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 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 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                         printk(KERN_WARNING "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 (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1933                 /* Use that stateid */
1934         } else if (state != NULL) {
1935                 nfs4_select_rw_stateid(&arg.stateid, state, current->files, current->tgid);
1936         } else
1937                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
1938
1939         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1940         if (status == 0 && state != NULL)
1941                 renew_lease(server, timestamp);
1942         return status;
1943 }
1944
1945 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1946                            struct nfs_fattr *fattr, struct iattr *sattr,
1947                            struct nfs4_state *state)
1948 {
1949         struct nfs_server *server = NFS_SERVER(inode);
1950         struct nfs4_exception exception = {
1951                 .state = state,
1952                 .inode = inode,
1953         };
1954         int err;
1955         do {
1956                 err = nfs4_handle_exception(server,
1957                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1958                                 &exception);
1959         } while (exception.retry);
1960         return err;
1961 }
1962
1963 struct nfs4_closedata {
1964         struct inode *inode;
1965         struct nfs4_state *state;
1966         struct nfs_closeargs arg;
1967         struct nfs_closeres res;
1968         struct nfs_fattr fattr;
1969         unsigned long timestamp;
1970         bool roc;
1971         u32 roc_barrier;
1972 };
1973
1974 static void nfs4_free_closedata(void *data)
1975 {
1976         struct nfs4_closedata *calldata = data;
1977         struct nfs4_state_owner *sp = calldata->state->owner;
1978         struct super_block *sb = calldata->state->inode->i_sb;
1979
1980         if (calldata->roc)
1981                 pnfs_roc_release(calldata->state->inode);
1982         nfs4_put_open_state(calldata->state);
1983         nfs_free_seqid(calldata->arg.seqid);
1984         nfs4_put_state_owner(sp);
1985         nfs_sb_deactive(sb);
1986         kfree(calldata);
1987 }
1988
1989 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1990                 fmode_t fmode)
1991 {
1992         spin_lock(&state->owner->so_lock);
1993         if (!(fmode & FMODE_READ))
1994                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1995         if (!(fmode & FMODE_WRITE))
1996                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1997         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1998         spin_unlock(&state->owner->so_lock);
1999 }
2000
2001 static void nfs4_close_done(struct rpc_task *task, void *data)
2002 {
2003         struct nfs4_closedata *calldata = data;
2004         struct nfs4_state *state = calldata->state;
2005         struct nfs_server *server = NFS_SERVER(calldata->inode);
2006
2007         dprintk("%s: begin!\n", __func__);
2008         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2009                 return;
2010         /* hmm. we are done with the inode, and in the process of freeing
2011          * the state_owner. we keep this around to process errors
2012          */
2013         switch (task->tk_status) {
2014                 case 0:
2015                         if (calldata->roc)
2016                                 pnfs_roc_set_barrier(state->inode,
2017                                                      calldata->roc_barrier);
2018                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2019                         renew_lease(server, calldata->timestamp);
2020                         nfs4_close_clear_stateid_flags(state,
2021                                         calldata->arg.fmode);
2022                         break;
2023                 case -NFS4ERR_STALE_STATEID:
2024                 case -NFS4ERR_OLD_STATEID:
2025                 case -NFS4ERR_BAD_STATEID:
2026                 case -NFS4ERR_EXPIRED:
2027                         if (calldata->arg.fmode == 0)
2028                                 break;
2029                 default:
2030                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2031                                 rpc_restart_call_prepare(task);
2032         }
2033         nfs_release_seqid(calldata->arg.seqid);
2034         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2035         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2036 }
2037
2038 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2039 {
2040         struct nfs4_closedata *calldata = data;
2041         struct nfs4_state *state = calldata->state;
2042         int call_close = 0;
2043
2044         dprintk("%s: begin!\n", __func__);
2045         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2046                 return;
2047
2048         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2049         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2050         spin_lock(&state->owner->so_lock);
2051         /* Calculate the change in open mode */
2052         if (state->n_rdwr == 0) {
2053                 if (state->n_rdonly == 0) {
2054                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2055                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2056                         calldata->arg.fmode &= ~FMODE_READ;
2057                 }
2058                 if (state->n_wronly == 0) {
2059                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2060                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2061                         calldata->arg.fmode &= ~FMODE_WRITE;
2062                 }
2063         }
2064         spin_unlock(&state->owner->so_lock);
2065
2066         if (!call_close) {
2067                 /* Note: exit _without_ calling nfs4_close_done */
2068                 task->tk_action = NULL;
2069                 goto out;
2070         }
2071
2072         if (calldata->arg.fmode == 0) {
2073                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2074                 if (calldata->roc &&
2075                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2076                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2077                                      task, NULL);
2078                         goto out;
2079                 }
2080         }
2081
2082         nfs_fattr_init(calldata->res.fattr);
2083         calldata->timestamp = jiffies;
2084         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2085                                 &calldata->arg.seq_args,
2086                                 &calldata->res.seq_res,
2087                                 task))
2088                 goto out;
2089         rpc_call_start(task);
2090 out:
2091         dprintk("%s: done!\n", __func__);
2092 }
2093
2094 static const struct rpc_call_ops nfs4_close_ops = {
2095         .rpc_call_prepare = nfs4_close_prepare,
2096         .rpc_call_done = nfs4_close_done,
2097         .rpc_release = nfs4_free_closedata,
2098 };
2099
2100 /*
2101  * It is possible for data to be read/written from a mem-mapped file
2102  * after the sys_close call (which hits the vfs layer as a flush).
2103  * This means that we can't safely call nfsv4 close on a file until
2104  * the inode is cleared. This in turn means that we are not good
2105  * NFSv4 citizens - we do not indicate to the server to update the file's
2106  * share state even when we are done with one of the three share
2107  * stateid's in the inode.
2108  *
2109  * NOTE: Caller must be holding the sp->so_owner semaphore!
2110  */
2111 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2112 {
2113         struct nfs_server *server = NFS_SERVER(state->inode);
2114         struct nfs4_closedata *calldata;
2115         struct nfs4_state_owner *sp = state->owner;
2116         struct rpc_task *task;
2117         struct rpc_message msg = {
2118                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2119                 .rpc_cred = state->owner->so_cred,
2120         };
2121         struct rpc_task_setup task_setup_data = {
2122                 .rpc_client = server->client,
2123                 .rpc_message = &msg,
2124                 .callback_ops = &nfs4_close_ops,
2125                 .workqueue = nfsiod_workqueue,
2126                 .flags = RPC_TASK_ASYNC,
2127         };
2128         int status = -ENOMEM;
2129
2130         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2131         if (calldata == NULL)
2132                 goto out;
2133         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2134         calldata->inode = state->inode;
2135         calldata->state = state;
2136         calldata->arg.fh = NFS_FH(state->inode);
2137         calldata->arg.stateid = &state->open_stateid;
2138         /* Serialization for the sequence id */
2139         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2140         if (calldata->arg.seqid == NULL)
2141                 goto out_free_calldata;
2142         calldata->arg.fmode = 0;
2143         calldata->arg.bitmask = server->cache_consistency_bitmask;
2144         calldata->res.fattr = &calldata->fattr;
2145         calldata->res.seqid = calldata->arg.seqid;
2146         calldata->res.server = server;
2147         calldata->roc = roc;
2148         nfs_sb_active(calldata->inode->i_sb);
2149
2150         msg.rpc_argp = &calldata->arg;
2151         msg.rpc_resp = &calldata->res;
2152         task_setup_data.callback_data = calldata;
2153         task = rpc_run_task(&task_setup_data);
2154         if (IS_ERR(task))
2155                 return PTR_ERR(task);
2156         status = 0;
2157         if (wait)
2158                 status = rpc_wait_for_completion_task(task);
2159         rpc_put_task(task);
2160         return status;
2161 out_free_calldata:
2162         kfree(calldata);
2163 out:
2164         if (roc)
2165                 pnfs_roc_release(state->inode);
2166         nfs4_put_open_state(state);
2167         nfs4_put_state_owner(sp);
2168         return status;
2169 }
2170
2171 static struct inode *
2172 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2173 {
2174         struct nfs4_state *state;
2175
2176         /* Protect against concurrent sillydeletes */
2177         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2178         if (IS_ERR(state))
2179                 return ERR_CAST(state);
2180         ctx->state = state;
2181         return igrab(state->inode);
2182 }
2183
2184 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2185 {
2186         if (ctx->state == NULL)
2187                 return;
2188         if (is_sync)
2189                 nfs4_close_sync(ctx->state, ctx->mode);
2190         else
2191                 nfs4_close_state(ctx->state, ctx->mode);
2192 }
2193
2194 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2195 {
2196         struct nfs4_server_caps_arg args = {
2197                 .fhandle = fhandle,
2198         };
2199         struct nfs4_server_caps_res res = {};
2200         struct rpc_message msg = {
2201                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2202                 .rpc_argp = &args,
2203                 .rpc_resp = &res,
2204         };
2205         int status;
2206
2207         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2208         if (status == 0) {
2209                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2210                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2211                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2212                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2213                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2214                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2215                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2216                         server->caps |= NFS_CAP_ACLS;
2217                 if (res.has_links != 0)
2218                         server->caps |= NFS_CAP_HARDLINKS;
2219                 if (res.has_symlinks != 0)
2220                         server->caps |= NFS_CAP_SYMLINKS;
2221                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2222                         server->caps |= NFS_CAP_FILEID;
2223                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2224                         server->caps |= NFS_CAP_MODE;
2225                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2226                         server->caps |= NFS_CAP_NLINK;
2227                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2228                         server->caps |= NFS_CAP_OWNER;
2229                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2230                         server->caps |= NFS_CAP_OWNER_GROUP;
2231                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2232                         server->caps |= NFS_CAP_ATIME;
2233                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2234                         server->caps |= NFS_CAP_CTIME;
2235                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2236                         server->caps |= NFS_CAP_MTIME;
2237
2238                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2239                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2240                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2241                 server->acl_bitmask = res.acl_bitmask;
2242                 server->fh_expire_type = res.fh_expire_type;
2243         }
2244
2245         return status;
2246 }
2247
2248 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2249 {
2250         struct nfs4_exception exception = { };
2251         int err;
2252         do {
2253                 err = nfs4_handle_exception(server,
2254                                 _nfs4_server_capabilities(server, fhandle),
2255                                 &exception);
2256         } while (exception.retry);
2257         return err;
2258 }
2259
2260 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2261                 struct nfs_fsinfo *info)
2262 {
2263         struct nfs4_lookup_root_arg args = {
2264                 .bitmask = nfs4_fattr_bitmap,
2265         };
2266         struct nfs4_lookup_res res = {
2267                 .server = server,
2268                 .fattr = info->fattr,
2269                 .fh = fhandle,
2270         };
2271         struct rpc_message msg = {
2272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2273                 .rpc_argp = &args,
2274                 .rpc_resp = &res,
2275         };
2276
2277         nfs_fattr_init(info->fattr);
2278         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2279 }
2280
2281 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2282                 struct nfs_fsinfo *info)
2283 {
2284         struct nfs4_exception exception = { };
2285         int err;
2286         do {
2287                 err = _nfs4_lookup_root(server, fhandle, info);
2288                 switch (err) {
2289                 case 0:
2290                 case -NFS4ERR_WRONGSEC:
2291                         break;
2292                 default:
2293                         err = nfs4_handle_exception(server, err, &exception);
2294                 }
2295         } while (exception.retry);
2296         return err;
2297 }
2298
2299 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2300                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2301 {
2302         struct rpc_auth *auth;
2303         int ret;
2304
2305         auth = rpcauth_create(flavor, server->client);
2306         if (!auth) {
2307                 ret = -EIO;
2308                 goto out;
2309         }
2310         ret = nfs4_lookup_root(server, fhandle, info);
2311 out:
2312         return ret;
2313 }
2314
2315 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2316                               struct nfs_fsinfo *info)
2317 {
2318         int i, len, status = 0;
2319         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2320
2321         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2322         flav_array[len] = RPC_AUTH_NULL;
2323         len += 1;
2324
2325         for (i = 0; i < len; i++) {
2326                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2327                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2328                         continue;
2329                 break;
2330         }
2331         /*
2332          * -EACCESS could mean that the user doesn't have correct permissions
2333          * to access the mount.  It could also mean that we tried to mount
2334          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2335          * existing mount programs don't handle -EACCES very well so it should
2336          * be mapped to -EPERM instead.
2337          */
2338         if (status == -EACCES)
2339                 status = -EPERM;
2340         return status;
2341 }
2342
2343 /*
2344  * get the file handle for the "/" directory on the server
2345  */
2346 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2347                               struct nfs_fsinfo *info)
2348 {
2349         int minor_version = server->nfs_client->cl_minorversion;
2350         int status = nfs4_lookup_root(server, fhandle, info);
2351         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2352                 /*
2353                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2354                  * by nfs4_map_errors() as this function exits.
2355                  */
2356                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2357         if (status == 0)
2358                 status = nfs4_server_capabilities(server, fhandle);
2359         if (status == 0)
2360                 status = nfs4_do_fsinfo(server, fhandle, info);
2361         return nfs4_map_errors(status);
2362 }
2363
2364 /*
2365  * Get locations and (maybe) other attributes of a referral.
2366  * Note that we'll actually follow the referral later when
2367  * we detect fsid mismatch in inode revalidation
2368  */
2369 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2370                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2371 {
2372         int status = -ENOMEM;
2373         struct page *page = NULL;
2374         struct nfs4_fs_locations *locations = NULL;
2375
2376         page = alloc_page(GFP_KERNEL);
2377         if (page == NULL)
2378                 goto out;
2379         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2380         if (locations == NULL)
2381                 goto out;
2382
2383         status = nfs4_proc_fs_locations(dir, name, locations, page);
2384         if (status != 0)
2385                 goto out;
2386         /* Make sure server returned a different fsid for the referral */
2387         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2388                 dprintk("%s: server did not return a different fsid for"
2389                         " a referral at %s\n", __func__, name->name);
2390                 status = -EIO;
2391                 goto out;
2392         }
2393         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2394         nfs_fixup_referral_attributes(&locations->fattr);
2395
2396         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2397         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2398         memset(fhandle, 0, sizeof(struct nfs_fh));
2399 out:
2400         if (page)
2401                 __free_page(page);
2402         kfree(locations);
2403         return status;
2404 }
2405
2406 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2407 {
2408         struct nfs4_getattr_arg args = {
2409                 .fh = fhandle,
2410                 .bitmask = server->attr_bitmask,
2411         };
2412         struct nfs4_getattr_res res = {
2413                 .fattr = fattr,
2414                 .server = server,
2415         };
2416         struct rpc_message msg = {
2417                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2418                 .rpc_argp = &args,
2419                 .rpc_resp = &res,
2420         };
2421
2422         nfs_fattr_init(fattr);
2423         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2424 }
2425
2426 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2427 {
2428         struct nfs4_exception exception = { };
2429         int err;
2430         do {
2431                 err = nfs4_handle_exception(server,
2432                                 _nfs4_proc_getattr(server, fhandle, fattr),
2433                                 &exception);
2434         } while (exception.retry);
2435         return err;
2436 }
2437
2438 /*
2439  * The file is not closed if it is opened due to the a request to change
2440  * the size of the file. The open call will not be needed once the
2441  * VFS layer lookup-intents are implemented.
2442  *
2443  * Close is called when the inode is destroyed.
2444  * If we haven't opened the file for O_WRONLY, we
2445  * need to in the size_change case to obtain a stateid.
2446  *
2447  * Got race?
2448  * Because OPEN is always done by name in nfsv4, it is
2449  * possible that we opened a different file by the same
2450  * name.  We can recognize this race condition, but we
2451  * can't do anything about it besides returning an error.
2452  *
2453  * This will be fixed with VFS changes (lookup-intent).
