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 void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  85                             struct nfs_fattr *fattr, struct iattr *sattr,
  86                             struct nfs4_state *state);
  87 #ifdef CONFIG_NFS_V4_1
  88 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *);
  89 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *);
  90 #endif
  91 /* Prevent leaks of NFSv4 errors into userland */
  92 static int nfs4_map_errors(int err)
  93 {
  94         if (err >= -1000)
  95                 return err;
  96         switch (err) {
  97         case -NFS4ERR_RESOURCE:
  98                 return -EREMOTEIO;
  99         case -NFS4ERR_WRONGSEC:
 100                 return -EPERM;
 101         case -NFS4ERR_BADOWNER:
 102         case -NFS4ERR_BADNAME:
 103                 return -EINVAL;
 104         default:
 105                 dprintk("%s could not handle NFSv4 error %d\n",
 106                                 __func__, -err);
 107                 break;
 108         }
 109         return -EIO;
 110 }
 111
 112 /*
 113  * This is our standard bitmap for GETATTR requests.
 114  */
 115 const u32 nfs4_fattr_bitmap[2] = {
 116         FATTR4_WORD0_TYPE
 117         | FATTR4_WORD0_CHANGE
 118         | FATTR4_WORD0_SIZE
 119         | FATTR4_WORD0_FSID
 120         | FATTR4_WORD0_FILEID,
 121         FATTR4_WORD1_MODE
 122         | FATTR4_WORD1_NUMLINKS
 123         | FATTR4_WORD1_OWNER
 124         | FATTR4_WORD1_OWNER_GROUP
 125         | FATTR4_WORD1_RAWDEV
 126         | FATTR4_WORD1_SPACE_USED
 127         | FATTR4_WORD1_TIME_ACCESS
 128         | FATTR4_WORD1_TIME_METADATA
 129         | FATTR4_WORD1_TIME_MODIFY
 130 };
 131
 132 const u32 nfs4_statfs_bitmap[2] = {
 133         FATTR4_WORD0_FILES_AVAIL
 134         | FATTR4_WORD0_FILES_FREE
 135         | FATTR4_WORD0_FILES_TOTAL,
 136         FATTR4_WORD1_SPACE_AVAIL
 137         | FATTR4_WORD1_SPACE_FREE
 138         | FATTR4_WORD1_SPACE_TOTAL
 139 };
 140
 141 const u32 nfs4_pathconf_bitmap[2] = {
 142         FATTR4_WORD0_MAXLINK
 143         | FATTR4_WORD0_MAXNAME,
 144         0
 145 };
 146
 147 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 148                         | FATTR4_WORD0_MAXREAD
 149                         | FATTR4_WORD0_MAXWRITE
 150                         | FATTR4_WORD0_LEASE_TIME,
 151                         FATTR4_WORD1_TIME_DELTA
 152                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 153                         FATTR4_WORD2_LAYOUT_BLKSIZE
 154 };
 155
 156 const u32 nfs4_fs_locations_bitmap[2] = {
 157         FATTR4_WORD0_TYPE
 158         | FATTR4_WORD0_CHANGE
 159         | FATTR4_WORD0_SIZE
 160         | FATTR4_WORD0_FSID
 161         | FATTR4_WORD0_FILEID
 162         | FATTR4_WORD0_FS_LOCATIONS,
 163         FATTR4_WORD1_MODE
 164         | FATTR4_WORD1_NUMLINKS
 165         | FATTR4_WORD1_OWNER
 166         | FATTR4_WORD1_OWNER_GROUP
 167         | FATTR4_WORD1_RAWDEV
 168         | FATTR4_WORD1_SPACE_USED
 169         | FATTR4_WORD1_TIME_ACCESS
 170         | FATTR4_WORD1_TIME_METADATA
 171         | FATTR4_WORD1_TIME_MODIFY
 172         | FATTR4_WORD1_MOUNTED_ON_FILEID
 173 };
 174
 175 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 176                 struct nfs4_readdir_arg *readdir)
 177 {
 178         __be32 *start, *p;
 179
 180         BUG_ON(readdir->count < 80);
 181         if (cookie > 2) {
 182                 readdir->cookie = cookie;
 183                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 184                 return;
 185         }
 186
 187         readdir->cookie = 0;
 188         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 189         if (cookie == 2)
 190                 return;
 191
 192         /*
 193          * NFSv4 servers do not return entries for '.' and '..'
 194          * Therefore, we fake these entries here.  We let '.'
 195          * have cookie 0 and '..' have cookie 1.  Note that
 196          * when talking to the server, we always send cookie 0
 197          * instead of 1 or 2.
 198          */
 199         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 200
 201         if (cookie == 0) {
 202                 *p++ = xdr_one;                                  /* next */
 203                 *p++ = xdr_zero;                   /* cookie, first word */
 204                 *p++ = xdr_one;                   /* cookie, second word */
 205                 *p++ = xdr_one;                             /* entry len */
 206                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 207                 p++;
 208                 *p++ = xdr_one;                         /* bitmap length */
 209                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 210                 *p++ = htonl(8);              /* attribute buffer length */
 211                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 212         }
 213
 214         *p++ = xdr_one;                                  /* next */
 215         *p++ = xdr_zero;                   /* cookie, first word */
 216         *p++ = xdr_two;                   /* cookie, second word */
 217         *p++ = xdr_two;                             /* entry len */
 218         memcpy(p, "..\0\0", 4);                         /* entry */
 219         p++;
 220         *p++ = xdr_one;                         /* bitmap length */
 221         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 222         *p++ = htonl(8);              /* attribute buffer length */
 223         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 224
 225         readdir->pgbase = (char *)p - (char *)start;
 226         readdir->count -= readdir->pgbase;
 227         kunmap_atomic(start, KM_USER0);
 228 }
 229
 230 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 231 {
 232         int res;
 233
 234         might_sleep();
 235
 236         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 237                         nfs_wait_bit_killable, TASK_KILLABLE);
 238         return res;
 239 }
 240
 241 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 242 {
 243         int res = 0;
 244
 245         might_sleep();
 246
 247         if (*timeout <= 0)
 248                 *timeout = NFS4_POLL_RETRY_MIN;
 249         if (*timeout > NFS4_POLL_RETRY_MAX)
 250                 *timeout = NFS4_POLL_RETRY_MAX;
 251         freezable_schedule_timeout_killable(*timeout);
 252         if (fatal_signal_pending(current))
 253                 res = -ERESTARTSYS;
 254         *timeout <<= 1;
 255         return res;
 256 }
 257
 258 /* This is the error handling routine for processes that are allowed
 259  * to sleep.
 260  */
 261 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 262 {
 263         struct nfs_client *clp = server->nfs_client;
 264         struct nfs4_state *state = exception->state;
 265         struct inode *inode = exception->inode;
 266         int ret = errorcode;
 267
 268         exception->retry = 0;
 269         switch(errorcode) {
 270                 case 0:
 271                         return 0;
 272                 case -NFS4ERR_OPENMODE:
 273                         if (nfs_have_delegation(inode, FMODE_READ)) {
 274                                 nfs_inode_return_delegation(inode);
 275                                 exception->retry = 1;
 276                                 return 0;
 277                         }
 278                         if (state == NULL)
 279                                 break;
 280                         nfs4_schedule_stateid_recovery(server, state);
 281                         goto wait_on_recovery;
 282                 case -NFS4ERR_DELEG_REVOKED:
 283                 case -NFS4ERR_ADMIN_REVOKED:
 284                 case -NFS4ERR_BAD_STATEID:
 285                         if (state != NULL)
 286                                 nfs_remove_bad_delegation(state->inode);
 287                         if (state == NULL)
 288                                 break;
 289                         nfs4_schedule_stateid_recovery(server, state);
 290                         goto wait_on_recovery;
 291                 case -NFS4ERR_EXPIRED:
 292                         if (state != NULL)
 293                                 nfs4_schedule_stateid_recovery(server, state);
 294                 case -NFS4ERR_STALE_STATEID:
 295                 case -NFS4ERR_STALE_CLIENTID:
 296                         nfs4_schedule_lease_recovery(clp);
 297                         goto wait_on_recovery;
 298 #if defined(CONFIG_NFS_V4_1)
 299                 case -NFS4ERR_BADSESSION:
 300                 case -NFS4ERR_BADSLOT:
 301                 case -NFS4ERR_BAD_HIGH_SLOT:
 302                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 303                 case -NFS4ERR_DEADSESSION:
 304                 case -NFS4ERR_SEQ_FALSE_RETRY:
 305                 case -NFS4ERR_SEQ_MISORDERED:
 306                         dprintk("%s ERROR: %d Reset session\n", __func__,
 307                                 errorcode);
 308                         nfs4_schedule_session_recovery(clp->cl_session);
 309                         exception->retry = 1;
 310                         break;
 311 #endif /* defined(CONFIG_NFS_V4_1) */
 312                 case -NFS4ERR_FILE_OPEN:
 313                         if (exception->timeout > HZ) {
 314                                 /* We have retried a decent amount, time to
 315                                  * fail
 316                                  */
 317                                 ret = -EBUSY;
 318                                 break;
 319                         }
 320                 case -NFS4ERR_GRACE:
 321                 case -NFS4ERR_DELAY:
 322                 case -EKEYEXPIRED:
 323                         ret = nfs4_delay(server->client, &exception->timeout);
 324                         if (ret != 0)
 325                                 break;
 326                 case -NFS4ERR_RETRY_UNCACHED_REP:
 327                 case -NFS4ERR_OLD_STATEID:
 328                         exception->retry = 1;
 329                         break;
 330                 case -NFS4ERR_BADOWNER:
 331                         /* The following works around a Linux server bug! */
 332                 case -NFS4ERR_BADNAME:
 333                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 334                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 335                                 exception->retry = 1;
 336                                 printk(KERN_WARNING "NFS: v4 server %s "
 337                                                 "does not accept raw "
 338                                                 "uid/gids. "
 339                                                 "Reenabling the idmapper.\n",
 340                                                 server->nfs_client->cl_hostname);
 341                         }
 342         }
 343         /* We failed to handle the error */
 344         return nfs4_map_errors(ret);
 345 wait_on_recovery:
 346         ret = nfs4_wait_clnt_recover(clp);
 347         if (ret == 0)
 348                 exception->retry = 1;
 349         return ret;
 350 }
 351
 352
 353 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 354 {
 355         spin_lock(&clp->cl_lock);
 356         if (time_before(clp->cl_last_renewal,timestamp))
 357                 clp->cl_last_renewal = timestamp;
 358         spin_unlock(&clp->cl_lock);
 359 }
 360
 361 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 362 {
 363         do_renew_lease(server->nfs_client, timestamp);
 364 }
 365
 366 #if defined(CONFIG_NFS_V4_1)
 367
 368 /*
 369  * nfs4_free_slot - free a slot and efficiently update slot table.
 370  *
 371  * freeing a slot is trivially done by clearing its respective bit
 372  * in the bitmap.
 373  * If the freed slotid equals highest_used_slotid we want to update it
 374  * so that the server would be able to size down the slot table if needed,
 375  * otherwise we know that the highest_used_slotid is still in use.
 376  * When updating highest_used_slotid there may be "holes" in the bitmap
 377  * so we need to scan down from highest_used_slotid to 0 looking for the now
 378  * highest slotid in use.
 379  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 380  *
 381  * Must be called while holding tbl->slot_tbl_lock
 382  */
 383 static void
 384 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 385 {
 386         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 387         /* clear used bit in bitmap */
 388         __clear_bit(slotid, tbl->used_slots);
 389
 390         /* update highest_used_slotid when it is freed */
 391         if (slotid == tbl->highest_used_slotid) {
 392                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 393                 if (slotid < tbl->max_slots)
 394                         tbl->highest_used_slotid = slotid;
 395                 else
 396                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 397         }
 398         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 399                 slotid, tbl->highest_used_slotid);
 400 }
 401
 402 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 403 {
 404         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 405         return true;
 406 }
 407
 408 /*
 409  * Signal state manager thread if session fore channel is drained
 410  */
 411 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 412 {
 413         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 414                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 415                                 nfs4_set_task_privileged, NULL);
 416                 return;
 417         }
 418
 419         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 420                 return;
 421
 422         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 423         complete(&ses->fc_slot_table.complete);
 424 }
 425
 426 /*
 427  * Signal state manager thread if session back channel is drained
 428  */
 429 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 430 {
 431         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 432             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 433                 return;
 434         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 435         complete(&ses->bc_slot_table.complete);
 436 }
 437
 438 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 439 {
 440         struct nfs4_slot_table *tbl;
 441
 442         tbl = &res->sr_session->fc_slot_table;
 443         if (!res->sr_slot) {
 444                 /* just wake up the next guy waiting since
 445                  * we may have not consumed a slot after all */
 446                 dprintk("%s: No slot\n", __func__);
 447                 return;
 448         }
 449
 450         spin_lock(&tbl->slot_tbl_lock);
 451         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 452         nfs4_check_drain_fc_complete(res->sr_session);
 453         spin_unlock(&tbl->slot_tbl_lock);
 454         res->sr_slot = NULL;
 455 }
 456
 457 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 458 {
 459         unsigned long timestamp;
 460         struct nfs_client *clp;
 461
 462         /*
 463          * sr_status remains 1 if an RPC level error occurred. The server
 464          * may or may not have processed the sequence operation..
 465          * Proceed as if the server received and processed the sequence
 466          * operation.
 467          */
 468         if (res->sr_status == 1)
 469                 res->sr_status = NFS_OK;
 470
 471         /* don't increment the sequence number if the task wasn't sent */
 472         if (!RPC_WAS_SENT(task))
 473                 goto out;
 474
 475         /* Check the SEQUENCE operation status */
 476         switch (res->sr_status) {
 477         case 0:
 478                 /* Update the slot's sequence and clientid lease timer */
 479                 ++res->sr_slot->seq_nr;
 480                 timestamp = res->sr_renewal_time;
 481                 clp = res->sr_session->clp;
 482                 do_renew_lease(clp, timestamp);
 483                 /* Check sequence flags */
 484                 if (res->sr_status_flags != 0)
 485                         nfs4_schedule_lease_recovery(clp);
 486                 break;
 487         case -NFS4ERR_DELAY:
 488                 /* The server detected a resend of the RPC call and
 489                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 490                  * of RFC5661.
 491                  */
 492                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 493                         __func__,
 494                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 495                         res->sr_slot->seq_nr);
 496                 goto out_retry;
 497         default:
 498                 /* Just update the slot sequence no. */
 499                 ++res->sr_slot->seq_nr;
 500         }
 501 out:
 502         /* The session may be reset by one of the error handlers. */
 503         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 504         nfs41_sequence_free_slot(res);
 505         return 1;
 506 out_retry:
 507         if (!rpc_restart_call(task))
 508                 goto out;
 509         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 510         return 0;
 511 }
 512
 513 static int nfs4_sequence_done(struct rpc_task *task,
 514                                struct nfs4_sequence_res *res)
 515 {
 516         if (res->sr_session == NULL)
 517                 return 1;
 518         return nfs41_sequence_done(task, res);
 519 }
 520
 521 /*
 522  * nfs4_find_slot - efficiently look for a free slot
 523  *
 524  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 525  * If found, we mark the slot as used, update the highest_used_slotid,
 526  * and respectively set up the sequence operation args.
 527  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 528  *
 529  * Note: must be called with under the slot_tbl_lock.
 530  */
 531 static u32
 532 nfs4_find_slot(struct nfs4_slot_table *tbl)
 533 {
 534         u32 slotid;
 535         u32 ret_id = NFS4_NO_SLOT;
 536
 537         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 538                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 539                 tbl->max_slots);
 540         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 541         if (slotid >= tbl->max_slots)
 542                 goto out;
 543         __set_bit(slotid, tbl->used_slots);
 544         if (slotid > tbl->highest_used_slotid ||
 545                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 546                 tbl->highest_used_slotid = slotid;
 547         ret_id = slotid;
 548 out:
 549         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 550                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 551         return ret_id;
 552 }
 553
 554 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 555                 struct nfs4_sequence_res *res, int cache_reply)
 556 {
 557         args->sa_session = NULL;
 558         args->sa_cache_this = 0;
 559         if (cache_reply)
 560                 args->sa_cache_this = 1;
 561         res->sr_session = NULL;
 562         res->sr_slot = NULL;
 563 }
 564
 565 int nfs41_setup_sequence(struct nfs4_session *session,
 566                                 struct nfs4_sequence_args *args,
 567                                 struct nfs4_sequence_res *res,
 568                                 struct rpc_task *task)
 569 {
 570         struct nfs4_slot *slot;
 571         struct nfs4_slot_table *tbl;
 572         u32 slotid;
 573
 574         dprintk("--> %s\n", __func__);
 575         /* slot already allocated? */
 576         if (res->sr_slot != NULL)
 577                 return 0;
 578
 579         tbl = &session->fc_slot_table;
 580
 581         spin_lock(&tbl->slot_tbl_lock);
 582         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 583             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 584                 /* The state manager will wait until the slot table is empty */
 585                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 586                 spin_unlock(&tbl->slot_tbl_lock);
 587                 dprintk("%s session is draining\n", __func__);
 588                 return -EAGAIN;
 589         }
 590
 591         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 592             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 593                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 594                 spin_unlock(&tbl->slot_tbl_lock);
 595                 dprintk("%s enforce FIFO order\n", __func__);
 596                 return -EAGAIN;
 597         }
 598
 599         slotid = nfs4_find_slot(tbl);
 600         if (slotid == NFS4_NO_SLOT) {
 601                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 602                 spin_unlock(&tbl->slot_tbl_lock);
 603                 dprintk("<-- %s: no free slots\n", __func__);
 604                 return -EAGAIN;
 605         }
 606         spin_unlock(&tbl->slot_tbl_lock);
 607
 608         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 609         slot = tbl->slots + slotid;
 610         args->sa_session = session;
 611         args->sa_slotid = slotid;
 612
 613         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 614
 615         res->sr_session = session;
 616         res->sr_slot = slot;
 617         res->sr_renewal_time = jiffies;
 618         res->sr_status_flags = 0;
 619         /*
 620          * sr_status is only set in decode_sequence, and so will remain
 621          * set to 1 if an rpc level failure occurs.
 622          */
 623         res->sr_status = 1;
 624         return 0;
 625 }
 626 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 627
 628 int nfs4_setup_sequence(const struct nfs_server *server,
 629                         struct nfs4_sequence_args *args,
 630                         struct nfs4_sequence_res *res,
 631                         struct rpc_task *task)
 632 {
 633         struct nfs4_session *session = nfs4_get_session(server);
 634         int ret = 0;
 635
 636         if (session == NULL)
 637                 goto out;
 638
 639         dprintk("--> %s clp %p session %p sr_slot %td\n",
 640                 __func__, session->clp, session, res->sr_slot ?
