_REP:
5682                 return -EAGAIN;
5683         default:
5684                 nfs4_schedule_lease_recovery(clp);
5685         }
5686         return 0;
5687 }
5688
5689 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5690 {
5691         struct nfs4_reclaim_complete_data *calldata = data;
5692         struct nfs_client *clp = calldata->clp;
5693         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5694
5695         dprintk("--> %s\n", __func__);
5696         if (!nfs41_sequence_done(task, res))
5697                 return;
5698
5699         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5700                 rpc_restart_call_prepare(task);
5701                 return;
5702         }
5703         dprintk("<-- %s\n", __func__);
5704 }
5705
5706 static void nfs4_free_reclaim_complete_data(void *data)
5707 {
5708         struct nfs4_reclaim_complete_data *calldata = data;
5709
5710         kfree(calldata);
5711 }
5712
5713 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5714         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5715         .rpc_call_done = nfs4_reclaim_complete_done,
5716         .rpc_release = nfs4_free_reclaim_complete_data,
5717 };
5718
5719 /*
5720  * Issue a global reclaim complete.
5721  */
5722 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5723 {
5724         struct nfs4_reclaim_complete_data *calldata;
5725         struct rpc_task *task;
5726         struct rpc_message msg = {
5727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5728         };
5729         struct rpc_task_setup task_setup_data = {
5730                 .rpc_client = clp->cl_rpcclient,
5731                 .rpc_message = &msg,
5732                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5733                 .flags = RPC_TASK_ASYNC,
5734         };
5735         int status = -ENOMEM;
5736
5737         dprintk("--> %s\n", __func__);
5738         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5739         if (calldata == NULL)
5740                 goto out;
5741         calldata->clp = clp;
5742         calldata->arg.one_fs = 0;
5743
5744         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5745         msg.rpc_argp = &calldata->arg;
5746         msg.rpc_resp = &calldata->res;
5747         task_setup_data.callback_data = calldata;
5748         task = rpc_run_task(&task_setup_data);
5749         if (IS_ERR(task)) {
5750                 status = PTR_ERR(task);
5751                 goto out;
5752         }
5753         status = nfs4_wait_for_completion_rpc_task(task);
5754         if (status == 0)
5755                 status = task->tk_status;
5756         rpc_put_task(task);
5757         return 0;
5758 out:
5759         dprintk("<-- %s status=%d\n", __func__, status);
5760         return status;
5761 }
5762
5763 static void
5764 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5765 {
5766         struct nfs4_layoutget *lgp = calldata;
5767         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5768
5769         dprintk("--> %s\n", __func__);
5770         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5771          * right now covering the LAYOUTGET we are about to send.
5772          * However, that is not so catastrophic, and there seems
5773          * to be no way to prevent it completely.
5774          */
5775         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5776                                 &lgp->res.seq_res, task))
5777                 return;
5778         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5779                                           NFS_I(lgp->args.inode)->layout,
5780                                           lgp->args.ctx->state)) {
5781                 rpc_exit(task, NFS4_OK);
5782                 return;
5783         }
5784         rpc_call_start(task);
5785 }
5786
5787 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5788 {
5789         struct nfs4_layoutget *lgp = calldata;
5790         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5791
5792         dprintk("--> %s\n", __func__);
5793
5794         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5795                 return;
5796
5797         switch (task->tk_status) {
5798         case 0:
5799                 break;
5800         case -NFS4ERR_LAYOUTTRYLATER:
5801         case -NFS4ERR_RECALLCONFLICT:
5802                 task->tk_status = -NFS4ERR_DELAY;
5803                 /* Fall through */
5804         default:
5805                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5806                         rpc_restart_call_prepare(task);
5807                         return;
5808                 }
5809         }
5810         dprintk("<-- %s\n", __func__);
5811 }
5812
5813 static void nfs4_layoutget_release(void *calldata)
5814 {
5815         struct nfs4_layoutget *lgp = calldata;
5816
5817         dprintk("--> %s\n", __func__);