2454  */
2455 static int
2456 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2457                   struct iattr *sattr)
2458 {
2459         struct inode *inode = dentry->d_inode;
2460         struct rpc_cred *cred = NULL;
2461         struct nfs4_state *state = NULL;
2462         int status;
2463
2464         if (pnfs_ld_layoutret_on_setattr(inode))
2465                 pnfs_return_layout(inode);
2466
2467         nfs_fattr_init(fattr);
2468
2469         /* Search for an existing open(O_WRITE) file */
2470         if (sattr->ia_valid & ATTR_FILE) {
2471                 struct nfs_open_context *ctx;
2472
2473                 ctx = nfs_file_open_context(sattr->ia_file);
2474                 if (ctx) {
2475                         cred = ctx->cred;
2476                         state = ctx->state;
2477                 }
2478         }
2479
2480         /* Deal with open(O_TRUNC) */
2481         if (sattr->ia_valid & ATTR_OPEN)
2482                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2483
2484         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2485         if (status == 0)
2486                 nfs_setattr_update_inode(inode, sattr);
2487         return status;
2488 }
2489
2490 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2491                 const struct qstr *name, struct nfs_fh *fhandle,
2492                 struct nfs_fattr *fattr)
2493 {
2494         struct nfs_server *server = NFS_SERVER(dir);
2495         int                    status;
2496         struct nfs4_lookup_arg args = {
2497                 .bitmask = server->attr_bitmask,
2498                 .dir_fh = NFS_FH(dir),
2499                 .name = name,
2500         };
2501         struct nfs4_lookup_res res = {
2502                 .server = server,
2503                 .fattr = fattr,
2504                 .fh = fhandle,
2505         };
2506         struct rpc_message msg = {
2507                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2508                 .rpc_argp = &args,
2509                 .rpc_resp = &res,
2510         };
2511
2512         nfs_fattr_init(fattr);
2513
2514         dprintk("NFS call  lookup %s\n", name->name);
2515         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2516         dprintk("NFS reply lookup: %d\n", status);
2517         return status;
2518 }
2519
2520 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2521 {
2522         memset(fh, 0, sizeof(struct nfs_fh));
2523         fattr->fsid.major = 1;
2524         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2525                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2526         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2527         fattr->nlink = 2;
2528 }
2529
2530 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2531                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2532 {
2533         struct nfs4_exception exception = { };
2534         int err;
2535         do {
2536                 int status;
2537
2538                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2539                 switch (status) {
2540                 case -NFS4ERR_BADNAME:
2541                         return -ENOENT;
2542                 case -NFS4ERR_MOVED:
2543                         return nfs4_get_referral(dir, name, fattr, fhandle);
2544                 case -NFS4ERR_WRONGSEC:
2545                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2546                 }
2547                 err = nfs4_handle_exception(NFS_SERVER(dir),
2548                                 status, &exception);
2549         } while (exception.retry);
2550         return err;
2551 }
2552
2553 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2554 {
2555         struct nfs_server *server = NFS_SERVER(inode);
2556         struct nfs4_accessargs args = {
2557                 .fh = NFS_FH(inode),
2558                 .bitmask = server->cache_consistency_bitmask,
2559         };
2560         struct nfs4_accessres res = {
2561                 .server = server,
2562         };
2563         struct rpc_message msg = {
2564                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2565                 .rpc_argp = &args,
2566                 .rpc_resp = &res,
2567                 .rpc_cred = entry->cred,
2568         };
2569         int mode = entry->mask;
2570         int status;
2571
2572         /*
2573          * Determine which access bits we want to ask for...
2574          */
2575         if (mode & MAY_READ)
2576                 args.access |= NFS4_ACCESS_READ;
2577         if (S_ISDIR(inode->i_mode)) {
2578                 if (mode & MAY_WRITE)
2579                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2580                 if (mode & MAY_EXEC)
2581                         args.access |= NFS4_ACCESS_LOOKUP;
2582         } else {
2583                 if (mode & MAY_WRITE)
2584                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2585                 if (mode & MAY_EXEC)
2586                         args.access |= NFS4_ACCESS_EXECUTE;
2587         }
2588
2589         res.fattr = nfs_alloc_fattr();
2590         if (res.fattr == NULL)
2591                 return -ENOMEM;
2592
2593         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2594         if (!status) {
2595                 entry->mask = 0;
2596                 if (res.access & NFS4_ACCESS_READ)
2597                         entry->mask |= MAY_READ;
2598                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2599                         entry->mask |= MAY_WRITE;
2600                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2601                         entry->mask |= MAY_EXEC;
2602                 nfs_refresh_inode(inode, res.fattr);
2603         }
2604         nfs_free_fattr(res.fattr);
2605         return status;
2606 }
2607
2608 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2609 {
2610         struct nfs4_exception exception = { };
2611         int err;
2612         do {
2613                 err = nfs4_handle_exception(NFS_SERVER(inode),
2614                                 _nfs4_proc_access(inode, entry),
2615                                 &exception);
2616         } while (exception.retry);
2617         return err;
2618 }
2619
2620 /*
2621  * TODO: For the time being, we don't try to get any attributes
2622  * along with any of the zero-copy operations READ, READDIR,
2623  * READLINK, WRITE.
2624  *
2625  * In the case of the first three, we want to put the GETATTR
2626  * after the read-type operation -- this is because it is hard
2627  * to predict the length of a GETATTR response in v4, and thus
2628  * align the READ data correctly.  This means that the GETATTR
2629  * may end up partially falling into the page cache, and we should
2630  * shift it into the 'tail' of the xdr_buf before processing.
2631  * To do this efficiently, we need to know the total length
2632  * of data received, which doesn't seem to be available outside
2633  * of the RPC layer.
2634  *
2635  * In the case of WRITE, we also want to put the GETATTR after
2636  * the operation -- in this case because we want to make sure
2637  * we get the post-operation mtime and size.  This means that
2638  * we can't use xdr_encode_pages() as written: we need a variant
2639  * of it which would leave room in the 'tail' iovec.
2640  *
2641  * Both of these changes to the XDR layer would in fact be quite
2642  * minor, but I decided to leave them for a subsequent patch.
2643  */
2644 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2645                 unsigned int pgbase, unsigned int pglen)
2646 {
2647         struct nfs4_readlink args = {
2648                 .fh       = NFS_FH(inode),
2649                 .pgbase   = pgbase,
2650                 .pglen    = pglen,
2651                 .pages    = &page,
2652         };
2653         struct nfs4_readlink_res res;
2654         struct rpc_message msg = {
2655                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2656                 .rpc_argp = &args,
2657                 .rpc_resp = &res,
2658         };
2659
2660         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2661 }
2662
2663 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2664                 unsigned int pgbase, unsigned int pglen)
2665 {
2666         struct nfs4_exception exception = { };
2667         int err;
2668         do {
2669                 err = nfs4_handle_exception(NFS_SERVER(inode),
2670                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2671                                 &exception);
2672         } while (exception.retry);
2673         return err;
2674 }
2675
2676 /*
2677  * Got race?
2678  * We will need to arrange for the VFS layer to provide an atomic open.
2679  * Until then, this create/open method is prone to inefficiency and race
2680  * conditions due to the lookup, create, and open VFS calls from sys_open()
2681  * placed on the wire.
2682  *
2683  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2684  * The file will be opened again in the subsequent VFS open call
2685  * (nfs4_proc_file_open).
2686  *
2687  * The open for read will just hang around to be used by any process that
2688  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2689  */
2690
2691 static int
2692 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2693                  int flags, struct nfs_open_context *ctx)
2694 {
2695         struct dentry *de = dentry;
2696         struct nfs4_state *state;
2697         struct rpc_cred *cred = NULL;
2698         fmode_t fmode = 0;
2699         int status = 0;
2700
2701         if (ctx != NULL) {
2702                 cred = ctx->cred;
2703                 de = ctx->dentry;
2704                 fmode = ctx->mode;
2705         }
2706         sattr->ia_mode &= ~current_umask();
2707         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2708         d_drop(dentry);
2709         if (IS_ERR(state)) {
2710                 status = PTR_ERR(state);
2711                 goto out;
2712         }
2713         d_add(dentry, igrab(state->inode));
2714         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2715         if (ctx != NULL)
2716                 ctx->state = state;
2717         else
2718                 nfs4_close_sync(state, fmode);
2719 out:
2720         return status;
2721 }
2722
2723 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2724 {
2725         struct nfs_server *server = NFS_SERVER(dir);
2726         struct nfs_removeargs args = {
2727                 .fh = NFS_FH(dir),
2728                 .name.len = name->len,
2729                 .name.name = name->name,
2730                 .bitmask = server->attr_bitmask,
2731         };
2732         struct nfs_removeres res = {
2733                 .server = server,
2734         };
2735         struct rpc_message msg = {
2736                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2737                 .rpc_argp = &args,
2738                 .rpc_resp = &res,
2739         };
2740         int status = -ENOMEM;
2741
2742         res.dir_attr = nfs_alloc_fattr();
2743         if (res.dir_attr == NULL)
2744                 goto out;
2745
2746         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2747         if (status == 0) {
2748                 update_changeattr(dir, &res.cinfo);
2749                 nfs_post_op_update_inode(dir, res.dir_attr);
2750         }
2751         nfs_free_fattr(res.dir_attr);
2752 out:
2753         return status;
2754 }
2755
2756 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2757 {
2758         struct nfs4_exception exception = { };
2759         int err;
2760         do {
2761                 err = nfs4_handle_exception(NFS_SERVER(dir),
2762                                 _nfs4_proc_remove(dir, name),
2763                                 &exception);
2764         } while (exception.retry);
2765         return err;
2766 }
2767
2768 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2769 {
2770         struct nfs_server *server = NFS_SERVER(dir);
2771         struct nfs_removeargs *args = msg->rpc_argp;
2772         struct nfs_removeres *res = msg->rpc_resp;
2773
2774         args->bitmask = server->cache_consistency_bitmask;
2775         res->server = server;
2776         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2777         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2778 }
2779
2780 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2781 {
2782         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2783
2784         if (!nfs4_sequence_done(task, &res->seq_res))
2785                 return 0;
2786         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2787                 return 0;
2788         update_changeattr(dir, &res->cinfo);
2789         nfs_post_op_update_inode(dir, res->dir_attr);
2790         return 1;
2791 }
2792
2793 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2794 {
2795         struct nfs_server *server = NFS_SERVER(dir);
2796         struct nfs_renameargs *arg = msg->rpc_argp;
2797         struct nfs_renameres *res = msg->rpc_resp;
2798
2799         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2800         arg->bitmask = server->attr_bitmask;
2801         res->server = server;
2802         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2803 }
2804
2805 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2806                                  struct inode *new_dir)
2807 {
2808         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2809
2810         if (!nfs4_sequence_done(task, &res->seq_res))
2811                 return 0;
2812         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2813                 return 0;
2814
2815         update_changeattr(old_dir, &res->old_cinfo);
2816         nfs_post_op_update_inode(old_dir, res->old_fattr);
2817         update_changeattr(new_dir, &res->new_cinfo);
2818         nfs_post_op_update_inode(new_dir, res->new_fattr);
2819         return 1;
2820 }
2821
2822 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2823                 struct inode *new_dir, struct qstr *new_name)
2824 {
2825         struct nfs_server *server = NFS_SERVER(old_dir);
2826         struct nfs_renameargs arg = {
2827                 .old_dir = NFS_FH(old_dir),
2828                 .new_dir = NFS_FH(new_dir),
2829                 .old_name = old_name,
2830                 .new_name = new_name,
2831                 .bitmask = server->attr_bitmask,
2832         };
2833         struct nfs_renameres res = {
2834                 .server = server,
2835         };
2836         struct rpc_message msg = {
2837                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2838                 .rpc_argp = &arg,
2839                 .rpc_resp = &res,
2840         };
2841         int status = -ENOMEM;
2842
2843         res.old_fattr = nfs_alloc_fattr();
2844         res.new_fattr = nfs_alloc_fattr();
2845         if (res.old_fattr == NULL || res.new_fattr == NULL)
2846                 goto out;
2847
2848         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2849         if (!status) {
2850                 update_changeattr(old_dir, &res.old_cinfo);
2851                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2852                 update_changeattr(new_dir, &res.new_cinfo);
2853                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2854         }
2855 out:
2856         nfs_free_fattr(res.new_fattr);
2857         nfs_free_fattr(res.old_fattr);
2858         return status;
2859 }
2860
2861 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2862                 struct inode *new_dir, struct qstr *new_name)
2863 {
2864         struct nfs4_exception exception = { };
2865         int err;
2866         do {
2867                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2868                                 _nfs4_proc_rename(old_dir, old_name,
2869                                         new_dir, new_name),
2870                                 &exception);
2871         } while (exception.retry);
2872         return err;
2873 }
2874
2875 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2876 {
2877         struct nfs_server *server = NFS_SERVER(inode);
2878         struct nfs4_link_arg arg = {
2879                 .fh     = NFS_FH(inode),
2880                 .dir_fh = NFS_FH(dir),
2881                 .name   = name,
2882                 .bitmask = server->attr_bitmask,
2883         };
2884         struct nfs4_link_res res = {
2885                 .server = server,
2886         };
2887         struct rpc_message msg = {
2888                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2889                 .rpc_argp = &arg,
2890                 .rpc_resp = &res,
2891         };
2892         int status = -ENOMEM;
2893
2894         res.fattr = nfs_alloc_fattr();
2895         res.dir_attr = nfs_alloc_fattr();
2896         if (res.fattr == NULL || res.dir_attr == NULL)
2897                 goto out;
2898
2899         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2900         if (!status) {
2901                 update_changeattr(dir, &res.cinfo);
2902                 nfs_post_op_update_inode(dir, res.dir_attr);
2903                 nfs_post_op_update_inode(inode, res.fattr);
2904         }
2905 out:
2906         nfs_free_fattr(res.dir_attr);
2907         nfs_free_fattr(res.fattr);
2908         return status;
2909 }
2910
2911 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2912 {
2913         struct nfs4_exception exception = { };
2914         int err;
2915         do {
2916                 err = nfs4_handle_exception(NFS_SERVER(inode),
2917                                 _nfs4_proc_link(inode, dir, name),
2918                                 &exception);
2919         } while (exception.retry);
2920         return err;
2921 }
2922
2923 struct nfs4_createdata {
2924         struct rpc_message msg;
2925         struct nfs4_create_arg arg;
2926         struct nfs4_create_res res;
2927         struct nfs_fh fh;
2928         struct nfs_fattr fattr;
2929         struct nfs_fattr dir_fattr;
2930 };
2931
2932 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2933                 struct qstr *name, struct iattr *sattr, u32 ftype)
2934 {
2935         struct nfs4_createdata *data;
2936
2937         data = kzalloc(sizeof(*data), GFP_KERNEL);
2938         if (data != NULL) {
2939                 struct nfs_server *server = NFS_SERVER(dir);
2940
2941                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2942                 data->msg.rpc_argp = &data->arg;
2943                 data->msg.rpc_resp = &data->res;
2944                 data->arg.dir_fh = NFS_FH(dir);
2945                 data->arg.server = server;
2946                 data->arg.name = name;
2947                 data->arg.attrs = sattr;
2948                 data->arg.ftype = ftype;
2949                 data->arg.bitmask = server->attr_bitmask;
2950                 data->res.server = server;
2951                 data->res.fh = &data->fh;
2952                 data->res.fattr = &data->fattr;
2953                 data->res.dir_fattr = &data->dir_fattr;
2954                 nfs_fattr_init(data->res.fattr);
2955                 nfs_fattr_init(data->res.dir_fattr);
2956         }
2957         return data;
2958 }
2959
2960 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2961 {
2962         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2963                                     &data->arg.seq_args, &data->res.seq_res, 1);
2964         if (status == 0) {
2965                 update_changeattr(dir, &data->res.dir_cinfo);
2966                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2967                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2968         }
2969         return status;
2970 }
2971
2972 static void nfs4_free_createdata(struct nfs4_createdata *data)
2973 {
2974         kfree(data);
2975 }
2976
2977 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2978                 struct page *page, unsigned int len, struct iattr *sattr)
2979 {
2980         struct nfs4_createdata *data;
2981         int status = -ENAMETOOLONG;
2982
2983         if (len > NFS4_MAXPATHLEN)
2984                 goto out;
2985
2986         status = -ENOMEM;
2987         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2988         if (data == NULL)
2989                 goto out;
2990
2991         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2992         data->arg.u.symlink.pages = &page;
2993         data->arg.u.symlink.len = len;
2994
2995         status = nfs4_do_create(dir, dentry, data);
2996
2997         nfs4_free_createdata(data);
2998 out:
2999         return status;
3000 }
3001
3002 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3003                 struct page *page, unsigned int len, struct iattr *sattr)
3004 {
3005         struct nfs4_exception exception = { };
3006         int err;
3007         do {
3008                 err = nfs4_handle_exception(NFS_SERVER(dir),
3009                                 _nfs4_proc_symlink(dir, dentry, page,
3010                                                         len, sattr),
3011                                 &exception);