 641                         res->sr_slot - session->fc_slot_table.slots : -1);
 642
 643         ret = nfs41_setup_sequence(session, args, res, task);
 644 out:
 645         dprintk("<-- %s status=%d\n", __func__, ret);
 646         return ret;
 647 }
 648
 649 struct nfs41_call_sync_data {
 650         const struct nfs_server *seq_server;
 651         struct nfs4_sequence_args *seq_args;
 652         struct nfs4_sequence_res *seq_res;
 653 };
 654
 655 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 656 {
 657         struct nfs41_call_sync_data *data = calldata;
 658
 659         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 660
 661         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 662                                 data->seq_res, task))
 663                 return;
 664         rpc_call_start(task);
 665 }
 666
 667 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 668 {
 669         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 670         nfs41_call_sync_prepare(task, calldata);
 671 }
 672
 673 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 674 {
 675         struct nfs41_call_sync_data *data = calldata;
 676
 677         nfs41_sequence_done(task, data->seq_res);
 678 }
 679
 680 static const struct rpc_call_ops nfs41_call_sync_ops = {
 681         .rpc_call_prepare = nfs41_call_sync_prepare,
 682         .rpc_call_done = nfs41_call_sync_done,
 683 };
 684
 685 static const struct rpc_call_ops nfs41_call_priv_sync_ops = {
 686         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 687         .rpc_call_done = nfs41_call_sync_done,
 688 };
 689
 690 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 691                                    struct nfs_server *server,
 692                                    struct rpc_message *msg,
 693                                    struct nfs4_sequence_args *args,
 694                                    struct nfs4_sequence_res *res,
 695                                    int privileged)
 696 {
 697         int ret;
 698         struct rpc_task *task;
 699         struct nfs41_call_sync_data data = {
 700                 .seq_server = server,
 701                 .seq_args = args,
 702                 .seq_res = res,
 703         };
 704         struct rpc_task_setup task_setup = {
 705                 .rpc_client = clnt,
 706                 .rpc_message = msg,
 707                 .callback_ops = &nfs41_call_sync_ops,
 708                 .callback_data = &data
 709         };
 710
 711         if (privileged)
 712                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 713         task = rpc_run_task(&task_setup);
 714         if (IS_ERR(task))
 715                 ret = PTR_ERR(task);
 716         else {
 717                 ret = task->tk_status;
 718                 rpc_put_task(task);
 719         }
 720         return ret;
 721 }
 722
 723 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 724                             struct nfs_server *server,
 725                             struct rpc_message *msg,
 726                             struct nfs4_sequence_args *args,
 727                             struct nfs4_sequence_res *res,
 728                             int cache_reply)
 729 {
 730         nfs41_init_sequence(args, res, cache_reply);
 731         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 732 }
 733
 734 #else
 735 static inline
 736 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 737                 struct nfs4_sequence_res *res, int cache_reply)
 738 {
 739 }
 740
 741 static int nfs4_sequence_done(struct rpc_task *task,
 742                                struct nfs4_sequence_res *res)
 743 {
 744         return 1;
 745 }
 746 #endif /* CONFIG_NFS_V4_1 */
 747
 748 int _nfs4_call_sync(struct rpc_clnt *clnt,
 749                     struct nfs_server *server,
 750                     struct rpc_message *msg,
 751                     struct nfs4_sequence_args *args,
 752                     struct nfs4_sequence_res *res,
 753                     int cache_reply)
 754 {
 755         nfs41_init_sequence(args, res, cache_reply);
 756         return rpc_call_sync(clnt, msg, 0);
 757 }
 758
 759 static inline
 760 int nfs4_call_sync(struct rpc_clnt *clnt,
 761                    struct nfs_server *server,
 762                    struct rpc_message *msg,
 763                    struct nfs4_sequence_args *args,
 764                    struct nfs4_sequence_res *res,
 765                    int cache_reply)
 766 {
 767         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 768                                                 args, res, cache_reply);
 769 }
 770
 771 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 772 {
 773         struct nfs_inode *nfsi = NFS_I(dir);
 774
 775         spin_lock(&dir->i_lock);
 776         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 777         if (!cinfo->atomic || cinfo->before != dir->i_version)
 778                 nfs_force_lookup_revalidate(dir);
 779         dir->i_version = cinfo->after;
 780         spin_unlock(&dir->i_lock);
 781 }
 782
 783 struct nfs4_opendata {
 784         struct kref kref;
 785         struct nfs_openargs o_arg;
 786         struct nfs_openres o_res;
 787         struct nfs_open_confirmargs c_arg;
 788         struct nfs_open_confirmres c_res;
 789         struct nfs4_string owner_name;
 790         struct nfs4_string group_name;
 791         struct nfs_fattr f_attr;
 792         struct nfs_fattr dir_attr;
 793         struct dentry *dir;
 794         struct dentry *dentry;
 795         struct nfs4_state_owner *owner;
 796         struct nfs4_state *state;
 797         struct iattr attrs;
 798         unsigned long timestamp;
 799         unsigned int rpc_done : 1;
 800         int rpc_status;
 801         int cancelled;
 802 };
 803
 804
 805 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 806 {
 807         p->o_res.f_attr = &p->f_attr;
 808         p->o_res.dir_attr = &p->dir_attr;
 809         p->o_res.seqid = p->o_arg.seqid;
 810         p->c_res.seqid = p->c_arg.seqid;
 811         p->o_res.server = p->o_arg.server;
 812         nfs_fattr_init(&p->f_attr);
 813         nfs_fattr_init(&p->dir_attr);
 814         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 815 }
 816
 817 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 818                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 819                 const struct iattr *attrs,
 820                 gfp_t gfp_mask)
 821 {
 822         struct dentry *parent = dget_parent(dentry);
 823         struct inode *dir = parent->d_inode;
 824         struct nfs_server *server = NFS_SERVER(dir);
 825         struct nfs4_opendata *p;
 826
 827         p = kzalloc(sizeof(*p), gfp_mask);
 828         if (p == NULL)
 829                 goto err;
 830         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 831         if (p->o_arg.seqid == NULL)
 832                 goto err_free;
 833         nfs_sb_active(dentry->d_sb);
 834         p->dentry = dget(dentry);
 835         p->dir = parent;
 836         p->owner = sp;
 837         atomic_inc(&sp->so_count);
 838         p->o_arg.fh = NFS_FH(dir);
 839         p->o_arg.open_flags = flags;
 840         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 841         p->o_arg.clientid = server->nfs_client->cl_clientid;
 842         p->o_arg.id = sp->so_seqid.owner_id;
 843         p->o_arg.name = &dentry->d_name;
 844         p->o_arg.server = server;
 845         p->o_arg.bitmask = server->attr_bitmask;
 846         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 847         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 848         if (attrs != NULL && attrs->ia_valid != 0) {
 849                 __be32 verf[2];
 850
 851                 p->o_arg.u.attrs = &p->attrs;
 852                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 853
 854                 verf[0] = jiffies;
 855                 verf[1] = current->pid;
 856                 memcpy(p->o_arg.u.verifier.data, verf,
 857                                 sizeof(p->o_arg.u.verifier.data));
 858         }
 859         p->c_arg.fh = &p->o_res.fh;
 860         p->c_arg.stateid = &p->o_res.stateid;
 861         p->c_arg.seqid = p->o_arg.seqid;
 862         nfs4_init_opendata_res(p);
 863         kref_init(&p->kref);
 864         return p;
 865 err_free:
 866         kfree(p);
 867 err:
 868         dput(parent);
 869         return NULL;
 870 }
 871
 872 static void nfs4_opendata_free(struct kref *kref)
 873 {
 874         struct nfs4_opendata *p = container_of(kref,
 875                         struct nfs4_opendata, kref);
 876         struct super_block *sb = p->dentry->d_sb;
 877
 878         nfs_free_seqid(p->o_arg.seqid);
 879         if (p->state != NULL)
 880                 nfs4_put_open_state(p->state);
 881         nfs4_put_state_owner(p->owner);
 882         dput(p->dir);
 883         dput(p->dentry);
 884         nfs_sb_deactive(sb);
 885         nfs_fattr_free_names(&p->f_attr);
 886         kfree(p);
 887 }
 888
 889 static void nfs4_opendata_put(struct nfs4_opendata *p)
 890 {
 891         if (p != NULL)
 892                 kref_put(&p->kref, nfs4_opendata_free);
 893 }
 894
 895 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 896 {
 897         int ret;
 898
 899         ret = rpc_wait_for_completion_task(task);
 900         return ret;
 901 }
 902
 903 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 904 {
 905         int ret = 0;
 906
 907         if (open_mode & (O_EXCL|O_TRUNC))
 908                 goto out;
 909         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 910                 case FMODE_READ:
 911                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 912                                 && state->n_rdonly != 0;
 913                         break;
 914                 case FMODE_WRITE:
 915                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 916                                 && state->n_wronly != 0;
 917                         break;
 918                 case FMODE_READ|FMODE_WRITE:
 919                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 920                                 && state->n_rdwr != 0;
 921         }
 922 out:
 923         return ret;
 924 }
 925
 926 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 927 {
 928         if (delegation == NULL)
 929                 return 0;
 930         if ((delegation->type & fmode) != fmode)
 931                 return 0;
 932         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 933                 return 0;
 934         nfs_mark_delegation_referenced(delegation);
 935         return 1;
 936 }
 937
 938 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 939 {
 940         switch (fmode) {
 941                 case FMODE_WRITE:
 942                         state->n_wronly++;
 943                         break;
 944                 case FMODE_READ:
 945                         state->n_rdonly++;
 946                         break;
 947                 case FMODE_READ|FMODE_WRITE:
 948                         state->n_rdwr++;
 949         }
 950         nfs4_state_set_mode_locked(state, state->state | fmode);
 951 }
 952
 953 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 954 {
 955         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 956                 nfs4_stateid_copy(&state->stateid, stateid);
 957         nfs4_stateid_copy(&state->open_stateid, stateid);
 958         switch (fmode) {
 959                 case FMODE_READ:
 960                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 961                         break;
 962                 case FMODE_WRITE:
 963                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 964                         break;
 965                 case FMODE_READ|FMODE_WRITE:
 966                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 967         }
 968 }
 969
 970 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 971 {
 972         write_seqlock(&state->seqlock);
 973         nfs_set_open_stateid_locked(state, stateid, fmode);
 974         write_sequnlock(&state->seqlock);
 975 }
 976
 977 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 978 {
 979         /*
 980          * Protect the call to nfs4_state_set_mode_locked and
 981          * serialise the stateid update
 982          */
 983         write_seqlock(&state->seqlock);
 984         if (deleg_stateid != NULL) {
 985                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
 986                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 987         }
 988         if (open_stateid != NULL)
 989                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 990         write_sequnlock(&state->seqlock);
 991         spin_lock(&state->owner->so_lock);
 992         update_open_stateflags(state, fmode);
 993         spin_unlock(&state->owner->so_lock);
 994 }
 995
 996 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 997 {
 998         struct nfs_inode *nfsi = NFS_I(state->inode);
 999         struct nfs_delegation *deleg_cur;
1000         int ret = 0;
1001
1002         fmode &= (FMODE_READ|FMODE_WRITE);
1003
1004         rcu_read_lock();
1005         deleg_cur = rcu_dereference(nfsi->delegation);
1006         if (deleg_cur == NULL)
1007                 goto no_delegation;
1008
1009         spin_lock(&deleg_cur->lock);
1010         if (nfsi->delegation != deleg_cur ||
1011             (deleg_cur->type & fmode) != fmode)
1012                 goto no_delegation_unlock;
1013
1014         if (delegation == NULL)
1015                 delegation = &deleg_cur->stateid;
1016         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1017                 goto no_delegation_unlock;
1018
1019         nfs_mark_delegation_referenced(deleg_cur);
1020         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1021         ret = 1;
1022 no_delegation_unlock:
1023         spin_unlock(&deleg_cur->lock);
1024 no_delegation:
1025         rcu_read_unlock();
1026
1027         if (!ret && open_stateid != NULL) {
1028                 __update_open_stateid(state, open_stateid, NULL, fmode);
1029                 ret = 1;
1030         }
1031
1032         return ret;
1033 }
1034
1035
1036 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1037 {
1038         struct nfs_delegation *delegation;
1039
1040         rcu_read_lock();
1041         delegation = rcu_dereference(NFS_I(inode)->delegation);
1042         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1043                 rcu_read_unlock();
1044                 return;
1045         }
1046         rcu_read_unlock();
1047         nfs_inode_return_delegation(inode);
1048 }
1049
1050 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1051 {
1052         struct nfs4_state *state = opendata->state;
1053         struct nfs_inode *nfsi = NFS_I(state->inode);
1054         struct nfs_delegation *delegation;
1055         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1056         fmode_t fmode = opendata->o_arg.fmode;
1057         nfs4_stateid stateid;
1058         int ret = -EAGAIN;
1059
1060         for (;;) {
1061                 if (can_open_cached(state, fmode, open_mode)) {
1062                         spin_lock(&state->owner->so_lock);
1063                         if (can_open_cached(state, fmode, open_mode)) {
1064                                 update_open_stateflags(state, fmode);
1065                                 spin_unlock(&state->owner->so_lock);
1066                                 goto out_return_state;
1067                         }
1068                         spin_unlock(&state->owner->so_lock);
1069                 }
1070                 rcu_read_lock();
1071                 delegation = rcu_dereference(nfsi->delegation);
1072                 if (!can_open_delegated(delegation, fmode)) {
1073                         rcu_read_unlock();
1074                         break;
1075                 }
1076                 /* Save the delegation */
1077                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1078                 rcu_read_unlock();
1079                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1080                 if (ret != 0)
1081                         goto out;
1082                 ret = -EAGAIN;
1083
1084                 /* Try to update the stateid using the delegation */
1085                 if (update_open_stateid(state, NULL, &stateid, fmode))
1086                         goto out_return_state;
1087         }
1088 out:
1089         return ERR_PTR(ret);
1090 out_return_state:
1091         atomic_inc(&state->count);
1092         return state;
1093 }
1094
1095 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1096 {
1097         struct inode *inode;
1098         struct nfs4_state *state = NULL;
1099         struct nfs_delegation *delegation;
1100         int ret;
1101
1102         if (!data->rpc_done) {
1103                 state = nfs4_try_open_cached(data);
1104                 goto out;
1105         }
1106
1107         ret = -EAGAIN;
1108         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1109                 goto err;
1110         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1111         ret = PTR_ERR(inode);
1112         if (IS_ERR(inode))
1113                 goto err;
1114         ret = -ENOMEM;
1115         state = nfs4_get_open_state(inode, data->owner);
1116         if (state == NULL)
1117                 goto err_put_inode;
1118         if (data->o_res.delegation_type != 0) {
1119                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1120                 int delegation_flags = 0;
1121
1122                 rcu_read_lock();
1123                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1124                 if (delegation)
1125                         delegation_flags = delegation->flags;
1126                 rcu_read_unlock();
1127                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1128                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1129                                         "returning a delegation for "
1130                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1131                                         clp->cl_hostname);
1132                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1133                         nfs_inode_set_delegation(state->inode,
1134                                         data->owner->so_cred,
1135                                         &data->o_res);
1136                 else
1137                         nfs_inode_reclaim_delegation(state->inode,
1138                                         data->owner->so_cred,
1139                                         &data->o_res);
1140         }
1141
1142         update_open_stateid(state, &data->o_res.stateid, NULL,
1143                         data->o_arg.fmode);
1144         iput(inode);
1145 out:
1146         return state;
1147 err_put_inode:
1148         iput(inode);
1149 err:
1150         return ERR_PTR(ret);
1151 }
1152
1153 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1154 {
1155         struct nfs_inode *nfsi = NFS_I(state->inode);
1156         struct nfs_open_context *ctx;
1157
1158         spin_lock(&state->inode->i_lock);
1159         list_for_each_entry(ctx, &nfsi->open_files, list) {
1160                 if (ctx->state != state)
1161                         continue;
1162                 get_nfs_open_context(ctx);
1163                 spin_unlock(&state->inode->i_lock);
1164                 return ctx;
1165         }
1166         spin_unlock(&state->inode->i_lock);
1167         return ERR_PTR(-ENOENT);
1168 }
1169
1170 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1171 {
1172         struct nfs4_opendata *opendata;
1173
1174         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1175         if (opendata == NULL)
1176                 return ERR_PTR(-ENOMEM);
1177         opendata->state = state;
1178         atomic_inc(&state->count);
1179         return opendata;
1180 }
1181
1182 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1183 {
1184         struct nfs4_state *newstate;
1185         int ret;
1186
1187         opendata->o_arg.open_flags = 0;
1188         opendata->o_arg.fmode = fmode;
1189         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1190         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1191         nfs4_init_opendata_res(opendata);
1192         ret = _nfs4_recover_proc_open(opendata);
1193         if (ret != 0)
1194                 return ret;
1195         newstate = nfs4_opendata_to_nfs4_state(opendata);
1196         if (IS_ERR(newstate))
1197                 return PTR_ERR(newstate);
1198         nfs4_close_state(newstate, fmode);
1199         *res = newstate;
1200         return 0;
1201 }
1202
1203 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1204 {
1205         struct nfs4_state *newstate;
1206         int ret;
1207
1208         /* memory barrier prior to reading state->n_* */
1209         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1210         smp_rmb();
1211         if (state->n_rdwr != 0) {
1212                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1213                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1214                 if (ret != 0)
1215                         return ret;
1216                 if (newstate != state)
1217                         return -ESTALE;
1218         }
1219         if (state->n_wronly != 0) {
1220                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1221                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1222                 if (ret != 0)
1223                         return ret;
1224                 if (newstate != state)
1225                         return -ESTALE;
1226         }
1227         if (state->n_rdonly != 0) {
1228                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1229                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1230                 if (ret != 0)
1231                         return ret;
1232                 if (newstate != state)
1233                         return -ESTALE;
1234         }
1235         /*
1236          * We may have performed cached opens for all three recoveries.
1237          * Check if we need to update the current stateid.
1238          */
1239         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1240             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1241                 write_seqlock(&state->seqlock);
1242                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1243                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1244                 write_sequnlock(&state->seqlock);
1245         }
1246         return 0;
1247 }
1248
1249 /*
1250  * OPEN_RECLAIM:
1251  *      reclaim state on the server after a reboot.
1252  */
1253 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1254 {
1255         struct nfs_delegation *delegation;
1256         struct nfs4_opendata *opendata;
1257         fmode_t delegation_type = 0;
1258         int status;
1259
1260         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1261         if (IS_ERR(opendata))
1262                 return PTR_ERR(opendata);
1263         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1264         opendata->o_arg.fh = NFS_FH(state->inode);
1265         rcu_read_lock();
1266         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1267         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1268                 delegation_type = delegation->type;
1269         rcu_read_unlock();
1270         opendata->o_arg.u.delegation_type = delegation_type;
1271         status = nfs4_open_recover(opendata, state);
1272         nfs4_opendata_put(opendata);
1273         return status;
1274 }
1275
1276 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1277 {
1278         struct nfs_server *server = NFS_SERVER(state->inode);
1279         struct nfs4_exception exception = { };
1280         int err;
1281         do {
1282                 err = _nfs4_do_open_reclaim(ctx, state);
1283                 if (err != -NFS4ERR_DELAY)
1284                         break;
1285                 nfs4_handle_exception(server, err, &exception);
1286         } while (exception.retry);
1287         return err;
1288 }
1289
1290 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1291 {
1292         struct nfs_open_context *ctx;
1293         int ret;
1294
1295         ctx = nfs4_state_find_open_context(state);
1296         if (IS_ERR(ctx))
1297                 return PTR_ERR(ctx);
1298         ret = nfs4_do_open_reclaim(ctx, state);
1299         put_nfs_open_context(ctx);
1300         return ret;
1301 }
1302
1303 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1304 {
1305         struct nfs4_opendata *opendata;
1306         int ret;
1307
1308         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1309         if (IS_ERR(opendata))
1310                 return PTR_ERR(opendata);
1311         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1312         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1313         ret = nfs4_open_recover(opendata, state);
1314         nfs4_opendata_put(opendata);
1315         return ret;
1316 }
1317
1318 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1319 {
1320         struct nfs4_exception exception = { };
1321         struct nfs_server *server = NFS_SERVER(state->inode);
1322         int err;
1323         do {
1324                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1325                 switch (err) {
1326                         case 0:
1327                         case -ENOENT:
1328                         case -ESTALE:
1329                                 goto out;
1330                         case -NFS4ERR_BADSESSION:
1331                         case -NFS4ERR_BADSLOT:
1332                         case -NFS4ERR_BAD_HIGH_SLOT:
1333                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1334                         case -NFS4ERR_DEADSESSION:
1335                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1336                                 goto out;
1337                         case -NFS4ERR_STALE_CLIENTID:
1338                         case -NFS4ERR_STALE_STATEID:
1339                         case -NFS4ERR_EXPIRED:
1340                                 /* Don't recall a delegation if it was lost */
1341                                 nfs4_schedule_lease_recovery(server->nfs_client);
1342                                 goto out;
1343                         case -ERESTARTSYS:
1344                                 /*
1345                                  * The show must go on: exit, but mark the
1346                                  * stateid as needing recovery.
1347                                  */
1348                         case -NFS4ERR_DELEG_REVOKED:
1349                         case -NFS4ERR_ADMIN_REVOKED:
1350                         case -NFS4ERR_BAD_STATEID:
1351                                 nfs_inode_find_state_and_recover(state->inode,
1352                                                 stateid);
1353                                 nfs4_schedule_stateid_recovery(server, state);
1354                         case -EKEYEXPIRED:
1355                                 /*
1356                                  * User RPCSEC_GSS context has expired.
1357                                  * We cannot recover this stateid now, so
1358                                  * skip it and allow recovery thread to
1359                                  * proceed.
1360                                  */
1361                         case -ENOMEM:
1362                                 err = 0;
1363                                 goto out;
1364                 }
1365                 err = nfs4_handle_exception(server, err, &exception);
1366         } while (exception.retry);
1367 out:
1368         return err;
1369 }
1370
1371 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1372 {
1373         struct nfs4_opendata *data = calldata;
1374
1375         data->rpc_status = task->tk_status;
1376         if (data->rpc_status == 0) {
1377                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1378                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1379                 renew_lease(data->o_res.server, data->timestamp);
1380                 data->rpc_done = 1;
1381         }
1382 }
1383
1384 static void nfs4_open_confirm_release(void *calldata)
1385 {
1386         struct nfs4_opendata *data = calldata;
1387         struct nfs4_state *state = NULL;
1388
1389         /* If this request hasn't been cancelled, do nothing */
1390         if (data->cancelled == 0)
1391                 goto out_free;
1392         /* In case of error, no cleanup! */
1393         if (!data->rpc_done)
1394                 goto out_free;
1395         state = nfs4_opendata_to_nfs4_state(data);
1396         if (!IS_ERR(state))
1397                 nfs4_close_state(state, data->o_arg.fmode);
1398 out_free:
1399         nfs4_opendata_put(data);
1400 }
1401
1402 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1403         .rpc_call_done = nfs4_open_confirm_done,
1404         .rpc_release = nfs4_open_confirm_release,
1405 };
1406
1407 /*
1408  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1409  */
1410 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1411 {
1412         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1413         struct rpc_task *task;
1414         struct  rpc_message msg = {
1415                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1416                 .rpc_argp = &data->c_arg,
1417                 .rpc_resp = &data->c_res,
1418                 .rpc_cred = data->owner->so_cred,
1419         };
1420         struct rpc_task_setup task_setup_data = {
1421                 .rpc_client = server->client,
1422                 .rpc_message = &msg,
1423                 .callback_ops = &nfs4_open_confirm_ops,
1424                 .callback_data = data,
1425                 .workqueue = nfsiod_workqueue,
1426                 .flags = RPC_TASK_ASYNC,
1427         };
1428         int status;
1429
1430         kref_get(&data->kref);
1431         data->rpc_done = 0;
1432         data->rpc_status = 0;
1433         data->timestamp = jiffies;
1434         task = rpc_run_task(&task_setup_data);
1435         if (IS_ERR(task))
1436                 return PTR_ERR(task);
1437         status = nfs4_wait_for_completion_rpc_task(task);
1438         if (status != 0) {
1439                 data->cancelled = 1;
1440                 smp_wmb();
1441         } else
1442                 status = data->rpc_status;
1443         rpc_put_task(task);
1444         return status;
1445 }
1446
1447 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1448 {
1449         struct nfs4_opendata *data = calldata;
1450         struct nfs4_state_owner *sp = data->owner;
1451
1452         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1453                 return;
1454         /*
1455          * Check if we still need to send an OPEN call, or if we can use
1456          * a delegation instead.
1457          */
1458         if (data->state != NULL) {
1459                 struct nfs_delegation *delegation;
1460
1461                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1462                         goto out_no_action;
1463                 rcu_read_lock();
1464                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1465                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1466                     can_open_delegated(delegation, data->o_arg.fmode))
1467                         goto unlock_no_action;
1468                 rcu_read_unlock();
1469         }
1470         /* Update sequence id. */
1471         data->o_arg.id = sp->so_seqid.owner_id;
1472         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1473         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1474                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1475                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1476         }
1477         data->timestamp = jiffies;
1478         if (nfs4_setup_sequence(data->o_arg.server,
1479                                 &data->o_arg.seq_args,
1480                                 &data->o_res.seq_res, task))
1481                 return;
1482         rpc_call_start(task);
1483         return;
1484 unlock_no_action:
1485         rcu_read_unlock();
1486 out_no_action:
1487         task->tk_action = NULL;
1488
1489 }
1490
1491 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1492 {
1493         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1494         nfs4_open_prepare(task, calldata);
1495 }
1496
1497 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1498 {
1499         struct nfs4_opendata *data = calldata;
1500
1501         data->rpc_status = task->tk_status;
1502
1503         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1504                 return;
1505
1506         if (task->tk_status == 0) {
1507                 switch (data->o_res.f_attr->mode & S_IFMT) {
1508                         case S_IFREG:
1509                                 break;
1510                         case S_IFLNK:
1511                                 data->rpc_status = -ELOOP;
1512                                 break;
1513                         case S_IFDIR:
1514                                 data->rpc_status = -EISDIR;
1515                                 break;
1516                         default:
1517                                 data->rpc_status = -ENOTDIR;
1518                 }
1519                 renew_lease(data->o_res.server, data->timestamp);
1520                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1521                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1522         }
1523         data->rpc_done = 1;
1524 }
1525
1526 static void nfs4_open_release(void *calldata)
1527 {
1528         struct nfs4_opendata *data = calldata;
1529         struct nfs4_state *state = NULL;
1530
1531         /* If this request hasn't been cancelled, do nothing */
1532         if (data->cancelled == 0)
1533                 goto out_free;
1534         /* In case of error, no cleanup! */
1535         if (data->rpc_status != 0 || !data->rpc_done)
1536                 goto out_free;
1537         /* In case we need an open_confirm, no cleanup! */
1538         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1539                 goto out_free;
1540         state = nfs4_opendata_to_nfs4_state(data);
1541         if (!IS_ERR(state))
1542                 nfs4_close_state(state, data->o_arg.fmode);
1543 out_free:
1544         nfs4_opendata_put(data);
1545 }
1546
1547 static const struct rpc_call_ops nfs4_open_ops = {
1548         .rpc_call_prepare = nfs4_open_prepare,
1549         .rpc_call_done = nfs4_open_done,
1550         .rpc_release = nfs4_open_release,
1551 };
1552
1553 static const struct rpc_call_ops nfs4_recover_open_ops = {
1554         .rpc_call_prepare = nfs4_recover_open_prepare,
1555         .rpc_call_done = nfs4_open_done,
1556         .rpc_release = nfs4_open_release,
1557 };
1558
1559 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1560 {
1561         struct inode *dir = data->dir->d_inode;
1562         struct nfs_server *server = NFS_SERVER(dir);
1563         struct nfs_openargs *o_arg = &data->o_arg;
1564         struct nfs_openres *o_res = &data->o_res;
1565         struct rpc_task *task;
1566         struct rpc_message msg = {
1567                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1568                 .rpc_argp = o_arg,
1569                 .rpc_resp = o_res,
1570                 .rpc_cred = data->owner->so_cred,
1571         };
1572         struct rpc_task_setup task_setup_data = {
1573                 .rpc_client = server->client,
1574                 .rpc_message = &msg,
1575                 .callback_ops = &nfs4_open_ops,
1576                 .callback_data = data,
1577                 .workqueue = nfsiod_workqueue,
1578                 .flags = RPC_TASK_ASYNC,
1579         };
1580         int status;
1581
1582         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1583         kref_get(&data->kref);
1584         data->rpc_done = 0;
1585         data->rpc_status = 0;
1586         data->cancelled = 0;
1587         if (isrecover)
1588                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1589         task = rpc_run_task(&task_setup_data);
1590         if (IS_ERR(task))
1591                 return PTR_ERR(task);
1592         status = nfs4_wait_for_completion_rpc_task(task);
1593         if (status != 0) {
1594                 data->cancelled = 1;
1595                 smp_wmb();
1596         } else
1597                 status = data->rpc_status;
1598         rpc_put_task(task);
1599
1600         return status;
1601 }
1602
1603 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1604 {
1605         struct inode *dir = data->dir->d_inode;
1606         struct nfs_openres *o_res = &data->o_res;
1607         int status;
1608
1609         status = nfs4_run_open_task(data, 1);
1610         if (status != 0 || !data->rpc_done)
1611                 return status;
1612
1613         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1614
1615         nfs_refresh_inode(dir, o_res->dir_attr);
1616
1617         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1618                 status = _nfs4_proc_open_confirm(data);
1619                 if (status != 0)
1620                         return status;
1621         }
1622
1623         return status;
1624 }
1625
1626 /*
1627  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1628  */
1629 static int _nfs4_proc_open(struct nfs4_opendata *data)
1630 {
1631         struct inode *dir = data->dir->d_inode;
1632         struct nfs_server *server = NFS_SERVER(dir);
1633         struct nfs_openargs *o_arg = &data->o_arg;
1634         struct nfs_openres *o_res = &data->o_res;
1635         int status;
1636
1637         status = nfs4_run_open_task(data, 0);
1638         if (!data->rpc_done)
1639                 return status;
1640         if (status != 0) {
1641                 if (status == -NFS4ERR_BADNAME &&
1642                                 !(o_arg->open_flags & O_CREAT))
1643                         return -ENOENT;
1644                 return status;
1645         }
1646
1647         nfs_fattr_map_and_free_names(server, &data->f_attr);
1648
1649         if (o_arg->open_flags & O_CREAT) {
1650                 update_changeattr(dir, &o_res->cinfo);
1651                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1652         } else
1653                 nfs_refresh_inode(dir, o_res->dir_attr);
1654         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1655                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1656         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1657                 status = _nfs4_proc_open_confirm(data);
1658                 if (status != 0)
1659                         return status;
1660         }
1661         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1662                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1663         return 0;
1664 }
1665
1666 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1667 {
1668         unsigned int loop;
1669         int ret;
1670
1671         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1672                 ret = nfs4_wait_clnt_recover(clp);
1673                 if (ret != 0)
1674                         break;
1675                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1676                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1677                         break;
1678                 nfs4_schedule_state_manager(clp);
1679                 ret = -EIO;
1680         }
1681         return ret;
1682 }
1683
1684 static int nfs4_recover_expired_lease(struct nfs_server *server)
1685 {
1686         return nfs4_client_recover_expired_lease(server->nfs_client);
1687 }
1688
1689 /*
1690  * OPEN_EXPIRED:
1691  *      reclaim state on the server after a network partition.
