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         int ret = errorcode;
 266
 267         exception->retry = 0;
 268         switch(errorcode) {
 269                 case 0:
 270                         return 0;
 271                 case -NFS4ERR_ADMIN_REVOKED:
 272                 case -NFS4ERR_BAD_STATEID:
 273                 case -NFS4ERR_OPENMODE:
 274                         if (state == NULL)
 275                                 break;
 276                         nfs4_schedule_stateid_recovery(server, state);
 277                         goto wait_on_recovery;
 278                 case -NFS4ERR_EXPIRED:
 279                         if (state != NULL)
 280                                 nfs4_schedule_stateid_recovery(server, state);
 281                 case -NFS4ERR_STALE_STATEID:
 282                 case -NFS4ERR_STALE_CLIENTID:
 283                         nfs4_schedule_lease_recovery(clp);
 284                         goto wait_on_recovery;
 285 #if defined(CONFIG_NFS_V4_1)
 286                 case -NFS4ERR_BADSESSION:
 287                 case -NFS4ERR_BADSLOT:
 288                 case -NFS4ERR_BAD_HIGH_SLOT:
 289                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 290                 case -NFS4ERR_DEADSESSION:
 291                 case -NFS4ERR_SEQ_FALSE_RETRY:
 292                 case -NFS4ERR_SEQ_MISORDERED:
 293                         dprintk("%s ERROR: %d Reset session\n", __func__,
 294                                 errorcode);
 295                         nfs4_schedule_session_recovery(clp->cl_session);
 296                         exception->retry = 1;
 297                         break;
 298 #endif /* defined(CONFIG_NFS_V4_1) */
 299                 case -NFS4ERR_FILE_OPEN:
 300                         if (exception->timeout > HZ) {
 301                                 /* We have retried a decent amount, time to
 302                                  * fail
 303                                  */
 304                                 ret = -EBUSY;
 305                                 break;
 306                         }
 307                 case -NFS4ERR_GRACE:
 308                 case -NFS4ERR_DELAY:
 309                 case -EKEYEXPIRED:
 310                         ret = nfs4_delay(server->client, &exception->timeout);
 311                         if (ret != 0)
 312                                 break;
 313                 case -NFS4ERR_RETRY_UNCACHED_REP:
 314                 case -NFS4ERR_OLD_STATEID:
 315                         exception->retry = 1;
 316                         break;
 317                 case -NFS4ERR_BADOWNER:
 318                         /* The following works around a Linux server bug! */
 319                 case -NFS4ERR_BADNAME:
 320                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 321                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 322                                 exception->retry = 1;
 323                                 printk(KERN_WARNING "NFS: v4 server %s "
 324                                                 "does not accept raw "
 325                                                 "uid/gids. "
 326                                                 "Reenabling the idmapper.\n",
 327                                                 server->nfs_client->cl_hostname);
 328                         }
 329         }
 330         /* We failed to handle the error */
 331         return nfs4_map_errors(ret);
 332 wait_on_recovery:
 333         ret = nfs4_wait_clnt_recover(clp);
 334         if (ret == 0)
 335                 exception->retry = 1;
 336         return ret;
 337 }
 338
 339
 340 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 341 {
 342         spin_lock(&clp->cl_lock);
 343         if (time_before(clp->cl_last_renewal,timestamp))
 344                 clp->cl_last_renewal = timestamp;
 345         spin_unlock(&clp->cl_lock);
 346 }
 347
 348 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 349 {
 350         do_renew_lease(server->nfs_client, timestamp);
 351 }
 352
 353 #if defined(CONFIG_NFS_V4_1)
 354
 355 /*
 356  * nfs4_free_slot - free a slot and efficiently update slot table.
 357  *
 358  * freeing a slot is trivially done by clearing its respective bit
 359  * in the bitmap.
 360  * If the freed slotid equals highest_used_slotid we want to update it
 361  * so that the server would be able to size down the slot table if needed,
 362  * otherwise we know that the highest_used_slotid is still in use.
 363  * When updating highest_used_slotid there may be "holes" in the bitmap
 364  * so we need to scan down from highest_used_slotid to 0 looking for the now
 365  * highest slotid in use.
 366  * If none found, highest_used_slotid is set to NFS4_NO_SLOT.