3012         } while (exception.retry);
3013         return err;
3014 }
3015
3016 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3017                 struct iattr *sattr)
3018 {
3019         struct nfs4_createdata *data;
3020         int status = -ENOMEM;
3021
3022         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3023         if (data == NULL)
3024                 goto out;
3025
3026         status = nfs4_do_create(dir, dentry, data);
3027
3028         nfs4_free_createdata(data);
3029 out:
3030         return status;
3031 }
3032
3033 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3034                 struct iattr *sattr)
3035 {
3036         struct nfs4_exception exception = { };
3037         int err;
3038
3039         sattr->ia_mode &= ~current_umask();
3040         do {
3041                 err = nfs4_handle_exception(NFS_SERVER(dir),
3042                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3043                                 &exception);
3044         } while (exception.retry);
3045         return err;
3046 }
3047
3048 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3049                 u64 cookie, struct page **pages, unsigned int count, int plus)
3050 {
3051         struct inode            *dir = dentry->d_inode;
3052         struct nfs4_readdir_arg args = {
3053                 .fh = NFS_FH(dir),
3054                 .pages = pages,
3055                 .pgbase = 0,
3056                 .count = count,
3057                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3058                 .plus = plus,
3059         };
3060         struct nfs4_readdir_res res;
3061         struct rpc_message msg = {
3062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3063                 .rpc_argp = &args,
3064                 .rpc_resp = &res,
3065                 .rpc_cred = cred,
3066         };
3067         int                     status;
3068
3069         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3070                         dentry->d_parent->d_name.name,
3071                         dentry->d_name.name,
3072                         (unsigned long long)cookie);
3073         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3074         res.pgbase = args.pgbase;
3075         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3076         if (status >= 0) {
3077                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3078                 status += args.pgbase;
3079         }
3080
3081         nfs_invalidate_atime(dir);
3082
3083         dprintk("%s: returns %d\n", __func__, status);
3084         return status;
3085 }
3086
3087 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3088                 u64 cookie, struct page **pages, unsigned int count, int plus)
3089 {
3090         struct nfs4_exception exception = { };
3091         int err;
3092         do {
3093                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3094                                 _nfs4_proc_readdir(dentry, cred, cookie,
3095                                         pages, count, plus),
3096                                 &exception);
3097         } while (exception.retry);
3098         return err;
3099 }
3100
3101 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3102                 struct iattr *sattr, dev_t rdev)
3103 {
3104         struct nfs4_createdata *data;
3105         int mode = sattr->ia_mode;
3106         int status = -ENOMEM;
3107
3108         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3109         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3110
3111         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3112         if (data == NULL)
3113                 goto out;
3114
3115         if (S_ISFIFO(mode))
3116                 data->arg.ftype = NF4FIFO;
3117         else if (S_ISBLK(mode)) {
3118                 data->arg.ftype = NF4BLK;
3119                 data->arg.u.device.specdata1 = MAJOR(rdev);
3120                 data->arg.u.device.specdata2 = MINOR(rdev);
3121         }
3122         else if (S_ISCHR(mode)) {
3123                 data->arg.ftype = NF4CHR;
3124                 data->arg.u.device.specdata1 = MAJOR(rdev);
3125                 data->arg.u.device.specdata2 = MINOR(rdev);
3126         }
3127
3128         status = nfs4_do_create(dir, dentry, data);
3129
3130         nfs4_free_createdata(data);
3131 out:
3132         return status;
3133 }
3134
3135 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3136                 struct iattr *sattr, dev_t rdev)
3137 {
3138         struct nfs4_exception exception = { };
3139         int err;
3140
3141         sattr->ia_mode &= ~current_umask();
3142         do {
3143                 err = nfs4_handle_exception(NFS_SERVER(dir),
3144                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3145                                 &exception);
3146         } while (exception.retry);
3147         return err;
3148 }
3149
3150 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3151                  struct nfs_fsstat *fsstat)
3152 {
3153         struct nfs4_statfs_arg args = {
3154                 .fh = fhandle,
3155                 .bitmask = server->attr_bitmask,
3156         };
3157         struct nfs4_statfs_res res = {
3158                 .fsstat = fsstat,
3159         };
3160         struct rpc_message msg = {
3161                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3162                 .rpc_argp = &args,
3163                 .rpc_resp = &res,
3164         };
3165
3166         nfs_fattr_init(fsstat->fattr);
3167         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3168 }
3169
3170 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3171 {
3172         struct nfs4_exception exception = { };
3173         int err;
3174         do {
3175                 err = nfs4_handle_exception(server,
3176                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3177                                 &exception);
3178         } while (exception.retry);
3179         return err;
3180 }
3181
3182 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3183                 struct nfs_fsinfo *fsinfo)
3184 {
3185         struct nfs4_fsinfo_arg args = {
3186                 .fh = fhandle,
3187                 .bitmask = server->attr_bitmask,
3188         };
3189         struct nfs4_fsinfo_res res = {
3190                 .fsinfo = fsinfo,
3191         };
3192         struct rpc_message msg = {
3193                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3194                 .rpc_argp = &args,
3195                 .rpc_resp = &res,
3196         };
3197
3198         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3199 }
3200
3201 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3202 {
3203         struct nfs4_exception exception = { };
3204         int err;
3205
3206         do {
3207                 err = nfs4_handle_exception(server,
3208                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3209                                 &exception);
3210         } while (exception.retry);
3211         return err;
3212 }
3213
3214 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3215 {
3216         nfs_fattr_init(fsinfo->fattr);
3217         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3218 }
3219
3220 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3221                 struct nfs_pathconf *pathconf)
3222 {
3223         struct nfs4_pathconf_arg args = {
3224                 .fh = fhandle,
3225                 .bitmask = server->attr_bitmask,
3226         };
3227         struct nfs4_pathconf_res res = {
3228                 .pathconf = pathconf,
3229         };
3230         struct rpc_message msg = {
3231                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3232                 .rpc_argp = &args,
3233                 .rpc_resp = &res,
3234         };
3235
3236         /* None of the pathconf attributes are mandatory to implement */
3237         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3238                 memset(pathconf, 0, sizeof(*pathconf));
3239                 return 0;
3240         }
3241
3242         nfs_fattr_init(pathconf->fattr);
3243         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3244 }
3245
3246 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3247                 struct nfs_pathconf *pathconf)
3248 {
3249         struct nfs4_exception exception = { };
3250         int err;
3251
3252         do {
3253                 err = nfs4_handle_exception(server,
3254                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3255                                 &exception);
3256         } while (exception.retry);
3257         return err;
3258 }
3259
3260 void __nfs4_read_done_cb(struct nfs_read_data *data)
3261 {
3262         nfs_invalidate_atime(data->inode);
3263 }
3264
3265 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3266 {
3267         struct nfs_server *server = NFS_SERVER(data->inode);
3268
3269         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3270                 rpc_restart_call_prepare(task);
3271                 return -EAGAIN;
3272         }
3273
3274         __nfs4_read_done_cb(data);
3275         if (task->tk_status > 0)
3276                 renew_lease(server, data->timestamp);
3277         return 0;
3278 }
3279
3280 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3281 {
3282
3283         dprintk("--> %s\n", __func__);
3284
3285         if (!nfs4_sequence_done(task, &data->res.seq_res))
3286                 return -EAGAIN;
3287
3288         return data->read_done_cb ? data->read_done_cb(task, data) :
3289                                     nfs4_read_done_cb(task, data);
3290 }
3291
3292 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3293 {
3294         data->timestamp   = jiffies;
3295         data->read_done_cb = nfs4_read_done_cb;
3296         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3297         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3298 }
3299
3300 /* Reset the the nfs_read_data to send the read to the MDS. */
3301 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3302 {
3303         dprintk("%s Reset task for i/o through\n", __func__);
3304         put_lseg(data->lseg);
3305         data->lseg = NULL;
3306         /* offsets will differ in the dense stripe case */
3307         data->args.offset = data->mds_offset;
3308         data->ds_clp = NULL;
3309         data->args.fh     = NFS_FH(data->inode);
3310         data->read_done_cb = nfs4_read_done_cb;
3311         task->tk_ops = data->mds_ops;
3312         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3313 }
3314 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3315
3316 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3317 {
3318         struct inode *inode = data->inode;
3319
3320         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3321                 rpc_restart_call_prepare(task);
3322                 return -EAGAIN;
3323         }
3324         if (task->tk_status >= 0) {
3325                 renew_lease(NFS_SERVER(inode), data->timestamp);
3326                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3327         }
3328         return 0;
3329 }
3330
3331 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3332 {
3333         if (!nfs4_sequence_done(task, &data->res.seq_res))
3334                 return -EAGAIN;
3335         return data->write_done_cb ? data->write_done_cb(task, data) :
3336                 nfs4_write_done_cb(task, data);
3337 }
3338
3339 /* Reset the the nfs_write_data to send the write to the MDS. */
3340 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3341 {
3342         dprintk("%s Reset task for i/o through\n", __func__);
3343         put_lseg(data->lseg);
3344         data->lseg          = NULL;
3345         data->ds_clp        = NULL;
3346         data->write_done_cb = nfs4_write_done_cb;
3347         data->args.fh       = NFS_FH(data->inode);
3348         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3349         data->args.offset   = data->mds_offset;
3350         data->res.fattr     = &data->fattr;
3351         task->tk_ops        = data->mds_ops;
3352         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3353 }
3354 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3355
3356 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3357 {
3358         struct nfs_server *server = NFS_SERVER(data->inode);
3359
3360         if (data->lseg) {
3361                 data->args.bitmask = NULL;
3362                 data->res.fattr = NULL;
3363         } else
3364                 data->args.bitmask = server->cache_consistency_bitmask;
3365         if (!data->write_done_cb)
3366                 data->write_done_cb = nfs4_write_done_cb;
3367         data->res.server = server;
3368         data->timestamp   = jiffies;
3369
3370         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3371         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3372 }
3373
3374 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3375 {
3376         struct inode *inode = data->inode;
3377
3378         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3379                 rpc_restart_call_prepare(task);
3380                 return -EAGAIN;
3381         }
3382         nfs_refresh_inode(inode, data->res.fattr);
3383         return 0;
3384 }
3385
3386 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3387 {
3388         if (!nfs4_sequence_done(task, &data->res.seq_res))
3389                 return -EAGAIN;
3390         return data->write_done_cb(task, data);
3391 }
3392
3393 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3394 {
3395         struct nfs_server *server = NFS_SERVER(data->inode);
3396
3397         if (data->lseg) {
3398                 data->args.bitmask = NULL;
3399                 data->res.fattr = NULL;
3400         } else
3401                 data->args.bitmask = server->cache_consistency_bitmask;
3402         if (!data->write_done_cb)
3403                 data->write_done_cb = nfs4_commit_done_cb;
3404         data->res.server = server;
3405         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3406         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3407 }
3408
3409 struct nfs4_renewdata {
3410         struct nfs_client       *client;
3411         unsigned long           timestamp;
3412 };
3413
3414 /*
3415  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3416  * standalone procedure for queueing an asynchronous RENEW.
3417  */
3418 static void nfs4_renew_release(void *calldata)
3419 {
3420         struct nfs4_renewdata *data = calldata;
3421         struct nfs_client *clp = data->client;
3422
3423         if (atomic_read(&clp->cl_count) > 1)
3424                 nfs4_schedule_state_renewal(clp);
3425         nfs_put_client(clp);
3426         kfree(data);
3427 }
3428
3429 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3430 {
3431         struct nfs4_renewdata *data = calldata;
3432         struct nfs_client *clp = data->client;
3433         unsigned long timestamp = data->timestamp;
3434
3435         if (task->tk_status < 0) {
3436                 /* Unless we're shutting down, schedule state recovery! */
3437                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3438                         return;
3439                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3440                         nfs4_schedule_lease_recovery(clp);
3441                         return;
3442                 }
3443                 nfs4_schedule_path_down_recovery(clp);
3444         }
3445         do_renew_lease(clp, timestamp);
3446 }
3447
3448 static const struct rpc_call_ops nfs4_renew_ops = {
3449         .rpc_call_done = nfs4_renew_done,
3450         .rpc_release = nfs4_renew_release,
3451 };
3452
3453 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3454 {
3455         struct rpc_message msg = {
3456                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3457                 .rpc_argp       = clp,
3458                 .rpc_cred       = cred,
3459         };
3460         struct nfs4_renewdata *data;
3461
3462         if (renew_flags == 0)
3463                 return 0;
3464         if (!atomic_inc_not_zero(&clp->cl_count))
3465                 return -EIO;
3466         data = kmalloc(sizeof(*data), GFP_NOFS);
3467         if (data == NULL)
3468                 return -ENOMEM;
3469         data->client = clp;
3470         data->timestamp = jiffies;
3471         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3472                         &nfs4_renew_ops, data);
3473 }
3474
3475 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3476 {
3477         struct rpc_message msg = {
3478                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3479                 .rpc_argp       = clp,
3480                 .rpc_cred       = cred,
3481         };
3482         unsigned long now = jiffies;
3483         int status;
3484
3485         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3486         if (status < 0)
3487                 return status;
3488         do_renew_lease(clp, now);
3489         return 0;
3490 }
3491
3492 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3493 {
3494         return (server->caps & NFS_CAP_ACLS)
3495                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3496                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3497 }
3498
3499 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3500  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3501  * the stack.
3502  */
3503 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3504
3505 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3506                 struct page **pages, unsigned int *pgbase)
3507 {
3508         struct page *newpage, **spages;
3509         int rc = 0;
3510         size_t len;
3511         spages = pages;
3512
3513         do {
3514                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3515                 newpage = alloc_page(GFP_KERNEL);
3516
3517                 if (newpage == NULL)
3518                         goto unwind;
3519                 memcpy(page_address(newpage), buf, len);
3520                 buf += len;
3521                 buflen -= len;
3522                 *pages++ = newpage;
3523                 rc++;
3524         } while (buflen != 0);
3525
3526         return rc;
3527
3528 unwind:
3529         for(; rc > 0; rc--)
3530                 __free_page(spages[rc-1]);
3531         return -ENOMEM;
3532 }
3533
3534 struct nfs4_cached_acl {
3535         int cached;
3536         size_t len;
3537         char data[0];
3538 };
3539
3540 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3541 {
3542         struct nfs_inode *nfsi = NFS_I(inode);
3543
3544         spin_lock(&inode->i_lock);
3545         kfree(nfsi->nfs4_acl);
3546         nfsi->nfs4_acl = acl;
3547         spin_unlock(&inode->i_lock);
3548 }
3549
3550 static void nfs4_zap_acl_attr(struct inode *inode)
3551 {
3552         nfs4_set_cached_acl(inode, NULL);
3553 }
3554
3555 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3556 {
3557         struct nfs_inode *nfsi = NFS_I(inode);
3558         struct nfs4_cached_acl *acl;
3559         int ret = -ENOENT;
3560
3561         spin_lock(&inode->i_lock);
3562         acl = nfsi->nfs4_acl;
3563         if (acl == NULL)
3564                 goto out;
3565         if (buf == NULL) /* user is just asking for length */
3566                 goto out_len;
3567         if (acl->cached == 0)
3568                 goto out;
3569         ret = -ERANGE; /* see getxattr(2) man page */
3570         if (acl->len > buflen)
3571                 goto out;
3572         memcpy(buf, acl->data, acl->len);
3573 out_len:
3574         ret = acl->len;
3575 out:
3576         spin_unlock(&inode->i_lock);
3577         return ret;
3578 }
3579
3580 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3581 {
3582         struct nfs4_cached_acl *acl;
3583
3584         if (buf && acl_len <= PAGE_SIZE) {
3585                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3586                 if (acl == NULL)
3587                         goto out;
3588                 acl->cached = 1;
3589                 memcpy(acl->data, buf, acl_len);
3590         } else {
3591                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3592                 if (acl == NULL)
3593                         goto out;
3594                 acl->cached = 0;
3595         }
3596         acl->len = acl_len;
3597 out:
3598         nfs4_set_cached_acl(inode, acl);
3599 }
3600
3601 /*
3602  * The getxattr API returns the required buffer length when called with a
3603  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3604  * the required buf.  On a NULL buf, we send a page of data to the server
3605  * guessing that the ACL request can be serviced by a page. If so, we cache
3606  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3607  * the cache. If not so, we throw away the page, and cache the required
3608  * length. The next getxattr call will then produce another round trip to
3609  * the server, this time with the input buf of the required size.