5818         put_nfs_open_context(lgp->args.ctx);
5819         kfree(calldata);
5820         dprintk("<-- %s\n", __func__);
5821 }
5822
5823 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5824         .rpc_call_prepare = nfs4_layoutget_prepare,
5825         .rpc_call_done = nfs4_layoutget_done,
5826         .rpc_release = nfs4_layoutget_release,
5827 };
5828
5829 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5830 {
5831         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5832         struct rpc_task *task;
5833         struct rpc_message msg = {
5834                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5835                 .rpc_argp = &lgp->args,
5836                 .rpc_resp = &lgp->res,
5837         };
5838         struct rpc_task_setup task_setup_data = {
5839                 .rpc_client = server->client,
5840                 .rpc_message = &msg,
5841                 .callback_ops = &nfs4_layoutget_call_ops,
5842                 .callback_data = lgp,
5843                 .flags = RPC_TASK_ASYNC,
5844         };
5845         int status = 0;
5846
5847         dprintk("--> %s\n", __func__);
5848
5849         lgp->res.layoutp = &lgp->args.layout;
5850         lgp->res.seq_res.sr_slot = NULL;
5851         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5852         task = rpc_run_task(&task_setup_data);
5853         if (IS_ERR(task))
5854                 return PTR_ERR(task);
5855         status = nfs4_wait_for_completion_rpc_task(task);
5856         if (status == 0)
5857                 status = task->tk_status;
5858         if (status == 0)
5859                 status = pnfs_layout_process(lgp);
5860         rpc_put_task(task);
5861         dprintk("<-- %s status=%d\n", __func__, status);
5862         return status;
5863 }
5864
5865 static void
5866 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5867 {
5868         struct nfs4_layoutreturn *lrp = calldata;
5869
5870         dprintk("--> %s\n", __func__);
5871         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5872                                 &lrp->res.seq_res, task))
5873                 return;
5874         rpc_call_start(task);
5875 }
5876
5877 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5878 {
5879         struct nfs4_layoutreturn *lrp = calldata;
5880         struct nfs_server *server;
5881         struct pnfs_layout_hdr *lo = lrp->args.layout;
5882
5883         dprintk("--> %s\n", __func__);
5884
5885         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5886                 return;
5887
5888         server = NFS_SERVER(lrp->args.inode);
5889         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5890                 rpc_restart_call_prepare(task);
5891                 return;
5892         }
5893         spin_lock(&lo->plh_inode->i_lock);
5894         if (task->tk_status == 0) {
5895                 if (lrp->res.lrs_present) {
5896                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5897                 } else
5898                         BUG_ON(!list_empty(&lo->plh_segs));
5899         }
5900         lo->plh_block_lgets--;
5901         spin_unlock(&lo->plh_inode->i_lock);
5902         dprintk("<-- %s\n", __func__);
5903 }
5904
5905 static void nfs4_layoutreturn_release(void *calldata)
5906 {
5907         struct nfs4_layoutreturn *lrp = calldata;
5908
5909         dprintk("--> %s\n", __func__);
5910         put_layout_hdr(lrp->args.layout);
5911         kfree(calldata);
5912         dprintk("<-- %s\n", __func__);
5913 }
5914
5915 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5916         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5917         .rpc_call_done = nfs4_layoutreturn_done,
5918         .rpc_release = nfs4_layoutreturn_release,
5919 };
5920
5921 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5922 {
5923         struct rpc_task *task;
5924         struct rpc_message msg = {
5925                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5926                 .rpc_argp = &lrp->args,
5927                 .rpc_resp = &lrp->res,
5928         };
5929         struct rpc_task_setup task_setup_data = {
5930                 .rpc_client = lrp->clp->cl_rpcclient,
5931                 .rpc_message = &msg,
5932                 .callback_ops = &nfs4_layoutreturn_call_ops,
5933                 .callback_data = lrp,
5934         };
5935         int status;
5936
5937         dprintk("--> %s\n", __func__);
5938         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5939         task = rpc_run_task(&task_setup_data);
5940         if (IS_ERR(task))
5941                 return PTR_ERR(task);
5942         status = task->tk_status;
5943         dprintk("<-- %s status=%d\n", __func__, status);
5944         rpc_put_task(task);
5945         return status;
5946 }
5947
5948 /*
5949  * Retrieve the list of Data Server devices from the MDS.