1692  *      Assumes caller holds the appropriate lock
1693  */
1694 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1695 {
1696         struct nfs4_opendata *opendata;
1697         int ret;
1698
1699         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1700         if (IS_ERR(opendata))
1701                 return PTR_ERR(opendata);
1702         ret = nfs4_open_recover(opendata, state);
1703         if (ret == -ESTALE)
1704                 d_drop(ctx->dentry);
1705         nfs4_opendata_put(opendata);
1706         return ret;
1707 }
1708
1709 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1710 {
1711         struct nfs_server *server = NFS_SERVER(state->inode);
1712         struct nfs4_exception exception = { };
1713         int err;
1714
1715         do {
1716                 err = _nfs4_open_expired(ctx, state);
1717                 switch (err) {
1718                 default:
1719                         goto out;
1720                 case -NFS4ERR_GRACE:
1721                 case -NFS4ERR_DELAY:
1722                         nfs4_handle_exception(server, err, &exception);
1723                         err = 0;
1724                 }
1725         } while (exception.retry);
1726 out:
1727         return err;
1728 }
1729
1730 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1731 {
1732         struct nfs_open_context *ctx;
1733         int ret;
1734
1735         ctx = nfs4_state_find_open_context(state);
1736         if (IS_ERR(ctx))
1737                 return PTR_ERR(ctx);
1738         ret = nfs4_do_open_expired(ctx, state);
1739         put_nfs_open_context(ctx);
1740         return ret;
1741 }
1742
1743 #if defined(CONFIG_NFS_V4_1)
1744 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1745 {
1746         int status = NFS_OK;
1747         struct nfs_server *server = NFS_SERVER(state->inode);
1748
1749         if (state->flags & flags) {
1750                 status = nfs41_test_stateid(server, stateid);
1751                 if (status != NFS_OK) {
1752                         nfs41_free_stateid(server, stateid);
1753                         state->flags &= ~flags;
1754                 }
1755         }
1756         return status;
1757 }
1758
1759 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1760 {
1761         int deleg_status, open_status;
1762         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1763         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1764
1765         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1766         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1767
1768         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1769                 return NFS_OK;
1770         return nfs4_open_expired(sp, state);
1771 }
1772 #endif
1773
1774 /*
1775  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1776  * fields corresponding to attributes that were used to store the verifier.
1777  * Make sure we clobber those fields in the later setattr call
1778  */
1779 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1780 {
1781         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1782             !(sattr->ia_valid & ATTR_ATIME_SET))
1783                 sattr->ia_valid |= ATTR_ATIME;
1784
1785         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1786             !(sattr->ia_valid & ATTR_MTIME_SET))
1787                 sattr->ia_valid |= ATTR_MTIME;
1788 }
1789
1790 /*
1791  * Returns a referenced nfs4_state
1792  */
1793 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)
1794 {
1795         struct nfs4_state_owner  *sp;
1796         struct nfs4_state     *state = NULL;
1797         struct nfs_server       *server = NFS_SERVER(dir);
1798         struct nfs4_opendata *opendata;
1799         int status;
1800
1801         /* Protect against reboot recovery conflicts */
1802         status = -ENOMEM;
1803         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1804         if (sp == NULL) {
1805                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1806                 goto out_err;
1807         }
1808         status = nfs4_recover_expired_lease(server);
1809         if (status != 0)
1810                 goto err_put_state_owner;
1811         if (dentry->d_inode != NULL)
1812                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1813         status = -ENOMEM;
1814         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1815         if (opendata == NULL)
1816                 goto err_put_state_owner;
1817
1818         if (dentry->d_inode != NULL)
1819                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1820
1821         status = _nfs4_proc_open(opendata);
1822         if (status != 0)
1823                 goto err_opendata_put;
1824
1825         state = nfs4_opendata_to_nfs4_state(opendata);
1826         status = PTR_ERR(state);
1827         if (IS_ERR(state))
1828                 goto err_opendata_put;
1829         if (server->caps & NFS_CAP_POSIX_LOCK)
1830                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1831
1832         if (opendata->o_arg.open_flags & O_EXCL) {
1833                 nfs4_exclusive_attrset(opendata, sattr);
1834
1835                 nfs_fattr_init(opendata->o_res.f_attr);
1836                 status = nfs4_do_setattr(state->inode, cred,
1837                                 opendata->o_res.f_attr, sattr,
1838                                 state);
1839                 if (status == 0)
1840                         nfs_setattr_update_inode(state->inode, sattr);
1841                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1842         }
1843         nfs4_opendata_put(opendata);
1844         nfs4_put_state_owner(sp);
1845         *res = state;
1846         return 0;
1847 err_opendata_put:
1848         nfs4_opendata_put(opendata);
1849 err_put_state_owner:
1850         nfs4_put_state_owner(sp);
1851 out_err:
1852         *res = NULL;
1853         return status;
1854 }
1855
1856
1857 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)
1858 {
1859         struct nfs4_exception exception = { };
1860         struct nfs4_state *res;
1861         int status;
1862
1863         do {
1864                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1865                 if (status == 0)
1866                         break;
1867                 /* NOTE: BAD_SEQID means the server and client disagree about the
1868                  * book-keeping w.r.t. state-changing operations
1869                  * (OPEN/CLOSE/LOCK/LOCKU...)
1870                  * It is actually a sign of a bug on the client or on the server.
1871                  *
1872                  * If we receive a BAD_SEQID error in the particular case of
1873                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1874                  * have unhashed the old state_owner for us, and that we can
1875                  * therefore safely retry using a new one. We should still warn
1876                  * the user though...
1877                  */
1878                 if (status == -NFS4ERR_BAD_SEQID) {
1879                         pr_warn_ratelimited("NFS: v4 server %s "
1880                                         " returned a bad sequence-id error!\n",
1881                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1882                         exception.retry = 1;
1883                         continue;
1884                 }
1885                 /*
1886                  * BAD_STATEID on OPEN means that the server cancelled our
1887                  * state before it received the OPEN_CONFIRM.
1888                  * Recover by retrying the request as per the discussion
1889                  * on Page 181 of RFC3530.
1890                  */
1891                 if (status == -NFS4ERR_BAD_STATEID) {
1892                         exception.retry = 1;
1893                         continue;
1894                 }
1895                 if (status == -EAGAIN) {
1896                         /* We must have found a delegation */
1897                         exception.retry = 1;
1898                         continue;
1899                 }
1900                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1901                                         status, &exception));
1902         } while (exception.retry);
1903         return res;
1904 }
1905
1906 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1907                             struct nfs_fattr *fattr, struct iattr *sattr,
1908                             struct nfs4_state *state)
1909 {
1910         struct nfs_server *server = NFS_SERVER(inode);
1911         struct nfs_setattrargs  arg = {
1912                 .fh             = NFS_FH(inode),
1913                 .iap            = sattr,
1914                 .server         = server,
1915                 .bitmask = server->attr_bitmask,
1916         };
1917         struct nfs_setattrres  res = {
1918                 .fattr          = fattr,
1919                 .server         = server,
1920         };
1921         struct rpc_message msg = {
1922                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1923                 .rpc_argp       = &arg,
1924                 .rpc_resp       = &res,
1925                 .rpc_cred       = cred,
1926         };
1927         unsigned long timestamp = jiffies;
1928         int status;
1929
1930         nfs_fattr_init(fattr);
1931
1932         if (state != NULL) {
1933                 nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
1934                                 current->files, current->tgid);
1935         } else if (nfs4_copy_delegation_stateid(&arg.stateid, inode,
1936                                 FMODE_WRITE)) {
1937                 /* Use that stateid */
1938         } else
1939                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1940
1941         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1942         if (status == 0 && state != NULL)
1943                 renew_lease(server, timestamp);
1944         return status;
1945 }
1946
1947 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1948                            struct nfs_fattr *fattr, struct iattr *sattr,
1949                            struct nfs4_state *state)
1950 {
1951         struct nfs_server *server = NFS_SERVER(inode);
1952         struct nfs4_exception exception = {
1953                 .state = state,
1954                 .inode = inode,
1955         };
1956         int err;
1957         do {
1958                 err = nfs4_handle_exception(server,
1959                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1960                                 &exception);
1961         } while (exception.retry);
1962         return err;
1963 }
1964
1965 struct nfs4_closedata {
1966         struct inode *inode;
1967         struct nfs4_state *state;
1968         struct nfs_closeargs arg;
1969         struct nfs_closeres res;
1970         struct nfs_fattr fattr;
1971         unsigned long timestamp;
1972         bool roc;
1973         u32 roc_barrier;
1974 };
1975
1976 static void nfs4_free_closedata(void *data)
1977 {
1978         struct nfs4_closedata *calldata = data;
1979         struct nfs4_state_owner *sp = calldata->state->owner;
1980         struct super_block *sb = calldata->state->inode->i_sb;
1981
1982         if (calldata->roc)
1983                 pnfs_roc_release(calldata->state->inode);
1984         nfs4_put_open_state(calldata->state);
1985         nfs_free_seqid(calldata->arg.seqid);
1986         nfs4_put_state_owner(sp);
1987         nfs_sb_deactive(sb);
1988         kfree(calldata);
1989 }
1990
1991 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1992                 fmode_t fmode)
1993 {
1994         spin_lock(&state->owner->so_lock);
1995         if (!(fmode & FMODE_READ))
1996                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1997         if (!(fmode & FMODE_WRITE))
1998                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1999         clear_bit(NFS_O_RDWR_STATE, &state->flags);
2000         spin_unlock(&state->owner->so_lock);
2001 }
2002
2003 static void nfs4_close_done(struct rpc_task *task, void *data)
2004 {
2005         struct nfs4_closedata *calldata = data;
2006         struct nfs4_state *state = calldata->state;
2007         struct nfs_server *server = NFS_SERVER(calldata->inode);
2008
2009         dprintk("%s: begin!\n", __func__);
2010         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2011                 return;
2012         /* hmm. we are done with the inode, and in the process of freeing
2013          * the state_owner. we keep this around to process errors
2014          */
2015         switch (task->tk_status) {
2016                 case 0:
2017                         if (calldata->roc)
2018                                 pnfs_roc_set_barrier(state->inode,
2019                                                      calldata->roc_barrier);
2020                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2021                         renew_lease(server, calldata->timestamp);
2022                         nfs4_close_clear_stateid_flags(state,
2023                                         calldata->arg.fmode);
2024                         break;
2025                 case -NFS4ERR_STALE_STATEID:
2026                 case -NFS4ERR_OLD_STATEID:
2027                 case -NFS4ERR_BAD_STATEID:
2028                 case -NFS4ERR_EXPIRED:
2029                         if (calldata->arg.fmode == 0)
2030                                 break;
2031                 default:
2032                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2033                                 rpc_restart_call_prepare(task);
2034         }
2035         nfs_release_seqid(calldata->arg.seqid);
2036         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2037         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2038 }
2039
2040 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2041 {
2042         struct nfs4_closedata *calldata = data;
2043         struct nfs4_state *state = calldata->state;
2044         int call_close = 0;
2045
2046         dprintk("%s: begin!\n", __func__);
2047         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2048                 return;
2049
2050         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2051         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2052         spin_lock(&state->owner->so_lock);
2053         /* Calculate the change in open mode */
2054         if (state->n_rdwr == 0) {
2055                 if (state->n_rdonly == 0) {
2056                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2057                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2058                         calldata->arg.fmode &= ~FMODE_READ;
2059                 }
2060                 if (state->n_wronly == 0) {
2061                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2062                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2063                         calldata->arg.fmode &= ~FMODE_WRITE;
2064                 }
2065         }
2066         spin_unlock(&state->owner->so_lock);
2067
2068         if (!call_close) {
2069                 /* Note: exit _without_ calling nfs4_close_done */
2070                 task->tk_action = NULL;
2071                 goto out;
2072         }
2073
2074         if (calldata->arg.fmode == 0) {
2075                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2076                 if (calldata->roc &&
2077                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2078                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2079                                      task, NULL);
2080                         goto out;
2081                 }
2082         }
2083
2084         nfs_fattr_init(calldata->res.fattr);
2085         calldata->timestamp = jiffies;
2086         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2087                                 &calldata->arg.seq_args,
2088                                 &calldata->res.seq_res,
2089                                 task))
2090                 goto out;
2091         rpc_call_start(task);
2092 out:
2093         dprintk("%s: done!\n", __func__);
2094 }
2095
2096 static const struct rpc_call_ops nfs4_close_ops = {
2097         .rpc_call_prepare = nfs4_close_prepare,
2098         .rpc_call_done = nfs4_close_done,
2099         .rpc_release = nfs4_free_closedata,
2100 };
2101
2102 /*
2103  * It is possible for data to be read/written from a mem-mapped file
2104  * after the sys_close call (which hits the vfs layer as a flush).
2105  * This means that we can't safely call nfsv4 close on a file until
2106  * the inode is cleared. This in turn means that we are not good
2107  * NFSv4 citizens - we do not indicate to the server to update the file's
2108  * share state even when we are done with one of the three share
2109  * stateid's in the inode.
2110  *
2111  * NOTE: Caller must be holding the sp->so_owner semaphore!
2112  */
2113 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2114 {
2115         struct nfs_server *server = NFS_SERVER(state->inode);
2116         struct nfs4_closedata *calldata;
2117         struct nfs4_state_owner *sp = state->owner;
2118         struct rpc_task *task;
2119         struct rpc_message msg = {
2120                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2121                 .rpc_cred = state->owner->so_cred,
2122         };
2123         struct rpc_task_setup task_setup_data = {
2124                 .rpc_client = server->client,
2125                 .rpc_message = &msg,
2126                 .callback_ops = &nfs4_close_ops,
2127                 .workqueue = nfsiod_workqueue,
2128                 .flags = RPC_TASK_ASYNC,
2129         };
2130         int status = -ENOMEM;
2131
2132         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2133         if (calldata == NULL)
2134                 goto out;
2135         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2136         calldata->inode = state->inode;
2137         calldata->state = state;
2138         calldata->arg.fh = NFS_FH(state->inode);
2139         calldata->arg.stateid = &state->open_stateid;
2140         /* Serialization for the sequence id */
2141         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2142         if (calldata->arg.seqid == NULL)
2143                 goto out_free_calldata;
2144         calldata->arg.fmode = 0;
2145         calldata->arg.bitmask = server->cache_consistency_bitmask;
2146         calldata->res.fattr = &calldata->fattr;
2147         calldata->res.seqid = calldata->arg.seqid;
2148         calldata->res.server = server;
2149         calldata->roc = roc;
2150         nfs_sb_active(calldata->inode->i_sb);
2151
2152         msg.rpc_argp = &calldata->arg;
2153         msg.rpc_resp = &calldata->res;
2154         task_setup_data.callback_data = calldata;
2155         task = rpc_run_task(&task_setup_data);
2156         if (IS_ERR(task))
2157                 return PTR_ERR(task);
2158         status = 0;
2159         if (wait)
2160                 status = rpc_wait_for_completion_task(task);
2161         rpc_put_task(task);
2162         return status;
2163 out_free_calldata:
2164         kfree(calldata);
2165 out:
2166         if (roc)
2167                 pnfs_roc_release(state->inode);
2168         nfs4_put_open_state(state);
2169         nfs4_put_state_owner(sp);
2170         return status;
2171 }
2172
2173 static struct inode *
2174 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2175 {
2176         struct nfs4_state *state;
2177
2178         /* Protect against concurrent sillydeletes */
2179         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2180         if (IS_ERR(state))
2181                 return ERR_CAST(state);
2182         ctx->state = state;
2183         return igrab(state->inode);
2184 }
2185
2186 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2187 {
2188         if (ctx->state == NULL)
2189                 return;
2190         if (is_sync)
2191                 nfs4_close_sync(ctx->state, ctx->mode);
2192         else
2193                 nfs4_close_state(ctx->state, ctx->mode);
2194 }
2195
2196 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2197 {
2198         struct nfs4_server_caps_arg args = {
2199                 .fhandle = fhandle,
2200         };
2201         struct nfs4_server_caps_res res = {};
2202         struct rpc_message msg = {
2203                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2204                 .rpc_argp = &args,
2205                 .rpc_resp = &res,
2206         };
2207         int status;
2208
2209         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2210         if (status == 0) {
2211                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2212                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2213                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2214                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2215                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2216                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2217                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2218                         server->caps |= NFS_CAP_ACLS;
2219                 if (res.has_links != 0)
2220                         server->caps |= NFS_CAP_HARDLINKS;
2221                 if (res.has_symlinks != 0)
2222                         server->caps |= NFS_CAP_SYMLINKS;
2223                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2224                         server->caps |= NFS_CAP_FILEID;
2225                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2226                         server->caps |= NFS_CAP_MODE;
2227                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2228                         server->caps |= NFS_CAP_NLINK;
2229                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2230                         server->caps |= NFS_CAP_OWNER;
2231                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2232                         server->caps |= NFS_CAP_OWNER_GROUP;
2233                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2234                         server->caps |= NFS_CAP_ATIME;
2235                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2236                         server->caps |= NFS_CAP_CTIME;
2237                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2238                         server->caps |= NFS_CAP_MTIME;
2239
2240                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2241                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2242                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2243                 server->acl_bitmask = res.acl_bitmask;
2244                 server->fh_expire_type = res.fh_expire_type;
2245         }
2246
2247         return status;
2248 }
2249
2250 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2251 {
2252         struct nfs4_exception exception = { };
2253         int err;
2254         do {
2255                 err = nfs4_handle_exception(server,
2256                                 _nfs4_server_capabilities(server, fhandle),
2257                                 &exception);
2258         } while (exception.retry);
2259         return err;
2260 }
2261
2262 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2263                 struct nfs_fsinfo *info)
2264 {
2265         struct nfs4_lookup_root_arg args = {
2266                 .bitmask = nfs4_fattr_bitmap,
2267         };
2268         struct nfs4_lookup_res res = {
2269                 .server = server,
2270                 .fattr = info->fattr,
2271                 .fh = fhandle,
2272         };
2273         struct rpc_message msg = {
2274                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2275                 .rpc_argp = &args,
2276                 .rpc_resp = &res,
2277         };
2278
2279         nfs_fattr_init(info->fattr);
2280         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2281 }
2282
2283 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2284                 struct nfs_fsinfo *info)
2285 {
2286         struct nfs4_exception exception = { };
2287         int err;
2288         do {
2289                 err = _nfs4_lookup_root(server, fhandle, info);
2290                 switch (err) {
2291                 case 0:
2292                 case -NFS4ERR_WRONGSEC:
2293                         break;
2294                 default:
2295                         err = nfs4_handle_exception(server, err, &exception);
2296                 }
2297         } while (exception.retry);
2298         return err;
2299 }
2300
2301 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2302                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2303 {
2304         struct rpc_auth *auth;
2305         int ret;
2306
2307         auth = rpcauth_create(flavor, server->client);
2308         if (!auth) {
2309                 ret = -EIO;
2310                 goto out;
2311         }
2312         ret = nfs4_lookup_root(server, fhandle, info);
2313 out:
2314         return ret;
2315 }
2316
2317 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2318                               struct nfs_fsinfo *info)
2319 {
2320         int i, len, status = 0;
2321         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2322
2323         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2324         flav_array[len] = RPC_AUTH_NULL;
2325         len += 1;
2326
2327         for (i = 0; i < len; i++) {
2328                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2329                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2330                         continue;
2331                 break;
2332         }
2333         /*
2334          * -EACCESS could mean that the user doesn't have correct permissions
2335          * to access the mount.  It could also mean that we tried to mount
2336          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2337          * existing mount programs don't handle -EACCES very well so it should
2338          * be mapped to -EPERM instead.
2339          */
2340         if (status == -EACCES)
2341                 status = -EPERM;
2342         return status;
2343 }
2344
2345 /*
2346  * get the file handle for the "/" directory on the server
2347  */
2348 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2349                               struct nfs_fsinfo *info)
2350 {
2351         int minor_version = server->nfs_client->cl_minorversion;
2352         int status = nfs4_lookup_root(server, fhandle, info);
2353         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2354                 /*
2355                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2356                  * by nfs4_map_errors() as this function exits.
2357                  */
2358                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2359         if (status == 0)
2360                 status = nfs4_server_capabilities(server, fhandle);
2361         if (status == 0)
2362                 status = nfs4_do_fsinfo(server, fhandle, info);
2363         return nfs4_map_errors(status);
2364 }
2365
2366 /*
2367  * Get locations and (maybe) other attributes of a referral.
2368  * Note that we'll actually follow the referral later when
2369  * we detect fsid mismatch in inode revalidation
2370  */
2371 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2372                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2373 {
2374         int status = -ENOMEM;
2375         struct page *page = NULL;
2376         struct nfs4_fs_locations *locations = NULL;
2377
2378         page = alloc_page(GFP_KERNEL);
2379         if (page == NULL)
2380                 goto out;
2381         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2382         if (locations == NULL)
2383                 goto out;
2384
2385         status = nfs4_proc_fs_locations(dir, name, locations, page);
2386         if (status != 0)
2387                 goto out;
2388         /* Make sure server returned a different fsid for the referral */
2389         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2390                 dprintk("%s: server did not return a different fsid for"
2391                         " a referral at %s\n", __func__, name->name);
2392                 status = -EIO;
2393                 goto out;
2394         }
2395         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2396         nfs_fixup_referral_attributes(&locations->fattr);
2397
2398         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2399         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2400         memset(fhandle, 0, sizeof(struct nfs_fh));
2401 out:
2402         if (page)
2403                 __free_page(page);
2404         kfree(locations);
2405         return status;
2406 }
2407
2408 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2409 {
2410         struct nfs4_getattr_arg args = {
2411                 .fh = fhandle,
2412                 .bitmask = server->attr_bitmask,
2413         };
2414         struct nfs4_getattr_res res = {
2415                 .fattr = fattr,
2416                 .server = server,
2417         };
2418         struct rpc_message msg = {
2419                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2420                 .rpc_argp = &args,
2421                 .rpc_resp = &res,
2422         };
2423
2424         nfs_fattr_init(fattr);
2425         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2426 }
2427
2428 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2429 {
2430         struct nfs4_exception exception = { };
2431         int err;
2432         do {
2433                 err = nfs4_handle_exception(server,
2434                                 _nfs4_proc_getattr(server, fhandle, fattr),
2435                                 &exception);
2436         } while (exception.retry);
2437         return err;
2438 }
2439
2440 /*
2441  * The file is not closed if it is opened due to the a request to change
2442  * the size of the file. The open call will not be needed once the
2443  * VFS layer lookup-intents are implemented.