3610  */
3611 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3612 {
3613         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3614         struct nfs_getaclargs args = {
3615                 .fh = NFS_FH(inode),
3616                 .acl_pages = pages,
3617                 .acl_len = buflen,
3618         };
3619         struct nfs_getaclres res = {
3620                 .acl_len = buflen,
3621         };
3622         void *resp_buf;
3623         struct rpc_message msg = {
3624                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3625                 .rpc_argp = &args,
3626                 .rpc_resp = &res,
3627         };
3628         int ret = -ENOMEM, npages, i, acl_len = 0;
3629
3630         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3631         /* As long as we're doing a round trip to the server anyway,
3632          * let's be prepared for a page of acl data. */
3633         if (npages == 0)
3634                 npages = 1;
3635
3636         for (i = 0; i < npages; i++) {
3637                 pages[i] = alloc_page(GFP_KERNEL);
3638                 if (!pages[i])
3639                         goto out_free;
3640         }
3641         if (npages > 1) {
3642                 /* for decoding across pages */
3643                 res.acl_scratch = alloc_page(GFP_KERNEL);
3644                 if (!res.acl_scratch)
3645                         goto out_free;
3646         }
3647         args.acl_len = npages * PAGE_SIZE;
3648         args.acl_pgbase = 0;
3649         /* Let decode_getfacl know not to fail if the ACL data is larger than
3650          * the page we send as a guess */
3651         if (buf == NULL)
3652                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3653         resp_buf = page_address(pages[0]);
3654
3655         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3656                 __func__, buf, buflen, npages, args.acl_len);
3657         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3658                              &msg, &args.seq_args, &res.seq_res, 0);
3659         if (ret)
3660                 goto out_free;
3661
3662         acl_len = res.acl_len - res.acl_data_offset;
3663         if (acl_len > args.acl_len)
3664                 nfs4_write_cached_acl(inode, NULL, acl_len);
3665         else
3666                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3667                                       acl_len);
3668         if (buf) {
3669                 ret = -ERANGE;
3670                 if (acl_len > buflen)
3671                         goto out_free;
3672                 _copy_from_pages(buf, pages, res.acl_data_offset,
3673                                 res.acl_len);
3674         }
3675         ret = acl_len;
3676 out_free:
3677         for (i = 0; i < npages; i++)
3678                 if (pages[i])
3679                         __free_page(pages[i]);
3680         if (res.acl_scratch)
3681                 __free_page(res.acl_scratch);
3682         return ret;
3683 }
3684
3685 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3686 {
3687         struct nfs4_exception exception = { };
3688         ssize_t ret;
3689         do {
3690                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3691                 if (ret >= 0)
3692                         break;
3693                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3694         } while (exception.retry);
3695         return ret;
3696 }
3697
3698 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3699 {
3700         struct nfs_server *server = NFS_SERVER(inode);
3701         int ret;
3702
3703         if (!nfs4_server_supports_acls(server))
3704                 return -EOPNOTSUPP;
3705         ret = nfs_revalidate_inode(server, inode);
3706         if (ret < 0)
3707                 return ret;
3708         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3709                 nfs_zap_acl_cache(inode);
3710         ret = nfs4_read_cached_acl(inode, buf, buflen);
3711         if (ret != -ENOENT)
3712                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3713                  * but no cached acl data, just the acl length */
3714                 return ret;
3715         return nfs4_get_acl_uncached(inode, buf, buflen);
3716 }
3717
3718 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3719 {
3720         struct nfs_server *server = NFS_SERVER(inode);
3721         struct page *pages[NFS4ACL_MAXPAGES];
3722         struct nfs_setaclargs arg = {
3723                 .fh             = NFS_FH(inode),
3724                 .acl_pages      = pages,
3725                 .acl_len        = buflen,
3726         };
3727         struct nfs_setaclres res;
3728         struct rpc_message msg = {
3729                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3730                 .rpc_argp       = &arg,
3731                 .rpc_resp       = &res,
3732         };
3733         int ret, i;
3734
3735         if (!nfs4_server_supports_acls(server))
3736                 return -EOPNOTSUPP;
3737         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3738         if (i < 0)
3739                 return i;
3740         nfs_inode_return_delegation(inode);
3741         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3742
3743         /*
3744          * Free each page after tx, so the only ref left is
3745          * held by the network stack
3746          */
3747         for (; i > 0; i--)
3748                 put_page(pages[i-1]);
3749
3750         /*
3751          * Acl update can result in inode attribute update.
3752          * so mark the attribute cache invalid.
3753          */
3754         spin_lock(&inode->i_lock);
3755         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3756         spin_unlock(&inode->i_lock);
3757         nfs_access_zap_cache(inode);
3758         nfs_zap_acl_cache(inode);
3759         return ret;
3760 }
3761
3762 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3763 {
3764         struct nfs4_exception exception = { };
3765         int err;
3766         do {
3767                 err = nfs4_handle_exception(NFS_SERVER(inode),
3768                                 __nfs4_proc_set_acl(inode, buf, buflen),
3769                                 &exception);
3770         } while (exception.retry);
3771         return err;
3772 }
3773
3774 static int
3775 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3776 {
3777         struct nfs_client *clp = server->nfs_client;
3778
3779         if (task->tk_status >= 0)
3780                 return 0;
3781         switch(task->tk_status) {
3782                 case -NFS4ERR_DELEG_REVOKED:
3783                 case -NFS4ERR_ADMIN_REVOKED:
3784                 case -NFS4ERR_BAD_STATEID:
3785                         if (state != NULL)
3786                                 nfs_remove_bad_delegation(state->inode);
3787                 case -NFS4ERR_OPENMODE:
3788                         if (state == NULL)
3789                                 break;
3790                         nfs4_schedule_stateid_recovery(server, state);
3791                         goto wait_on_recovery;
3792                 case -NFS4ERR_EXPIRED:
3793                         if (state != NULL)
3794                                 nfs4_schedule_stateid_recovery(server, state);
3795                 case -NFS4ERR_STALE_STATEID:
3796                 case -NFS4ERR_STALE_CLIENTID:
3797                         nfs4_schedule_lease_recovery(clp);
3798                         goto wait_on_recovery;
3799 #if defined(CONFIG_NFS_V4_1)
3800                 case -NFS4ERR_BADSESSION:
3801                 case -NFS4ERR_BADSLOT:
3802                 case -NFS4ERR_BAD_HIGH_SLOT:
3803                 case -NFS4ERR_DEADSESSION:
3804                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3805                 case -NFS4ERR_SEQ_FALSE_RETRY:
3806                 case -NFS4ERR_SEQ_MISORDERED:
3807                         dprintk("%s ERROR %d, Reset session\n", __func__,
3808                                 task->tk_status);
3809                         nfs4_schedule_session_recovery(clp->cl_session);
3810                         task->tk_status = 0;
3811                         return -EAGAIN;
3812 #endif /* CONFIG_NFS_V4_1 */
3813                 case -NFS4ERR_DELAY:
3814                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3815                 case -NFS4ERR_GRACE:
3816                 case -EKEYEXPIRED:
3817                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3818                         task->tk_status = 0;
3819                         return -EAGAIN;
3820                 case -NFS4ERR_RETRY_UNCACHED_REP:
3821                 case -NFS4ERR_OLD_STATEID:
3822                         task->tk_status = 0;
3823                         return -EAGAIN;
3824         }
3825         task->tk_status = nfs4_map_errors(task->tk_status);
3826         return 0;
3827 wait_on_recovery:
3828         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3829         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3830                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3831         task->tk_status = 0;
3832         return -EAGAIN;
3833 }
3834
3835 static void nfs4_construct_boot_verifier(struct nfs_client *clp,
3836                                          nfs4_verifier *bootverf)
3837 {
3838         __be32 verf[2];
3839
3840         verf[0] = htonl((u32)clp->cl_boot_time.tv_sec);
3841         verf[1] = htonl((u32)clp->cl_boot_time.tv_nsec);
3842         memcpy(bootverf->data, verf, sizeof(bootverf->data));
3843 }
3844
3845 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3846                 unsigned short port, struct rpc_cred *cred,
3847                 struct nfs4_setclientid_res *res)
3848 {
3849         nfs4_verifier sc_verifier;
3850         struct nfs4_setclientid setclientid = {
3851                 .sc_verifier = &sc_verifier,
3852                 .sc_prog = program,
3853                 .sc_cb_ident = clp->cl_cb_ident,
3854         };
3855         struct rpc_message msg = {
3856                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3857                 .rpc_argp = &setclientid,
3858                 .rpc_resp = res,
3859                 .rpc_cred = cred,
3860         };
3861         int loop = 0;
3862         int status;
3863
3864         nfs4_construct_boot_verifier(clp, &sc_verifier);
3865
3866         for(;;) {
3867                 rcu_read_lock();
3868                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3869                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3870                                 clp->cl_ipaddr,
3871                                 rpc_peeraddr2str(clp->cl_rpcclient,
3872                                                         RPC_DISPLAY_ADDR),
3873                                 rpc_peeraddr2str(clp->cl_rpcclient,
3874                                                         RPC_DISPLAY_PROTO),
3875                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3876                                 clp->cl_id_uniquifier);
3877                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3878                                 sizeof(setclientid.sc_netid),
3879                                 rpc_peeraddr2str(clp->cl_rpcclient,
3880                                                         RPC_DISPLAY_NETID));
3881                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3882                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3883                                 clp->cl_ipaddr, port >> 8, port & 255);
3884                 rcu_read_unlock();
3885
3886                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3887                 if (status != -NFS4ERR_CLID_INUSE)
3888                         break;
3889                 if (loop != 0) {
3890                         ++clp->cl_id_uniquifier;
3891                         break;
3892                 }
3893                 ++loop;
3894                 ssleep(clp->cl_lease_time / HZ + 1);
3895         }
3896         return status;
3897 }
3898
3899 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3900                 struct nfs4_setclientid_res *arg,
3901                 struct rpc_cred *cred)
3902 {
3903         struct nfs_fsinfo fsinfo;
3904         struct rpc_message msg = {
3905                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3906                 .rpc_argp = arg,
3907                 .rpc_resp = &fsinfo,
3908                 .rpc_cred = cred,
3909         };
3910         unsigned long now;
3911         int status;
3912
3913         now = jiffies;
3914         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3915         if (status == 0) {
3916                 spin_lock(&clp->cl_lock);
3917                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3918                 clp->cl_last_renewal = now;
3919                 spin_unlock(&clp->cl_lock);
3920         }
3921         return status;
3922 }
3923
3924 struct nfs4_delegreturndata {
3925         struct nfs4_delegreturnargs args;
3926         struct nfs4_delegreturnres res;
3927         struct nfs_fh fh;
3928         nfs4_stateid stateid;
3929         unsigned long timestamp;
3930         struct nfs_fattr fattr;
3931         int rpc_status;
3932 };
3933
3934 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3935 {
3936         struct nfs4_delegreturndata *data = calldata;
3937
3938         if (!nfs4_sequence_done(task, &data->res.seq_res))
3939                 return;
3940
3941         switch (task->tk_status) {
3942         case -NFS4ERR_STALE_STATEID:
3943         case -NFS4ERR_EXPIRED:
3944         case 0:
3945                 renew_lease(data->res.server, data->timestamp);
3946                 break;
3947         default:
3948                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3949                                 -EAGAIN) {
3950                         rpc_restart_call_prepare(task);
3951                         return;
3952                 }
3953         }
3954         data->rpc_status = task->tk_status;
3955 }
3956
3957 static void nfs4_delegreturn_release(void *calldata)
3958 {
3959         kfree(calldata);
3960 }
3961
3962 #if defined(CONFIG_NFS_V4_1)
3963 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3964 {
3965         struct nfs4_delegreturndata *d_data;
3966
3967         d_data = (struct nfs4_delegreturndata *)data;
3968
3969         if (nfs4_setup_sequence(d_data->res.server,
3970                                 &d_data->args.seq_args,
3971                                 &d_data->res.seq_res, task))
3972                 return;
3973         rpc_call_start(task);
3974 }
3975 #endif /* CONFIG_NFS_V4_1 */
3976
3977 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3978 #if defined(CONFIG_NFS_V4_1)
3979         .rpc_call_prepare = nfs4_delegreturn_prepare,
3980 #endif /* CONFIG_NFS_V4_1 */
3981         .rpc_call_done = nfs4_delegreturn_done,
3982         .rpc_release = nfs4_delegreturn_release,
3983 };
3984
3985 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3986 {
3987         struct nfs4_delegreturndata *data;
3988         struct nfs_server *server = NFS_SERVER(inode);
3989         struct rpc_task *task;
3990         struct rpc_message msg = {
3991                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3992                 .rpc_cred = cred,
3993         };
3994         struct rpc_task_setup task_setup_data = {
3995                 .rpc_client = server->client,
3996                 .rpc_message = &msg,
3997                 .callback_ops = &nfs4_delegreturn_ops,
3998                 .flags = RPC_TASK_ASYNC,
3999         };
4000         int status = 0;
4001
4002         data = kzalloc(sizeof(*data), GFP_NOFS);
4003         if (data == NULL)
4004                 return -ENOMEM;
4005         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4006         data->args.fhandle = &data->fh;
4007         data->args.stateid = &data->stateid;
4008         data->args.bitmask = server->attr_bitmask;
4009         nfs_copy_fh(&data->fh, NFS_FH(inode));
4010         nfs4_stateid_copy(&data->stateid, stateid);
4011         data->res.fattr = &data->fattr;
4012         data->res.server = server;
4013         nfs_fattr_init(data->res.fattr);
4014         data->timestamp = jiffies;
4015         data->rpc_status = 0;
4016
4017         task_setup_data.callback_data = data;
4018         msg.rpc_argp = &data->args;
4019         msg.rpc_resp = &data->res;
4020         task = rpc_run_task(&task_setup_data);
4021         if (IS_ERR(task))
4022                 return PTR_ERR(task);
4023         if (!issync)
4024                 goto out;
4025         status = nfs4_wait_for_completion_rpc_task(task);
4026         if (status != 0)
4027                 goto out;
4028         status = data->rpc_status;
4029         if (status != 0)
4030                 goto out;
4031         nfs_refresh_inode(inode, &data->fattr);
4032 out:
4033         rpc_put_task(task);
4034         return status;
4035 }
4036
4037 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4038 {
4039         struct nfs_server *server = NFS_SERVER(inode);
4040         struct nfs4_exception exception = { };
4041         int err;
4042         do {
4043                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4044                 switch (err) {
4045                         case -NFS4ERR_STALE_STATEID:
4046                         case -NFS4ERR_EXPIRED:
4047                         case 0:
4048                                 return 0;
4049                 }
4050                 err = nfs4_handle_exception(server, err, &exception);
4051         } while (exception.retry);
4052         return err;
4053 }
4054
4055 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4056 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4057
4058 /*
4059  * sleep, with exponential backoff, and retry the LOCK operation.
4060  */
4061 static unsigned long
4062 nfs4_set_lock_task_retry(unsigned long timeout)
4063 {
4064         freezable_schedule_timeout_killable(timeout);
4065         timeout <<= 1;
4066         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4067                 return NFS4_LOCK_MAXTIMEOUT;
4068         return timeout;
4069 }
4070
4071 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4072 {
4073         struct inode *inode = state->inode;
4074         struct nfs_server *server = NFS_SERVER(inode);
4075         struct nfs_client *clp = server->nfs_client;
4076         struct nfs_lockt_args arg = {
4077                 .fh = NFS_FH(inode),
4078                 .fl = request,
4079         };
4080         struct nfs_lockt_res res = {
4081                 .denied = request,
4082         };
4083         struct rpc_message msg = {
4084                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4085                 .rpc_argp       = &arg,
4086                 .rpc_resp       = &res,
4087                 .rpc_cred       = state->owner->so_cred,
4088         };
4089         struct nfs4_lock_state *lsp;
4090         int status;
4091
4092         arg.lock_owner.clientid = clp->cl_clientid;
4093         status = nfs4_set_lock_state(state, request);
4094         if (status != 0)
4095                 goto out;
4096         lsp = request->fl_u.nfs4_fl.owner;
4097         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4098         arg.lock_owner.s_dev = server->s_dev;
4099         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4100         switch (status) {
4101                 case 0:
4102                         request->fl_type = F_UNLCK;
4103                         break;
4104                 case -NFS4ERR_DENIED:
4105                         status = 0;
4106         }
4107         request->fl_ops->fl_release_private(request);
4108 out:
4109         return status;
4110 }
4111
4112 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4113 {
4114         struct nfs4_exception exception = { };
4115         int err;
4116
4117         do {
4118                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4119                                 _nfs4_proc_getlk(state, cmd, request),
4120                                 &exception);
4121         } while (exception.retry);
4122         return err;
4123 }
4124
4125 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4126 {
4127         int res = 0;
4128         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4129                 case FL_POSIX:
4130                         res = posix_lock_file_wait(file, fl);
4131                         break;
4132                 case FL_FLOCK:
4133                         res = flock_lock_file_wait(file, fl);
4134                         break;
4135                 default:
4136                         BUG();
4137         }
4138         return res;
4139 }
4140
4141 struct nfs4_unlockdata {
4142         struct nfs_locku_args arg;
4143         struct nfs_locku_res res;
4144         struct nfs4_lock_state *lsp;
4145         struct nfs_open_context *ctx;
4146         struct file_lock fl;
4147         const struct nfs_server *server;
4148         unsigned long timestamp;
4149 };
4150
4151 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4152                 struct nfs_open_context *ctx,
4153                 struct nfs4_lock_state *lsp,
4154                 struct nfs_seqid *seqid)
4155 {
4156         struct nfs4_unlockdata *p;
4157         struct inode *inode = lsp->ls_state->inode;
4158
4159         p = kzalloc(sizeof(*p), GFP_NOFS);
4160         if (p == NULL)
4161                 return NULL;
4162         p->arg.fh = NFS_FH(inode);
4163         p->arg.fl = &p->fl;
4164         p->arg.seqid = seqid;
4165         p->res.seqid = seqid;
4166         p->arg.stateid = &lsp->ls_stateid;
4167         p->lsp = lsp;
4168         atomic_inc(&lsp->ls_count);
4169         /* Ensure we don't close file until we're done freeing locks! */
4170         p->ctx = get_nfs_open_context(ctx);
4171         memcpy(&p->fl, fl, sizeof(p->fl));
4172         p->server = NFS_SERVER(inode);
4173         return p;
4174 }
4175
4176 static void nfs4_locku_release_calldata(void *data)
4177 {
4178         struct nfs4_unlockdata *calldata = data;
4179         nfs_free_seqid(calldata->arg.seqid);
4180         nfs4_put_lock_state(calldata->lsp);
4181         put_nfs_open_context(calldata->ctx);
4182         kfree(calldata);
4183 }
4184
4185 static void nfs4_locku_done(struct rpc_task *task, void *data)
4186 {
4187         struct nfs4_unlockdata *calldata = data;
4188
4189         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4190                 return;
4191         switch (task->tk_status) {
4192                 case 0:
4193                         nfs4_stateid_copy(&calldata->lsp->ls_stateid,
4194                                         &calldata->res.stateid);
4195                         renew_lease(calldata->server, calldata->timestamp);
4196                         break;
4197                 case -NFS4ERR_BAD_STATEID:
4198                 case -NFS4ERR_OLD_STATEID:
4199                 case -NFS4ERR_STALE_STATEID:
4200                 case -NFS4ERR_EXPIRED:
4201                         break;
4202                 default:
4203                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4204                                 rpc_restart_call_prepare(task);
4205         }
4206 }
4207
4208 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4209 {
4210         struct nfs4_unlockdata *calldata = data;
4211
4212         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4213                 return;
4214         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4215                 /* Note: exit _without_ running nfs4_locku_done */
4216                 task->tk_action = NULL;
4217                 return;
4218         }
4219         calldata->timestamp = jiffies;
4220         if (nfs4_setup_sequence(calldata->server,
4221                                 &calldata->arg.seq_args,
4222                                 &calldata->res.seq_res, task))
4223                 return;
4224         rpc_call_start(task);
4225 }
4226
4227 static const struct rpc_call_ops nfs4_locku_ops = {
4228         .rpc_call_prepare = nfs4_locku_prepare,
4229         .rpc_call_done = nfs4_locku_done,
4230         .rpc_release = nfs4_locku_release_calldata,
4231 };
4232
4233 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4234                 struct nfs_open_context *ctx,
4235                 struct nfs4_lock_state *lsp,
4236                 struct nfs_seqid *seqid)
4237 {
4238         struct nfs4_unlockdata *data;
4239         struct rpc_message msg = {
4240                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4241                 .rpc_cred = ctx->cred,
4242         };
4243         struct rpc_task_setup task_setup_data = {
4244                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4245                 .rpc_message = &msg,
4246                 .callback_ops = &nfs4_locku_ops,
4247                 .workqueue = nfsiod_workqueue,
4248                 .flags = RPC_TASK_ASYNC,
4249         };
4250
4251         /* Ensure this is an unlock - when canceling a lock, the
4252          * canceled lock is passed in, and it won't be an unlock.