5950  */
5951 static int _nfs4_getdevicelist(struct nfs_server *server,
5952                                     const struct nfs_fh *fh,
5953                                     struct pnfs_devicelist *devlist)
5954 {
5955         struct nfs4_getdevicelist_args args = {
5956                 .fh = fh,
5957                 .layoutclass = server->pnfs_curr_ld->id,
5958         };
5959         struct nfs4_getdevicelist_res res = {
5960                 .devlist = devlist,
5961         };
5962         struct rpc_message msg = {
5963                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5964                 .rpc_argp = &args,
5965                 .rpc_resp = &res,
5966         };
5967         int status;
5968
5969         dprintk("--> %s\n", __func__);
5970         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5971                                 &res.seq_res, 0);
5972         dprintk("<-- %s status=%d\n", __func__, status);
5973         return status;
5974 }
5975
5976 int nfs4_proc_getdevicelist(struct nfs_server *server,
5977                             const struct nfs_fh *fh,
5978                             struct pnfs_devicelist *devlist)
5979 {
5980         struct nfs4_exception exception = { };
5981         int err;
5982
5983         do {
5984                 err = nfs4_handle_exception(server,
5985                                 _nfs4_getdevicelist(server, fh, devlist),
5986                                 &exception);
5987         } while (exception.retry);
5988
5989         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5990                 err, devlist->num_devs);
5991
5992         return err;
5993 }
5994 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5995
5996 static int
5997 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5998 {
5999         struct nfs4_getdeviceinfo_args args = {
6000                 .pdev = pdev,
6001         };
6002         struct nfs4_getdeviceinfo_res res = {
6003                 .pdev = pdev,
6004         };
6005         struct rpc_message msg = {
6006                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6007                 .rpc_argp = &args,
6008                 .rpc_resp = &res,
6009         };
6010         int status;
6011
6012         dprintk("--> %s\n", __func__);
6013         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6014         dprintk("<-- %s status=%d\n", __func__, status);
6015
6016         return status;
6017 }
6018
6019 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6020 {
6021         struct nfs4_exception exception = { };
6022         int err;
6023
6024         do {
6025                 err = nfs4_handle_exception(server,
6026                                         _nfs4_proc_getdeviceinfo(server, pdev),
6027                                         &exception);
6028         } while (exception.retry);
6029         return err;
6030 }
6031 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6032
6033 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6034 {
6035         struct nfs4_layoutcommit_data *data = calldata;
6036         struct nfs_server *server = NFS_SERVER(data->args.inode);
6037
6038         if (nfs4_setup_sequence(server, &data->args.seq_args,
6039                                 &data->res.seq_res, task))
6040                 return;
6041         rpc_call_start(task);
6042 }
6043
6044 static void
6045 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6046 {
6047         struct nfs4_layoutcommit_data *data = calldata;
6048         struct nfs_server *server = NFS_SERVER(data->args.inode);
6049
6050         if (!nfs4_sequence_done(task, &data->res.seq_res))
6051                 return;
6052
6053         switch (task->tk_status) { /* Just ignore these failures */
6054         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6055         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6056         case NFS4ERR_BADLAYOUT:     /* no layout */
6057         case NFS4ERR_GRACE:         /* loca_recalim always false */
6058                 task->tk_status = 0;
6059         }
6060
6061         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6062                 rpc_restart_call_prepare(task);
6063                 return;
6064         }
6065
6066         if (task->tk_status == 0)
6067                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6068                                                    data->res.fattr);
6069 }
6070
6071 static void nfs4_layoutcommit_release(void *calldata)
6072 {
6073         struct nfs4_layoutcommit_data *data = calldata;
6074         struct pnfs_layout_segment *lseg, *tmp;
6075         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6076