2444  *
2445  * Close is called when the inode is destroyed.
2446  * If we haven't opened the file for O_WRONLY, we
2447  * need to in the size_change case to obtain a stateid.
2448  *
2449  * Got race?
2450  * Because OPEN is always done by name in nfsv4, it is
2451  * possible that we opened a different file by the same
2452  * name.  We can recognize this race condition, but we
2453  * can't do anything about it besides returning an error.
2454  *
2455  * This will be fixed with VFS changes (lookup-intent).
2456  */
2457 static int
2458 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2459                   struct iattr *sattr)
2460 {
2461         struct inode *inode = dentry->d_inode;
2462         struct rpc_cred *cred = NULL;
2463         struct nfs4_state *state = NULL;
2464         int status;
2465
2466         if (pnfs_ld_layoutret_on_setattr(inode))
2467                 pnfs_return_layout(inode);
2468
2469         nfs_fattr_init(fattr);
2470
2471         /* Search for an existing open(O_WRITE) file */
2472         if (sattr->ia_valid & ATTR_FILE) {
2473                 struct nfs_open_context *ctx;
2474
2475                 ctx = nfs_file_open_context(sattr->ia_file);
2476                 if (ctx) {
2477                         cred = ctx->cred;
2478                         state = ctx->state;
2479                 }
2480         }
2481
2482         /* Deal with open(O_TRUNC) */
2483         if (sattr->ia_valid & ATTR_OPEN)
2484                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2485
2486         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2487         if (status == 0)
2488                 nfs_setattr_update_inode(inode, sattr);
2489         return status;
2490 }
2491
2492 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2493                 const struct qstr *name, struct nfs_fh *fhandle,
2494                 struct nfs_fattr *fattr)
2495 {
2496         struct nfs_server *server = NFS_SERVER(dir);
2497         int                    status;
2498         struct nfs4_lookup_arg args = {
2499                 .bitmask = server->attr_bitmask,
2500                 .dir_fh = NFS_FH(dir),
2501                 .name = name,
2502         };
2503         struct nfs4_lookup_res res = {
2504                 .server = server,
2505                 .fattr = fattr,
2506                 .fh = fhandle,
2507         };
2508         struct rpc_message msg = {
2509                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2510                 .rpc_argp = &args,
2511                 .rpc_resp = &res,
2512         };
2513
2514         nfs_fattr_init(fattr);
2515
2516         dprintk("NFS call  lookup %s\n", name->name);
2517         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2518         dprintk("NFS reply lookup: %d\n", status);
2519         return status;
2520 }
2521
2522 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2523 {
2524         memset(fh, 0, sizeof(struct nfs_fh));
2525         fattr->fsid.major = 1;
2526         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2527                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2528         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2529         fattr->nlink = 2;
2530 }
2531
2532 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2533                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2534 {
2535         struct nfs4_exception exception = { };
2536         int err;
2537         do {
2538                 int status;
2539
2540                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2541                 switch (status) {
2542                 case -NFS4ERR_BADNAME:
2543                         return -ENOENT;
2544                 case -NFS4ERR_MOVED:
2545                         return nfs4_get_referral(dir, name, fattr, fhandle);
2546                 case -NFS4ERR_WRONGSEC:
2547                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2548                 }
2549                 err = nfs4_handle_exception(NFS_SERVER(dir),
2550                                 status, &exception);
2551         } while (exception.retry);
2552         return err;
2553 }
2554
2555 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2556 {
2557         struct nfs_server *server = NFS_SERVER(inode);
2558         struct nfs4_accessargs args = {
2559                 .fh = NFS_FH(inode),
2560                 .bitmask = server->cache_consistency_bitmask,
2561         };
2562         struct nfs4_accessres res = {
2563                 .server = server,
2564         };
2565         struct rpc_message msg = {
2566                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2567                 .rpc_argp = &args,
2568                 .rpc_resp = &res,
2569                 .rpc_cred = entry->cred,
2570         };
2571         int mode = entry->mask;
2572         int status;
2573
2574         /*
2575          * Determine which access bits we want to ask for...
2576          */
2577         if (mode & MAY_READ)
2578                 args.access |= NFS4_ACCESS_READ;
2579         if (S_ISDIR(inode->i_mode)) {
2580                 if (mode & MAY_WRITE)
2581                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2582                 if (mode & MAY_EXEC)
2583                         args.access |= NFS4_ACCESS_LOOKUP;
2584         } else {
2585                 if (mode & MAY_WRITE)
2586                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2587                 if (mode & MAY_EXEC)
2588                         args.access |= NFS4_ACCESS_EXECUTE;
2589         }
2590
2591         res.fattr = nfs_alloc_fattr();
2592         if (res.fattr == NULL)
2593                 return -ENOMEM;
2594
2595         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2596         if (!status) {
2597                 entry->mask = 0;
2598                 if (res.access & NFS4_ACCESS_READ)
2599                         entry->mask |= MAY_READ;
2600                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2601                         entry->mask |= MAY_WRITE;
2602                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2603                         entry->mask |= MAY_EXEC;
2604                 nfs_refresh_inode(inode, res.fattr);
2605         }
2606         nfs_free_fattr(res.fattr);
2607         return status;
2608 }
2609
2610 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2611 {
2612         struct nfs4_exception exception = { };
2613         int err;
2614         do {
2615                 err = nfs4_handle_exception(NFS_SERVER(inode),
2616                                 _nfs4_proc_access(inode, entry),
2617                                 &exception);
2618         } while (exception.retry);
2619         return err;
2620 }
2621
2622 /*
2623  * TODO: For the time being, we don't try to get any attributes
2624  * along with any of the zero-copy operations READ, READDIR,
2625  * READLINK, WRITE.
2626  *
2627  * In the case of the first three, we want to put the GETATTR
2628  * after the read-type operation -- this is because it is hard
2629  * to predict the length of a GETATTR response in v4, and thus
2630  * align the READ data correctly.  This means that the GETATTR
2631  * may end up partially falling into the page cache, and we should
2632  * shift it into the 'tail' of the xdr_buf before processing.
2633  * To do this efficiently, we need to know the total length
2634  * of data received, which doesn't seem to be available outside
2635  * of the RPC layer.
2636  *
2637  * In the case of WRITE, we also want to put the GETATTR after
2638  * the operation -- in this case because we want to make sure
2639  * we get the post-operation mtime and size.  This means that
2640  * we can't use xdr_encode_pages() as written: we need a variant
2641  * of it which would leave room in the 'tail' iovec.
2642  *
2643  * Both of these changes to the XDR layer would in fact be quite
2644  * minor, but I decided to leave them for a subsequent patch.
2645  */
2646 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2647                 unsigned int pgbase, unsigned int pglen)
2648 {
2649         struct nfs4_readlink args = {
2650                 .fh       = NFS_FH(inode),
2651                 .pgbase   = pgbase,
2652                 .pglen    = pglen,
2653                 .pages    = &page,
2654         };
2655         struct nfs4_readlink_res res;
2656         struct rpc_message msg = {
2657                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2658                 .rpc_argp = &args,
2659                 .rpc_resp = &res,
2660         };
2661
2662         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2663 }
2664
2665 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2666                 unsigned int pgbase, unsigned int pglen)
2667 {
2668         struct nfs4_exception exception = { };
2669         int err;
2670         do {
2671                 err = nfs4_handle_exception(NFS_SERVER(inode),
2672                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2673                                 &exception);
2674         } while (exception.retry);
2675         return err;
2676 }
2677
2678 /*
2679  * Got race?
2680  * We will need to arrange for the VFS layer to provide an atomic open.
2681  * Until then, this create/open method is prone to inefficiency and race
2682  * conditions due to the lookup, create, and open VFS calls from sys_open()
2683  * placed on the wire.
2684  *
2685  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2686  * The file will be opened again in the subsequent VFS open call
2687  * (nfs4_proc_file_open).
2688  *
2689  * The open for read will just hang around to be used by any process that
2690  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2691  */
2692
2693 static int
2694 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2695                  int flags, struct nfs_open_context *ctx)
2696 {
2697         struct dentry *de = dentry;
2698         struct nfs4_state *state;
2699         struct rpc_cred *cred = NULL;
2700         fmode_t fmode = 0;
2701         int status = 0;
2702
2703         if (ctx != NULL) {
2704                 cred = ctx->cred;
2705                 de = ctx->dentry;
2706                 fmode = ctx->mode;
2707         }
2708         sattr->ia_mode &= ~current_umask();
2709         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2710         d_drop(dentry);
2711         if (IS_ERR(state)) {
2712                 status = PTR_ERR(state);
2713                 goto out;
2714         }
2715         d_add(dentry, igrab(state->inode));
2716         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2717         if (ctx != NULL)
2718                 ctx->state = state;
2719         else
2720                 nfs4_close_sync(state, fmode);
2721 out:
2722         return status;
2723 }
2724
2725 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2726 {
2727         struct nfs_server *server = NFS_SERVER(dir);
2728         struct nfs_removeargs args = {
2729                 .fh = NFS_FH(dir),
2730                 .name.len = name->len,
2731                 .name.name = name->name,
2732                 .bitmask = server->attr_bitmask,
2733         };
2734         struct nfs_removeres res = {
2735                 .server = server,
2736         };
2737         struct rpc_message msg = {
2738                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2739                 .rpc_argp = &args,
2740                 .rpc_resp = &res,
2741         };
2742         int status = -ENOMEM;
2743
2744         res.dir_attr = nfs_alloc_fattr();
2745         if (res.dir_attr == NULL)
2746                 goto out;
2747
2748         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2749         if (status == 0) {
2750                 update_changeattr(dir, &res.cinfo);
2751                 nfs_post_op_update_inode(dir, res.dir_attr);
2752         }
2753         nfs_free_fattr(res.dir_attr);
2754 out:
2755         return status;
2756 }
2757
2758 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2759 {
2760         struct nfs4_exception exception = { };
2761         int err;
2762         do {
2763                 err = nfs4_handle_exception(NFS_SERVER(dir),
2764                                 _nfs4_proc_remove(dir, name),
2765                                 &exception);
2766         } while (exception.retry);
2767         return err;
2768 }
2769
2770 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2771 {
2772         struct nfs_server *server = NFS_SERVER(dir);
2773         struct nfs_removeargs *args = msg->rpc_argp;
2774         struct nfs_removeres *res = msg->rpc_resp;
2775
2776         args->bitmask = server->cache_consistency_bitmask;
2777         res->server = server;
2778         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2779         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2780 }
2781
2782 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2783 {
2784         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2785
2786         if (!nfs4_sequence_done(task, &res->seq_res))
2787                 return 0;
2788         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2789                 return 0;
2790         update_changeattr(dir, &res->cinfo);
2791         nfs_post_op_update_inode(dir, res->dir_attr);
2792         return 1;
2793 }
2794
2795 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2796 {
2797         struct nfs_server *server = NFS_SERVER(dir);
2798         struct nfs_renameargs *arg = msg->rpc_argp;
2799         struct nfs_renameres *res = msg->rpc_resp;
2800
2801         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2802         arg->bitmask = server->attr_bitmask;
2803         res->server = server;
2804         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2805 }
2806
2807 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2808                                  struct inode *new_dir)
2809 {
2810         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2811
2812         if (!nfs4_sequence_done(task, &res->seq_res))
2813                 return 0;
2814         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2815                 return 0;
2816
2817         update_changeattr(old_dir, &res->old_cinfo);
2818         nfs_post_op_update_inode(old_dir, res->old_fattr);
2819         update_changeattr(new_dir, &res->new_cinfo);
2820         nfs_post_op_update_inode(new_dir, res->new_fattr);
2821         return 1;
2822 }
2823
2824 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2825                 struct inode *new_dir, struct qstr *new_name)
2826 {
2827         struct nfs_server *server = NFS_SERVER(old_dir);
2828         struct nfs_renameargs arg = {
2829                 .old_dir = NFS_FH(old_dir),
2830                 .new_dir = NFS_FH(new_dir),
2831                 .old_name = old_name,
2832                 .new_name = new_name,
2833                 .bitmask = server->attr_bitmask,
2834         };
2835         struct nfs_renameres res = {
2836                 .server = server,
2837         };
2838         struct rpc_message msg = {
2839                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2840                 .rpc_argp = &arg,
2841                 .rpc_resp = &res,
2842         };
2843         int status = -ENOMEM;
2844
2845         res.old_fattr = nfs_alloc_fattr();
2846         res.new_fattr = nfs_alloc_fattr();
2847         if (res.old_fattr == NULL || res.new_fattr == NULL)
2848                 goto out;
2849
2850         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2851         if (!status) {
2852                 update_changeattr(old_dir, &res.old_cinfo);
2853                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2854                 update_changeattr(new_dir, &res.new_cinfo);
2855                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2856         }
2857 out:
2858         nfs_free_fattr(res.new_fattr);
2859         nfs_free_fattr(res.old_fattr);
2860         return status;
2861 }
2862
2863 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2864                 struct inode *new_dir, struct qstr *new_name)
2865 {
2866         struct nfs4_exception exception = { };
2867         int err;
2868         do {
2869                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2870                                 _nfs4_proc_rename(old_dir, old_name,
2871                                         new_dir, new_name),
2872                                 &exception);
2873         } while (exception.retry);
2874         return err;
2875 }
2876
2877 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2878 {
2879         struct nfs_server *server = NFS_SERVER(inode);
2880         struct nfs4_link_arg arg = {
2881                 .fh     = NFS_FH(inode),
2882                 .dir_fh = NFS_FH(dir),
2883                 .name   = name,
2884                 .bitmask = server->attr_bitmask,
2885         };
2886         struct nfs4_link_res res = {
2887                 .server = server,
2888         };
2889         struct rpc_message msg = {
2890                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2891                 .rpc_argp = &arg,
2892                 .rpc_resp = &res,
2893         };
2894         int status = -ENOMEM;
2895
2896         res.fattr = nfs_alloc_fattr();
2897         res.dir_attr = nfs_alloc_fattr();
2898         if (res.fattr == NULL || res.dir_attr == NULL)
2899                 goto out;
2900
2901         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2902         if (!status) {
2903                 update_changeattr(dir, &res.cinfo);
2904                 nfs_post_op_update_inode(dir, res.dir_attr);
2905                 nfs_post_op_update_inode(inode, res.fattr);
2906         }
2907 out:
2908         nfs_free_fattr(res.dir_attr);
2909         nfs_free_fattr(res.fattr);
2910         return status;
2911 }
2912
2913 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2914 {
2915         struct nfs4_exception exception = { };
2916         int err;
2917         do {
2918                 err = nfs4_handle_exception(NFS_SERVER(inode),
2919                                 _nfs4_proc_link(inode, dir, name),
2920                                 &exception);
2921         } while (exception.retry);
2922         return err;
2923 }
2924
2925 struct nfs4_createdata {
2926         struct rpc_message msg;
2927         struct nfs4_create_arg arg;
2928         struct nfs4_create_res res;
2929         struct nfs_fh fh;
2930         struct nfs_fattr fattr;
2931         struct nfs_fattr dir_fattr;
2932 };
2933
2934 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2935                 struct qstr *name, struct iattr *sattr, u32 ftype)
2936 {
2937         struct nfs4_createdata *data;
2938
2939         data = kzalloc(sizeof(*data), GFP_KERNEL);
2940         if (data != NULL) {
2941                 struct nfs_server *server = NFS_SERVER(dir);
2942
2943                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2944                 data->msg.rpc_argp = &data->arg;
2945                 data->msg.rpc_resp = &data->res;
2946                 data->arg.dir_fh = NFS_FH(dir);
2947                 data->arg.server = server;
2948                 data->arg.name = name;
2949                 data->arg.attrs = sattr;
2950                 data->arg.ftype = ftype;
2951                 data->arg.bitmask = server->attr_bitmask;
2952                 data->res.server = server;
2953                 data->res.fh = &data->fh;
2954                 data->res.fattr = &data->fattr;
2955                 data->res.dir_fattr = &data->dir_fattr;
2956                 nfs_fattr_init(data->res.fattr);
2957                 nfs_fattr_init(data->res.dir_fattr);
2958         }
2959         return data;
2960 }
2961
2962 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2963 {
2964         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2965                                     &data->arg.seq_args, &data->res.seq_res, 1);
2966         if (status == 0) {
2967                 update_changeattr(dir, &data->res.dir_cinfo);
2968                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2969                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2970         }
2971         return status;
2972 }
2973
2974 static void nfs4_free_createdata(struct nfs4_createdata *data)
2975 {
2976         kfree(data);
2977 }
2978
2979 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2980                 struct page *page, unsigned int len, struct iattr *sattr)
2981 {
2982         struct nfs4_createdata *data;
2983         int status = -ENAMETOOLONG;
2984
2985         if (len > NFS4_MAXPATHLEN)
2986                 goto out;
2987
2988         status = -ENOMEM;
2989         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2990         if (data == NULL)
2991                 goto out;
2992
2993         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2994         data->arg.u.symlink.pages = &page;
2995         data->arg.u.symlink.len = len;
2996
2997         status = nfs4_do_create(dir, dentry, data);
2998
2999         nfs4_free_createdata(data);
3000 out:
3001         return status;
3002 }
3003
3004 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3005                 struct page *page, unsigned int len, struct iattr *sattr)
3006 {
3007         struct nfs4_exception exception = { };
3008         int err;
3009         do {
3010                 err = nfs4_handle_exception(NFS_SERVER(dir),
3011                                 _nfs4_proc_symlink(dir, dentry, page,
3012                                                         len, sattr),
3013                                 &exception);
3014         } while (exception.retry);
3015         return err;
3016 }
3017
3018 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3019                 struct iattr *sattr)
3020 {
3021         struct nfs4_createdata *data;
3022         int status = -ENOMEM;
3023
3024         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3025         if (data == NULL)
3026                 goto out;
3027
3028         status = nfs4_do_create(dir, dentry, data);
3029
3030         nfs4_free_createdata(data);
3031 out:
3032         return status;
3033 }
3034
3035 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3036                 struct iattr *sattr)
3037 {
3038         struct nfs4_exception exception = { };
3039         int err;
3040
3041         sattr->ia_mode &= ~current_umask();
3042         do {
3043                 err = nfs4_handle_exception(NFS_SERVER(dir),
3044                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3045                                 &exception);
3046         } while (exception.retry);
3047         return err;
3048 }
3049
3050 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3051                 u64 cookie, struct page **pages, unsigned int count, int plus)
3052 {
3053         struct inode            *dir = dentry->d_inode;
3054         struct nfs4_readdir_arg args = {
3055                 .fh = NFS_FH(dir),
3056                 .pages = pages,
3057                 .pgbase = 0,
3058                 .count = count,
3059                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3060                 .plus = plus,
3061         };
3062         struct nfs4_readdir_res res;
3063         struct rpc_message msg = {
3064                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3065                 .rpc_argp = &args,
3066                 .rpc_resp = &res,
3067                 .rpc_cred = cred,
3068         };
3069         int                     status;
3070
3071         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3072                         dentry->d_parent->d_name.name,
3073                         dentry->d_name.name,
3074                         (unsigned long long)cookie);
3075         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3076         res.pgbase = args.pgbase;
3077         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3078         if (status >= 0) {
3079                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3080                 status += args.pgbase;
3081         }
3082
3083         nfs_invalidate_atime(dir);
3084
3085         dprintk("%s: returns %d\n", __func__, status);
3086         return status;
3087 }
3088
3089 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3090                 u64 cookie, struct page **pages, unsigned int count, int plus)
3091 {
3092         struct nfs4_exception exception = { };
3093         int err;
3094         do {
3095                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3096                                 _nfs4_proc_readdir(dentry, cred, cookie,
3097                                         pages, count, plus),
3098                                 &exception);
3099         } while (exception.retry);
3100         return err;
3101 }
3102
3103 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3104                 struct iattr *sattr, dev_t rdev)
3105 {
3106         struct nfs4_createdata *data;
3107         int mode = sattr->ia_mode;
3108         int status = -ENOMEM;
3109
3110         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3111         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3112
3113         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3114         if (data == NULL)
3115                 goto out;
3116
3117         if (S_ISFIFO(mode))
3118                 data->arg.ftype = NF4FIFO;
3119         else if (S_ISBLK(mode)) {
3120                 data->arg.ftype = NF4BLK;
3121                 data->arg.u.device.specdata1 = MAJOR(rdev);
3122                 data->arg.u.device.specdata2 = MINOR(rdev);
3123         }
3124         else if (S_ISCHR(mode)) {
3125                 data->arg.ftype = NF4CHR;
3126                 data->arg.u.device.specdata1 = MAJOR(rdev);
3127                 data->arg.u.device.specdata2 = MINOR(rdev);
3128         }
3129
3130         status = nfs4_do_create(dir, dentry, data);
3131
3132         nfs4_free_createdata(data);
3133 out:
3134         return status;
3135 }
3136
3137 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3138                 struct iattr *sattr, dev_t rdev)
3139 {
3140         struct nfs4_exception exception = { };
3141         int err;
3142
3143         sattr->ia_mode &= ~current_umask();
3144         do {
3145                 err = nfs4_handle_exception(NFS_SERVER(dir),
3146                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3147                                 &exception);
3148         } while (exception.retry);
3149         return err;
3150 }
3151
3152 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3153                  struct nfs_fsstat *fsstat)
3154 {
3155         struct nfs4_statfs_arg args = {
3156                 .fh = fhandle,
3157                 .bitmask = server->attr_bitmask,
3158         };
3159         struct nfs4_statfs_res res = {
3160                 .fsstat = fsstat,
3161         };
3162         struct rpc_message msg = {
3163                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3164                 .rpc_argp = &args,
3165                 .rpc_resp = &res,
3166         };
3167
3168         nfs_fattr_init(fsstat->fattr);
3169         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3170 }
3171
3172 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3173 {
3174         struct nfs4_exception exception = { };
3175         int err;
3176         do {
3177                 err = nfs4_handle_exception(server,
3178                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3179                                 &exception);
3180         } while (exception.retry);
3181         return err;
3182 }
3183
3184 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3185                 struct nfs_fsinfo *fsinfo)
3186 {
3187         struct nfs4_fsinfo_arg args = {
3188                 .fh = fhandle,
3189                 .bitmask = server->attr_bitmask,
3190         };
3191         struct nfs4_fsinfo_res res = {
3192                 .fsinfo = fsinfo,
3193         };
3194         struct rpc_message msg = {
3195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3196                 .rpc_argp = &args,
3197                 .rpc_resp = &res,
3198         };
3199
3200         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3201 }
3202
3203 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3204 {
3205         struct nfs4_exception exception = { };
3206         int err;
3207
3208         do {
3209                 err = nfs4_handle_exception(server,
3210                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3211                                 &exception);
3212         } while (exception.retry);
3213         return err;
3214 }
3215
3216 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3217 {
3218         nfs_fattr_init(fsinfo->fattr);
3219         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3220 }
3221
3222 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3223                 struct nfs_pathconf *pathconf)
3224 {
3225         struct nfs4_pathconf_arg args = {
3226                 .fh = fhandle,
3227                 .bitmask = server->attr_bitmask,
3228         };
3229         struct nfs4_pathconf_res res = {
3230                 .pathconf = pathconf,
3231         };
3232         struct rpc_message msg = {
3233                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3234                 .rpc_argp = &args,
3235                 .rpc_resp = &res,
3236         };
3237
3238         /* None of the pathconf attributes are mandatory to implement */
3239         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3240                 memset(pathconf, 0, sizeof(*pathconf));
3241                 return 0;
3242         }
3243
3244         nfs_fattr_init(pathconf->fattr);
3245         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3246 }
3247
3248 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3249                 struct nfs_pathconf *pathconf)
3250 {
3251         struct nfs4_exception exception = { };
3252         int err;
3253
3254         do {
3255                 err = nfs4_handle_exception(server,
3256                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3257                                 &exception);
3258         } while (exception.retry);
3259         return err;
3260 }
3261
3262 void __nfs4_read_done_cb(struct nfs_read_data *data)
3263 {
3264         nfs_invalidate_atime(data->inode);
3265 }
3266
3267 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3268 {
3269         struct nfs_server *server = NFS_SERVER(data->inode);
3270
3271         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3272                 rpc_restart_call_prepare(task);
3273                 return -EAGAIN;
3274         }
3275
3276         __nfs4_read_done_cb(data);
3277         if (task->tk_status > 0)
3278                 renew_lease(server, data->timestamp);
3279         return 0;
3280 }
3281
3282 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3283 {
3284
3285         dprintk("--> %s\n", __func__);
3286
3287         if (!nfs4_sequence_done(task, &data->res.seq_res))
3288                 return -EAGAIN;
3289
3290         return data->read_done_cb ? data->read_done_cb(task, data) :
3291                                     nfs4_read_done_cb(task, data);
3292 }
3293
3294 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3295 {
3296         data->timestamp   = jiffies;
3297         data->read_done_cb = nfs4_read_done_cb;
3298         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3299         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3300 }
3301
3302 static void nfs4_proc_read_rpc_prepare(struct rpc_task *task, struct nfs_read_data *data)
3303 {
3304         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3305                                 &data->args.seq_args,
3306                                 &data->res.seq_res,
3307                                 task))
3308                 return;
3309         rpc_call_start(task);
3310 }
3311
3312 /* Reset the the nfs_read_data to send the read to the MDS. */
3313 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3314 {
3315         dprintk("%s Reset task for i/o through\n", __func__);
3316         put_lseg(data->lseg);
3317         data->lseg = NULL;
3318         /* offsets will differ in the dense stripe case */
3319         data->args.offset = data->mds_offset;
3320         data->ds_clp = NULL;
3321         data->args.fh     = NFS_FH(data->inode);
3322         data->read_done_cb = nfs4_read_done_cb;
3323         task->tk_ops = data->mds_ops;
3324         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3325 }
3326 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3327
3328 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3329 {
3330         struct inode *inode = data->inode;
3331
3332         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3333                 rpc_restart_call_prepare(task);
3334                 return -EAGAIN;
3335         }
3336         if (task->tk_status >= 0) {
3337                 renew_lease(NFS_SERVER(inode), data->timestamp);
3338                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3339         }
3340         return 0;
3341 }
3342
3343 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3344 {
3345         if (!nfs4_sequence_done(task, &data->res.seq_res))
3346                 return -EAGAIN;
3347         return data->write_done_cb ? data->write_done_cb(task, data) :
3348                 nfs4_write_done_cb(task, data);
3349 }
3350
3351 /* Reset the the nfs_write_data to send the write to the MDS. */
3352 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3353 {
3354         dprintk("%s Reset task for i/o through\n", __func__);
3355         put_lseg(data->lseg);
3356         data->lseg          = NULL;
3357         data->ds_clp        = NULL;
3358         data->write_done_cb = nfs4_write_done_cb;
3359         data->args.fh       = NFS_FH(data->inode);
3360         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3361         data->args.offset   = data->mds_offset;
3362         data->res.fattr     = &data->fattr;
3363         task->tk_ops        = data->mds_ops;
3364         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3365 }
3366 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3367
3368 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3369 {
3370         struct nfs_server *server = NFS_SERVER(data->inode);
3371
3372         if (data->lseg) {
3373                 data->args.bitmask = NULL;
3374                 data->res.fattr = NULL;
3375         } else
3376                 data->args.bitmask = server->cache_consistency_bitmask;
3377         if (!data->write_done_cb)
3378                 data->write_done_cb = nfs4_write_done_cb;
3379         data->res.server = server;
3380         data->timestamp   = jiffies;
3381
3382         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3383         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3384 }
3385
3386 static void nfs4_proc_write_rpc_prepare(struct rpc_task *task, struct nfs_write_data *data)
3387 {
3388         if (nfs4_setup_sequence(NFS_SERVER(data->inode),
3389                                 &data->args.seq_args,
3390                                 &data->res.seq_res,
3391                                 task))
3392                 return;
3393         rpc_call_start(task);
3394 }
3395
3396 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3397 {
3398         struct inode *inode = data->inode;
3399
3400         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3401                 rpc_restart_call_prepare(task);
3402                 return -EAGAIN;
3403         }
3404         nfs_refresh_inode(inode, data->res.fattr);
3405         return 0;
3406 }
3407
3408 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3409 {
3410         if (!nfs4_sequence_done(task, &data->res.seq_res))
3411                 return -EAGAIN;
3412         return data->write_done_cb(task, data);
3413 }
3414
3415 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3416 {
3417         struct nfs_server *server = NFS_SERVER(data->inode);
3418
3419         if (data->lseg) {
3420                 data->args.bitmask = NULL;
3421                 data->res.fattr = NULL;
3422         } else
3423                 data->args.bitmask = server->cache_consistency_bitmask;
3424         if (!data->write_done_cb)
3425                 data->write_done_cb = nfs4_commit_done_cb;
3426         data->res.server = server;
3427         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3428         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3429 }
3430
3431 struct nfs4_renewdata {
3432         struct nfs_client       *client;
3433         unsigned long           timestamp;
3434 };
3435
3436 /*
3437  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3438  * standalone procedure for queueing an asynchronous RENEW.