4253          */
4254         fl->fl_type = F_UNLCK;
4255
4256         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4257         if (data == NULL) {
4258                 nfs_free_seqid(seqid);
4259                 return ERR_PTR(-ENOMEM);
4260         }
4261
4262         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4263         msg.rpc_argp = &data->arg;
4264         msg.rpc_resp = &data->res;
4265         task_setup_data.callback_data = data;
4266         return rpc_run_task(&task_setup_data);
4267 }
4268
4269 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4270 {
4271         struct nfs_inode *nfsi = NFS_I(state->inode);
4272         struct nfs_seqid *seqid;
4273         struct nfs4_lock_state *lsp;
4274         struct rpc_task *task;
4275         int status = 0;
4276         unsigned char fl_flags = request->fl_flags;
4277
4278         status = nfs4_set_lock_state(state, request);
4279         /* Unlock _before_ we do the RPC call */
4280         request->fl_flags |= FL_EXISTS;
4281         down_read(&nfsi->rwsem);
4282         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4283                 up_read(&nfsi->rwsem);
4284                 goto out;
4285         }
4286         up_read(&nfsi->rwsem);
4287         if (status != 0)
4288                 goto out;
4289         /* Is this a delegated lock? */
4290         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4291                 goto out;
4292         lsp = request->fl_u.nfs4_fl.owner;
4293         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4294         status = -ENOMEM;
4295         if (seqid == NULL)
4296                 goto out;
4297         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4298         status = PTR_ERR(task);
4299         if (IS_ERR(task))
4300                 goto out;
4301         status = nfs4_wait_for_completion_rpc_task(task);
4302         rpc_put_task(task);
4303 out:
4304         request->fl_flags = fl_flags;
4305         return status;
4306 }
4307
4308 struct nfs4_lockdata {
4309         struct nfs_lock_args arg;
4310         struct nfs_lock_res res;
4311         struct nfs4_lock_state *lsp;
4312         struct nfs_open_context *ctx;
4313         struct file_lock fl;
4314         unsigned long timestamp;
4315         int rpc_status;
4316         int cancelled;
4317         struct nfs_server *server;
4318 };
4319
4320 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4321                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4322                 gfp_t gfp_mask)
4323 {
4324         struct nfs4_lockdata *p;
4325         struct inode *inode = lsp->ls_state->inode;
4326         struct nfs_server *server = NFS_SERVER(inode);
4327
4328         p = kzalloc(sizeof(*p), gfp_mask);
4329         if (p == NULL)
4330                 return NULL;
4331
4332         p->arg.fh = NFS_FH(inode);
4333         p->arg.fl = &p->fl;
4334         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4335         if (p->arg.open_seqid == NULL)
4336                 goto out_free;
4337         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4338         if (p->arg.lock_seqid == NULL)
4339                 goto out_free_seqid;
4340         p->arg.lock_stateid = &lsp->ls_stateid;
4341         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4342         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4343         p->arg.lock_owner.s_dev = server->s_dev;
4344         p->res.lock_seqid = p->arg.lock_seqid;
4345         p->lsp = lsp;
4346         p->server = server;
4347         atomic_inc(&lsp->ls_count);
4348         p->ctx = get_nfs_open_context(ctx);
4349         memcpy(&p->fl, fl, sizeof(p->fl));
4350         return p;
4351 out_free_seqid:
4352         nfs_free_seqid(p->arg.open_seqid);
4353 out_free:
4354         kfree(p);
4355         return NULL;
4356 }
4357
4358 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4359 {
4360         struct nfs4_lockdata *data = calldata;
4361         struct nfs4_state *state = data->lsp->ls_state;
4362
4363         dprintk("%s: begin!\n", __func__);
4364         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4365                 return;
4366         /* Do we need to do an open_to_lock_owner? */
4367         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4368                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4369                         return;
4370                 data->arg.open_stateid = &state->stateid;
4371                 data->arg.new_lock_owner = 1;
4372                 data->res.open_seqid = data->arg.open_seqid;
4373         } else
4374                 data->arg.new_lock_owner = 0;
4375         data->timestamp = jiffies;
4376         if (nfs4_setup_sequence(data->server,
4377                                 &data->arg.seq_args,
4378                                 &data->res.seq_res, task))
4379                 return;
4380         rpc_call_start(task);
4381         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4382 }
4383
4384 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4385 {
4386         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4387         nfs4_lock_prepare(task, calldata);
4388 }
4389
4390 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4391 {
4392         struct nfs4_lockdata *data = calldata;
4393
4394         dprintk("%s: begin!\n", __func__);
4395
4396         if (!nfs4_sequence_done(task, &data->res.seq_res))
4397                 return;
4398
4399         data->rpc_status = task->tk_status;
4400         if (data->arg.new_lock_owner != 0) {
4401                 if (data->rpc_status == 0)
4402                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4403                 else
4404                         goto out;
4405         }
4406         if (data->rpc_status == 0) {
4407                 nfs4_stateid_copy(&data->lsp->ls_stateid, &data->res.stateid);
4408                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4409                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4410         }
4411 out:
4412         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4413 }
4414
4415 static void nfs4_lock_release(void *calldata)
4416 {
4417         struct nfs4_lockdata *data = calldata;
4418
4419         dprintk("%s: begin!\n", __func__);
4420         nfs_free_seqid(data->arg.open_seqid);
4421         if (data->cancelled != 0) {
4422                 struct rpc_task *task;
4423                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4424                                 data->arg.lock_seqid);
4425                 if (!IS_ERR(task))
4426                         rpc_put_task_async(task);
4427                 dprintk("%s: cancelling lock!\n", __func__);
4428         } else
4429                 nfs_free_seqid(data->arg.lock_seqid);
4430         nfs4_put_lock_state(data->lsp);
4431         put_nfs_open_context(data->ctx);
4432         kfree(data);
4433         dprintk("%s: done!\n", __func__);
4434 }
4435
4436 static const struct rpc_call_ops nfs4_lock_ops = {
4437         .rpc_call_prepare = nfs4_lock_prepare,
4438         .rpc_call_done = nfs4_lock_done,
4439         .rpc_release = nfs4_lock_release,
4440 };
4441
4442 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4443         .rpc_call_prepare = nfs4_recover_lock_prepare,
4444         .rpc_call_done = nfs4_lock_done,
4445         .rpc_release = nfs4_lock_release,
4446 };
4447
4448 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4449 {
4450         switch (error) {
4451         case -NFS4ERR_ADMIN_REVOKED:
4452         case -NFS4ERR_BAD_STATEID:
4453                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4454                 if (new_lock_owner != 0 ||
4455                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4456                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4457                 break;
4458         case -NFS4ERR_STALE_STATEID:
4459                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4460         case -NFS4ERR_EXPIRED:
4461                 nfs4_schedule_lease_recovery(server->nfs_client);
4462         };
4463 }
4464
4465 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4466 {
4467         struct nfs4_lockdata *data;
4468         struct rpc_task *task;
4469         struct rpc_message msg = {
4470                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4471                 .rpc_cred = state->owner->so_cred,
4472         };
4473         struct rpc_task_setup task_setup_data = {
4474                 .rpc_client = NFS_CLIENT(state->inode),
4475                 .rpc_message = &msg,
4476                 .callback_ops = &nfs4_lock_ops,
4477                 .workqueue = nfsiod_workqueue,
4478                 .flags = RPC_TASK_ASYNC,
4479         };
4480         int ret;
4481
4482         dprintk("%s: begin!\n", __func__);
4483         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4484                         fl->fl_u.nfs4_fl.owner,
4485                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4486         if (data == NULL)
4487                 return -ENOMEM;
4488         if (IS_SETLKW(cmd))
4489                 data->arg.block = 1;
4490         if (recovery_type > NFS_LOCK_NEW) {
4491                 if (recovery_type == NFS_LOCK_RECLAIM)
4492                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4493                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4494         }
4495         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4496         msg.rpc_argp = &data->arg;
4497         msg.rpc_resp = &data->res;
4498         task_setup_data.callback_data = data;
4499         task = rpc_run_task(&task_setup_data);
4500         if (IS_ERR(task))
4501                 return PTR_ERR(task);
4502         ret = nfs4_wait_for_completion_rpc_task(task);
4503         if (ret == 0) {
4504                 ret = data->rpc_status;
4505                 if (ret)
4506                         nfs4_handle_setlk_error(data->server, data->lsp,
4507                                         data->arg.new_lock_owner, ret);
4508         } else
4509                 data->cancelled = 1;
4510         rpc_put_task(task);
4511         dprintk("%s: done, ret = %d!\n", __func__, ret);
4512         return ret;
4513 }
4514
4515 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4516 {
4517         struct nfs_server *server = NFS_SERVER(state->inode);
4518         struct nfs4_exception exception = { };
4519         int err;
4520
4521         do {
4522                 /* Cache the lock if possible... */
4523                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4524                         return 0;
4525                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4526                 if (err != -NFS4ERR_DELAY)
4527                         break;
4528                 nfs4_handle_exception(server, err, &exception);
4529         } while (exception.retry);
4530         return err;
4531 }
4532
4533 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4534 {
4535         struct nfs_server *server = NFS_SERVER(state->inode);
4536         struct nfs4_exception exception = { };
4537         int err;
4538
4539         err = nfs4_set_lock_state(state, request);
4540         if (err != 0)
4541                 return err;
4542         do {
4543                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4544                         return 0;
4545                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4546                 switch (err) {
4547                 default:
4548                         goto out;
4549                 case -NFS4ERR_GRACE:
4550                 case -NFS4ERR_DELAY:
4551                         nfs4_handle_exception(server, err, &exception);
4552                         err = 0;
4553                 }
4554         } while (exception.retry);
4555 out:
4556         return err;
4557 }
4558
4559 #if defined(CONFIG_NFS_V4_1)
4560 static int nfs41_check_expired_locks(struct nfs4_state *state)
4561 {
4562         int status, ret = NFS_OK;
4563         struct nfs4_lock_state *lsp;
4564         struct nfs_server *server = NFS_SERVER(state->inode);
4565
4566         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4567                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4568                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4569                         if (status != NFS_OK) {
4570                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4571                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4572                                 ret = status;
4573                         }
4574                 }
4575         };
4576
4577         return ret;
4578 }
4579
4580 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4581 {
4582         int status = NFS_OK;
4583
4584         if (test_bit(LK_STATE_IN_USE, &state->flags))
4585                 status = nfs41_check_expired_locks(state);
4586         if (status == NFS_OK)
4587                 return status;
4588         return nfs4_lock_expired(state, request);
4589 }
4590 #endif
4591
4592 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4593 {
4594         struct nfs_inode *nfsi = NFS_I(state->inode);
4595         unsigned char fl_flags = request->fl_flags;
4596         int status = -ENOLCK;
4597
4598         if ((fl_flags & FL_POSIX) &&
4599                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4600                 goto out;
4601         /* Is this a delegated open? */
4602         status = nfs4_set_lock_state(state, request);
4603         if (status != 0)
4604                 goto out;
4605         request->fl_flags |= FL_ACCESS;
4606         status = do_vfs_lock(request->fl_file, request);
4607         if (status < 0)
4608                 goto out;
4609         down_read(&nfsi->rwsem);
4610         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4611                 /* Yes: cache locks! */
4612                 /* ...but avoid races with delegation recall... */
4613                 request->fl_flags = fl_flags & ~FL_SLEEP;
4614                 status = do_vfs_lock(request->fl_file, request);
4615                 goto out_unlock;
4616         }
4617         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4618         if (status != 0)
4619                 goto out_unlock;
4620         /* Note: we always want to sleep here! */
4621         request->fl_flags = fl_flags | FL_SLEEP;
4622         if (do_vfs_lock(request->fl_file, request) < 0)
4623                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4624                         "manager!\n", __func__);
4625 out_unlock:
4626         up_read(&nfsi->rwsem);
4627 out:
4628         request->fl_flags = fl_flags;
4629         return status;
4630 }
4631
4632 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4633 {
4634         struct nfs4_exception exception = {
4635                 .state = state,
4636         };
4637         int err;
4638
4639         do {
4640                 err = _nfs4_proc_setlk(state, cmd, request);
4641                 if (err == -NFS4ERR_DENIED)
4642                         err = -EAGAIN;
4643                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4644                                 err, &exception);
4645         } while (exception.retry);
4646         return err;
4647 }
4648
4649 static int
4650 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4651 {
4652         struct nfs_open_context *ctx;
4653         struct nfs4_state *state;
4654         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4655         int status;
4656
4657         /* verify open state */
4658         ctx = nfs_file_open_context(filp);
4659         state = ctx->state;
4660
4661         if (request->fl_start < 0 || request->fl_end < 0)
4662                 return -EINVAL;
4663
4664         if (IS_GETLK(cmd)) {
4665                 if (state != NULL)
4666                         return nfs4_proc_getlk(state, F_GETLK, request);
4667                 return 0;
4668         }
4669
4670         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4671                 return -EINVAL;
4672
4673         if (request->fl_type == F_UNLCK) {
4674                 if (state != NULL)
4675                         return nfs4_proc_unlck(state, cmd, request);
4676                 return 0;
4677         }
4678
4679         if (state == NULL)
4680                 return -ENOLCK;
4681         do {
4682                 status = nfs4_proc_setlk(state, cmd, request);
4683                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4684                         break;
4685                 timeout = nfs4_set_lock_task_retry(timeout);
4686                 status = -ERESTARTSYS;
4687                 if (signalled())
4688                         break;
4689         } while(status < 0);
4690         return status;
4691 }
4692
4693 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4694 {
4695         struct nfs_server *server = NFS_SERVER(state->inode);
4696         struct nfs4_exception exception = { };
4697         int err;
4698
4699         err = nfs4_set_lock_state(state, fl);
4700         if (err != 0)
4701                 goto out;
4702         do {
4703                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4704                 switch (err) {
4705                         default:
4706                                 printk(KERN_ERR "NFS: %s: unhandled error "
4707                                         "%d.\n", __func__, err);
4708                         case 0:
4709                         case -ESTALE:
4710                                 goto out;
4711                         case -NFS4ERR_EXPIRED:
4712                                 nfs4_schedule_stateid_recovery(server, state);
4713                         case -NFS4ERR_STALE_CLIENTID:
4714                         case -NFS4ERR_STALE_STATEID:
4715                                 nfs4_schedule_lease_recovery(server->nfs_client);
4716                                 goto out;
4717                         case -NFS4ERR_BADSESSION:
4718                         case -NFS4ERR_BADSLOT:
4719                         case -NFS4ERR_BAD_HIGH_SLOT:
4720                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4721                         case -NFS4ERR_DEADSESSION:
4722                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4723                                 goto out;
4724                         case -ERESTARTSYS:
4725                                 /*
4726                                  * The show must go on: exit, but mark the
4727                                  * stateid as needing recovery.
4728                                  */
4729                         case -NFS4ERR_DELEG_REVOKED:
4730                         case -NFS4ERR_ADMIN_REVOKED:
4731                         case -NFS4ERR_BAD_STATEID:
4732                         case -NFS4ERR_OPENMODE:
4733                                 nfs4_schedule_stateid_recovery(server, state);
4734                                 err = 0;
4735                                 goto out;
4736                         case -EKEYEXPIRED:
4737                                 /*
4738                                  * User RPCSEC_GSS context has expired.