 367  *
 368  * Must be called while holding tbl->slot_tbl_lock
 369  */
 370 static void
 371 nfs4_free_slot(struct nfs4_slot_table *tbl, u32 slotid)
 372 {
 373         BUG_ON(slotid >= NFS4_MAX_SLOT_TABLE);
 374         /* clear used bit in bitmap */
 375         __clear_bit(slotid, tbl->used_slots);
 376
 377         /* update highest_used_slotid when it is freed */
 378         if (slotid == tbl->highest_used_slotid) {
 379                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 380                 if (slotid < tbl->max_slots)
 381                         tbl->highest_used_slotid = slotid;
 382                 else
 383                         tbl->highest_used_slotid = NFS4_NO_SLOT;
 384         }
 385         dprintk("%s: slotid %u highest_used_slotid %d\n", __func__,
 386                 slotid, tbl->highest_used_slotid);
 387 }
 388
 389 bool nfs4_set_task_privileged(struct rpc_task *task, void *dummy)
 390 {
 391         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 392         return true;
 393 }
 394
 395 /*
 396  * Signal state manager thread if session fore channel is drained
 397  */
 398 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 399 {
 400         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 401                 rpc_wake_up_first(&ses->fc_slot_table.slot_tbl_waitq,
 402                                 nfs4_set_task_privileged, NULL);
 403                 return;
 404         }
 405
 406         if (ses->fc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 407                 return;
 408
 409         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 410         complete(&ses->fc_slot_table.complete);
 411 }
 412
 413 /*
 414  * Signal state manager thread if session back channel is drained
 415  */
 416 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 417 {
 418         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 419             ses->bc_slot_table.highest_used_slotid != NFS4_NO_SLOT)
 420                 return;
 421         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 422         complete(&ses->bc_slot_table.complete);
 423 }
 424
 425 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 426 {
 427         struct nfs4_slot_table *tbl;
 428
 429         tbl = &res->sr_session->fc_slot_table;
 430         if (!res->sr_slot) {
 431                 /* just wake up the next guy waiting since
 432                  * we may have not consumed a slot after all */
 433                 dprintk("%s: No slot\n", __func__);
 434                 return;
 435         }
 436
 437         spin_lock(&tbl->slot_tbl_lock);
 438         nfs4_free_slot(tbl, res->sr_slot - tbl->slots);
 439         nfs4_check_drain_fc_complete(res->sr_session);
 440         spin_unlock(&tbl->slot_tbl_lock);
 441         res->sr_slot = NULL;
 442 }
 443
 444 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 445 {
 446         unsigned long timestamp;
 447         struct nfs_client *clp;
 448
 449         /*
 450          * sr_status remains 1 if an RPC level error occurred. The server
 451          * may or may not have processed the sequence operation..
 452          * Proceed as if the server received and processed the sequence
 453          * operation.
 454          */
 455         if (res->sr_status == 1)
 456                 res->sr_status = NFS_OK;
 457
 458         /* don't increment the sequence number if the task wasn't sent */
 459         if (!RPC_WAS_SENT(task))
 460                 goto out;
 461
 462         /* Check the SEQUENCE operation status */
 463         switch (res->sr_status) {
 464         case 0:
 465                 /* Update the slot's sequence and clientid lease timer */
 466                 ++res->sr_slot->seq_nr;
 467                 timestamp = res->sr_renewal_time;
 468                 clp = res->sr_session->clp;
 469                 do_renew_lease(clp, timestamp);
 470                 /* Check sequence flags */
 471                 if (res->sr_status_flags != 0)
 472                         nfs4_schedule_lease_recovery(clp);
 473                 break;
 474         case -NFS4ERR_DELAY:
 475                 /* The server detected a resend of the RPC call and
 476                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 477                  * of RFC5661.
 478                  */
 479                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 480                         __func__,
 481                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 482                         res->sr_slot->seq_nr);
 483                 goto out_retry;
 484         default:
 485                 /* Just update the slot sequence no. */
 486                 ++res->sr_slot->seq_nr;
 487         }
 488 out:
 489         /* The session may be reset by one of the error handlers. */
 490         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 491         nfs41_sequence_free_slot(res);
 492         return 1;
 493 out_retry:
 494         if (!rpc_restart_call(task))
 495                 goto out;
 496         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 497         return 0;
 498 }
 499
 500 static int nfs4_sequence_done(struct rpc_task *task,
 501                                struct nfs4_sequence_res *res)
 502 {
 503         if (res->sr_session == NULL)
 504                 return 1;
 505         return nfs41_sequence_done(task, res);
 506 }
 507
 508 /*
 509  * nfs4_find_slot - efficiently look for a free slot
 510  *
 511  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 512  * If found, we mark the slot as used, update the highest_used_slotid,
 513  * and respectively set up the sequence operation args.
 514  * The slot number is returned if found, or NFS4_NO_SLOT otherwise.