3439  */
3440 static void nfs4_renew_release(void *calldata)
3441 {
3442         struct nfs4_renewdata *data = calldata;
3443         struct nfs_client *clp = data->client;
3444
3445         if (atomic_read(&clp->cl_count) > 1)
3446                 nfs4_schedule_state_renewal(clp);
3447         nfs_put_client(clp);
3448         kfree(data);
3449 }
3450
3451 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3452 {
3453         struct nfs4_renewdata *data = calldata;
3454         struct nfs_client *clp = data->client;
3455         unsigned long timestamp = data->timestamp;
3456
3457         if (task->tk_status < 0) {
3458                 /* Unless we're shutting down, schedule state recovery! */
3459                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3460                         return;
3461                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3462                         nfs4_schedule_lease_recovery(clp);
3463                         return;
3464                 }
3465                 nfs4_schedule_path_down_recovery(clp);
3466         }
3467         do_renew_lease(clp, timestamp);
3468 }
3469
3470 static const struct rpc_call_ops nfs4_renew_ops = {
3471         .rpc_call_done = nfs4_renew_done,
3472         .rpc_release = nfs4_renew_release,
3473 };
3474
3475 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3476 {
3477         struct rpc_message msg = {
3478                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3479                 .rpc_argp       = clp,
3480                 .rpc_cred       = cred,
3481         };
3482         struct nfs4_renewdata *data;
3483
3484         if (renew_flags == 0)
3485                 return 0;
3486         if (!atomic_inc_not_zero(&clp->cl_count))
3487                 return -EIO;
3488         data = kmalloc(sizeof(*data), GFP_NOFS);
3489         if (data == NULL)
3490                 return -ENOMEM;
3491         data->client = clp;
3492         data->timestamp = jiffies;
3493         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3494                         &nfs4_renew_ops, data);
3495 }
3496
3497 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3498 {
3499         struct rpc_message msg = {
3500                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3501                 .rpc_argp       = clp,
3502                 .rpc_cred       = cred,
3503         };
3504         unsigned long now = jiffies;
3505         int status;
3506
3507         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3508         if (status < 0)
3509                 return status;
3510         do_renew_lease(clp, now);
3511         return 0;
3512 }
3513
3514 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3515 {
3516         return (server->caps & NFS_CAP_ACLS)
3517                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3518                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3519 }
3520
3521 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3522  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3523  * the stack.
3524  */
3525 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3526
3527 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3528                 struct page **pages, unsigned int *pgbase)
3529 {
3530         struct page *newpage, **spages;
3531         int rc = 0;
3532         size_t len;
3533         spages = pages;
3534
3535         do {
3536                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3537                 newpage = alloc_page(GFP_KERNEL);
3538
3539                 if (newpage == NULL)
3540                         goto unwind;
3541                 memcpy(page_address(newpage), buf, len);
3542                 buf += len;
3543                 buflen -= len;
3544                 *pages++ = newpage;
3545                 rc++;
3546         } while (buflen != 0);
3547
3548         return rc;
3549
3550 unwind:
3551         for(; rc > 0; rc--)
3552                 __free_page(spages[rc-1]);
3553         return -ENOMEM;
3554 }
3555
3556 struct nfs4_cached_acl {
3557         int cached;
3558         size_t len;
3559         char data[0];
3560 };
3561
3562 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3563 {
3564         struct nfs_inode *nfsi = NFS_I(inode);
3565
3566         spin_lock(&inode->i_lock);
3567         kfree(nfsi->nfs4_acl);
3568         nfsi->nfs4_acl = acl;
3569         spin_unlock(&inode->i_lock);
3570 }
3571
3572 static void nfs4_zap_acl_attr(struct inode *inode)
3573 {
3574         nfs4_set_cached_acl(inode, NULL);
3575 }
3576
3577 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3578 {
3579         struct nfs_inode *nfsi = NFS_I(inode);
3580         struct nfs4_cached_acl *acl;
3581         int ret = -ENOENT;
3582
3583         spin_lock(&inode->i_lock);
3584         acl = nfsi->nfs4_acl;
3585         if (acl == NULL)
3586                 goto out;
3587         if (buf == NULL) /* user is just asking for length */
3588                 goto out_len;
3589         if (acl->cached == 0)
3590                 goto out;
3591         ret = -ERANGE; /* see getxattr(2) man page */
3592         if (acl->len > buflen)
3593                 goto out;
3594         memcpy(buf, acl->data, acl->len);
3595 out_len:
3596         ret = acl->len;
3597 out:
3598         spin_unlock(&inode->i_lock);
3599         return ret;
3600 }
3601
3602 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3603 {
3604         struct nfs4_cached_acl *acl;
3605
3606         if (buf && acl_len <= PAGE_SIZE) {
3607                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3608                 if (acl == NULL)
3609                         goto out;
3610                 acl->cached = 1;
3611                 memcpy(acl->data, buf, acl_len);
3612         } else {
3613                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3614                 if (acl == NULL)
3615                         goto out;
3616                 acl->cached = 0;
3617         }
3618         acl->len = acl_len;
3619 out:
3620         nfs4_set_cached_acl(inode, acl);
3621 }
3622
3623 /*
3624  * The getxattr API returns the required buffer length when called with a
3625  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3626  * the required buf.  On a NULL buf, we send a page of data to the server
3627  * guessing that the ACL request can be serviced by a page. If so, we cache
3628  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3629  * the cache. If not so, we throw away the page, and cache the required
3630  * length. The next getxattr call will then produce another round trip to
3631  * the server, this time with the input buf of the required size.
3632  */
3633 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3634 {
3635         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3636         struct nfs_getaclargs args = {
3637                 .fh = NFS_FH(inode),
3638                 .acl_pages = pages,
3639                 .acl_len = buflen,
3640         };
3641         struct nfs_getaclres res = {
3642                 .acl_len = buflen,
3643         };
3644         void *resp_buf;
3645         struct rpc_message msg = {
3646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3647                 .rpc_argp = &args,
3648                 .rpc_resp = &res,
3649         };
3650         int ret = -ENOMEM, npages, i, acl_len = 0;
3651
3652         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3653         /* As long as we're doing a round trip to the server anyway,
3654          * let's be prepared for a page of acl data. */
3655         if (npages == 0)
3656                 npages = 1;
3657
3658         for (i = 0; i < npages; i++) {
3659                 pages[i] = alloc_page(GFP_KERNEL);
3660                 if (!pages[i])
3661                         goto out_free;
3662         }
3663         if (npages > 1) {
3664                 /* for decoding across pages */
3665                 res.acl_scratch = alloc_page(GFP_KERNEL);
3666                 if (!res.acl_scratch)
3667                         goto out_free;
3668         }
3669         args.acl_len = npages * PAGE_SIZE;
3670         args.acl_pgbase = 0;
3671         /* Let decode_getfacl know not to fail if the ACL data is larger than
3672          * the page we send as a guess */
3673         if (buf == NULL)
3674                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3675         resp_buf = page_address(pages[0]);
3676
3677         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3678                 __func__, buf, buflen, npages, args.acl_len);
3679         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3680                              &msg, &args.seq_args, &res.seq_res, 0);
3681         if (ret)
3682                 goto out_free;
3683
3684         acl_len = res.acl_len - res.acl_data_offset;
3685         if (acl_len > args.acl_len)
3686                 nfs4_write_cached_acl(inode, NULL, acl_len);
3687         else
3688                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3689                                       acl_len);
3690         if (buf) {
3691                 ret = -ERANGE;
3692                 if (acl_len > buflen)
3693                         goto out_free;
3694                 _copy_from_pages(buf, pages, res.acl_data_offset,
3695                                 res.acl_len);
3696         }
3697         ret = acl_len;
3698 out_free:
3699         for (i = 0; i < npages; i++)
3700                 if (pages[i])
3701                         __free_page(pages[i]);
3702         if (res.acl_scratch)
3703                 __free_page(res.acl_scratch);
3704         return ret;
3705 }
3706
3707 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3708 {
3709         struct nfs4_exception exception = { };
3710         ssize_t ret;
3711         do {
3712                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3713                 if (ret >= 0)
3714                         break;
3715                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3716         } while (exception.retry);
3717         return ret;
3718 }
3719
3720 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3721 {
3722         struct nfs_server *server = NFS_SERVER(inode);
3723         int ret;
3724
3725         if (!nfs4_server_supports_acls(server))
3726                 return -EOPNOTSUPP;
3727         ret = nfs_revalidate_inode(server, inode);
3728         if (ret < 0)
3729                 return ret;
3730         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3731                 nfs_zap_acl_cache(inode);
3732         ret = nfs4_read_cached_acl(inode, buf, buflen);
3733         if (ret != -ENOENT)
3734                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3735                  * but no cached acl data, just the acl length */
3736                 return ret;
3737         return nfs4_get_acl_uncached(inode, buf, buflen);
3738 }
3739
3740 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3741 {
3742         struct nfs_server *server = NFS_SERVER(inode);
3743         struct page *pages[NFS4ACL_MAXPAGES];
3744         struct nfs_setaclargs arg = {
3745                 .fh             = NFS_FH(inode),
3746                 .acl_pages      = pages,
3747                 .acl_len        = buflen,
3748         };
3749         struct nfs_setaclres res;
3750         struct rpc_message msg = {
3751                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3752                 .rpc_argp       = &arg,
3753                 .rpc_resp       = &res,
3754         };
3755         int ret, i;
3756
3757         if (!nfs4_server_supports_acls(server))
3758                 return -EOPNOTSUPP;
3759         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3760         if (i < 0)
3761                 return i;
3762         nfs_inode_return_delegation(inode);
3763         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3764
3765         /*
3766          * Free each page after tx, so the only ref left is
3767          * held by the network stack
3768          */
3769         for (; i > 0; i--)
3770                 put_page(pages[i-1]);
3771
3772         /*
3773          * Acl update can result in inode attribute update.
3774          * so mark the attribute cache invalid.
3775          */
3776         spin_lock(&inode->i_lock);
3777         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3778         spin_unlock(&inode->i_lock);
3779         nfs_access_zap_cache(inode);
3780         nfs_zap_acl_cache(inode);
3781         return ret;
3782 }
3783
3784 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3785 {
3786         struct nfs4_exception exception = { };
3787         int err;
3788         do {
3789                 err = nfs4_handle_exception(NFS_SERVER(inode),
3790                                 __nfs4_proc_set_acl(inode, buf, buflen),
3791                                 &exception);
3792         } while (exception.retry);
3793         return err;
3794 }
3795
3796 static int
3797 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3798 {
3799         struct nfs_client *clp = server->nfs_client;
3800
3801         if (task->tk_status >= 0)
3802                 return 0;
3803         switch(task->tk_status) {
3804                 case -NFS4ERR_DELEG_REVOKED:
3805                 case -NFS4ERR_ADMIN_REVOKED:
3806                 case -NFS4ERR_BAD_STATEID:
3807                         if (state != NULL)
3808                                 nfs_remove_bad_delegation(state->inode);
3809                 case -NFS4ERR_OPENMODE:
3810                         if (state == NULL)
3811                                 break;
3812                         nfs4_schedule_stateid_recovery(server, state);
3813                         goto wait_on_recovery;
3814                 case -NFS4ERR_EXPIRED:
3815                         if (state != NULL)
3816                                 nfs4_schedule_stateid_recovery(server, state);
3817                 case -NFS4ERR_STALE_STATEID:
3818                 case -NFS4ERR_STALE_CLIENTID:
3819                         nfs4_schedule_lease_recovery(clp);
3820                         goto wait_on_recovery;
3821 #if defined(CONFIG_NFS_V4_1)
3822                 case -NFS4ERR_BADSESSION:
3823                 case -NFS4ERR_BADSLOT:
3824                 case -NFS4ERR_BAD_HIGH_SLOT:
3825                 case -NFS4ERR_DEADSESSION:
3826                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3827                 case -NFS4ERR_SEQ_FALSE_RETRY:
3828                 case -NFS4ERR_SEQ_MISORDERED:
3829                         dprintk("%s ERROR %d, Reset session\n", __func__,
3830                                 task->tk_status);
3831                         nfs4_schedule_session_recovery(clp->cl_session);
3832                         task->tk_status = 0;
3833                         return -EAGAIN;
3834 #endif /* CONFIG_NFS_V4_1 */
3835                 case -NFS4ERR_DELAY:
3836                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3837                 case -NFS4ERR_GRACE:
3838                 case -EKEYEXPIRED:
3839                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3840                         task->tk_status = 0;
3841                         return -EAGAIN;
3842                 case -NFS4ERR_RETRY_UNCACHED_REP:
3843                 case -NFS4ERR_OLD_STATEID:
3844                         task->tk_status = 0;
3845                         return -EAGAIN;
3846         }
3847         task->tk_status = nfs4_map_errors(task->tk_status);
3848         return 0;
3849 wait_on_recovery:
3850         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3851         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3852                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3853         task->tk_status = 0;
3854         return -EAGAIN;
3855 }
3856
3857 static void nfs4_construct_boot_verifier(struct nfs_client *clp,
3858                                          nfs4_verifier *bootverf)
3859 {
3860         __be32 verf[2];
3861
3862         verf[0] = htonl((u32)clp->cl_boot_time.tv_sec);
3863         verf[1] = htonl((u32)clp->cl_boot_time.tv_nsec);
3864         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3865 }
3866
3867 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3868                 unsigned short port, struct rpc_cred *cred,
3869                 struct nfs4_setclientid_res *res)
3870 {
3871         nfs4_verifier sc_verifier;
3872         struct nfs4_setclientid setclientid = {
3873                 .sc_verifier = &sc_verifier,
3874                 .sc_prog = program,
3875                 .sc_cb_ident = clp->cl_cb_ident,
3876         };
3877         struct rpc_message msg = {
3878                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3879                 .rpc_argp = &setclientid,
3880                 .rpc_resp = res,
3881                 .rpc_cred = cred,
3882         };
3883         int loop = 0;
3884         int status;
3885
3886         nfs4_construct_boot_verifier(clp, &sc_verifier);
3887
3888         for(;;) {
3889                 rcu_read_lock();
3890                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3891                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3892                                 clp->cl_ipaddr,
3893                                 rpc_peeraddr2str(clp->cl_rpcclient,
3894                                                         RPC_DISPLAY_ADDR),
3895                                 rpc_peeraddr2str(clp->cl_rpcclient,
3896                                                         RPC_DISPLAY_PROTO),
3897                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3898                                 clp->cl_id_uniquifier);
3899                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3900                                 sizeof(setclientid.sc_netid),
3901                                 rpc_peeraddr2str(clp->cl_rpcclient,
3902                                                         RPC_DISPLAY_NETID));
3903                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3904                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3905                                 clp->cl_ipaddr, port >> 8, port & 255);
3906                 rcu_read_unlock();
3907
3908                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3909                 if (status != -NFS4ERR_CLID_INUSE)
3910                         break;
3911                 if (loop != 0) {
3912                         ++clp->cl_id_uniquifier;
3913                         break;
3914                 }
3915                 ++loop;
3916                 ssleep(clp->cl_lease_time / HZ + 1);
3917         }
3918         return status;
3919 }
3920
3921 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3922                 struct nfs4_setclientid_res *arg,
3923                 struct rpc_cred *cred)
3924 {
3925         struct nfs_fsinfo fsinfo;
3926         struct rpc_message msg = {
3927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3928                 .rpc_argp = arg,
3929                 .rpc_resp = &fsinfo,
3930                 .rpc_cred = cred,
3931         };
3932         unsigned long now;
3933         int status;
3934
3935         now = jiffies;
3936         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3937         if (status == 0) {
3938                 spin_lock(&clp->cl_lock);
3939                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3940                 clp->cl_last_renewal = now;
3941                 spin_unlock(&clp->cl_lock);
3942         }
3943         return status;
3944 }
3945
3946 struct nfs4_delegreturndata {
3947         struct nfs4_delegreturnargs args;
3948         struct nfs4_delegreturnres res;
3949         struct nfs_fh fh;
3950         nfs4_stateid stateid;
3951         unsigned long timestamp;
3952         struct nfs_fattr fattr;
3953         int rpc_status;
3954 };
3955
3956 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3957 {
3958         struct nfs4_delegreturndata *data = calldata;
3959
3960         if (!nfs4_sequence_done(task, &data->res.seq_res))
3961                 return;
3962
3963         switch (task->tk_status) {
3964         case -NFS4ERR_STALE_STATEID:
3965         case -NFS4ERR_EXPIRED:
3966         case 0:
3967                 renew_lease(data->res.server, data->timestamp);
3968                 break;
3969         default:
3970                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3971                                 -EAGAIN) {
3972                         rpc_restart_call_prepare(task);
3973                         return;
3974                 }
3975         }
3976         data->rpc_status = task->tk_status;
3977 }
3978
3979 static void nfs4_delegreturn_release(void *calldata)
3980 {
3981         kfree(calldata);
3982 }
3983
3984 #if defined(CONFIG_NFS_V4_1)
3985 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3986 {
3987         struct nfs4_delegreturndata *d_data;
3988
3989         d_data = (struct nfs4_delegreturndata *)data;
3990
3991         if (nfs4_setup_sequence(d_data->res.server,
3992                                 &d_data->args.seq_args,
3993                                 &d_data->res.seq_res, task))
3994                 return;
3995         rpc_call_start(task);
3996 }
3997 #endif /* CONFIG_NFS_V4_1 */
3998
3999 static const struct rpc_call_ops nfs4_delegreturn_ops = {
4000 #if defined(CONFIG_NFS_V4_1)
4001         .rpc_call_prepare = nfs4_delegreturn_prepare,
4002 #endif /* CONFIG_NFS_V4_1 */
4003         .rpc_call_done = nfs4_delegreturn_done,
4004         .rpc_release = nfs4_delegreturn_release,
4005 };
4006
4007 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4008 {
4009         struct nfs4_delegreturndata *data;
4010         struct nfs_server *server = NFS_SERVER(inode);
4011         struct rpc_task *task;
4012         struct rpc_message msg = {
4013                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
4014                 .rpc_cred = cred,
4015         };
4016         struct rpc_task_setup task_setup_data = {
4017                 .rpc_client = server->client,
4018                 .rpc_message = &msg,
4019                 .callback_ops = &nfs4_delegreturn_ops,
4020                 .flags = RPC_TASK_ASYNC,
4021         };
4022         int status = 0;
4023
4024         data = kzalloc(sizeof(*data), GFP_NOFS);
4025         if (data == NULL)
4026                 return -ENOMEM;
4027         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4028         data->args.fhandle = &data->fh;
4029         data->args.stateid = &data->stateid;
4030         data->args.bitmask = server->attr_bitmask;
4031         nfs_copy_fh(&data->fh, NFS_FH(inode));
4032         nfs4_stateid_copy(&data->stateid, stateid);
4033         data->res.fattr = &data->fattr;
4034         data->res.server = server;
4035         nfs_fattr_init(data->res.fattr);
4036         data->timestamp = jiffies;
4037         data->rpc_status = 0;
4038
4039         task_setup_data.callback_data = data;
4040         msg.rpc_argp = &data->args;
4041         msg.rpc_resp = &data->res;
4042         task = rpc_run_task(&task_setup_data);
4043         if (IS_ERR(task))
4044                 return PTR_ERR(task);
4045         if (!issync)
4046                 goto out;
4047         status = nfs4_wait_for_completion_rpc_task(task);
4048         if (status != 0)
4049                 goto out;
4050         status = data->rpc_status;
4051         if (status != 0)
4052                 goto out;
4053         nfs_refresh_inode(inode, &data->fattr);
4054 out:
4055         rpc_put_task(task);
4056         return status;
4057 }
4058
4059 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4060 {
4061         struct nfs_server *server = NFS_SERVER(inode);
4062         struct nfs4_exception exception = { };
4063         int err;
4064         do {
4065                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4066                 switch (err) {
4067                         case -NFS4ERR_STALE_STATEID:
4068                         case -NFS4ERR_EXPIRED:
4069                         case 0:
4070                                 return 0;
4071                 }
4072                 err = nfs4_handle_exception(server, err, &exception);
4073         } while (exception.retry);
4074         return err;
4075 }
4076
4077 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4078 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4079
4080 /*
4081  * sleep, with exponential backoff, and retry the LOCK operation.