4739                                  * We cannot recover this stateid now, so
4740                                  * skip it and allow recovery thread to
4741                                  * proceed.
4742                                  */
4743                                 err = 0;
4744                                 goto out;
4745                         case -ENOMEM:
4746                         case -NFS4ERR_DENIED:
4747                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4748                                 err = 0;
4749                                 goto out;
4750                         case -NFS4ERR_DELAY:
4751                                 break;
4752                 }
4753                 err = nfs4_handle_exception(server, err, &exception);
4754         } while (exception.retry);
4755 out:
4756         return err;
4757 }
4758
4759 struct nfs_release_lockowner_data {
4760         struct nfs4_lock_state *lsp;
4761         struct nfs_release_lockowner_args args;
4762 };
4763
4764 static void nfs4_release_lockowner_release(void *calldata)
4765 {
4766         struct nfs_release_lockowner_data *data = calldata;
4767         nfs4_free_lock_state(data->lsp);
4768         kfree(calldata);
4769 }
4770
4771 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4772         .rpc_release = nfs4_release_lockowner_release,
4773 };
4774
4775 int nfs4_release_lockowner(struct nfs4_lock_state *lsp)
4776 {
4777         struct nfs_server *server = lsp->ls_state->owner->so_server;
4778         struct nfs_release_lockowner_data *data;
4779         struct rpc_message msg = {
4780                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4781         };
4782
4783         if (server->nfs_client->cl_mvops->minor_version != 0)
4784                 return -EINVAL;
4785         data = kmalloc(sizeof(*data), GFP_NOFS);
4786         if (!data)
4787                 return -ENOMEM;
4788         data->lsp = lsp;
4789         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
4790         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
4791         data->args.lock_owner.s_dev = server->s_dev;
4792         msg.rpc_argp = &data->args;
4793         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
4794         return 0;
4795 }
4796
4797 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4798
4799 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4800                                    const void *buf, size_t buflen,
4801                                    int flags, int type)
4802 {
4803         if (strcmp(key, "") != 0)
4804                 return -EINVAL;
4805
4806         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4807 }
4808
4809 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4810                                    void *buf, size_t buflen, int type)
4811 {
4812         if (strcmp(key, "") != 0)
4813                 return -EINVAL;
4814
4815         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4816 }
4817
4818 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4819                                        size_t list_len, const char *name,
4820                                        size_t name_len, int type)
4821 {
4822         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4823
4824         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4825                 return 0;
4826
4827         if (list && len <= list_len)
4828                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4829         return len;
4830 }
4831
4832 /*
4833  * nfs_fhget will use either the mounted_on_fileid or the fileid
4834  */
4835 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4836 {
4837         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4838                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4839               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4840               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4841                 return;
4842
4843         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4844                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4845         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4846         fattr->nlink = 2;
4847 }
4848
4849 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4850                 struct nfs4_fs_locations *fs_locations, struct page *page)
4851 {
4852         struct nfs_server *server = NFS_SERVER(dir);
4853         u32 bitmask[2] = {
4854                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4855         };
4856         struct nfs4_fs_locations_arg args = {
4857                 .dir_fh = NFS_FH(dir),
4858                 .name = name,
4859                 .page = page,
4860                 .bitmask = bitmask,
4861         };
4862         struct nfs4_fs_locations_res res = {
4863                 .fs_locations = fs_locations,
4864         };
4865         struct rpc_message msg = {
4866                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4867                 .rpc_argp = &args,
4868                 .rpc_resp = &res,
4869         };
4870         int status;
4871
4872         dprintk("%s: start\n", __func__);
4873
4874         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4875          * is not supported */
4876         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4877                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4878         else
4879                 bitmask[0] |= FATTR4_WORD0_FILEID;
4880
4881         nfs_fattr_init(&fs_locations->fattr);
4882         fs_locations->server = server;
4883         fs_locations->nlocations = 0;
4884         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4885         dprintk("%s: returned status = %d\n", __func__, status);
4886         return status;
4887 }
4888
4889 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4890 {
4891         int status;
4892         struct nfs4_secinfo_arg args = {
4893                 .dir_fh = NFS_FH(dir),
4894                 .name   = name,
4895         };
4896         struct nfs4_secinfo_res res = {
4897                 .flavors     = flavors,
4898         };
4899         struct rpc_message msg = {
4900                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4901                 .rpc_argp = &args,
4902                 .rpc_resp = &res,
4903         };
4904
4905         dprintk("NFS call  secinfo %s\n", name->name);
4906         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4907         dprintk("NFS reply  secinfo: %d\n", status);
4908         return status;
4909 }
4910
4911 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4912 {
4913         struct nfs4_exception exception = { };
4914         int err;
4915         do {
4916                 err = nfs4_handle_exception(NFS_SERVER(dir),
4917                                 _nfs4_proc_secinfo(dir, name, flavors),
4918                                 &exception);
4919         } while (exception.retry);
4920         return err;
4921 }
4922
4923 #ifdef CONFIG_NFS_V4_1
4924 /*
4925  * Check the exchange flags returned by the server for invalid flags, having
4926  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4927  * DS flags set.
4928  */
4929 static int nfs4_check_cl_exchange_flags(u32 flags)
4930 {
4931         if (flags & ~EXCHGID4_FLAG_MASK_R)
4932                 goto out_inval;
4933         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4934             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4935                 goto out_inval;
4936         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4937                 goto out_inval;
4938         return NFS_OK;
4939 out_inval:
4940         return -NFS4ERR_INVAL;
4941 }
4942
4943 static bool
4944 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4945 {
4946         if (a->server_scope_sz == b->server_scope_sz &&
4947             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4948                 return true;
4949
4950         return false;
4951 }
4952
4953 /*
4954  * nfs4_proc_exchange_id()
4955  *
4956  * Since the clientid has expired, all compounds using sessions
4957  * associated with the stale clientid will be returning
4958  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4959  * be in some phase of session reset.
4960  */
4961 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4962 {
4963         nfs4_verifier verifier;
4964         struct nfs41_exchange_id_args args = {
4965                 .verifier = &verifier,
4966                 .client = clp,
4967                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4968         };
4969         struct nfs41_exchange_id_res res = {
4970                 .client = clp,
4971         };
4972         int status;
4973         struct rpc_message msg = {
4974                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4975                 .rpc_argp = &args,
4976                 .rpc_resp = &res,
4977                 .rpc_cred = cred,
4978         };
4979
4980         dprintk("--> %s\n", __func__);
4981         BUG_ON(clp == NULL);
4982
4983         nfs4_construct_boot_verifier(clp, &verifier);
4984
4985         args.id_len = scnprintf(args.id, sizeof(args.id),
4986                                 "%s/%s.%s/%u",
4987                                 clp->cl_ipaddr,
4988                                 init_utsname()->nodename,
4989                                 init_utsname()->domainname,
4990                                 clp->cl_rpcclient->cl_auth->au_flavor);
4991
4992         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4993         if (unlikely(!res.server_scope)) {
4994                 status = -ENOMEM;
4995                 goto out;
4996         }
4997
4998         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
4999         if (unlikely(!res.impl_id)) {
5000                 status = -ENOMEM;
5001                 goto out_server_scope;
5002         }
5003
5004         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5005         if (!status)
5006                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
5007
5008         if (!status) {
5009                 /* use the most recent implementation id */
5010                 kfree(clp->impl_id);
5011                 clp->impl_id = res.impl_id;
5012         } else
5013                 kfree(res.impl_id);
5014
5015         if (!status) {
5016                 if (clp->server_scope &&
5017                     !nfs41_same_server_scope(clp->server_scope,
5018                                              res.server_scope)) {
5019                         dprintk("%s: server_scope mismatch detected\n",
5020                                 __func__);
5021                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
5022                         kfree(clp->server_scope);
5023                         clp->server_scope = NULL;
5024                 }
5025
5026                 if (!clp->server_scope) {
5027                         clp->server_scope = res.server_scope;
5028                         goto out;
5029                 }
5030         }
5031
5032 out_server_scope:
5033         kfree(res.server_scope);
5034 out:
5035         if (clp->impl_id)
5036                 dprintk("%s: Server Implementation ID: "
5037                         "domain: %s, name: %s, date: %llu,%u\n",
5038                         __func__, clp->impl_id->domain, clp->impl_id->name,
5039                         clp->impl_id->date.seconds,
5040                         clp->impl_id->date.nseconds);
5041         dprintk("<-- %s status= %d\n", __func__, status);
5042         return status;
5043 }
5044
5045 struct nfs4_get_lease_time_data {
5046         struct nfs4_get_lease_time_args *args;
5047         struct nfs4_get_lease_time_res *res;
5048         struct nfs_client *clp;
5049 };
5050
5051 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5052                                         void *calldata)
5053 {
5054         int ret;
5055         struct nfs4_get_lease_time_data *data =
5056                         (struct nfs4_get_lease_time_data *)calldata;
5057
5058         dprintk("--> %s\n", __func__);
5059         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5060         /* just setup sequence, do not trigger session recovery
5061            since we're invoked within one */
5062         ret = nfs41_setup_sequence(data->clp->cl_session,
5063                                    &data->args->la_seq_args,
5064                                    &data->res->lr_seq_res, task);
5065
5066         BUG_ON(ret == -EAGAIN);
5067         rpc_call_start(task);
5068         dprintk("<-- %s\n", __func__);
5069 }
5070
5071 /*
5072  * Called from nfs4_state_manager thread for session setup, so don't recover
5073  * from sequence operation or clientid errors.
5074  */
5075 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5076 {
5077         struct nfs4_get_lease_time_data *data =
5078                         (struct nfs4_get_lease_time_data *)calldata;
5079
5080         dprintk("--> %s\n", __func__);
5081         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5082                 return;
5083         switch (task->tk_status) {
5084         case -NFS4ERR_DELAY:
5085         case -NFS4ERR_GRACE:
5086                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5087                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5088                 task->tk_status = 0;
5089                 /* fall through */
5090         case -NFS4ERR_RETRY_UNCACHED_REP:
5091                 rpc_restart_call_prepare(task);
5092                 return;
5093         }
5094         dprintk("<-- %s\n", __func__);
5095 }
5096
5097 struct rpc_call_ops nfs4_get_lease_time_ops = {
5098         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5099         .rpc_call_done = nfs4_get_lease_time_done,
5100 };
5101
5102 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5103 {
5104         struct rpc_task *task;
5105         struct nfs4_get_lease_time_args args;
5106         struct nfs4_get_lease_time_res res = {
5107                 .lr_fsinfo = fsinfo,
5108         };
5109         struct nfs4_get_lease_time_data data = {
5110                 .args = &args,
5111                 .res = &res,
5112                 .clp = clp,
5113         };
5114         struct rpc_message msg = {
5115                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5116                 .rpc_argp = &args,
5117                 .rpc_resp = &res,
5118         };
5119         struct rpc_task_setup task_setup = {
5120                 .rpc_client = clp->cl_rpcclient,
5121                 .rpc_message = &msg,
5122                 .callback_ops = &nfs4_get_lease_time_ops,
5123                 .callback_data = &data,
5124                 .flags = RPC_TASK_TIMEOUT,
5125         };
5126         int status;
5127
5128         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5129         dprintk("--> %s\n", __func__);
5130         task = rpc_run_task(&task_setup);
5131
5132         if (IS_ERR(task))
5133                 status = PTR_ERR(task);
5134         else {
5135                 status = task->tk_status;
5136                 rpc_put_task(task);
5137         }
5138         dprintk("<-- %s return %d\n", __func__, status);
5139
5140         return status;
5141 }
5142
5143 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5144 {
5145         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5146 }
5147
5148 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5149                 struct nfs4_slot *new,
5150                 u32 max_slots,
5151                 u32 ivalue)
5152 {
5153         struct nfs4_slot *old = NULL;
5154         u32 i;
5155
5156         spin_lock(&tbl->slot_tbl_lock);
5157         if (new) {
5158                 old = tbl->slots;
5159                 tbl->slots = new;
5160                 tbl->max_slots = max_slots;
5161         }
5162         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5163         for (i = 0; i < tbl->max_slots; i++)
5164                 tbl->slots[i].seq_nr = ivalue;
5165         spin_unlock(&tbl->slot_tbl_lock);
5166         kfree(old);
5167 }
5168
5169 /*
5170  * (re)Initialise a slot table
5171  */
5172 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5173                                  u32 ivalue)
5174 {
5175         struct nfs4_slot *new = NULL;
5176         int ret = -ENOMEM;
5177
5178         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5179                 max_reqs, tbl->max_slots);
5180
5181         /* Does the newly negotiated max_reqs match the existing slot table? */
5182         if (max_reqs != tbl->max_slots) {
5183                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5184                 if (!new)
5185                         goto out;
5186         }
5187         ret = 0;
5188
5189         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5190         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5191                 tbl, tbl->slots, tbl->max_slots);
5192 out:
5193         dprintk("<-- %s: return %d\n", __func__, ret);
5194         return ret;
5195 }
5196
5197 /* Destroy the slot table */
5198 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5199 {
5200         if (session->fc_slot_table.slots != NULL) {
5201                 kfree(session->fc_slot_table.slots);
5202                 session->fc_slot_table.slots = NULL;
5203         }
5204         if (session->bc_slot_table.slots != NULL) {
5205                 kfree(session->bc_slot_table.slots);
5206                 session->bc_slot_table.slots = NULL;
5207         }
5208         return;
5209 }
5210
5211 /*
5212  * Initialize or reset the forechannel and backchannel tables
5213  */
5214 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5215 {
5216         struct nfs4_slot_table *tbl;
5217         int status;
5218
5219         dprintk("--> %s\n", __func__);
5220         /* Fore channel */
5221         tbl = &ses->fc_slot_table;
5222         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5223         if (status) /* -ENOMEM */
5224                 return status;
5225         /* Back channel */
5226         tbl = &ses->bc_slot_table;
5227         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5228         if (status && tbl->slots == NULL)
5229                 /* Fore and back channel share a connection so get
5230                  * both slot tables or neither */
5231                 nfs4_destroy_slot_tables(ses);
5232         return status;
5233 }
5234
5235 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5236 {
5237         struct nfs4_session *session;
5238         struct nfs4_slot_table *tbl;
5239
5240         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5241         if (!session)
5242                 return NULL;
5243
5244         tbl = &session->fc_slot_table;
5245         tbl->highest_used_slotid = NFS4_NO_SLOT;
5246         spin_lock_init(&tbl->slot_tbl_lock);
5247         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5248         init_completion(&tbl->complete);
5249
5250         tbl = &session->bc_slot_table;
5251         tbl->highest_used_slotid = NFS4_NO_SLOT;
5252         spin_lock_init(&tbl->slot_tbl_lock);
5253         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5254         init_completion(&tbl->complete);
5255
5256         session->session_state = 1<<NFS4_SESSION_INITING;
5257
5258         session->clp = clp;
5259         return session;
5260 }
5261
5262 void nfs4_destroy_session(struct nfs4_session *session)
5263 {
5264         struct rpc_xprt *xprt;
5265
5266         nfs4_proc_destroy_session(session);
5267
5268         rcu_read_lock();
5269         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5270         rcu_read_unlock();
5271         dprintk("%s Destroy backchannel for xprt %p\n",
5272                 __func__, xprt);
5273         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5274         nfs4_destroy_slot_tables(session);
5275         kfree(session);
5276 }
5277
5278 /*
5279  * Initialize the values to be used by the client in CREATE_SESSION
5280  * If nfs4_init_session set the fore channel request and response sizes,
5281  * use them.
5282  *
5283  * Set the back channel max_resp_sz_cached to zero to force the client to
5284  * always set csa_cachethis to FALSE because the current implementation
5285  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5286  */
5287 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5288 {
5289         struct nfs4_session *session = args->client->cl_session;
5290         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5291                      mxresp_sz = session->fc_attrs.max_resp_sz;
5292
5293         if (mxrqst_sz == 0)
5294                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5295         if (mxresp_sz == 0)
5296                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5297         /* Fore channel attributes */
5298         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5299         args->fc_attrs.max_resp_sz = mxresp_sz;
5300         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5301         args->fc_attrs.max_reqs = max_session_slots;
5302
5303         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5304                 "max_ops=%u max_reqs=%u\n",
5305                 __func__,
5306                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5307                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5308
5309         /* Back channel attributes */
5310         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5311         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5312         args->bc_attrs.max_resp_sz_cached = 0;
5313         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5314         args->bc_attrs.max_reqs = 1;
5315
5316         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5317                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5318                 __func__,
5319                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5320                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5321                 args->bc_attrs.max_reqs);
5322 }
5323
5324 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5325 {
5326         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5327         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5328
5329         if (rcvd->max_resp_sz > sent->max_resp_sz)
5330                 return -EINVAL;
5331         /*
5332          * Our requested max_ops is the minimum we need; we're not
5333          * prepared to break up compounds into smaller pieces than that.