 515  *
 516  * Note: must be called with under the slot_tbl_lock.
 517  */
 518 static u32
 519 nfs4_find_slot(struct nfs4_slot_table *tbl)
 520 {
 521         u32 slotid;
 522         u32 ret_id = NFS4_NO_SLOT;
 523
 524         dprintk("--> %s used_slots=%04lx highest_used=%u max_slots=%u\n",
 525                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 526                 tbl->max_slots);
 527         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 528         if (slotid >= tbl->max_slots)
 529                 goto out;
 530         __set_bit(slotid, tbl->used_slots);
 531         if (slotid > tbl->highest_used_slotid ||
 532                         tbl->highest_used_slotid == NFS4_NO_SLOT)
 533                 tbl->highest_used_slotid = slotid;
 534         ret_id = slotid;
 535 out:
 536         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 537                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 538         return ret_id;
 539 }
 540
 541 static void nfs41_init_sequence(struct nfs4_sequence_args *args,
 542                 struct nfs4_sequence_res *res, int cache_reply)
 543 {
 544         args->sa_session = NULL;
 545         args->sa_cache_this = 0;
 546         if (cache_reply)
 547                 args->sa_cache_this = 1;
 548         res->sr_session = NULL;
 549         res->sr_slot = NULL;
 550 }
 551
 552 int nfs41_setup_sequence(struct nfs4_session *session,
 553                                 struct nfs4_sequence_args *args,
 554                                 struct nfs4_sequence_res *res,
 555                                 struct rpc_task *task)
 556 {
 557         struct nfs4_slot *slot;
 558         struct nfs4_slot_table *tbl;
 559         u32 slotid;
 560
 561         dprintk("--> %s\n", __func__);
 562         /* slot already allocated? */
 563         if (res->sr_slot != NULL)
 564                 return 0;
 565
 566         tbl = &session->fc_slot_table;
 567
 568         spin_lock(&tbl->slot_tbl_lock);
 569         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 570             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 571                 /* The state manager will wait until the slot table is empty */
 572                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 573                 spin_unlock(&tbl->slot_tbl_lock);
 574                 dprintk("%s session is draining\n", __func__);
 575                 return -EAGAIN;
 576         }
 577
 578         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 579             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 580                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 581                 spin_unlock(&tbl->slot_tbl_lock);
 582                 dprintk("%s enforce FIFO order\n", __func__);
 583                 return -EAGAIN;
 584         }
 585
 586         slotid = nfs4_find_slot(tbl);
 587         if (slotid == NFS4_NO_SLOT) {
 588                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 589                 spin_unlock(&tbl->slot_tbl_lock);
 590                 dprintk("<-- %s: no free slots\n", __func__);
 591                 return -EAGAIN;
 592         }
 593         spin_unlock(&tbl->slot_tbl_lock);
 594
 595         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 596         slot = tbl->slots + slotid;
 597         args->sa_session = session;
 598         args->sa_slotid = slotid;
 599
 600         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 601
 602         res->sr_session = session;
 603         res->sr_slot = slot;
 604         res->sr_renewal_time = jiffies;
 605         res->sr_status_flags = 0;
 606         /*
 607          * sr_status is only set in decode_sequence, and so will remain
 608          * set to 1 if an rpc level failure occurs.
 609          */
 610         res->sr_status = 1;
 611         return 0;
 612 }
 613 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 614
 615 int nfs4_setup_sequence(const struct nfs_server *server,
 616                         struct nfs4_sequence_args *args,
 617                         struct nfs4_sequence_res *res,
 618                         struct rpc_task *task)
 619 {
 620         struct nfs4_session *session = nfs4_get_session(server);
 621         int ret = 0;
 622
 623         if (session == NULL)
 624                 goto out;
 625
 626         dprintk("--> %s clp %p session %p sr_slot %td\n",
 627                 __func__, session->clp, session, res->sr_slot ?