4082  */
4083 static unsigned long
4084 nfs4_set_lock_task_retry(unsigned long timeout)
4085 {
4086         freezable_schedule_timeout_killable(timeout);
4087         timeout <<= 1;
4088         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4089                 return NFS4_LOCK_MAXTIMEOUT;
4090         return timeout;
4091 }
4092
4093 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4094 {
4095         struct inode *inode = state->inode;
4096         struct nfs_server *server = NFS_SERVER(inode);
4097         struct nfs_client *clp = server->nfs_client;
4098         struct nfs_lockt_args arg = {
4099                 .fh = NFS_FH(inode),
4100                 .fl = request,
4101         };
4102         struct nfs_lockt_res res = {
4103                 .denied = request,
4104         };
4105         struct rpc_message msg = {
4106                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4107                 .rpc_argp       = &arg,
4108                 .rpc_resp       = &res,
4109                 .rpc_cred       = state->owner->so_cred,
4110         };
4111         struct nfs4_lock_state *lsp;
4112         int status;
4113
4114         arg.lock_owner.clientid = clp->cl_clientid;
4115         status = nfs4_set_lock_state(state, request);
4116         if (status != 0)
4117                 goto out;
4118         lsp = request->fl_u.nfs4_fl.owner;
4119         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4120         arg.lock_owner.s_dev = server->s_dev;
4121         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4122         switch (status) {
4123                 case 0:
4124                         request->fl_type = F_UNLCK;
4125                         break;
4126                 case -NFS4ERR_DENIED:
4127                         status = 0;
4128         }
4129         request->fl_ops->fl_release_private(request);
4130 out:
4131         return status;
4132 }
4133
4134 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4135 {
4136         struct nfs4_exception exception = { };
4137         int err;
4138
4139         do {
4140                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4141                                 _nfs4_proc_getlk(state, cmd, request),
4142                                 &exception);
4143         } while (exception.retry);
4144         return err;
4145 }
4146
4147 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4148 {
4149         int res = 0;
4150         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4151                 case FL_POSIX:
4152                         res = posix_lock_file_wait(file, fl);
4153                         break;
4154                 case FL_FLOCK:
4155                         res = flock_lock_file_wait(file, fl);
4156                         break;
4157                 default:
4158                         BUG();
4159         }
4160         return res;
4161 }
4162
4163 struct nfs4_unlockdata {
4164         struct nfs_locku_args arg;
4165         struct nfs_locku_res res;
4166         struct nfs4_lock_state *lsp;
4167         struct nfs_open_context *ctx;
4168         struct file_lock fl;
4169         const struct nfs_server *server;
4170         unsigned long timestamp;
4171 };
4172
4173 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4174                 struct nfs_open_context *ctx,
4175                 struct nfs4_lock_state *lsp,
4176                 struct nfs_seqid *seqid)
4177 {
4178         struct nfs4_unlockdata *p;
4179         struct inode *inode = lsp->ls_state->inode;
4180
4181         p = kzalloc(sizeof(*p), GFP_NOFS);
4182         if (p == NULL)
4183                 return NULL;
4184         p->arg.fh = NFS_FH(inode);
4185         p->arg.fl = &p->fl;
4186         p->arg.seqid = seqid;
4187         p->res.seqid = seqid;
4188         p->arg.stateid = &lsp->ls_stateid;
4189         p->lsp = lsp;
4190         atomic_inc(&lsp->ls_count);
4191         /* Ensure we don't close file until we're done freeing locks! */
4192         p->ctx = get_nfs_open_context(ctx);
4193         memcpy(&p->fl, fl, sizeof(p->fl));
4194         p->server = NFS_SERVER(inode);
4195         return p;
4196 }
4197
4198 static void nfs4_locku_release_calldata(void *data)
4199 {
4200         struct nfs4_unlockdata *calldata = data;
4201         nfs_free_seqid(calldata->arg.seqid);
4202         nfs4_put_lock_state(calldata->lsp);
4203         put_nfs_open_context(calldata->ctx);
4204         kfree(calldata);
4205 }
4206
4207 static void nfs4_locku_done(struct rpc_task *task, void *data)
4208 {
4209         struct nfs4_unlockdata *calldata = data;
4210
4211         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4212                 return;
4213         switch (task->tk_status) {
4214                 case 0:
4215                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4216                                         &calldata->res.stateid);
4217                         renew_lease(calldata->server, calldata->timestamp);
4218                         break;
4219                 case -NFS4ERR_BAD_STATEID:
4220                 case -NFS4ERR_OLD_STATEID:
4221                 case -NFS4ERR_STALE_STATEID:
4222                 case -NFS4ERR_EXPIRED:
4223                         break;
4224                 default:
4225                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4226                                 rpc_restart_call_prepare(task);
4227         }
4228 }
4229
4230 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4231 {
4232         struct nfs4_unlockdata *calldata = data;
4233
4234         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4235                 return;
4236         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4237                 /* Note: exit _without_ running nfs4_locku_done */
4238                 task->tk_action = NULL;
4239                 return;
4240         }
4241         calldata->timestamp = jiffies;
4242         if (nfs4_setup_sequence(calldata->server,
4243                                 &calldata->arg.seq_args,
4244                                 &calldata->res.seq_res, task))
4245                 return;
4246         rpc_call_start(task);
4247 }
4248
4249 static const struct rpc_call_ops nfs4_locku_ops = {
4250         .rpc_call_prepare = nfs4_locku_prepare,
4251         .rpc_call_done = nfs4_locku_done,
4252         .rpc_release = nfs4_locku_release_calldata,
4253 };
4254
4255 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4256                 struct nfs_open_context *ctx,
4257                 struct nfs4_lock_state *lsp,
4258                 struct nfs_seqid *seqid)
4259 {
4260         struct nfs4_unlockdata *data;
4261         struct rpc_message msg = {
4262                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4263                 .rpc_cred = ctx->cred,
4264         };
4265         struct rpc_task_setup task_setup_data = {
4266                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4267                 .rpc_message = &msg,
4268                 .callback_ops = &nfs4_locku_ops,
4269                 .workqueue = nfsiod_workqueue,
4270                 .flags = RPC_TASK_ASYNC,
4271         };
4272
4273         /* Ensure this is an unlock - when canceling a lock, the
4274          * canceled lock is passed in, and it won't be an unlock.
4275          */
4276         fl->fl_type = F_UNLCK;
4277
4278         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4279         if (data == NULL) {
4280                 nfs_free_seqid(seqid);
4281                 return ERR_PTR(-ENOMEM);
4282         }
4283
4284         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4285         msg.rpc_argp = &data->arg;
4286         msg.rpc_resp = &data->res;
4287         task_setup_data.callback_data = data;
4288         return rpc_run_task(&task_setup_data);
4289 }
4290
4291 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4292 {
4293         struct nfs_inode *nfsi = NFS_I(state->inode);
4294         struct nfs_seqid *seqid;
4295         struct nfs4_lock_state *lsp;
4296         struct rpc_task *task;
4297         int status = 0;
4298         unsigned char fl_flags = request->fl_flags;
4299
4300         status = nfs4_set_lock_state(state, request);
4301         /* Unlock _before_ we do the RPC call */
4302         request->fl_flags |= FL_EXISTS;
4303         down_read(&nfsi->rwsem);
4304         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4305                 up_read(&nfsi->rwsem);
4306                 goto out;
4307         }
4308         up_read(&nfsi->rwsem);
4309         if (status != 0)
4310                 goto out;
4311         /* Is this a delegated lock? */
4312         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4313                 goto out;
4314         lsp = request->fl_u.nfs4_fl.owner;
4315         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4316         status = -ENOMEM;
4317         if (seqid == NULL)
4318                 goto out;
4319         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4320         status = PTR_ERR(task);
4321         if (IS_ERR(task))
4322                 goto out;
4323         status = nfs4_wait_for_completion_rpc_task(task);
4324         rpc_put_task(task);
4325 out:
4326         request->fl_flags = fl_flags;
4327         return status;
4328 }
4329
4330 struct nfs4_lockdata {
4331         struct nfs_lock_args arg;
4332         struct nfs_lock_res res;
4333         struct nfs4_lock_state *lsp;
4334         struct nfs_open_context *ctx;
4335         struct file_lock fl;
4336         unsigned long timestamp;
4337         int rpc_status;
4338         int cancelled;
4339         struct nfs_server *server;
4340 };
4341
4342 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4343                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4344                 gfp_t gfp_mask)
4345 {
4346         struct nfs4_lockdata *p;
4347         struct inode *inode = lsp->ls_state->inode;
4348         struct nfs_server *server = NFS_SERVER(inode);
4349
4350         p = kzalloc(sizeof(*p), gfp_mask);
4351         if (p == NULL)
4352                 return NULL;
4353
4354         p->arg.fh = NFS_FH(inode);
4355         p->arg.fl = &p->fl;
4356         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4357         if (p->arg.open_seqid == NULL)
4358                 goto out_free;
4359         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4360         if (p->arg.lock_seqid == NULL)
4361                 goto out_free_seqid;
4362         p->arg.lock_stateid = &lsp->ls_stateid;
4363         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4364         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4365         p->arg.lock_owner.s_dev = server->s_dev;
4366         p->res.lock_seqid = p->arg.lock_seqid;
4367         p->lsp = lsp;
4368         p->server = server;
4369         atomic_inc(&lsp->ls_count);
4370         p->ctx = get_nfs_open_context(ctx);
4371         memcpy(&p->fl, fl, sizeof(p->fl));
4372         return p;
4373 out_free_seqid:
4374         nfs_free_seqid(p->arg.open_seqid);
4375 out_free:
4376         kfree(p);
4377         return NULL;
4378 }
4379
4380 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4381 {
4382         struct nfs4_lockdata *data = calldata;
4383         struct nfs4_state *state = data->lsp->ls_state;
4384
4385         dprintk("%s: begin!\n", __func__);
4386         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4387                 return;
4388         /* Do we need to do an open_to_lock_owner? */
4389         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4390                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4391                         return;
4392                 data->arg.open_stateid = &state->stateid;
4393                 data->arg.new_lock_owner = 1;
4394                 data->res.open_seqid = data->arg.open_seqid;
4395         } else
4396                 data->arg.new_lock_owner = 0;
4397         data->timestamp = jiffies;
4398         if (nfs4_setup_sequence(data->server,
4399                                 &data->arg.seq_args,
4400                                 &data->res.seq_res, task))
4401                 return;
4402         rpc_call_start(task);
4403         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4404 }
4405
4406 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4407 {
4408         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4409         nfs4_lock_prepare(task, calldata);
4410 }
4411
4412 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4413 {
4414         struct nfs4_lockdata *data = calldata;
4415
4416         dprintk("%s: begin!\n", __func__);
4417
4418         if (!nfs4_sequence_done(task, &data->res.seq_res))
4419                 return;
4420
4421         data->rpc_status = task->tk_status;
4422         if (data->arg.new_lock_owner != 0) {
4423                 if (data->rpc_status == 0)
4424                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4425                 else
4426                         goto out;
4427         }
4428         if (data->rpc_status == 0) {
4429                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4430                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4431                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4432         }
4433 out:
4434         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4435 }
4436
4437 static void nfs4_lock_release(void *calldata)
4438 {
4439         struct nfs4_lockdata *data = calldata;
4440
4441         dprintk("%s: begin!\n", __func__);
4442         nfs_free_seqid(data->arg.open_seqid);
4443         if (data->cancelled != 0) {
4444                 struct rpc_task *task;
4445                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4446                                 data->arg.lock_seqid);
4447                 if (!IS_ERR(task))
4448                         rpc_put_task_async(task);
4449                 dprintk("%s: cancelling lock!\n", __func__);
4450         } else
4451                 nfs_free_seqid(data->arg.lock_seqid);
4452         nfs4_put_lock_state(data->lsp);
4453         put_nfs_open_context(data->ctx);
4454         kfree(data);
4455         dprintk("%s: done!\n", __func__);
4456 }
4457
4458 static const struct rpc_call_ops nfs4_lock_ops = {
4459         .rpc_call_prepare = nfs4_lock_prepare,
4460         .rpc_call_done = nfs4_lock_done,
4461         .rpc_release = nfs4_lock_release,
4462 };
4463
4464 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4465         .rpc_call_prepare = nfs4_recover_lock_prepare,
4466         .rpc_call_done = nfs4_lock_done,
4467         .rpc_release = nfs4_lock_release,
4468 };
4469
4470 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4471 {
4472         switch (error) {
4473         case -NFS4ERR_ADMIN_REVOKED:
4474         case -NFS4ERR_BAD_STATEID:
4475                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4476                 if (new_lock_owner != 0 ||
4477                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4478                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4479                 break;
4480         case -NFS4ERR_STALE_STATEID:
4481                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4482         case -NFS4ERR_EXPIRED:
4483                 nfs4_schedule_lease_recovery(server->nfs_client);
4484         };
4485 }
4486
4487 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4488 {
4489         struct nfs4_lockdata *data;
4490         struct rpc_task *task;
4491         struct rpc_message msg = {
4492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4493                 .rpc_cred = state->owner->so_cred,
4494         };
4495         struct rpc_task_setup task_setup_data = {
4496                 .rpc_client = NFS_CLIENT(state->inode),
4497                 .rpc_message = &msg,
4498                 .callback_ops = &nfs4_lock_ops,
4499                 .workqueue = nfsiod_workqueue,
4500                 .flags = RPC_TASK_ASYNC,
4501         };
4502         int ret;
4503
4504         dprintk("%s: begin!\n", __func__);
4505         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4506                         fl->fl_u.nfs4_fl.owner,
4507                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4508         if (data == NULL)
4509                 return -ENOMEM;
4510         if (IS_SETLKW(cmd))
4511                 data->arg.block = 1;
4512         if (recovery_type > NFS_LOCK_NEW) {
4513                 if (recovery_type == NFS_LOCK_RECLAIM)
4514                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4515                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4516         }
4517         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4518         msg.rpc_argp = &data->arg;
4519         msg.rpc_resp = &data->res;
4520         task_setup_data.callback_data = data;
4521         task = rpc_run_task(&task_setup_data);
4522         if (IS_ERR(task))
4523                 return PTR_ERR(task);
4524         ret = nfs4_wait_for_completion_rpc_task(task);
4525         if (ret == 0) {
4526                 ret = data->rpc_status;
4527                 if (ret)
4528                         nfs4_handle_setlk_error(data->server, data->lsp,
4529                                         data->arg.new_lock_owner, ret);
4530         } else
4531                 data->cancelled = 1;
4532         rpc_put_task(task);
4533         dprintk("%s: done, ret = %d!\n", __func__, ret);
4534         return ret;
4535 }
4536
4537 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4538 {
4539         struct nfs_server *server = NFS_SERVER(state->inode);
4540         struct nfs4_exception exception = { };
4541         int err;
4542
4543         do {
4544                 /* Cache the lock if possible... */
4545                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4546                         return 0;
4547                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4548                 if (err != -NFS4ERR_DELAY)
4549                         break;
4550                 nfs4_handle_exception(server, err, &exception);
4551         } while (exception.retry);
4552         return err;
4553 }
4554
4555 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4556 {
4557         struct nfs_server *server = NFS_SERVER(state->inode);
4558         struct nfs4_exception exception = { };
4559         int err;
4560
4561         err = nfs4_set_lock_state(state, request);
4562         if (err != 0)
4563                 return err;
4564         do {
4565                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4566                         return 0;
4567                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4568                 switch (err) {
4569                 default:
4570                         goto out;
4571                 case -NFS4ERR_GRACE:
4572                 case -NFS4ERR_DELAY:
4573                         nfs4_handle_exception(server, err, &exception);
4574                         err = 0;
4575                 }
4576         } while (exception.retry);
4577 out:
4578         return err;
4579 }
4580
4581 #if defined(CONFIG_NFS_V4_1)
4582 static int nfs41_check_expired_locks(struct nfs4_state *state)
4583 {
4584         int status, ret = NFS_OK;
4585         struct nfs4_lock_state *lsp;
4586         struct nfs_server *server = NFS_SERVER(state->inode);
4587
4588         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4589                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4590                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4591                         if (status != NFS_OK) {
4592                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4593                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4594                                 ret = status;
4595                         }
4596                 }
4597         };
4598
4599         return ret;
4600 }
4601
4602 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4603 {
4604         int status = NFS_OK;
4605
4606         if (test_bit(LK_STATE_IN_USE, &state->flags))
4607                 status = nfs41_check_expired_locks(state);
4608         if (status == NFS_OK)
4609                 return status;
4610         return nfs4_lock_expired(state, request);
4611 }
4612 #endif
4613
4614 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4615 {
4616         struct nfs_inode *nfsi = NFS_I(state->inode);
4617         unsigned char fl_flags = request->fl_flags;
4618         int status = -ENOLCK;
4619
4620         if ((fl_flags & FL_POSIX) &&
4621                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4622                 goto out;
4623         /* Is this a delegated open? */
4624         status = nfs4_set_lock_state(state, request);
4625         if (status != 0)
4626                 goto out;
4627         request->fl_flags |= FL_ACCESS;
4628         status = do_vfs_lock(request->fl_file, request);
4629         if (status < 0)
4630                 goto out;
4631         down_read(&nfsi->rwsem);
4632         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4633                 /* Yes: cache locks! */
4634                 /* ...but avoid races with delegation recall... */
4635                 request->fl_flags = fl_flags & ~FL_SLEEP;
4636                 status = do_vfs_lock(request->fl_file, request);
4637                 goto out_unlock;
4638         }
4639         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4640         if (status != 0)
4641                 goto out_unlock;
4642         /* Note: we always want to sleep here! */
4643         request->fl_flags = fl_flags | FL_SLEEP;
4644         if (do_vfs_lock(request->fl_file, request) < 0)
4645                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4646                         "manager!\n", __func__);
4647 out_unlock:
4648         up_read(&nfsi->rwsem);
4649 out:
4650         request->fl_flags = fl_flags;
4651         return status;
4652 }
4653
4654 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4655 {
4656         struct nfs4_exception exception = {
4657                 .state = state,
4658         };
4659         int err;
4660
4661         do {
4662                 err = _nfs4_proc_setlk(state, cmd, request);
4663                 if (err == -NFS4ERR_DENIED)
4664                         err = -EAGAIN;
4665                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4666                                 err, &exception);
4667         } while (exception.retry);
4668         return err;
4669 }
4670
4671 static int
4672 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4673 {
4674         struct nfs_open_context *ctx;
4675         struct nfs4_state *state;
4676         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4677         int status;
4678
4679         /* verify open state */
4680         ctx = nfs_file_open_context(filp);
4681         state = ctx->state;
4682
4683         if (request->fl_start < 0 || request->fl_end < 0)
4684                 return -EINVAL;
4685
4686         if (IS_GETLK(cmd)) {
4687                 if (state != NULL)
4688                         return nfs4_proc_getlk(state, F_GETLK, request);
4689                 return 0;
4690         }
4691
4692         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4693                 return -EINVAL;
4694
4695         if (request->fl_type == F_UNLCK) {
4696                 if (state != NULL)
4697                         return nfs4_proc_unlck(state, cmd, request);
4698                 return 0;
4699         }
4700
4701         if (state == NULL)
4702                 return -ENOLCK;
4703         do {
4704                 status = nfs4_proc_setlk(state, cmd, request);
4705                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4706                         break;
4707                 timeout = nfs4_set_lock_task_retry(timeout);
4708                 status = -ERESTARTSYS;
4709                 if (signalled())
4710                         break;
4711         } while(status < 0);
4712         return status;
4713 }
4714
4715 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4716 {
4717         struct nfs_server *server = NFS_SERVER(state->inode);
4718         struct nfs4_exception exception = { };
4719         int err;
4720
4721         err = nfs4_set_lock_state(state, fl);
4722         if (err != 0)
4723                 goto out;
4724         do {
4725                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4726                 switch (err) {
4727                         default:
4728                                 printk(KERN_ERR "NFS: %s: unhandled error "
4729                                         "%d.\n", __func__, err);
4730                         case 0:
4731                         case -ESTALE:
4732                                 goto out;
4733                         case -NFS4ERR_EXPIRED:
4734                                 nfs4_schedule_stateid_recovery(server, state);
4735                         case -NFS4ERR_STALE_CLIENTID:
4736                         case -NFS4ERR_STALE_STATEID:
4737                                 nfs4_schedule_lease_recovery(server->nfs_client);
4738                                 goto out;
4739                         case -NFS4ERR_BADSESSION:
4740                         case -NFS4ERR_BADSLOT:
4741                         case -NFS4ERR_BAD_HIGH_SLOT:
4742                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4743                         case -NFS4ERR_DEADSESSION:
4744                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4745                                 goto out;
4746                         case -ERESTARTSYS:
4747                                 /*
4748                                  * The show must go on: exit, but mark the
4749                                  * stateid as needing recovery.
4750                                  */
4751                         case -NFS4ERR_DELEG_REVOKED:
4752                         case -NFS4ERR_ADMIN_REVOKED:
4753                         case -NFS4ERR_BAD_STATEID:
4754                         case -NFS4ERR_OPENMODE:
4755                                 nfs4_schedule_stateid_recovery(server, state);
4756                                 err = 0;
4757                                 goto out;
4758                         case -EKEYEXPIRED:
4759                                 /*
4760                                  * User RPCSEC_GSS context has expired.
4761                                  * We cannot recover this stateid now, so
4762                                  * skip it and allow recovery thread to
4763                                  * proceed.