5334          * So, no point even trying to continue if the server won't
5335          * cooperate:
5336          */
5337         if (rcvd->max_ops < sent->max_ops)
5338                 return -EINVAL;
5339         if (rcvd->max_reqs == 0)
5340                 return -EINVAL;
5341         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5342                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5343         return 0;
5344 }
5345
5346 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5347 {
5348         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5349         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5350
5351         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5352                 return -EINVAL;
5353         if (rcvd->max_resp_sz < sent->max_resp_sz)
5354                 return -EINVAL;
5355         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5356                 return -EINVAL;
5357         /* These would render the backchannel useless: */
5358         if (rcvd->max_ops != sent->max_ops)
5359                 return -EINVAL;
5360         if (rcvd->max_reqs != sent->max_reqs)
5361                 return -EINVAL;
5362         return 0;
5363 }
5364
5365 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5366                                      struct nfs4_session *session)
5367 {
5368         int ret;
5369
5370         ret = nfs4_verify_fore_channel_attrs(args, session);
5371         if (ret)
5372                 return ret;
5373         return nfs4_verify_back_channel_attrs(args, session);
5374 }
5375
5376 static int _nfs4_proc_create_session(struct nfs_client *clp)
5377 {
5378         struct nfs4_session *session = clp->cl_session;
5379         struct nfs41_create_session_args args = {
5380                 .client = clp,
5381                 .cb_program = NFS4_CALLBACK,
5382         };
5383         struct nfs41_create_session_res res = {
5384                 .client = clp,
5385         };
5386         struct rpc_message msg = {
5387                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5388                 .rpc_argp = &args,
5389                 .rpc_resp = &res,
5390         };
5391         int status;
5392
5393         nfs4_init_channel_attrs(&args);
5394         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5395
5396         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5397
5398         if (!status)
5399                 /* Verify the session's negotiated channel_attrs values */
5400                 status = nfs4_verify_channel_attrs(&args, session);
5401         if (!status) {
5402                 /* Increment the clientid slot sequence id */
5403                 clp->cl_seqid++;
5404         }
5405
5406         return status;
5407 }
5408
5409 /*
5410  * Issues a CREATE_SESSION operation to the server.
5411  * It is the responsibility of the caller to verify the session is
5412  * expired before calling this routine.
5413  */
5414 int nfs4_proc_create_session(struct nfs_client *clp)
5415 {
5416         int status;
5417         unsigned *ptr;
5418         struct nfs4_session *session = clp->cl_session;
5419
5420         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5421
5422         status = _nfs4_proc_create_session(clp);
5423         if (status)
5424                 goto out;
5425
5426         /* Init or reset the session slot tables */
5427         status = nfs4_setup_session_slot_tables(session);
5428         dprintk("slot table setup returned %d\n", status);
5429         if (status)
5430                 goto out;
5431
5432         ptr = (unsigned *)&session->sess_id.data[0];
5433         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5434                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5435 out:
5436         dprintk("<-- %s\n", __func__);
5437         return status;
5438 }
5439
5440 /*
5441  * Issue the over-the-wire RPC DESTROY_SESSION.
5442  * The caller must serialize access to this routine.
5443  */
5444 int nfs4_proc_destroy_session(struct nfs4_session *session)
5445 {
5446         int status = 0;
5447         struct rpc_message msg;
5448
5449         dprintk("--> nfs4_proc_destroy_session\n");
5450
5451         /* session is still being setup */
5452         if (session->clp->cl_cons_state != NFS_CS_READY)
5453                 return status;
5454
5455         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5456         msg.rpc_argp = session;
5457         msg.rpc_resp = NULL;
5458         msg.rpc_cred = NULL;
5459         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5460
5461         if (status)
5462                 printk(KERN_WARNING
5463                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5464                         "Session has been destroyed regardless...\n", status);
5465
5466         dprintk("<-- nfs4_proc_destroy_session\n");
5467         return status;
5468 }
5469
5470 int nfs4_init_session(struct nfs_server *server)
5471 {
5472         struct nfs_client *clp = server->nfs_client;
5473         struct nfs4_session *session;
5474         unsigned int rsize, wsize;
5475         int ret;
5476
5477         if (!nfs4_has_session(clp))
5478                 return 0;
5479
5480         session = clp->cl_session;
5481         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5482                 return 0;
5483
5484         rsize = server->rsize;
5485         if (rsize == 0)
5486                 rsize = NFS_MAX_FILE_IO_SIZE;
5487         wsize = server->wsize;
5488         if (wsize == 0)
5489                 wsize = NFS_MAX_FILE_IO_SIZE;
5490
5491         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5492         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5493
5494         ret = nfs4_recover_expired_lease(server);
5495         if (!ret)
5496                 ret = nfs4_check_client_ready(clp);
5497         return ret;
5498 }
5499
5500 int nfs4_init_ds_session(struct nfs_client *clp)
5501 {
5502         struct nfs4_session *session = clp->cl_session;
5503         int ret;
5504
5505         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5506                 return 0;
5507
5508         ret = nfs4_client_recover_expired_lease(clp);
5509         if (!ret)
5510                 /* Test for the DS role */
5511                 if (!is_ds_client(clp))
5512                         ret = -ENODEV;
5513         if (!ret)
5514                 ret = nfs4_check_client_ready(clp);
5515         return ret;
5516
5517 }
5518 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5519
5520
5521 /*
5522  * Renew the cl_session lease.
5523  */
5524 struct nfs4_sequence_data {
5525         struct nfs_client *clp;
5526         struct nfs4_sequence_args args;
5527         struct nfs4_sequence_res res;
5528 };
5529
5530 static void nfs41_sequence_release(void *data)
5531 {
5532         struct nfs4_sequence_data *calldata = data;
5533         struct nfs_client *clp = calldata->clp;
5534
5535         if (atomic_read(&clp->cl_count) > 1)
5536                 nfs4_schedule_state_renewal(clp);
5537         nfs_put_client(clp);
5538         kfree(calldata);
5539 }
5540
5541 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5542 {
5543         switch(task->tk_status) {
5544         case -NFS4ERR_DELAY:
5545                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5546                 return -EAGAIN;
5547         default:
5548                 nfs4_schedule_lease_recovery(clp);
5549         }
5550         return 0;
5551 }
5552
5553 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5554 {
5555         struct nfs4_sequence_data *calldata = data;
5556         struct nfs_client *clp = calldata->clp;
5557
5558         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5559                 return;
5560
5561         if (task->tk_status < 0) {
5562                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5563                 if (atomic_read(&clp->cl_count) == 1)
5564                         goto out;
5565
5566                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5567                         rpc_restart_call_prepare(task);
5568                         return;
5569                 }
5570         }
5571         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5572 out:
5573         dprintk("<-- %s\n", __func__);
5574 }
5575
5576 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5577 {
5578         struct nfs4_sequence_data *calldata = data;
5579         struct nfs_client *clp = calldata->clp;
5580         struct nfs4_sequence_args *args;
5581         struct nfs4_sequence_res *res;
5582
5583         args = task->tk_msg.rpc_argp;
5584         res = task->tk_msg.rpc_resp;
5585
5586         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5587                 return;
5588         rpc_call_start(task);
5589 }
5590
5591 static const struct rpc_call_ops nfs41_sequence_ops = {
5592         .rpc_call_done = nfs41_sequence_call_done,
5593         .rpc_call_prepare = nfs41_sequence_prepare,
5594         .rpc_release = nfs41_sequence_release,
5595 };
5596
5597 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5598 {
5599         struct nfs4_sequence_data *calldata;
5600         struct rpc_message msg = {
5601                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5602                 .rpc_cred = cred,
5603         };
5604         struct rpc_task_setup task_setup_data = {
5605                 .rpc_client = clp->cl_rpcclient,
5606                 .rpc_message = &msg,
5607                 .callback_ops = &nfs41_sequence_ops,
5608                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5609         };
5610
5611         if (!atomic_inc_not_zero(&clp->cl_count))
5612                 return ERR_PTR(-EIO);
5613         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5614         if (calldata == NULL) {
5615                 nfs_put_client(clp);
5616                 return ERR_PTR(-ENOMEM);
5617         }
5618         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5619         msg.rpc_argp = &calldata->args;
5620         msg.rpc_resp = &calldata->res;
5621         calldata->clp = clp;
5622         task_setup_data.callback_data = calldata;
5623
5624         return rpc_run_task(&task_setup_data);
5625 }
5626
5627 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5628 {
5629         struct rpc_task *task;
5630         int ret = 0;
5631
5632         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5633                 return 0;
5634         task = _nfs41_proc_sequence(clp, cred);
5635         if (IS_ERR(task))
5636                 ret = PTR_ERR(task);
5637         else
5638                 rpc_put_task_async(task);
5639         dprintk("<-- %s status=%d\n", __func__, ret);
5640         return ret;
5641 }
5642
5643 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5644 {
5645         struct rpc_task *task;
5646         int ret;
5647
5648         task = _nfs41_proc_sequence(clp, cred);
5649         if (IS_ERR(task)) {
5650                 ret = PTR_ERR(task);
5651                 goto out;
5652         }
5653         ret = rpc_wait_for_completion_task(task);
5654         if (!ret) {
5655                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5656
5657                 if (task->tk_status == 0)
5658                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5659                 ret = task->tk_status;
5660         }
5661         rpc_put_task(task);
5662 out:
5663         dprintk("<-- %s status=%d\n", __func__, ret);
5664         return ret;
5665 }
5666
5667 struct nfs4_reclaim_complete_data {
5668         struct nfs_client *clp;
5669         struct nfs41_reclaim_complete_args arg;
5670         struct nfs41_reclaim_complete_res res;
5671 };
5672
5673 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5674 {
5675         struct nfs4_reclaim_complete_data *calldata = data;
5676
5677         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5678         if (nfs41_setup_sequence(calldata->clp->cl_session,
5679                                 &calldata->arg.seq_args,
5680                                 &calldata->res.seq_res, task))
5681                 return;
5682
5683         rpc_call_start(task);
5684 }
5685
5686 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5687 {
5688         switch(task->tk_status) {
5689         case 0:
5690         case -NFS4ERR_COMPLETE_ALREADY:
5691         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5692                 break;
5693         case -NFS4ERR_DELAY:
5694                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5695                 /* fall through */
5696         case -NFS4ERR_RETRY_UNCACHED_REP:
5697                 return -EAGAIN;
5698         default:
5699                 nfs4_schedule_lease_recovery(clp);
5700         }
5701         return 0;
5702 }
5703
5704 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5705 {
5706         struct nfs4_reclaim_complete_data *calldata = data;
5707         struct nfs_client *clp = calldata->clp;
5708         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5709
5710         dprintk("--> %s\n", __func__);
5711         if (!nfs41_sequence_done(task, res))
5712                 return;
5713
5714         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5715                 rpc_restart_call_prepare(task);
5716                 return;
5717         }
5718         dprintk("<-- %s\n", __func__);
5719 }
5720
5721 static void nfs4_free_reclaim_complete_data(void *data)
5722 {
5723         struct nfs4_reclaim_complete_data *calldata = data;
5724
5725         kfree(calldata);
5726 }
5727
5728 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5729         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5730         .rpc_call_done = nfs4_reclaim_complete_done,
5731         .rpc_release = nfs4_free_reclaim_complete_data,
5732 };
5733
5734 /*
5735  * Issue a global reclaim complete.
5736  */
5737 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5738 {
5739         struct nfs4_reclaim_complete_data *calldata;
5740         struct rpc_task *task;
5741         struct rpc_message msg = {
5742                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5743         };
5744         struct rpc_task_setup task_setup_data = {
5745                 .rpc_client = clp->cl_rpcclient,
5746                 .rpc_message = &msg,
5747                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5748                 .flags = RPC_TASK_ASYNC,
5749         };
5750         int status = -ENOMEM;
5751
5752         dprintk("--> %s\n", __func__);
5753         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5754         if (calldata == NULL)
5755                 goto out;
5756         calldata->clp = clp;
5757         calldata->arg.one_fs = 0;
5758
5759         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5760         msg.rpc_argp = &calldata->arg;
5761         msg.rpc_resp = &calldata->res;
5762         task_setup_data.callback_data = calldata;
5763         task = rpc_run_task(&task_setup_data);
5764         if (IS_ERR(task)) {
5765                 status = PTR_ERR(task);
5766                 goto out;
5767         }
5768         status = nfs4_wait_for_completion_rpc_task(task);
5769         if (status == 0)
5770                 status = task->tk_status;
5771         rpc_put_task(task);
5772         return 0;
5773 out:
5774         dprintk("<-- %s status=%d\n", __func__, status);
5775         return status;
5776 }
5777
5778 static void
5779 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5780 {
5781         struct nfs4_layoutget *lgp = calldata;
5782         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5783
5784         dprintk("--> %s\n", __func__);
5785         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5786          * right now covering the LAYOUTGET we are about to send.
5787          * However, that is not so catastrophic, and there seems
5788          * to be no way to prevent it completely.
5789          */
5790         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5791                                 &lgp->res.seq_res, task))
5792                 return;
5793         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5794                                           NFS_I(lgp->args.inode)->layout,
5795                                           lgp->args.ctx->state)) {
5796                 rpc_exit(task, NFS4_OK);
5797                 return;
5798         }
5799         rpc_call_start(task);
5800 }
5801
5802 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5803 {
5804         struct nfs4_layoutget *lgp = calldata;
5805         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5806
5807         dprintk("--> %s\n", __func__);
5808
5809         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5810                 return;
5811
5812         switch (task->tk_status) {
5813         case 0:
5814                 break;
5815         case -NFS4ERR_LAYOUTTRYLATER:
5816         case -NFS4ERR_RECALLCONFLICT:
5817                 task->tk_status = -NFS4ERR_DELAY;
5818                 /* Fall through */
5819         default:
5820                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5821                         rpc_restart_call_prepare(task);
5822                         return;
5823                 }
5824         }
5825         dprintk("<-- %s\n", __func__);
5826 }
5827
5828 static void nfs4_layoutget_release(void *calldata)
5829 {
5830         struct nfs4_layoutget *lgp = calldata;
5831
5832         dprintk("--> %s\n", __func__);
5833         put_nfs_open_context(lgp->args.ctx);
5834         kfree(calldata);
5835         dprintk("<-- %s\n", __func__);
5836 }
5837
5838 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5839         .rpc_call_prepare = nfs4_layoutget_prepare,
5840         .rpc_call_done = nfs4_layoutget_done,
5841         .rpc_release = nfs4_layoutget_release,
5842 };
5843
5844 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5845 {
5846         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5847         struct rpc_task *task;
5848         struct rpc_message msg = {
5849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5850                 .rpc_argp = &lgp->args,
5851                 .rpc_resp = &lgp->res,
5852         };
5853         struct rpc_task_setup task_setup_data = {
5854                 .rpc_client = server->client,
5855                 .rpc_message = &msg,
5856                 .callback_ops = &nfs4_layoutget_call_ops,
5857                 .callback_data = lgp,
5858                 .flags = RPC_TASK_ASYNC,
5859         };
5860         int status = 0;
5861
5862         dprintk("--> %s\n", __func__);
5863
5864         lgp->res.layoutp = &lgp->args.layout;
5865         lgp->res.seq_res.sr_slot = NULL;
5866         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5867         task = rpc_run_task(&task_setup_data);
5868         if (IS_ERR(task))
5869                 return PTR_ERR(task);
5870         status = nfs4_wait_for_completion_rpc_task(task);
5871         if (status == 0)
5872                 status = task->tk_status;
5873         if (status == 0)
5874                 status = pnfs_layout_process(lgp);
5875         rpc_put_task(task);
5876         dprintk("<-- %s status=%d\n", __func__, status);
5877         return status;
5878 }
5879
5880 static void
5881 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5882 {
5883         struct nfs4_layoutreturn *lrp = calldata;
5884
5885         dprintk("--> %s\n", __func__);
5886         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5887                                 &lrp->res.seq_res, task))
5888                 return;
5889         rpc_call_start(task);
5890 }
5891
5892 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5893 {
5894         struct nfs4_layoutreturn *lrp = calldata;
5895         struct nfs_server *server;
5896         struct pnfs_layout_hdr *lo = lrp->args.layout;
5897
5898         dprintk("--> %s\n", __func__);
5899
5900         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5901                 return;
5902
5903         server = NFS_SERVER(lrp->args.inode);
5904         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5905                 rpc_restart_call_prepare(task);
5906                 return;
5907         }
5908         spin_lock(&lo->plh_inode->i_lock);
5909         if (task->tk_status == 0) {
5910                 if (lrp->res.lrs_present) {
5911                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5912                 } else
5913                         BUG_ON(!list_empty(&lo->plh_segs));
5914         }
5915         lo->plh_block_lgets--;
5916         spin_unlock(&lo->plh_inode->i_lock);
5917         dprintk("<-- %s\n", __func__);
5918 }
5919
5920 static void nfs4_layoutreturn_release(void *calldata)
5921 {
5922         struct nfs4_layoutreturn *lrp = calldata;
5923
5924         dprintk("--> %s\n", __func__);
5925         put_layout_hdr(lrp->args.layout);
5926         kfree(calldata);
5927         dprintk("<-- %s\n", __func__);
5928 }
5929
5930 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5931         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5932         .rpc_call_done = nfs4_layoutreturn_done,
5933         .rpc_release = nfs4_layoutreturn_release,
5934 };
5935
5936 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5937 {
5938         struct rpc_task *task;
5939         struct rpc_message msg = {
5940                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5941                 .rpc_argp = &lrp->args,
5942                 .rpc_resp = &lrp->res,
5943         };
5944         struct rpc_task_setup task_setup_data = {
5945                 .rpc_client = lrp->clp->cl_rpcclient,
5946                 .rpc_message = &msg,
5947                 .callback_ops = &nfs4_layoutreturn_call_ops,
5948                 .callback_data = lrp,
5949         };
5950         int status;
5951
5952         dprintk("--> %s\n", __func__);
5953         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5954         task = rpc_run_task(&task_setup_data);
5955         if (IS_ERR(task))
5956                 return PTR_ERR(task);
5957         status = task->tk_status;
5958         dprintk("<-- %s status=%d\n", __func__, status);
5959         rpc_put_task(task);
5960         return status;
5961 }
5962
5963 /*
5964  * Retrieve the list of Data Server devices from the MDS.