 628                         res->sr_slot - session->fc_slot_table.slots : -1);
 629
 630         ret = nfs41_setup_sequence(session, args, res, task);
 631 out:
 632         dprintk("<-- %s status=%d\n", __func__, ret);
 633         return ret;
 634 }
 635
 636 struct nfs41_call_sync_data {
 637         const struct nfs_server *seq_server;
 638         struct nfs4_sequence_args *seq_args;
 639         struct nfs4_sequence_res *seq_res;
 640 };
 641
 642 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 643 {
 644         struct nfs41_call_sync_data *data = calldata;
 645
 646         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 647
 648         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 649                                 data->seq_res, task))
 650                 return;
 651         rpc_call_start(task);
 652 }
 653
 654 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 655 {
 656         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 657         nfs41_call_sync_prepare(task, calldata);
 658 }
 659
 660 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 661 {
 662         struct nfs41_call_sync_data *data = calldata;
 663
 664         nfs41_sequence_done(task, data->seq_res);
 665 }
 666
 667 struct rpc_call_ops nfs41_call_sync_ops = {
 668         .rpc_call_prepare = nfs41_call_sync_prepare,
 669         .rpc_call_done = nfs41_call_sync_done,
 670 };
 671
 672 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 673         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 674         .rpc_call_done = nfs41_call_sync_done,
 675 };
 676
 677 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 678                                    struct nfs_server *server,
 679                                    struct rpc_message *msg,
 680                                    struct nfs4_sequence_args *args,
 681                                    struct nfs4_sequence_res *res,
 682                                    int privileged)
 683 {
 684         int ret;
 685         struct rpc_task *task;
 686         struct nfs41_call_sync_data data = {
 687                 .seq_server = server,
 688                 .seq_args = args,
 689                 .seq_res = res,
 690         };
 691         struct rpc_task_setup task_setup = {
 692                 .rpc_client = clnt,
 693                 .rpc_message = msg,
 694                 .callback_ops = &nfs41_call_sync_ops,
 695                 .callback_data = &data
 696         };
 697
 698         if (privileged)
 699                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 700         task = rpc_run_task(&task_setup);
 701         if (IS_ERR(task))
 702                 ret = PTR_ERR(task);
 703         else {
 704                 ret = task->tk_status;
 705                 rpc_put_task(task);
 706         }
 707         return ret;
 708 }
 709
 710 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 711                             struct nfs_server *server,
 712                             struct rpc_message *msg,
 713                             struct nfs4_sequence_args *args,
 714                             struct nfs4_sequence_res *res,
 715                             int cache_reply)
 716 {
 717         nfs41_init_sequence(args, res, cache_reply);
 718         return nfs4_call_sync_sequence(clnt, server, msg, args, res, 0);
 719 }
 720
 721 #else
 722 static inline
 723 void nfs41_init_sequence(struct nfs4_sequence_args *args,
 724                 struct nfs4_sequence_res *res, int cache_reply)
 725 {
 726 }
 727
 728 static int nfs4_sequence_done(struct rpc_task *task,
 729                                struct nfs4_sequence_res *res)
 730 {
 731         return 1;
 732 }
 733 #endif /* CONFIG_NFS_V4_1 */
 734
 735 int _nfs4_call_sync(struct rpc_clnt *clnt,
 736                     struct nfs_server *server,
 737                     struct rpc_message *msg,
 738                     struct nfs4_sequence_args *args,
 739                     struct nfs4_sequence_res *res,
 740                     int cache_reply)
 741 {
 742         nfs41_init_sequence(args, res, cache_reply);
 743         return rpc_call_sync(clnt, msg, 0);
 744 }
 745
 746 static inline
 747 int nfs4_call_sync(struct rpc_clnt *clnt,
 748                    struct nfs_server *server,
 749                    struct rpc_message *msg,
 750                    struct nfs4_sequence_args *args,
 751                    struct nfs4_sequence_res *res,
 752                    int cache_reply)
 753 {
 754         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 755                                                 args, res, cache_reply);
 756 }
 757
 758 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 759 {
 760         struct nfs_inode *nfsi = NFS_I(dir);
 761
 762         spin_lock(&dir->i_lock);
 763         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 764         if (!cinfo->atomic || cinfo->before != dir->i_version)
 765                 nfs_force_lookup_revalidate(dir);
 766         dir->i_version = cinfo->after;
 767         spin_unlock(&dir->i_lock);
 768 }
 769
 770 struct nfs4_opendata {
 771         struct kref kref;
 772         struct nfs_openargs o_arg;
 773         struct nfs_openres o_res;
 774         struct nfs_open_confirmargs c_arg;
 775         struct nfs_open_confirmres c_res;
 776         struct nfs4_string owner_name;
 777         struct nfs4_string group_name;
 778         struct nfs_fattr f_attr;
 779         struct nfs_fattr dir_attr;
 780         struct dentry *dir;
 781         struct dentry *dentry;
 782         struct nfs4_state_owner *owner;
 783         struct nfs4_state *state;
 784         struct iattr attrs;
 785         unsigned long timestamp;