6077         pnfs_cleanup_layoutcommit(data);
6078         /* Matched by references in pnfs_set_layoutcommit */
6079         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6080                 list_del_init(&lseg->pls_lc_list);
6081                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6082                                        &lseg->pls_flags))
6083                         put_lseg(lseg);
6084         }
6085
6086         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6087         smp_mb__after_clear_bit();
6088         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6089
6090         put_rpccred(data->cred);
6091         kfree(data);
6092 }
6093
6094 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6095         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6096         .rpc_call_done = nfs4_layoutcommit_done,
6097         .rpc_release = nfs4_layoutcommit_release,
6098 };
6099
6100 int
6101 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6102 {
6103         struct rpc_message msg = {
6104                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6105                 .rpc_argp = &data->args,
6106                 .rpc_resp = &data->res,
6107                 .rpc_cred = data->cred,
6108         };
6109         struct rpc_task_setup task_setup_data = {
6110                 .task = &data->task,
6111                 .rpc_client = NFS_CLIENT(data->args.inode),
6112                 .rpc_message = &msg,
6113                 .callback_ops = &nfs4_layoutcommit_ops,
6114                 .callback_data = data,
6115                 .flags = RPC_TASK_ASYNC,
6116         };
6117         struct rpc_task *task;
6118         int status = 0;
6119
6120         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6121                 "lbw: %llu inode %lu\n",
6122                 data->task.tk_pid, sync,
6123                 data->args.lastbytewritten,
6124                 data->args.inode->i_ino);
6125
6126         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6127         task = rpc_run_task(&task_setup_data);
6128         if (IS_ERR(task))
6129                 return PTR_ERR(task);
6130         if (sync == false)
6131                 goto out;
6132         status = nfs4_wait_for_completion_rpc_task(task);
6133         if (status != 0)
6134                 goto out;
6135         status = task->tk_status;
6136 out:
6137         dprintk("%s: status %d\n", __func__, status);
6138         rpc_put_task(task);
6139         return status;
6140 }
6141
6142 static int
6143 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6144                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6145 {
6146         struct nfs41_secinfo_no_name_args args = {
6147                 .style = SECINFO_STYLE_CURRENT_FH,
6148         };
6149         struct nfs4_secinfo_res res = {
6150                 .flavors = flavors,
6151         };
6152         struct rpc_message msg = {
6153                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6154                 .rpc_argp = &args,
6155                 .rpc_resp = &res,
6156         };
6157         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6158 }
6159
6160 static int
6161 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6162                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6163 {
6164         struct nfs4_exception exception = { };
6165         int err;
6166         do {
6167                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6168                 switch (err) {
6169                 case 0:
6170                 case -NFS4ERR_WRONGSEC:
6171                 case -NFS4ERR_NOTSUPP:
6172                         break;
6173                 default:
6174                         err = nfs4_handle_exception(server, err, &exception);
6175                 }
6176         } while (exception.retry);
6177         return err;
6178 }
6179
6180 static int
6181 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6182                     struct nfs_fsinfo *info)
6183 {
6184         int err;
6185         struct page *page;
6186         rpc_authflavor_t flavor;
6187         struct nfs4_secinfo_flavors *flavors;
6188
6189         page = alloc_page(GFP_KERNEL);
6190         if (!page) {
6191                 err = -ENOMEM;
6192                 goto out;
6193         }
6194
6195         flavors = page_address(page);
6196         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6197
6198         /*
6199          * Fall back on "guess and check" method if
6200          * the server doesn't support SECINFO_NO_NAME
6201          */
6202         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6203                 err = nfs4_find_root_sec(server, fhandle, info);