4764                                  */
4765                                 err = 0;
4766                                 goto out;
4767                         case -ENOMEM:
4768                         case -NFS4ERR_DENIED:
4769                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4770                                 err = 0;
4771                                 goto out;
4772                         case -NFS4ERR_DELAY:
4773                                 break;
4774                 }
4775                 err = nfs4_handle_exception(server, err, &exception);
4776         } while (exception.retry);
4777 out:
4778         return err;
4779 }
4780
4781 struct nfs_release_lockowner_data {
4782         struct nfs4_lock_state *lsp;
4783         struct nfs_server *server;
4784         struct nfs_release_lockowner_args args;
4785 };
4786
4787 static void nfs4_release_lockowner_release(void *calldata)
4788 {
4789         struct nfs_release_lockowner_data *data = calldata;
4790         nfs4_free_lock_state(data->server, data->lsp);
4791         kfree(calldata);
4792 }
4793
4794 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
4795         .rpc_release = nfs4_release_lockowner_release,
4796 };
4797
4798 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4799 {
4800         struct nfs_server *server = lsp->ls_state->owner->so_server;
4801         struct nfs_release_lockowner_data *data;
4802         struct rpc_message msg = {
4803                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4804         };
4805
4806         if (server->nfs_client->cl_mvops->minor_version != 0)
4807                 return -EINVAL;
4808         data = kmalloc(sizeof(*data), GFP_NOFS);
4809         if (!data)
4810                 return -ENOMEM;
4811         data->lsp = lsp;
4812         data->server = server;
4813         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4814         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4815         data->args.lock_owner.s_dev = server->s_dev;
4816         msg.rpc_argp = &data->args;
4817         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4818         return 0;
4819 }
4820
4821 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4822
4823 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4824                                    const void *buf, size_t buflen,
4825                                    int flags, int type)
4826 {
4827         if (strcmp(key, "") != 0)
4828                 return -EINVAL;
4829
4830         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4831 }
4832
4833 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4834                                    void *buf, size_t buflen, int type)
4835 {
4836         if (strcmp(key, "") != 0)
4837                 return -EINVAL;
4838
4839         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4840 }
4841
4842 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4843                                        size_t list_len, const char *name,
4844                                        size_t name_len, int type)
4845 {
4846         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4847
4848         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4849                 return 0;
4850
4851         if (list && len <= list_len)
4852                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4853         return len;
4854 }
4855
4856 /*
4857  * nfs_fhget will use either the mounted_on_fileid or the fileid
4858  */
4859 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4860 {
4861         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4862                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4863               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4864               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4865                 return;
4866
4867         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4868                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4869         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4870         fattr->nlink = 2;
4871 }
4872
4873 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4874                 struct nfs4_fs_locations *fs_locations, struct page *page)
4875 {
4876         struct nfs_server *server = NFS_SERVER(dir);
4877         u32 bitmask[2] = {
4878                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4879         };
4880         struct nfs4_fs_locations_arg args = {
4881                 .dir_fh = NFS_FH(dir),
4882                 .name = name,
4883                 .page = page,
4884                 .bitmask = bitmask,
4885         };
4886         struct nfs4_fs_locations_res res = {
4887                 .fs_locations = fs_locations,
4888         };
4889         struct rpc_message msg = {
4890                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4891                 .rpc_argp = &args,
4892                 .rpc_resp = &res,
4893         };
4894         int status;
4895
4896         dprintk("%s: start\n", __func__);
4897
4898         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4899          * is not supported */
4900         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4901                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4902         else
4903                 bitmask[0] |= FATTR4_WORD0_FILEID;
4904
4905         nfs_fattr_init(&fs_locations->fattr);
4906         fs_locations->server = server;
4907         fs_locations->nlocations = 0;
4908         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4909         dprintk("%s: returned status = %d\n", __func__, status);
4910         return status;
4911 }
4912
4913 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4914 {
4915         int status;
4916         struct nfs4_secinfo_arg args = {
4917                 .dir_fh = NFS_FH(dir),
4918                 .name   = name,
4919         };
4920         struct nfs4_secinfo_res res = {
4921                 .flavors     = flavors,
4922         };
4923         struct rpc_message msg = {
4924                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4925                 .rpc_argp = &args,
4926                 .rpc_resp = &res,
4927         };
4928
4929         dprintk("NFS call  secinfo %s\n", name->name);
4930         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4931         dprintk("NFS reply  secinfo: %d\n", status);
4932         return status;
4933 }
4934
4935 static int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
4936                 struct nfs4_secinfo_flavors *flavors)
4937 {
4938         struct nfs4_exception exception = { };
4939         int err;
4940         do {
4941                 err = nfs4_handle_exception(NFS_SERVER(dir),
4942                                 _nfs4_proc_secinfo(dir, name, flavors),
4943                                 &exception);
4944         } while (exception.retry);
4945         return err;
4946 }
4947
4948 #ifdef CONFIG_NFS_V4_1
4949 /*
4950  * Check the exchange flags returned by the server for invalid flags, having
4951  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4952  * DS flags set.
4953  */
4954 static int nfs4_check_cl_exchange_flags(u32 flags)
4955 {
4956         if (flags & ~EXCHGID4_FLAG_MASK_R)
4957                 goto out_inval;
4958         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4959             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4960                 goto out_inval;
4961         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4962                 goto out_inval;
4963         return NFS_OK;
4964 out_inval:
4965         return -NFS4ERR_INVAL;
4966 }
4967
4968 static bool
4969 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4970 {
4971         if (a->server_scope_sz == b->server_scope_sz &&
4972             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4973                 return true;
4974
4975         return false;
4976 }
4977
4978 /*
4979  * nfs4_proc_exchange_id()
4980  *
4981  * Since the clientid has expired, all compounds using sessions
4982  * associated with the stale clientid will be returning
4983  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4984  * be in some phase of session reset.
4985  */
4986 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4987 {
4988         nfs4_verifier verifier;
4989         struct nfs41_exchange_id_args args = {
4990                 .verifier = &verifier,
4991                 .client = clp,
4992                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4993         };
4994         struct nfs41_exchange_id_res res = {
4995                 .client = clp,
4996         };
4997         int status;
4998         struct rpc_message msg = {
4999                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
5000                 .rpc_argp = &args,
5001                 .rpc_resp = &res,
5002                 .rpc_cred = cred,
5003         };
5004
5005         dprintk("--> %s\n", __func__);
5006         BUG_ON(clp == NULL);
5007
5008         nfs4_construct_boot_verifier(clp, &verifier);
5009
5010         args.id_len = scnprintf(args.id, sizeof(args.id),
5011                                 "%s/%s.%s/%u",
5012                                 clp->cl_ipaddr,
5013                                 init_utsname()->nodename,
5014                                 init_utsname()->domainname,
5015                                 clp->cl_rpcclient->cl_auth->au_flavor);
5016
5017         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
5018         if (unlikely(!res.server_scope)) {
5019                 status = -ENOMEM;
5020                 goto out;
5021         }
5022
5023         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
5024         if (unlikely(!res.impl_id)) {
5025                 status = -ENOMEM;
5026                 goto out_server_scope;
5027         }
5028
5029         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5030         if (!status)
5031                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
5032
5033         if (!status) {
5034                 /* use the most recent implementation id */
5035                 kfree(clp->impl_id);
5036                 clp->impl_id = res.impl_id;
5037         } else
5038                 kfree(res.impl_id);
5039
5040         if (!status) {
5041                 if (clp->server_scope &&
5042                     !nfs41_same_server_scope(clp->server_scope,
5043                                              res.server_scope)) {
5044                         dprintk("%s: server_scope mismatch detected\n",
5045                                 __func__);
5046                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5047                         kfree(clp->server_scope);
5048                         clp->server_scope = NULL;
5049                 }
5050
5051                 if (!clp->server_scope) {
5052                         clp->server_scope = res.server_scope;
5053                         goto out;
5054                 }
5055         }
5056
5057 out_server_scope:
5058         kfree(res.server_scope);
5059 out:
5060         if (clp->impl_id)
5061                 dprintk("%s: Server Implementation ID: "
5062                         "domain: %s, name: %s, date: %llu,%u\n",
5063                         __func__, clp->impl_id->domain, clp->impl_id->name,
5064                         clp->impl_id->date.seconds,
5065                         clp->impl_id->date.nseconds);
5066         dprintk("<-- %s status= %d\n", __func__, status);
5067         return status;
5068 }
5069
5070 struct nfs4_get_lease_time_data {
5071         struct nfs4_get_lease_time_args *args;
5072         struct nfs4_get_lease_time_res *res;
5073         struct nfs_client *clp;
5074 };
5075
5076 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5077                                         void *calldata)
5078 {
5079         int ret;
5080         struct nfs4_get_lease_time_data *data =
5081                         (struct nfs4_get_lease_time_data *)calldata;
5082
5083         dprintk("--> %s\n", __func__);
5084         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5085         /* just setup sequence, do not trigger session recovery
5086            since we're invoked within one */
5087         ret = nfs41_setup_sequence(data->clp->cl_session,
5088                                    &data->args->la_seq_args,
5089                                    &data->res->lr_seq_res, task);
5090
5091         BUG_ON(ret == -EAGAIN);
5092         rpc_call_start(task);
5093         dprintk("<-- %s\n", __func__);
5094 }
5095
5096 /*
5097  * Called from nfs4_state_manager thread for session setup, so don't recover
5098  * from sequence operation or clientid errors.
5099  */
5100 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5101 {
5102         struct nfs4_get_lease_time_data *data =
5103                         (struct nfs4_get_lease_time_data *)calldata;
5104
5105         dprintk("--> %s\n", __func__);
5106         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5107                 return;
5108         switch (task->tk_status) {
5109         case -NFS4ERR_DELAY:
5110         case -NFS4ERR_GRACE:
5111                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5112                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5113                 task->tk_status = 0;
5114                 /* fall through */
5115         case -NFS4ERR_RETRY_UNCACHED_REP:
5116                 rpc_restart_call_prepare(task);
5117                 return;
5118         }
5119         dprintk("<-- %s\n", __func__);
5120 }
5121
5122 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
5123         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5124         .rpc_call_done = nfs4_get_lease_time_done,
5125 };
5126
5127 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5128 {
5129         struct rpc_task *task;
5130         struct nfs4_get_lease_time_args args;
5131         struct nfs4_get_lease_time_res res = {
5132                 .lr_fsinfo = fsinfo,
5133         };
5134         struct nfs4_get_lease_time_data data = {
5135                 .args = &args,
5136                 .res = &res,
5137                 .clp = clp,
5138         };
5139         struct rpc_message msg = {
5140                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5141                 .rpc_argp = &args,
5142                 .rpc_resp = &res,
5143         };
5144         struct rpc_task_setup task_setup = {
5145                 .rpc_client = clp->cl_rpcclient,
5146                 .rpc_message = &msg,
5147                 .callback_ops = &nfs4_get_lease_time_ops,
5148                 .callback_data = &data,
5149                 .flags = RPC_TASK_TIMEOUT,
5150         };
5151         int status;
5152
5153         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5154         dprintk("--> %s\n", __func__);
5155         task = rpc_run_task(&task_setup);
5156
5157         if (IS_ERR(task))
5158                 status = PTR_ERR(task);
5159         else {
5160                 status = task->tk_status;
5161                 rpc_put_task(task);
5162         }
5163         dprintk("<-- %s return %d\n", __func__, status);
5164
5165         return status;
5166 }
5167
5168 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5169 {
5170         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5171 }
5172
5173 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5174                 struct nfs4_slot *new,
5175                 u32 max_slots,
5176                 u32 ivalue)
5177 {
5178         struct nfs4_slot *old = NULL;
5179         u32 i;
5180
5181         spin_lock(&tbl->slot_tbl_lock);
5182         if (new) {
5183                 old = tbl->slots;
5184                 tbl->slots = new;
5185                 tbl->max_slots = max_slots;
5186         }
5187         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5188         for (i = 0; i < tbl->max_slots; i++)
5189                 tbl->slots[i].seq_nr = ivalue;
5190         spin_unlock(&tbl->slot_tbl_lock);
5191         kfree(old);
5192 }
5193
5194 /*
5195  * (re)Initialise a slot table
5196  */
5197 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5198                                  u32 ivalue)
5199 {
5200         struct nfs4_slot *new = NULL;
5201         int ret = -ENOMEM;
5202
5203         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5204                 max_reqs, tbl->max_slots);
5205
5206         /* Does the newly negotiated max_reqs match the existing slot table? */
5207         if (max_reqs != tbl->max_slots) {
5208                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5209                 if (!new)
5210                         goto out;
5211         }
5212         ret = 0;
5213
5214         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5215         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5216                 tbl, tbl->slots, tbl->max_slots);
5217 out:
5218         dprintk("<-- %s: return %d\n", __func__, ret);
5219         return ret;
5220 }
5221
5222 /* Destroy the slot table */
5223 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5224 {
5225         if (session->fc_slot_table.slots != NULL) {
5226                 kfree(session->fc_slot_table.slots);
5227                 session->fc_slot_table.slots = NULL;
5228         }
5229         if (session->bc_slot_table.slots != NULL) {
5230                 kfree(session->bc_slot_table.slots);
5231                 session->bc_slot_table.slots = NULL;
5232         }
5233         return;
5234 }
5235
5236 /*
5237  * Initialize or reset the forechannel and backchannel tables
5238  */
5239 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5240 {
5241         struct nfs4_slot_table *tbl;
5242         int status;
5243
5244         dprintk("--> %s\n", __func__);
5245         /* Fore channel */
5246         tbl = &ses->fc_slot_table;
5247         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5248         if (status) /* -ENOMEM */
5249                 return status;
5250         /* Back channel */
5251         tbl = &ses->bc_slot_table;
5252         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5253         if (status && tbl->slots == NULL)
5254                 /* Fore and back channel share a connection so get
5255                  * both slot tables or neither */
5256                 nfs4_destroy_slot_tables(ses);
5257         return status;
5258 }
5259
5260 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5261 {
5262         struct nfs4_session *session;
5263         struct nfs4_slot_table *tbl;
5264
5265         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5266         if (!session)
5267                 return NULL;
5268
5269         tbl = &session->fc_slot_table;
5270         tbl->highest_used_slotid = NFS4_NO_SLOT;
5271         spin_lock_init(&tbl->slot_tbl_lock);
5272         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5273         init_completion(&tbl->complete);
5274
5275         tbl = &session->bc_slot_table;
5276         tbl->highest_used_slotid = NFS4_NO_SLOT;
5277         spin_lock_init(&tbl->slot_tbl_lock);
5278         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5279         init_completion(&tbl->complete);
5280
5281         session->session_state = 1<<NFS4_SESSION_INITING;
5282
5283         session->clp = clp;
5284         return session;
5285 }
5286
5287 void nfs4_destroy_session(struct nfs4_session *session)
5288 {
5289         struct rpc_xprt *xprt;
5290
5291         nfs4_proc_destroy_session(session);
5292
5293         rcu_read_lock();
5294         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5295         rcu_read_unlock();
5296         dprintk("%s Destroy backchannel for xprt %p\n",
5297                 __func__, xprt);
5298         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5299         nfs4_destroy_slot_tables(session);
5300         kfree(session);
5301 }
5302
5303 /*
5304  * Initialize the values to be used by the client in CREATE_SESSION
5305  * If nfs4_init_session set the fore channel request and response sizes,
5306  * use them.
5307  *
5308  * Set the back channel max_resp_sz_cached to zero to force the client to
5309  * always set csa_cachethis to FALSE because the current implementation
5310  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5311  */
5312 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5313 {
5314         struct nfs4_session *session = args->client->cl_session;
5315         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5316                      mxresp_sz = session->fc_attrs.max_resp_sz;
5317
5318         if (mxrqst_sz == 0)
5319                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5320         if (mxresp_sz == 0)
5321                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5322         /* Fore channel attributes */
5323         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5324         args->fc_attrs.max_resp_sz = mxresp_sz;
5325         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5326         args->fc_attrs.max_reqs = max_session_slots;
5327
5328         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5329                 "max_ops=%u max_reqs=%u\n",
5330                 __func__,
5331                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5332                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5333
5334         /* Back channel attributes */
5335         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5336         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5337         args->bc_attrs.max_resp_sz_cached = 0;
5338         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5339         args->bc_attrs.max_reqs = 1;
5340
5341         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5342                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5343                 __func__,
5344                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5345                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5346                 args->bc_attrs.max_reqs);
5347 }
5348
5349 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5350 {
5351         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5352         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5353
5354         if (rcvd->max_resp_sz > sent->max_resp_sz)
5355                 return -EINVAL;
5356         /*
5357          * Our requested max_ops is the minimum we need; we're not
5358          * prepared to break up compounds into smaller pieces than that.
5359          * So, no point even trying to continue if the server won't
5360          * cooperate:
5361          */
5362         if (rcvd->max_ops < sent->max_ops)
5363                 return -EINVAL;
5364         if (rcvd->max_reqs == 0)
5365                 return -EINVAL;
5366         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5367                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5368         return 0;
5369 }
5370
5371 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5372 {
5373         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5374         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5375
5376         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5377                 return -EINVAL;
5378         if (rcvd->max_resp_sz < sent->max_resp_sz)
5379                 return -EINVAL;
5380         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5381                 return -EINVAL;
5382         /* These would render the backchannel useless: */
5383         if (rcvd->max_ops != sent->max_ops)
5384                 return -EINVAL;
5385         if (rcvd->max_reqs != sent->max_reqs)
5386                 return -EINVAL;
5387         return 0;
5388 }
5389
5390 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5391                                      struct nfs4_session *session)
5392 {
5393         int ret;
5394
5395         ret = nfs4_verify_fore_channel_attrs(args, session);
5396         if (ret)
5397                 return ret;
5398         return nfs4_verify_back_channel_attrs(args, session);
5399 }
5400
5401 static int _nfs4_proc_create_session(struct nfs_client *clp)
5402 {
5403         struct nfs4_session *session = clp->cl_session;
5404         struct nfs41_create_session_args args = {
5405                 .client = clp,
5406                 .cb_program = NFS4_CALLBACK,
5407         };
5408         struct nfs41_create_session_res res = {
5409                 .client = clp,
5410         };
5411         struct rpc_message msg = {
5412                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5413                 .rpc_argp = &args,
5414                 .rpc_resp = &res,
5415         };
5416         int status;
5417
5418         nfs4_init_channel_attrs(&args);
5419         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5420
5421         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5422
5423         if (!status)
5424                 /* Verify the session's negotiated channel_attrs values */
5425                 status = nfs4_verify_channel_attrs(&args, session);
5426         if (!status) {
5427                 /* Increment the clientid slot sequence id */
5428                 clp->cl_seqid++;
5429         }
5430
5431         return status;
5432 }
5433
5434 /*
5435  * Issues a CREATE_SESSION operation to the server.
5436  * It is the responsibility of the caller to verify the session is
5437  * expired before calling this routine.
5438  */
5439 int nfs4_proc_create_session(struct nfs_client *clp)
5440 {
5441         int status;
5442         unsigned *ptr;
5443         struct nfs4_session *session = clp->cl_session;
5444
5445         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5446
5447         status = _nfs4_proc_create_session(clp);
5448         if (status)
5449                 goto out;
5450
5451         /* Init or reset the session slot tables */
5452         status = nfs4_setup_session_slot_tables(session);
5453         dprintk("slot table setup returned %d\n", status);
5454         if (status)
5455                 goto out;
5456
5457         ptr = (unsigned *)&session->sess_id.data[0];
5458         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5459                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5460 out:
5461         dprintk("<-- %s\n", __func__);
5462         return status;
5463 }
5464
5465 /*
5466  * Issue the over-the-wire RPC DESTROY_SESSION.
5467  * The caller must serialize access to this routine.
5468  */
5469 int nfs4_proc_destroy_session(struct nfs4_session *session)
5470 {
5471         int status = 0;
5472         struct rpc_message msg;
5473
5474         dprintk("--> nfs4_proc_destroy_session\n");
5475
5476         /* session is still being setup */
5477         if (session->clp->cl_cons_state != NFS_CS_READY)
5478                 return status;
5479
5480         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5481         msg.rpc_argp = session;
5482         msg.rpc_resp = NULL;
5483         msg.rpc_cred = NULL;
5484         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5485
5486         if (status)
5487                 printk(KERN_WARNING
5488                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5489                         "Session has been destroyed regardless...\n", status);
5490
5491         dprintk("<-- nfs4_proc_destroy_session\n");
5492         return status;
5493 }
5494
5495 int nfs4_init_session(struct nfs_server *server)
5496 {
5497         struct nfs_client *clp = server->nfs_client;
5498         struct nfs4_session *session;
5499         unsigned int rsize, wsize;
5500         int ret;
5501
5502         if (!nfs4_has_session(clp))
5503                 return 0;
5504
5505         session = clp->cl_session;
5506         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5507                 return 0;
5508
5509         rsize = server->rsize;
5510         if (rsize == 0)
5511                 rsize = NFS_MAX_FILE_IO_SIZE;
5512         wsize = server->wsize;
5513         if (wsize == 0)
5514                 wsize = NFS_MAX_FILE_IO_SIZE;
5515
5516         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5517         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5518
5519         ret = nfs4_recover_expired_lease(server);
5520         if (!ret)
5521                 ret = nfs4_check_client_ready(clp);
5522         return ret;
5523 }
5524
5525 int nfs4_init_ds_session(struct nfs_client *clp)
5526 {
5527         struct nfs4_session *session = clp->cl_session;
5528         int ret;
5529
5530         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5531                 return 0;
5532
5533         ret = nfs4_client_recover_expired_lease(clp);
5534         if (!ret)
5535                 /* Test for the DS role */
5536                 if (!is_ds_client(clp))
5537                         ret = -ENODEV;
5538         if (!ret)
5539                 ret = nfs4_check_client_ready(clp);
5540         return ret;
5541
5542 }
5543 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5544
5545
5546 /*
5547  * Renew the cl_session lease.
5548  */
5549 struct nfs4_sequence_data {
5550         struct nfs_client *clp;
5551         struct nfs4_sequence_args args;
5552         struct nfs4_sequence_res res;
5553 };
5554
5555 static void nfs41_sequence_release(void *data)
5556 {
5557         struct nfs4_sequence_data *calldata = data;
5558         struct nfs_client *clp = calldata->clp;
5559
5560         if (atomic_read(&clp->cl_count) > 1)
5561                 nfs4_schedule_state_renewal(clp);
5562         nfs_put_client(clp);
5563         kfree(calldata);
5564 }
5565
5566 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5567 {
5568         switch(task->tk_status) {
5569         case -NFS4ERR_DELAY:
5570                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5571                 return -EAGAIN;
5572         default:
5573                 nfs4_schedule_lease_recovery(clp);
5574         }
5575         return 0;
5576 }
5577
5578 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5579 {
5580         struct nfs4_sequence_data *calldata = data;
5581         struct nfs_client *clp = calldata->clp;
5582
5583         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5584                 return;
5585
5586         if (task->tk_status < 0) {
5587                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5588                 if (atomic_read(&clp->cl_count) == 1)
5589                         goto out;
5590
5591                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5592                         rpc_restart_call_prepare(task);
5593                         return;
5594                 }
5595         }
5596         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5597 out:
5598         dprintk("<-- %s\n", __func__);
5599 }
5600
5601 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5602 {
5603         struct nfs4_sequence_data *calldata = data;
5604         struct nfs_client *clp = calldata->clp;
5605         struct nfs4_sequence_args *args;
5606         struct nfs4_sequence_res *res;
5607
5608         args = task->tk_msg.rpc_argp;
5609         res = task->tk_msg.rpc_resp;
5610
5611         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5612                 return;
5613         rpc_call_start(task);
5614 }
5615
5616 static const struct rpc_call_ops nfs41_sequence_ops = {
5617         .rpc_call_done = nfs41_sequence_call_done,
5618         .rpc_call_prepare = nfs41_sequence_prepare,
5619         .rpc_release = nfs41_sequence_release,
5620 };
5621
5622 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5623 {
5624         struct nfs4_sequence_data *calldata;
5625         struct rpc_message msg = {
5626                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5627                 .rpc_cred = cred,
5628         };
5629         struct rpc_task_setup task_setup_data = {
5630                 .rpc_client = clp->cl_rpcclient,
5631                 .rpc_message = &msg,
5632                 .callback_ops = &nfs41_sequence_ops,
5633                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5634         };
5635
5636         if (!atomic_inc_not_zero(&clp->cl_count))
5637                 return ERR_PTR(-EIO);
5638         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5639         if (calldata == NULL) {
5640                 nfs_put_client(clp);
5641                 return ERR_PTR(-ENOMEM);
5642         }
5643         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5644         msg.rpc_argp = &calldata->args;
5645         msg.rpc_resp = &calldata->res;
5646         calldata->clp = clp;
5647         task_setup_data.callback_data = calldata;
5648
5649         return rpc_run_task(&task_setup_data);
5650 }
5651
5652 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5653 {
5654         struct rpc_task *task;
5655         int ret = 0;
5656
5657         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5658                 return 0;
5659         task = _nfs41_proc_sequence(clp, cred);
5660         if (IS_ERR(task))
5661                 ret = PTR_ERR(task);
5662         else
5663                 rpc_put_task_async(task);
5664         dprintk("<-- %s status=%d\n", __func__, ret);
5665         return ret;
5666 }
5667
5668 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5669 {
5670         struct rpc_task *task;
5671         int ret;
5672
5673         task = _nfs41_proc_sequence(clp, cred);
5674         if (IS_ERR(task)) {
5675                 ret = PTR_ERR(task);
5676                 goto out;
5677         }
5678         ret = rpc_wait_for_completion_task(task);
5679         if (!ret) {
5680                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5681
5682                 if (task->tk_status == 0)
5683                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5684                 ret = task->tk_status;
5685         }
5686         rpc_put_task(task);
5687 out:
5688         dprintk("<-- %s status=%d\n", __func__, ret);
5689         return ret;
5690 }
5691
5692 struct nfs4_reclaim_complete_data {
5693         struct nfs_client *clp;
5694         struct nfs41_reclaim_complete_args arg;
5695         struct nfs41_reclaim_complete_res res;
5696 };
5697
5698 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5699 {
5700         struct nfs4_reclaim_complete_data *calldata = data;
5701
5702         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5703         if (nfs41_setup_sequence(calldata->clp->cl_session,
5704                                 &calldata->arg.seq_args,
5705                                 &calldata->res.seq_res, task))
5706                 return;
5707
5708         rpc_call_start(task);
5709 }
5710
5711 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5712 {
5713         switch(task->tk_status) {
5714         case 0:
5715         case -NFS4ERR_COMPLETE_ALREADY:
5716         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5717                 break;
5718         case -NFS4ERR_DELAY:
5719                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5720                 /* fall through */
5721         case -NFS4ERR_RETRY_UNCACHED_REP:
5722                 return -EAGAIN;
5723         default:
5724                 nfs4_schedule_lease_recovery(clp);
5725         }
5726         return 0;
5727 }
5728
5729 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5730 {
5731         struct nfs4_reclaim_complete_data *calldata = data;
5732         struct nfs_client *clp = calldata->clp;
5733         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5734
5735         dprintk("--> %s\n", __func__);
5736         if (!nfs41_sequence_done(task, res))
5737                 return;
5738
5739         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5740                 rpc_restart_call_prepare(task);
5741                 return;
5742         }
5743         dprintk("<-- %s\n", __func__);
5744 }
5745
5746 static void nfs4_free_reclaim_complete_data(void *data)
5747 {
5748         struct nfs4_reclaim_complete_data *calldata = data;
5749
5750         kfree(calldata);
5751 }
5752
5753 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5754         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5755         .rpc_call_done = nfs4_reclaim_complete_done,
5756         .rpc_release = nfs4_free_reclaim_complete_data,
5757 };
5758
5759 /*
5760  * Issue a global reclaim complete.