 786         unsigned int rpc_done : 1;
 787         int rpc_status;
 788         int cancelled;
 789 };
 790
 791
 792 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 793 {
 794         p->o_res.f_attr = &p->f_attr;
 795         p->o_res.dir_attr = &p->dir_attr;
 796         p->o_res.seqid = p->o_arg.seqid;
 797         p->c_res.seqid = p->c_arg.seqid;
 798         p->o_res.server = p->o_arg.server;
 799         nfs_fattr_init(&p->f_attr);
 800         nfs_fattr_init(&p->dir_attr);
 801         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 802 }
 803
 804 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 805                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 806                 const struct iattr *attrs,
 807                 gfp_t gfp_mask)
 808 {
 809         struct dentry *parent = dget_parent(dentry);
 810         struct inode *dir = parent->d_inode;
 811         struct nfs_server *server = NFS_SERVER(dir);
 812         struct nfs4_opendata *p;
 813
 814         p = kzalloc(sizeof(*p), gfp_mask);
 815         if (p == NULL)
 816                 goto err;
 817         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 818         if (p->o_arg.seqid == NULL)
 819                 goto err_free;
 820         nfs_sb_active(dentry->d_sb);
 821         p->dentry = dget(dentry);
 822         p->dir = parent;
 823         p->owner = sp;
 824         atomic_inc(&sp->so_count);
 825         p->o_arg.fh = NFS_FH(dir);
 826         p->o_arg.open_flags = flags;
 827         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 828         p->o_arg.clientid = server->nfs_client->cl_clientid;
 829         p->o_arg.id = sp->so_seqid.owner_id;
 830         p->o_arg.name = &dentry->d_name;
 831         p->o_arg.server = server;
 832         p->o_arg.bitmask = server->attr_bitmask;
 833         p->o_arg.dir_bitmask = server->cache_consistency_bitmask;
 834         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 835         if (attrs != NULL && attrs->ia_valid != 0) {
 836                 u32 *s;
 837
 838                 p->o_arg.u.attrs = &p->attrs;
 839                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 840                 s = (u32 *) p->o_arg.u.verifier.data;
 841                 s[0] = jiffies;
 842                 s[1] = current->pid;
 843         }
 844         p->c_arg.fh = &p->o_res.fh;
 845         p->c_arg.stateid = &p->o_res.stateid;
 846         p->c_arg.seqid = p->o_arg.seqid;
 847         nfs4_init_opendata_res(p);
 848         kref_init(&p->kref);
 849         return p;
 850 err_free:
 851         kfree(p);
 852 err:
 853         dput(parent);
 854         return NULL;
 855 }
 856
 857 static void nfs4_opendata_free(struct kref *kref)
 858 {
 859         struct nfs4_opendata *p = container_of(kref,
 860                         struct nfs4_opendata, kref);
 861         struct super_block *sb = p->dentry->d_sb;
 862
 863         nfs_free_seqid(p->o_arg.seqid);
 864         if (p->state != NULL)
 865                 nfs4_put_open_state(p->state);
 866         nfs4_put_state_owner(p->owner);
 867         dput(p->dir);
 868         dput(p->dentry);
 869         nfs_sb_deactive(sb);
 870         nfs_fattr_free_names(&p->f_attr);
 871         kfree(p);
 872 }
 873
 874 static void nfs4_opendata_put(struct nfs4_opendata *p)
 875 {
 876         if (p != NULL)
 877                 kref_put(&p->kref, nfs4_opendata_free);
 878 }
 879
 880 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 881 {
 882         int ret;
 883
 884         ret = rpc_wait_for_completion_task(task);
 885         return ret;
 886 }
 887
 888 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 889 {
 890         int ret = 0;
 891
 892         if (open_mode & (O_EXCL|O_TRUNC))
 893                 goto out;
 894         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 895                 case FMODE_READ:
 896                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 897                                 && state->n_rdonly != 0;
 898                         break;
 899                 case FMODE_WRITE:
 900                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 901                                 && state->n_wronly != 0;
 902                         break;
 903                 case FMODE_READ|FMODE_WRITE:
 904                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 905                                 && state->n_rdwr != 0;
 906         }
 907 out:
 908         return ret;
 909 }
 910
 911 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 912 {
 913         if (delegation == NULL)
 914                 return 0;
 915         if ((delegation->type & fmode) != fmode)
 916                 return 0;
 917         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 918                 return 0;
 919         nfs_mark_delegation_referenced(delegation);
 920         return 1;
 921 }
 922
 923 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 924 {
 925         switch (fmode) {
 926                 case FMODE_WRITE:
 927                         state->n_wronly++;
 928                         break;
 929                 case FMODE_READ:
 930                         state->n_rdonly++;
 931                         break;
 932                 case FMODE_READ|FMODE_WRITE:
 933                         state->n_rdwr++;
 934         }
 935         nfs4_state_set_mode_locked(state, state->state | fmode);
 936 }
 937
 938 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 939 {
 940         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 941                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 942         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 943         switch (fmode) {