6204                 goto out_freepage;
6205         }
6206         if (err)
6207                 goto out_freepage;
6208
6209         flavor = nfs_find_best_sec(flavors);
6210         if (err == 0)
6211                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6212
6213 out_freepage:
6214         put_page(page);
6215         if (err == -EACCES)
6216                 return -EPERM;
6217 out:
6218         return err;
6219 }
6220
6221 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6222 {
6223         int status;
6224         struct nfs41_test_stateid_args args = {
6225                 .stateid = stateid,
6226         };
6227         struct nfs41_test_stateid_res res;
6228         struct rpc_message msg = {
6229                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6230                 .rpc_argp = &args,
6231                 .rpc_resp = &res,
6232         };
6233
6234         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6235         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6236
6237         if (status == NFS_OK)
6238                 return res.status;
6239         return status;
6240 }
6241
6242 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6243 {
6244         struct nfs4_exception exception = { };
6245         int err;
6246         do {
6247                 err = nfs4_handle_exception(server,
6248                                 _nfs41_test_stateid(server, stateid),
6249                                 &exception);
6250         } while (exception.retry);
6251         return err;
6252 }
6253
6254 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6255 {
6256         struct nfs41_free_stateid_args args = {
6257                 .stateid = stateid,
6258         };
6259         struct nfs41_free_stateid_res res;
6260         struct rpc_message msg = {
6261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6262                 .rpc_argp = &args,
6263                 .rpc_resp = &res,
6264         };
6265
6266         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6267         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6268 }
6269
6270 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6271 {
6272         struct nfs4_exception exception = { };
6273         int err;
6274         do {
6275                 err = nfs4_handle_exception(server,
6276                                 _nfs4_free_stateid(server, stateid),
6277                                 &exception);
6278         } while (exception.retry);
6279         return err;
6280 }
6281 #endif /* CONFIG_NFS_V4_1 */
6282
6283 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6284         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6285         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6286         .recover_open   = nfs4_open_reclaim,
6287         .recover_lock   = nfs4_lock_reclaim,
6288         .establish_clid = nfs4_init_clientid,
6289         .get_clid_cred  = nfs4_get_setclientid_cred,
6290 };
6291
6292 #if defined(CONFIG_NFS_V4_1)
6293 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6294         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6295         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6296         .recover_open   = nfs4_open_reclaim,
6297         .recover_lock   = nfs4_lock_reclaim,
6298         .establish_clid = nfs41_init_clientid,
6299         .get_clid_cred  = nfs4_get_exchange_id_cred,
6300         .reclaim_complete = nfs41_proc_reclaim_complete,
6301 };
6302 #endif /* CONFIG_NFS_V4_1 */
6303
6304 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6305         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6306         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6307         .recover_open   = nfs4_open_expired,
6308         .recover_lock   = nfs4_lock_expired,
6309         .establish_clid = nfs4_init_clientid,
6310         .get_clid_cred  = nfs4_get_setclientid_cred,
6311 };
6312
6313 #if defined(CONFIG_NFS_V4_1)
6314 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6315         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6316         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6317         .recover_open   = nfs41_open_expired,
6318         .recover_lock   = nfs41_lock_expired,
6319         .establish_clid = nfs41_init_clientid,
6320         .get_clid_cred  = nfs4_get_exchange_id_cred,
6321 };
6322 #endif /* CONFIG_NFS_V4_1 */
6323
6324 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6325         .sched_state_renewal = nfs4_proc_async_renew,
6326         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6327         .renew_lease = nfs4_proc_renew,
6328 };
6329