5761  */
5762 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5763 {
5764         struct nfs4_reclaim_complete_data *calldata;
5765         struct rpc_task *task;
5766         struct rpc_message msg = {
5767                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5768         };
5769         struct rpc_task_setup task_setup_data = {
5770                 .rpc_client = clp->cl_rpcclient,
5771                 .rpc_message = &msg,
5772                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5773                 .flags = RPC_TASK_ASYNC,
5774         };
5775         int status = -ENOMEM;
5776
5777         dprintk("--> %s\n", __func__);
5778         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5779         if (calldata == NULL)
5780                 goto out;
5781         calldata->clp = clp;
5782         calldata->arg.one_fs = 0;
5783
5784         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5785         msg.rpc_argp = &calldata->arg;
5786         msg.rpc_resp = &calldata->res;
5787         task_setup_data.callback_data = calldata;
5788         task = rpc_run_task(&task_setup_data);
5789         if (IS_ERR(task)) {
5790                 status = PTR_ERR(task);
5791                 goto out;
5792         }
5793         status = nfs4_wait_for_completion_rpc_task(task);
5794         if (status == 0)
5795                 status = task->tk_status;
5796         rpc_put_task(task);
5797         return 0;
5798 out:
5799         dprintk("<-- %s status=%d\n", __func__, status);
5800         return status;
5801 }
5802
5803 static void
5804 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5805 {
5806         struct nfs4_layoutget *lgp = calldata;
5807         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5808
5809         dprintk("--> %s\n", __func__);
5810         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5811          * right now covering the LAYOUTGET we are about to send.
5812          * However, that is not so catastrophic, and there seems
5813          * to be no way to prevent it completely.
5814          */
5815         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5816                                 &lgp->res.seq_res, task))
5817                 return;
5818         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5819                                           NFS_I(lgp->args.inode)->layout,
5820                                           lgp->args.ctx->state)) {
5821                 rpc_exit(task, NFS4_OK);
5822                 return;
5823         }
5824         rpc_call_start(task);
5825 }
5826
5827 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5828 {
5829         struct nfs4_layoutget *lgp = calldata;
5830         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5831
5832         dprintk("--> %s\n", __func__);
5833
5834         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5835                 return;
5836
5837         switch (task->tk_status) {
5838         case 0:
5839                 break;
5840         case -NFS4ERR_LAYOUTTRYLATER:
5841         case -NFS4ERR_RECALLCONFLICT:
5842                 task->tk_status = -NFS4ERR_DELAY;
5843                 /* Fall through */
5844         default:
5845                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5846                         rpc_restart_call_prepare(task);
5847                         return;
5848                 }
5849         }
5850         dprintk("<-- %s\n", __func__);
5851 }
5852
5853 static void nfs4_layoutget_release(void *calldata)
5854 {
5855         struct nfs4_layoutget *lgp = calldata;
5856
5857         dprintk("--> %s\n", __func__);
5858         put_nfs_open_context(lgp->args.ctx);
5859         kfree(calldata);
5860         dprintk("<-- %s\n", __func__);
5861 }
5862
5863 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5864         .rpc_call_prepare = nfs4_layoutget_prepare,
5865         .rpc_call_done = nfs4_layoutget_done,
5866         .rpc_release = nfs4_layoutget_release,
5867 };
5868
5869 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5870 {
5871         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5872         struct rpc_task *task;
5873         struct rpc_message msg = {
5874                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5875                 .rpc_argp = &lgp->args,
5876                 .rpc_resp = &lgp->res,
5877         };
5878         struct rpc_task_setup task_setup_data = {
5879                 .rpc_client = server->client,
5880                 .rpc_message = &msg,
5881                 .callback_ops = &nfs4_layoutget_call_ops,
5882                 .callback_data = lgp,
5883                 .flags = RPC_TASK_ASYNC,
5884         };
5885         int status = 0;
5886
5887         dprintk("--> %s\n", __func__);
5888
5889         lgp->res.layoutp = &lgp->args.layout;
5890         lgp->res.seq_res.sr_slot = NULL;
5891         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5892         task = rpc_run_task(&task_setup_data);
5893         if (IS_ERR(task))
5894                 return PTR_ERR(task);
5895         status = nfs4_wait_for_completion_rpc_task(task);
5896         if (status == 0)
5897                 status = task->tk_status;
5898         if (status == 0)
5899                 status = pnfs_layout_process(lgp);
5900         rpc_put_task(task);
5901         dprintk("<-- %s status=%d\n", __func__, status);
5902         return status;
5903 }
5904
5905 static void
5906 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5907 {
5908         struct nfs4_layoutreturn *lrp = calldata;
5909
5910         dprintk("--> %s\n", __func__);
5911         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5912                                 &lrp->res.seq_res, task))
5913                 return;
5914         rpc_call_start(task);
5915 }
5916
5917 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5918 {
5919         struct nfs4_layoutreturn *lrp = calldata;
5920         struct nfs_server *server;
5921         struct pnfs_layout_hdr *lo = lrp->args.layout;
5922
5923         dprintk("--> %s\n", __func__);
5924
5925         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5926                 return;
5927
5928         server = NFS_SERVER(lrp->args.inode);
5929         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5930                 rpc_restart_call_prepare(task);
5931                 return;
5932         }
5933         spin_lock(&lo->plh_inode->i_lock);
5934         if (task->tk_status == 0) {
5935                 if (lrp->res.lrs_present) {
5936                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5937                 } else
5938                         BUG_ON(!list_empty(&lo->plh_segs));
5939         }
5940         lo->plh_block_lgets--;
5941         spin_unlock(&lo->plh_inode->i_lock);
5942         dprintk("<-- %s\n", __func__);
5943 }
5944
5945 static void nfs4_layoutreturn_release(void *calldata)
5946 {
5947         struct nfs4_layoutreturn *lrp = calldata;
5948
5949         dprintk("--> %s\n", __func__);
5950         put_layout_hdr(lrp->args.layout);
5951         kfree(calldata);
5952         dprintk("<-- %s\n", __func__);
5953 }
5954
5955 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5956         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5957         .rpc_call_done = nfs4_layoutreturn_done,
5958         .rpc_release = nfs4_layoutreturn_release,
5959 };
5960
5961 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5962 {
5963         struct rpc_task *task;
5964         struct rpc_message msg = {
5965                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5966                 .rpc_argp = &lrp->args,
5967                 .rpc_resp = &lrp->res,
5968         };
5969         struct rpc_task_setup task_setup_data = {
5970                 .rpc_client = lrp->clp->cl_rpcclient,
5971                 .rpc_message = &msg,
5972                 .callback_ops = &nfs4_layoutreturn_call_ops,
5973                 .callback_data = lrp,
5974         };
5975         int status;
5976
5977         dprintk("--> %s\n", __func__);
5978         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5979         task = rpc_run_task(&task_setup_data);
5980         if (IS_ERR(task))
5981                 return PTR_ERR(task);
5982         status = task->tk_status;
5983         dprintk("<-- %s status=%d\n", __func__, status);
5984         rpc_put_task(task);
5985         return status;
5986 }
5987
5988 /*
5989  * Retrieve the list of Data Server devices from the MDS.
5990  */
5991 static int _nfs4_getdevicelist(struct nfs_server *server,
5992                                     const struct nfs_fh *fh,
5993                                     struct pnfs_devicelist *devlist)
5994 {
5995         struct nfs4_getdevicelist_args args = {
5996                 .fh = fh,
5997                 .layoutclass = server->pnfs_curr_ld->id,
5998         };
5999         struct nfs4_getdevicelist_res res = {
6000                 .devlist = devlist,
6001         };
6002         struct rpc_message msg = {
6003                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
6004                 .rpc_argp = &args,
6005                 .rpc_resp = &res,
6006         };
6007         int status;
6008
6009         dprintk("--> %s\n", __func__);
6010         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
6011                                 &res.seq_res, 0);
6012         dprintk("<-- %s status=%d\n", __func__, status);
6013         return status;
6014 }
6015
6016 int nfs4_proc_getdevicelist(struct nfs_server *server,
6017                             const struct nfs_fh *fh,
6018                             struct pnfs_devicelist *devlist)
6019 {
6020         struct nfs4_exception exception = { };
6021         int err;
6022
6023         do {
6024                 err = nfs4_handle_exception(server,
6025                                 _nfs4_getdevicelist(server, fh, devlist),
6026                                 &exception);
6027         } while (exception.retry);
6028
6029         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6030                 err, devlist->num_devs);
6031
6032         return err;
6033 }
6034 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6035
6036 static int
6037 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6038 {
6039         struct nfs4_getdeviceinfo_args args = {
6040                 .pdev = pdev,
6041         };
6042         struct nfs4_getdeviceinfo_res res = {
6043                 .pdev = pdev,
6044         };
6045         struct rpc_message msg = {
6046                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6047                 .rpc_argp = &args,
6048                 .rpc_resp = &res,
6049         };
6050         int status;
6051
6052         dprintk("--> %s\n", __func__);
6053         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6054         dprintk("<-- %s status=%d\n", __func__, status);
6055
6056         return status;
6057 }
6058
6059 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6060 {
6061         struct nfs4_exception exception = { };
6062         int err;
6063
6064         do {
6065                 err = nfs4_handle_exception(server,
6066                                         _nfs4_proc_getdeviceinfo(server, pdev),
6067                                         &exception);
6068         } while (exception.retry);
6069         return err;
6070 }
6071 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6072
6073 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6074 {
6075         struct nfs4_layoutcommit_data *data = calldata;
6076         struct nfs_server *server = NFS_SERVER(data->args.inode);
6077
6078         if (nfs4_setup_sequence(server, &data->args.seq_args,
6079                                 &data->res.seq_res, task))
6080                 return;
6081         rpc_call_start(task);
6082 }
6083
6084 static void
6085 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6086 {
6087         struct nfs4_layoutcommit_data *data = calldata;
6088         struct nfs_server *server = NFS_SERVER(data->args.inode);
6089
6090         if (!nfs4_sequence_done(task, &data->res.seq_res))
6091                 return;
6092
6093         switch (task->tk_status) { /* Just ignore these failures */
6094         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6095         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6096         case NFS4ERR_BADLAYOUT:     /* no layout */
6097         case NFS4ERR_GRACE:         /* loca_recalim always false */
6098                 task->tk_status = 0;
6099         }
6100
6101         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6102                 rpc_restart_call_prepare(task);
6103                 return;
6104         }
6105
6106         if (task->tk_status == 0)
6107                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6108                                                    data->res.fattr);
6109 }
6110
6111 static void nfs4_layoutcommit_release(void *calldata)
6112 {
6113         struct nfs4_layoutcommit_data *data = calldata;
6114         struct pnfs_layout_segment *lseg, *tmp;
6115         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6116
6117         pnfs_cleanup_layoutcommit(data);
6118         /* Matched by references in pnfs_set_layoutcommit */
6119         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6120                 list_del_init(&lseg->pls_lc_list);
6121                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6122                                        &lseg->pls_flags))
6123                         put_lseg(lseg);
6124         }
6125
6126         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6127         smp_mb__after_clear_bit();
6128         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6129
6130         put_rpccred(data->cred);
6131         kfree(data);
6132 }
6133
6134 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6135         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6136         .rpc_call_done = nfs4_layoutcommit_done,
6137         .rpc_release = nfs4_layoutcommit_release,
6138 };
6139
6140 int
6141 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6142 {
6143         struct rpc_message msg = {
6144                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6145                 .rpc_argp = &data->args,
6146                 .rpc_resp = &data->res,
6147                 .rpc_cred = data->cred,
6148         };
6149         struct rpc_task_setup task_setup_data = {
6150                 .task = &data->task,
6151                 .rpc_client = NFS_CLIENT(data->args.inode),
6152                 .rpc_message = &msg,
6153                 .callback_ops = &nfs4_layoutcommit_ops,
6154                 .callback_data = data,
6155                 .flags = RPC_TASK_ASYNC,
6156         };
6157         struct rpc_task *task;
6158         int status = 0;
6159
6160         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6161                 "lbw: %llu inode %lu\n",
6162                 data->task.tk_pid, sync,
6163                 data->args.lastbytewritten,
6164                 data->args.inode->i_ino);
6165
6166         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6167         task = rpc_run_task(&task_setup_data);
6168         if (IS_ERR(task))
6169                 return PTR_ERR(task);
6170         if (sync == false)
6171                 goto out;
6172         status = nfs4_wait_for_completion_rpc_task(task);
6173         if (status != 0)
6174                 goto out;
6175         status = task->tk_status;
6176 out:
6177         dprintk("%s: status %d\n", __func__, status);
6178         rpc_put_task(task);
6179         return status;
6180 }
6181
6182 static int
6183 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6184                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6185 {
6186         struct nfs41_secinfo_no_name_args args = {
6187                 .style = SECINFO_STYLE_CURRENT_FH,
6188         };
6189         struct nfs4_secinfo_res res = {
6190                 .flavors = flavors,
6191         };
6192         struct rpc_message msg = {
6193                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6194                 .rpc_argp = &args,
6195                 .rpc_resp = &res,
6196         };
6197         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6198 }
6199
6200 static int
6201 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6202                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6203 {
6204         struct nfs4_exception exception = { };
6205         int err;
6206         do {
6207                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6208                 switch (err) {
6209                 case 0:
6210                 case -NFS4ERR_WRONGSEC:
6211                 case -NFS4ERR_NOTSUPP:
6212                         break;
6213                 default:
6214                         err = nfs4_handle_exception(server, err, &exception);
6215                 }
6216         } while (exception.retry);
6217         return err;
6218 }
6219
6220 static int
6221 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6222                     struct nfs_fsinfo *info)
6223 {
6224         int err;
6225         struct page *page;
6226         rpc_authflavor_t flavor;
6227         struct nfs4_secinfo_flavors *flavors;
6228
6229         page = alloc_page(GFP_KERNEL);
6230         if (!page) {
6231                 err = -ENOMEM;
6232                 goto out;
6233         }
6234
6235         flavors = page_address(page);
6236         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6237
6238         /*
6239          * Fall back on "guess and check" method if
6240          * the server doesn't support SECINFO_NO_NAME
6241          */
6242         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6243                 err = nfs4_find_root_sec(server, fhandle, info);
6244                 goto out_freepage;
6245         }
6246         if (err)
6247                 goto out_freepage;
6248
6249         flavor = nfs_find_best_sec(flavors);
6250         if (err == 0)
6251                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6252
6253 out_freepage:
6254         put_page(page);
6255         if (err == -EACCES)
6256                 return -EPERM;
6257 out:
6258         return err;
6259 }
6260
6261 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6262 {
6263         int status;
6264         struct nfs41_test_stateid_args args = {
6265                 .stateid = stateid,
6266         };
6267         struct nfs41_test_stateid_res res;
6268         struct rpc_message msg = {
6269                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6270                 .rpc_argp = &args,
6271                 .rpc_resp = &res,
6272         };
6273
6274         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6275         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6276
6277         if (status == NFS_OK)
6278                 return res.status;
6279         return status;
6280 }
6281
6282 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6283 {
6284         struct nfs4_exception exception = { };
6285         int err;
6286         do {
6287                 err = nfs4_handle_exception(server,
6288                                 _nfs41_test_stateid(server, stateid),
6289                                 &exception);
6290         } while (exception.retry);
6291         return err;
6292 }
6293
6294 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6295 {
6296         struct nfs41_free_stateid_args args = {
6297                 .stateid = stateid,
6298         };
6299         struct nfs41_free_stateid_res res;
6300         struct rpc_message msg = {
6301                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6302                 .rpc_argp = &args,
6303                 .rpc_resp = &res,
6304         };
6305
6306         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6307         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6308 }
6309
6310 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6311 {
6312         struct nfs4_exception exception = { };
6313         int err;
6314         do {
6315                 err = nfs4_handle_exception(server,
6316                                 _nfs4_free_stateid(server, stateid),
6317                                 &exception);
6318         } while (exception.retry);
6319         return err;
6320 }
6321
6322 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6323                 const nfs4_stateid *s2)
6324 {
6325         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6326                 return false;
6327
6328         if (s1->seqid == s2->seqid)
6329                 return true;
6330         if (s1->seqid == 0 || s2->seqid == 0)
6331                 return true;
6332
6333         return false;
6334 }
6335
6336 #endif /* CONFIG_NFS_V4_1 */
6337
6338 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6339                 const nfs4_stateid *s2)
6340 {
6341         return nfs4_stateid_match(s1, s2);
6342 }
6343
6344
6345 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6346         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6347         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6348         .recover_open   = nfs4_open_reclaim,
6349         .recover_lock   = nfs4_lock_reclaim,
6350         .establish_clid = nfs4_init_clientid,
6351         .get_clid_cred  = nfs4_get_setclientid_cred,
6352 };
6353
6354 #if defined(CONFIG_NFS_V4_1)
6355 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6356         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6357         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6358         .recover_open   = nfs4_open_reclaim,
6359         .recover_lock   = nfs4_lock_reclaim,
6360         .establish_clid = nfs41_init_clientid,
6361         .get_clid_cred  = nfs4_get_exchange_id_cred,
6362         .reclaim_complete = nfs41_proc_reclaim_complete,
6363 };
6364 #endif /* CONFIG_NFS_V4_1 */
6365
6366 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6367         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6368         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6369         .recover_open   = nfs4_open_expired,
6370         .recover_lock   = nfs4_lock_expired,
6371         .establish_clid = nfs4_init_clientid,
6372         .get_clid_cred  = nfs4_get_setclientid_cred,
6373 };
6374
6375 #if defined(CONFIG_NFS_V4_1)
6376 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6377         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6378         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6379         .recover_open   = nfs41_open_expired,
6380         .recover_lock   = nfs41_lock_expired,
6381         .establish_clid = nfs41_init_clientid,
6382         .get_clid_cred  = nfs4_get_exchange_id_cred,
6383 };
6384 #endif /* CONFIG_NFS_V4_1 */
6385
6386 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6387         .sched_state_renewal = nfs4_proc_async_renew,
6388         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6389         .renew_lease = nfs4_proc_renew,
6390 };
6391
6392 #if defined(CONFIG_NFS_V4_1)
6393 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6394         .sched_state_renewal = nfs41_proc_async_sequence,
6395         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6396         .renew_lease = nfs4_proc_sequence,
6397 };
6398 #endif
6399
6400 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6401         .minor_version = 0,
6402         .call_sync = _nfs4_call_sync,
6403         .match_stateid = nfs4_match_stateid,
6404         .find_root_sec = nfs4_find_root_sec,
6405         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6406         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6407         .state_renewal_ops = &nfs40_state_renewal_ops,
6408 };
6409
6410 #if defined(CONFIG_NFS_V4_1)
6411 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6412         .minor_version = 1,
6413         .call_sync = _nfs4_call_sync_session,
6414         .match_stateid = nfs41_match_stateid,
6415         .find_root_sec = nfs41_find_root_sec,
6416         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6417         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6418         .state_renewal_ops = &nfs41_state_renewal_ops,
6419 };
6420 #endif
6421
6422 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6423         [0] = &nfs_v4_0_minor_ops,
6424 #if defined(CONFIG_NFS_V4_1)
6425         [1] = &nfs_v4_1_minor_ops,
6426 #endif
6427 };
6428
6429 static const struct inode_operations nfs4_file_inode_operations = {
6430         .permission     = nfs_permission,
6431         .getattr        = nfs_getattr,
6432         .setattr        = nfs_setattr,
6433         .getxattr       = generic_getxattr,
6434         .setxattr       = generic_setxattr,
6435         .listxattr      = generic_listxattr,
6436         .removexattr    = generic_removexattr,
6437 };
6438
6439 const struct nfs_rpc_ops nfs_v4_clientops = {
6440         .version        = 4,                    /* protocol version */
6441         .dentry_ops     = &nfs4_dentry_operations,
6442         .dir_inode_ops  = &nfs4_dir_inode_operations,
6443         .file_inode_ops = &nfs4_file_inode_operations,
6444         .file_ops       = &nfs4_file_operations,
6445         .getroot        = nfs4_proc_get_root,
6446         .getattr        = nfs4_proc_getattr,
6447         .setattr        = nfs4_proc_setattr,
6448         .lookup         = nfs4_proc_lookup,
6449         .access         = nfs4_proc_access,
6450         .readlink       = nfs4_proc_readlink,
6451         .create         = nfs4_proc_create,
6452         .remove         = nfs4_proc_remove,
6453         .unlink_setup   = nfs4_proc_unlink_setup,
6454         .unlink_done    = nfs4_proc_unlink_done,
6455         .rename         = nfs4_proc_rename,
6456         .rename_setup   = nfs4_proc_rename_setup,
6457         .rename_done    = nfs4_proc_rename_done,
6458         .link           = nfs4_proc_link,
6459         .symlink        = nfs4_proc_symlink,
6460         .mkdir          = nfs4_proc_mkdir,
6461         .rmdir          = nfs4_proc_remove,
6462         .readdir        = nfs4_proc_readdir,
6463         .mknod          = nfs4_proc_mknod,
6464         .statfs         = nfs4_proc_statfs,
6465         .fsinfo         = nfs4_proc_fsinfo,
6466         .pathconf       = nfs4_proc_pathconf,
6467         .set_capabilities = nfs4_server_capabilities,
6468         .decode_dirent  = nfs4_decode_dirent,
6469         .read_setup     = nfs4_proc_read_setup,
6470         .read_rpc_prepare = nfs4_proc_read_rpc_prepare,
6471         .read_done      = nfs4_read_done,
6472         .write_setup    = nfs4_proc_write_setup,
6473         .write_rpc_prepare = nfs4_proc_write_rpc_prepare,
6474         .write_done     = nfs4_write_done,
6475         .commit_setup   = nfs4_proc_commit_setup,
6476         .commit_done    = nfs4_commit_done,
6477         .lock           = nfs4_proc_lock,
6478         .clear_acl_cache = nfs4_zap_acl_attr,
6479         .close_context  = nfs4_close_context,
6480         .open_context   = nfs4_atomic_open,
6481         .init_client    = nfs4_init_client,
6482         .secinfo        = nfs4_proc_secinfo,
6483 };
6484
6485 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6486         .prefix = XATTR_NAME_NFSV4_ACL,
6487         .list   = nfs4_xattr_list_nfs4_acl,
6488         .get    = nfs4_xattr_get_nfs4_acl,
6489         .set    = nfs4_xattr_set_nfs4_acl,
6490 };
6491
6492 const struct xattr_handler *nfs4_xattr_handlers[] = {
6493         &nfs4_xattr_nfs4_acl_handler,
6494         NULL
6495 };
6496
6497 module_param(max_session_slots, ushort, 0644);
6498 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6499                 "requests the client will negotiate");
6500
6501 /*
6502  * Local variables:
6503  *  c-basic-offset: 8
6504  * End:
6505  */