 944                 case FMODE_READ:
 945                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 946                         break;
 947                 case FMODE_WRITE:
 948                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 949                         break;
 950                 case FMODE_READ|FMODE_WRITE:
 951                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 952         }
 953 }
 954
 955 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 956 {
 957         write_seqlock(&state->seqlock);
 958         nfs_set_open_stateid_locked(state, stateid, fmode);
 959         write_sequnlock(&state->seqlock);
 960 }
 961
 962 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 963 {
 964         /*
 965          * Protect the call to nfs4_state_set_mode_locked and
 966          * serialise the stateid update
 967          */
 968         write_seqlock(&state->seqlock);
 969         if (deleg_stateid != NULL) {
 970                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 971                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 972         }
 973         if (open_stateid != NULL)
 974                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 975         write_sequnlock(&state->seqlock);
 976         spin_lock(&state->owner->so_lock);
 977         update_open_stateflags(state, fmode);
 978         spin_unlock(&state->owner->so_lock);
 979 }
 980
 981 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 982 {
 983         struct nfs_inode *nfsi = NFS_I(state->inode);
 984         struct nfs_delegation *deleg_cur;
 985         int ret = 0;
 986
 987         fmode &= (FMODE_READ|FMODE_WRITE);
 988
 989         rcu_read_lock();
 990         deleg_cur = rcu_dereference(nfsi->delegation);
 991         if (deleg_cur == NULL)
 992                 goto no_delegation;
 993
 994         spin_lock(&deleg_cur->lock);
 995         if (nfsi->delegation != deleg_cur ||
 996             (deleg_cur->type & fmode) != fmode)
 997                 goto no_delegation_unlock;
 998
 999         if (delegation == NULL)
1000                 delegation = &deleg_cur->stateid;
1001         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
1002                 goto no_delegation_unlock;
1003
1004         nfs_mark_delegation_referenced(deleg_cur);
1005         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1006         ret = 1;
1007 no_delegation_unlock:
1008         spin_unlock(&deleg_cur->lock);
1009 no_delegation:
1010         rcu_read_unlock();
1011
1012         if (!ret && open_stateid != NULL) {
1013                 __update_open_stateid(state, open_stateid, NULL, fmode);
1014                 ret = 1;
1015         }
1016
1017         return ret;
1018 }
1019
1020
1021 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1022 {
1023         struct nfs_delegation *delegation;
1024
1025         rcu_read_lock();
1026         delegation = rcu_dereference(NFS_I(inode)->delegation);
1027         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1028                 rcu_read_unlock();
1029                 return;
1030         }
1031         rcu_read_unlock();
1032         nfs_inode_return_delegation(inode);
1033 }
1034
1035 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1036 {
1037         struct nfs4_state *state = opendata->state;
1038         struct nfs_inode *nfsi = NFS_I(state->inode);
1039         struct nfs_delegation *delegation;
1040         int open_mode = opendata->o_arg.open_flags & (O_EXCL|O_TRUNC);
1041         fmode_t fmode = opendata->o_arg.fmode;
1042         nfs4_stateid stateid;
1043         int ret = -EAGAIN;
1044
1045         for (;;) {
1046                 if (can_open_cached(state, fmode, open_mode)) {
1047                         spin_lock(&state->owner->so_lock);
1048                         if (can_open_cached(state, fmode, open_mode)) {
1049                                 update_open_stateflags(state, fmode);
1050                                 spin_unlock(&state->owner->so_lock);
1051                                 goto out_return_state;
1052                         }
1053                         spin_unlock(&state->owner->so_lock);
1054                 }
1055                 rcu_read_lock();
1056                 delegation = rcu_dereference(nfsi->delegation);
1057                 if (!can_open_delegated(delegation, fmode)) {
1058                         rcu_read_unlock();
1059                         break;
1060                 }
1061                 /* Save the delegation */
1062                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1063                 rcu_read_unlock();
1064                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1065                 if (ret != 0)
1066                         goto out;
1067                 ret = -EAGAIN;
1068
1069                 /* Try to update the stateid using the delegation */
1070                 if (update_open_stateid(state, NULL, &stateid, fmode))
1071                         goto out_return_state;
1072         }
1073 out:
1074         return ERR_PTR(ret);
1075 out_return_state:
1076         atomic_inc(&state->count);
1077         return state;
1078 }
1079
1080 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1081 {
1082         struct inode *inode;
1083         struct nfs4_state *state = NULL;
1084         struct nfs_delegation *delegation;
1085         int ret;
1086
1087         if (!data->rpc_done) {
1088                 state = nfs4_try_open_cached(data);
1089                 goto out;
1090         }
1091
1092         ret = -EAGAIN;
1093         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1094                 goto err;