6330 #if defined(CONFIG_NFS_V4_1)
6331 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6332         .sched_state_renewal = nfs41_proc_async_sequence,
6333         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6334         .renew_lease = nfs4_proc_sequence,
6335 };
6336 #endif
6337
6338 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6339         .minor_version = 0,
6340         .call_sync = _nfs4_call_sync,
6341         .validate_stateid = nfs4_validate_delegation_stateid,
6342         .find_root_sec = nfs4_find_root_sec,
6343         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6344         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6345         .state_renewal_ops = &nfs40_state_renewal_ops,
6346 };
6347
6348 #if defined(CONFIG_NFS_V4_1)
6349 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6350         .minor_version = 1,
6351         .call_sync = _nfs4_call_sync_session,
6352         .validate_stateid = nfs41_validate_delegation_stateid,
6353         .find_root_sec = nfs41_find_root_sec,
6354         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6355         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6356         .state_renewal_ops = &nfs41_state_renewal_ops,
6357 };
6358 #endif
6359
6360 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6361         [0] = &nfs_v4_0_minor_ops,
6362 #if defined(CONFIG_NFS_V4_1)
6363         [1] = &nfs_v4_1_minor_ops,
6364 #endif
6365 };
6366
6367 static const struct inode_operations nfs4_file_inode_operations = {
6368         .permission     = nfs_permission,
6369         .getattr        = nfs_getattr,
6370         .setattr        = nfs_setattr,
6371         .getxattr       = generic_getxattr,
6372         .setxattr       = generic_setxattr,
6373         .listxattr      = generic_listxattr,
6374         .removexattr    = generic_removexattr,
6375 };
6376
6377 const struct nfs_rpc_ops nfs_v4_clientops = {
6378         .version        = 4,                    /* protocol version */
6379         .dentry_ops     = &nfs4_dentry_operations,
6380         .dir_inode_ops  = &nfs4_dir_inode_operations,
6381         .file_inode_ops = &nfs4_file_inode_operations,
6382         .file_ops       = &nfs4_file_operations,
6383         .getroot        = nfs4_proc_get_root,
6384         .getattr        = nfs4_proc_getattr,
6385         .setattr        = nfs4_proc_setattr,
6386         .lookup         = nfs4_proc_lookup,
6387         .access         = nfs4_proc_access,
6388         .readlink       = nfs4_proc_readlink,
6389         .create         = nfs4_proc_create,
6390         .remove         = nfs4_proc_remove,
6391         .unlink_setup   = nfs4_proc_unlink_setup,
6392         .unlink_done    = nfs4_proc_unlink_done,
6393         .rename         = nfs4_proc_rename,
6394         .rename_setup   = nfs4_proc_rename_setup,
6395         .rename_done    = nfs4_proc_rename_done,
6396         .link           = nfs4_proc_link,
6397         .symlink        = nfs4_proc_symlink,
6398         .mkdir          = nfs4_proc_mkdir,
6399         .rmdir          = nfs4_proc_remove,
6400         .readdir        = nfs4_proc_readdir,
6401         .mknod          = nfs4_proc_mknod,
6402         .statfs         = nfs4_proc_statfs,
6403         .fsinfo         = nfs4_proc_fsinfo,
6404         .pathconf       = nfs4_proc_pathconf,
6405         .set_capabilities = nfs4_server_capabilities,
6406         .decode_dirent  = nfs4_decode_dirent,
6407         .read_setup     = nfs4_proc_read_setup,
6408         .read_done      = nfs4_read_done,
6409         .write_setup    = nfs4_proc_write_setup,
6410         .write_done     = nfs4_write_done,
6411         .commit_setup   = nfs4_proc_commit_setup,
6412         .commit_done    = nfs4_commit_done,
6413         .lock           = nfs4_proc_lock,
6414         .clear_acl_cache = nfs4_zap_acl_attr,
6415         .close_context  = nfs4_close_context,
6416         .open_context   = nfs4_atomic_open,
6417         .init_client    = nfs4_init_client,
6418         .secinfo        = nfs4_proc_secinfo,
6419 };
6420
6421 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6422         .prefix = XATTR_NAME_NFSV4_ACL,
6423         .list   = nfs4_xattr_list_nfs4_acl,
6424         .get    = nfs4_xattr_get_nfs4_acl,
6425         .set    = nfs4_xattr_set_nfs4_acl,
6426 };
6427
6428 const struct xattr_handler *nfs4_xattr_handlers[] = {
6429         &nfs4_xattr_nfs4_acl_handler,
6430         NULL
6431 };
6432
6433 module_param(max_session_slots, ushort, 0644);
6434 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6435                 "requests the client will negotiate");
6436
6437 /*
6438  * Local variables:
6439  *  c-basic-offset: 8
6440  * End:
6441  */