5965  */
5966 static int _nfs4_getdevicelist(struct nfs_server *server,
5967                                     const struct nfs_fh *fh,
5968                                     struct pnfs_devicelist *devlist)
5969 {
5970         struct nfs4_getdevicelist_args args = {
5971                 .fh = fh,
5972                 .layoutclass = server->pnfs_curr_ld->id,
5973         };
5974         struct nfs4_getdevicelist_res res = {
5975                 .devlist = devlist,
5976         };
5977         struct rpc_message msg = {
5978                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5979                 .rpc_argp = &args,
5980                 .rpc_resp = &res,
5981         };
5982         int status;
5983
5984         dprintk("--> %s\n", __func__);
5985         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5986                                 &res.seq_res, 0);
5987         dprintk("<-- %s status=%d\n", __func__, status);
5988         return status;
5989 }
5990
5991 int nfs4_proc_getdevicelist(struct nfs_server *server,
5992                             const struct nfs_fh *fh,
5993                             struct pnfs_devicelist *devlist)
5994 {
5995         struct nfs4_exception exception = { };
5996         int err;
5997
5998         do {
5999                 err = nfs4_handle_exception(server,
6000                                 _nfs4_getdevicelist(server, fh, devlist),
6001                                 &exception);
6002         } while (exception.retry);
6003
6004         dprintk("%s: err=%d, num_devs=%u\n", __func__,
6005                 err, devlist->num_devs);
6006
6007         return err;
6008 }
6009 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
6010
6011 static int
6012 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6013 {
6014         struct nfs4_getdeviceinfo_args args = {
6015                 .pdev = pdev,
6016         };
6017         struct nfs4_getdeviceinfo_res res = {
6018                 .pdev = pdev,
6019         };
6020         struct rpc_message msg = {
6021                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6022                 .rpc_argp = &args,
6023                 .rpc_resp = &res,
6024         };
6025         int status;
6026
6027         dprintk("--> %s\n", __func__);
6028         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6029         dprintk("<-- %s status=%d\n", __func__, status);
6030
6031         return status;
6032 }
6033
6034 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6035 {
6036         struct nfs4_exception exception = { };
6037         int err;
6038
6039         do {
6040                 err = nfs4_handle_exception(server,
6041                                         _nfs4_proc_getdeviceinfo(server, pdev),
6042                                         &exception);
6043         } while (exception.retry);
6044         return err;
6045 }
6046 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6047
6048 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6049 {
6050         struct nfs4_layoutcommit_data *data = calldata;
6051         struct nfs_server *server = NFS_SERVER(data->args.inode);
6052
6053         if (nfs4_setup_sequence(server, &data->args.seq_args,
6054                                 &data->res.seq_res, task))
6055                 return;
6056         rpc_call_start(task);
6057 }
6058
6059 static void
6060 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6061 {
6062         struct nfs4_layoutcommit_data *data = calldata;
6063         struct nfs_server *server = NFS_SERVER(data->args.inode);
6064
6065         if (!nfs4_sequence_done(task, &data->res.seq_res))
6066                 return;
6067
6068         switch (task->tk_status) { /* Just ignore these failures */
6069         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6070         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6071         case NFS4ERR_BADLAYOUT:     /* no layout */
6072         case NFS4ERR_GRACE:         /* loca_recalim always false */
6073                 task->tk_status = 0;
6074         }
6075
6076         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6077                 rpc_restart_call_prepare(task);
6078                 return;
6079         }
6080
6081         if (task->tk_status == 0)
6082                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6083                                                    data->res.fattr);
6084 }
6085
6086 static void nfs4_layoutcommit_release(void *calldata)
6087 {
6088         struct nfs4_layoutcommit_data *data = calldata;
6089         struct pnfs_layout_segment *lseg, *tmp;
6090         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6091
6092         pnfs_cleanup_layoutcommit(data);
6093         /* Matched by references in pnfs_set_layoutcommit */
6094         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6095                 list_del_init(&lseg->pls_lc_list);
6096                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6097                                        &lseg->pls_flags))
6098                         put_lseg(lseg);
6099         }
6100
6101         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6102         smp_mb__after_clear_bit();
6103         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6104
6105         put_rpccred(data->cred);
6106         kfree(data);
6107 }
6108
6109 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6110         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6111         .rpc_call_done = nfs4_layoutcommit_done,
6112         .rpc_release = nfs4_layoutcommit_release,
6113 };
6114
6115 int
6116 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6117 {
6118         struct rpc_message msg = {
6119                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6120                 .rpc_argp = &data->args,
6121                 .rpc_resp = &data->res,
6122                 .rpc_cred = data->cred,
6123         };
6124         struct rpc_task_setup task_setup_data = {
6125                 .task = &data->task,
6126                 .rpc_client = NFS_CLIENT(data->args.inode),
6127                 .rpc_message = &msg,
6128                 .callback_ops = &nfs4_layoutcommit_ops,
6129                 .callback_data = data,
6130                 .flags = RPC_TASK_ASYNC,
6131         };
6132         struct rpc_task *task;
6133         int status = 0;
6134
6135         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6136                 "lbw: %llu inode %lu\n",
6137                 data->task.tk_pid, sync,
6138                 data->args.lastbytewritten,
6139                 data->args.inode->i_ino);
6140
6141         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6142         task = rpc_run_task(&task_setup_data);
6143         if (IS_ERR(task))
6144                 return PTR_ERR(task);
6145         if (sync == false)
6146                 goto out;
6147         status = nfs4_wait_for_completion_rpc_task(task);
6148         if (status != 0)
6149                 goto out;
6150         status = task->tk_status;
6151 out:
6152         dprintk("%s: status %d\n", __func__, status);
6153         rpc_put_task(task);
6154         return status;
6155 }
6156
6157 static int
6158 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6159                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6160 {
6161         struct nfs41_secinfo_no_name_args args = {
6162                 .style = SECINFO_STYLE_CURRENT_FH,
6163         };
6164         struct nfs4_secinfo_res res = {
6165                 .flavors = flavors,
6166         };
6167         struct rpc_message msg = {
6168                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6169                 .rpc_argp = &args,
6170                 .rpc_resp = &res,
6171         };
6172         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6173 }
6174
6175 static int
6176 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6177                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6178 {
6179         struct nfs4_exception exception = { };
6180         int err;
6181         do {
6182                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6183                 switch (err) {
6184                 case 0:
6185                 case -NFS4ERR_WRONGSEC:
6186                 case -NFS4ERR_NOTSUPP:
6187                         break;
6188                 default:
6189                         err = nfs4_handle_exception(server, err, &exception);
6190                 }
6191         } while (exception.retry);
6192         return err;
6193 }
6194
6195 static int
6196 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6197                     struct nfs_fsinfo *info)
6198 {
6199         int err;
6200         struct page *page;
6201         rpc_authflavor_t flavor;
6202         struct nfs4_secinfo_flavors *flavors;
6203
6204         page = alloc_page(GFP_KERNEL);
6205         if (!page) {
6206                 err = -ENOMEM;
6207                 goto out;
6208         }
6209
6210         flavors = page_address(page);
6211         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6212
6213         /*
6214          * Fall back on "guess and check" method if
6215          * the server doesn't support SECINFO_NO_NAME
6216          */
6217         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6218                 err = nfs4_find_root_sec(server, fhandle, info);
6219                 goto out_freepage;
6220         }
6221         if (err)
6222                 goto out_freepage;
6223
6224         flavor = nfs_find_best_sec(flavors);
6225         if (err == 0)
6226                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6227
6228 out_freepage:
6229         put_page(page);
6230         if (err == -EACCES)
6231                 return -EPERM;
6232 out:
6233         return err;
6234 }
6235
6236 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6237 {
6238         int status;
6239         struct nfs41_test_stateid_args args = {
6240                 .stateid = stateid,
6241         };
6242         struct nfs41_test_stateid_res res;
6243         struct rpc_message msg = {
6244                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6245                 .rpc_argp = &args,
6246                 .rpc_resp = &res,
6247         };
6248
6249         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6250         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6251
6252         if (status == NFS_OK)
6253                 return res.status;
6254         return status;
6255 }
6256
6257 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6258 {
6259         struct nfs4_exception exception = { };
6260         int err;
6261         do {
6262                 err = nfs4_handle_exception(server,
6263                                 _nfs41_test_stateid(server, stateid),
6264                                 &exception);
6265         } while (exception.retry);
6266         return err;
6267 }
6268
6269 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6270 {
6271         struct nfs41_free_stateid_args args = {
6272                 .stateid = stateid,
6273         };
6274         struct nfs41_free_stateid_res res;
6275         struct rpc_message msg = {
6276                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6277                 .rpc_argp = &args,
6278                 .rpc_resp = &res,
6279         };
6280
6281         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6282         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6283 }
6284
6285 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6286 {
6287         struct nfs4_exception exception = { };
6288         int err;
6289         do {
6290                 err = nfs4_handle_exception(server,
6291                                 _nfs4_free_stateid(server, stateid),
6292                                 &exception);
6293         } while (exception.retry);
6294         return err;
6295 }
6296
6297 static bool nfs41_match_stateid(const nfs4_stateid *s1,
6298                 const nfs4_stateid *s2)
6299 {
6300         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
6301                 return false;
6302
6303         if (s1->seqid == s2->seqid)
6304                 return true;
6305         if (s1->seqid == 0 || s2->seqid == 0)
6306                 return true;
6307
6308         return false;
6309 }
6310
6311 #endif /* CONFIG_NFS_V4_1 */
6312
6313 static bool nfs4_match_stateid(const nfs4_stateid *s1,
6314                 const nfs4_stateid *s2)
6315 {
6316         return nfs4_stateid_match(s1, s2);
6317 }
6318
6319
6320 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6321         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6322         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6323         .recover_open   = nfs4_open_reclaim,
6324         .recover_lock   = nfs4_lock_reclaim,
6325         .establish_clid = nfs4_init_clientid,
6326         .get_clid_cred  = nfs4_get_setclientid_cred,
6327 };
6328
6329 #if defined(CONFIG_NFS_V4_1)
6330 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6331         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6332         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6333         .recover_open   = nfs4_open_reclaim,
6334         .recover_lock   = nfs4_lock_reclaim,
6335         .establish_clid = nfs41_init_clientid,
6336         .get_clid_cred  = nfs4_get_exchange_id_cred,
6337         .reclaim_complete = nfs41_proc_reclaim_complete,
6338 };
6339 #endif /* CONFIG_NFS_V4_1 */
6340
6341 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6342         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6343         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6344         .recover_open   = nfs4_open_expired,
6345         .recover_lock   = nfs4_lock_expired,
6346         .establish_clid = nfs4_init_clientid,
6347         .get_clid_cred  = nfs4_get_setclientid_cred,
6348 };
6349
6350 #if defined(CONFIG_NFS_V4_1)
6351 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6352         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6353         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6354         .recover_open   = nfs41_open_expired,
6355         .recover_lock   = nfs41_lock_expired,
6356         .establish_clid = nfs41_init_clientid,
6357         .get_clid_cred  = nfs4_get_exchange_id_cred,
6358 };
6359 #endif /* CONFIG_NFS_V4_1 */
6360
6361 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6362         .sched_state_renewal = nfs4_proc_async_renew,
6363         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6364         .renew_lease = nfs4_proc_renew,
6365 };
6366
6367 #if defined(CONFIG_NFS_V4_1)
6368 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6369         .sched_state_renewal = nfs41_proc_async_sequence,
6370         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6371         .renew_lease = nfs4_proc_sequence,
6372 };
6373 #endif
6374
6375 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6376         .minor_version = 0,
6377         .call_sync = _nfs4_call_sync,
6378         .match_stateid = nfs4_match_stateid,
6379         .find_root_sec = nfs4_find_root_sec,
6380         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6381         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6382         .state_renewal_ops = &nfs40_state_renewal_ops,
6383 };
6384
6385 #if defined(CONFIG_NFS_V4_1)
6386 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6387         .minor_version = 1,
6388         .call_sync = _nfs4_call_sync_session,
6389         .match_stateid = nfs41_match_stateid,
6390         .find_root_sec = nfs41_find_root_sec,
6391         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6392         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6393         .state_renewal_ops = &nfs41_state_renewal_ops,
6394 };
6395 #endif
6396
6397 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6398         [0] = &nfs_v4_0_minor_ops,
6399 #if defined(CONFIG_NFS_V4_1)
6400         [1] = &nfs_v4_1_minor_ops,
6401 #endif
6402 };
6403
6404 static const struct inode_operations nfs4_file_inode_operations = {
6405         .permission     = nfs_permission,
6406         .getattr        = nfs_getattr,
6407         .setattr        = nfs_setattr,
6408         .getxattr       = generic_getxattr,
6409         .setxattr       = generic_setxattr,
6410         .listxattr      = generic_listxattr,
6411         .removexattr    = generic_removexattr,
6412 };
6413
6414 const struct nfs_rpc_ops nfs_v4_clientops = {
6415         .version        = 4,                    /* protocol version */
6416         .dentry_ops     = &nfs4_dentry_operations,
6417         .dir_inode_ops  = &nfs4_dir_inode_operations,
6418         .file_inode_ops = &nfs4_file_inode_operations,
6419         .file_ops       = &nfs4_file_operations,
6420         .getroot        = nfs4_proc_get_root,
6421         .getattr        = nfs4_proc_getattr,
6422         .setattr        = nfs4_proc_setattr,
6423         .lookup         = nfs4_proc_lookup,
6424         .access         = nfs4_proc_access,
6425         .readlink       = nfs4_proc_readlink,
6426         .create         = nfs4_proc_create,
6427         .remove         = nfs4_proc_remove,
6428         .unlink_setup   = nfs4_proc_unlink_setup,
6429         .unlink_done    = nfs4_proc_unlink_done,
6430         .rename         = nfs4_proc_rename,
6431         .rename_setup   = nfs4_proc_rename_setup,
6432         .rename_done    = nfs4_proc_rename_done,
6433         .link           = nfs4_proc_link,
6434         .symlink        = nfs4_proc_symlink,
6435         .mkdir          = nfs4_proc_mkdir,
6436         .rmdir          = nfs4_proc_remove,
6437         .readdir        = nfs4_proc_readdir,
6438         .mknod          = nfs4_proc_mknod,
6439         .statfs         = nfs4_proc_statfs,
6440         .fsinfo         = nfs4_proc_fsinfo,
6441         .pathconf       = nfs4_proc_pathconf,
6442         .set_capabilities = nfs4_server_capabilities,
6443         .decode_dirent  = nfs4_decode_dirent,
6444         .read_setup     = nfs4_proc_read_setup,
6445         .read_done      = nfs4_read_done,
6446         .write_setup    = nfs4_proc_write_setup,
6447         .write_done     = nfs4_write_done,
6448         .commit_setup   = nfs4_proc_commit_setup,
6449         .commit_done    = nfs4_commit_done,
6450         .lock           = nfs4_proc_lock,
6451         .clear_acl_cache = nfs4_zap_acl_attr,
6452         .close_context  = nfs4_close_context,
6453         .open_context   = nfs4_atomic_open,
6454         .init_client    = nfs4_init_client,
6455         .secinfo        = nfs4_proc_secinfo,
6456 };
6457
6458 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6459         .prefix = XATTR_NAME_NFSV4_ACL,
6460         .list   = nfs4_xattr_list_nfs4_acl,
6461         .get    = nfs4_xattr_get_nfs4_acl,
6462         .set    = nfs4_xattr_set_nfs4_acl,
6463 };
6464
6465 const struct xattr_handler *nfs4_xattr_handlers[] = {
6466         &nfs4_xattr_nfs4_acl_handler,
6467         NULL
6468 };
6469
6470 module_param(max_session_slots, ushort, 0644);
6471 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6472                 "requests the client will negotiate");
6473
6474 /*
6475  * Local variables:
6476  *  c-basic-offset: 8
6477  * End:
6478  */