1095         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1096         ret = PTR_ERR(inode);
1097         if (IS_ERR(inode))
1098                 goto err;
1099         ret = -ENOMEM;
1100         state = nfs4_get_open_state(inode, data->owner);
1101         if (state == NULL)
1102                 goto err_put_inode;
1103         if (data->o_res.delegation_type != 0) {
1104                 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
1105                 int delegation_flags = 0;
1106
1107                 rcu_read_lock();
1108                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1109                 if (delegation)
1110                         delegation_flags = delegation->flags;
1111                 rcu_read_unlock();
1112                 if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1113                         pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1114                                         "returning a delegation for "
1115                                         "OPEN(CLAIM_DELEGATE_CUR)\n",
1116                                         clp->cl_hostname);
1117                 } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1118                         nfs_inode_set_delegation(state->inode,
1119                                         data->owner->so_cred,
1120                                         &data->o_res);
1121                 else
1122                         nfs_inode_reclaim_delegation(state->inode,
1123                                         data->owner->so_cred,
1124                                         &data->o_res);
1125         }
1126
1127         update_open_stateid(state, &data->o_res.stateid, NULL,
1128                         data->o_arg.fmode);
1129         iput(inode);
1130 out:
1131         return state;
1132 err_put_inode:
1133         iput(inode);
1134 err:
1135         return ERR_PTR(ret);
1136 }
1137
1138 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1139 {
1140         struct nfs_inode *nfsi = NFS_I(state->inode);
1141         struct nfs_open_context *ctx;
1142
1143         spin_lock(&state->inode->i_lock);
1144         list_for_each_entry(ctx, &nfsi->open_files, list) {
1145                 if (ctx->state != state)
1146                         continue;
1147                 get_nfs_open_context(ctx);
1148                 spin_unlock(&state->inode->i_lock);
1149                 return ctx;
1150         }
1151         spin_unlock(&state->inode->i_lock);
1152         return ERR_PTR(-ENOENT);
1153 }
1154
1155 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1156 {
1157         struct nfs4_opendata *opendata;
1158
1159         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0, NULL, GFP_NOFS);
1160         if (opendata == NULL)
1161                 return ERR_PTR(-ENOMEM);
1162         opendata->state = state;
1163         atomic_inc(&state->count);
1164         return opendata;
1165 }
1166
1167 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1168 {
1169         struct nfs4_state *newstate;
1170         int ret;
1171
1172         opendata->o_arg.open_flags = 0;
1173         opendata->o_arg.fmode = fmode;
1174         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1175         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1176         nfs4_init_opendata_res(opendata);
1177         ret = _nfs4_recover_proc_open(opendata);
1178         if (ret != 0)
1179                 return ret;
1180         newstate = nfs4_opendata_to_nfs4_state(opendata);
1181         if (IS_ERR(newstate))
1182                 return PTR_ERR(newstate);
1183         nfs4_close_state(newstate, fmode);
1184         *res = newstate;
1185         return 0;
1186 }
1187
1188 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1189 {
1190         struct nfs4_state *newstate;
1191         int ret;
1192
1193         /* memory barrier prior to reading state->n_* */
1194         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1195         smp_rmb();
1196         if (state->n_rdwr != 0) {
1197                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1198                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1199                 if (ret != 0)
1200                         return ret;
1201                 if (newstate != state)
1202                         return -ESTALE;
1203         }
1204         if (state->n_wronly != 0) {
1205                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1206                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1207                 if (ret != 0)
1208                         return ret;
1209                 if (newstate != state)
1210                         return -ESTALE;
1211         }
1212         if (state->n_rdonly != 0) {
1213                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1214                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1215                 if (ret != 0)
1216                         return ret;
1217                 if (newstate != state)
1218                         return -ESTALE;
1219         }
1220         /*
1221          * We may have performed cached opens for all three recoveries.
1222          * Check if we need to update the current stateid.
1223          */
1224         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1225             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1226                 write_seqlock(&state->seqlock);
1227                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1228                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1229                 write_sequnlock(&state->seqlock);
1230         }
1231         return 0;
1232 }
1233
1234 /*
1235  * OPEN_RECLAIM:
1236  *      reclaim state on the server after a reboot.
1237  */
1238 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1239 {
1240         struct nfs_delegation *delegation;
1241         struct nfs4_opendata *opendata;
1242         fmode_t delegation_type = 0;
1243         int status;
1244
1245         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1246         if (IS_ERR(opendata))
1247                 return PTR_ERR(opendata);
1248         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1249         opendata->o_arg.fh = NFS_FH(state->inode);
1250         rcu_read_lock();
1251         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1252         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1253                 delegation_type = delegation->type;
1254         rcu_read_unlock();
1255         opendata->o_arg.u.delegation_type = delegation_type;
1256         status = nfs4_open_recover(opendata, state);
1257         nfs4_opendata_put(opendata);
1258         return status;
1259 }
1260
1261 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1262 {
1263         struct nfs_server *server = NFS_SERVER(state->inode);
1264         struct nfs4_exception exception = { };
1265         int err;
1266         do {
1267                 err = _nfs4_do_open_reclaim(ctx, state);
1268                 if (err != -NFS4ERR_DELAY)
1269                         break;
1270                 nfs4_handle_exception(server, err, &exception);
1271         } while (exception.retry);
1272         return err;
1273 }
1274
1275 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1276 {
1277         struct nfs_open_context *ctx;
1278         int ret;
1279
1280         ctx = nfs4_state_find_open_context(state);
1281         if (IS_ERR(ctx))
1282                 return PTR_ERR(ctx);
1283         ret = nfs4_do_open_reclaim(ctx, state);
1284         put_nfs_open_context(ctx);
1285         return ret;
1286 }
1287
1288 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1289 {
1290         struct nfs4_opendata *opendata;
1291         int ret;
1292
1293         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1294         if (IS_ERR(opendata))
1295                 return PTR_ERR(opendata);
1296         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1297         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1298                         sizeof(opendata->o_arg.u.delegation.data));
1299         ret = nfs4_open_recover(opendata, state);
1300         nfs4_opendata_put(opendata);
1301         return ret;
1302 }
1303
1304 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1305 {
1306         struct nfs4_exception exception = { };
1307         struct nfs_server *server = NFS_SERVER(state->inode);
1308         int err;
1309         do {
1310                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1311                 switch (err) {
1312                         case 0:
1313                         case -ENOENT:
1314                         case -ESTALE:
1315                                 goto out;
1316                         case -NFS4ERR_BADSESSION:
1317                         case -NFS4ERR_BADSLOT:
1318                         case -NFS4ERR_BAD_HIGH_SLOT:
1319                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1320                         case -NFS4ERR_DEADSESSION:
1321                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1322                                 goto out;
1323                         case -NFS4ERR_STALE_CLIENTID:
1324                         case -NFS4ERR_STALE_STATEID:
1325                         case -NFS4ERR_EXPIRED:
1326                                 /* Don't recall a delegation if it was lost */
1327                                 nfs4_schedule_lease_recovery(server->nfs_client);
1328                                 goto out;
1329                         case -ERESTARTSYS:
1330                                 /*
1331                                  * The show must go on: exit, but mark the
1332                                  * stateid as needing recovery.
1333                                  */
1334                         case -NFS4ERR_ADMIN_REVOKED:
1335                         case -NFS4ERR_BAD_STATEID:
1336                                 nfs4_schedule_stateid_recovery(server, state);
1337                         case -EKEYEXPIRED:
1338                                 /*
1339                                  * User RPCSEC_GSS context has expired.
1340                                  * We cannot recover this stateid now, so
1341                                  * skip it and allow recovery thread to
1342                                  * proceed.
1343                                  */
1344                         case -ENOMEM:
1345                                 err = 0;
1346                                 goto out;
1347                 }
1348                 err = nfs4_handle_exception(server, err, &exception);
1349         } while (exception.retry);
1350 out:
1351         return err;
1352 }
1353
1354 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1355 {
1356         struct nfs4_opendata *data = calldata;
1357
1358         data->rpc_status = task->tk_status;
1359         if (data->rpc_status == 0) {
1360                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1361                                 sizeof(data->o_res.stateid.data));
1362                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1363                 renew_lease(data->o_res.server, data->timestamp);
1364                 data->rpc_done = 1;
1365         }
1366 }
1367
1368 static void nfs4_open_confirm_release(void *calldata)
1369 {
1370         struct nfs4_opendata *data = calldata;
1371         struct nfs4_state *state = NULL;
1372
1373         /* If this request hasn't been cancelled, do nothing */
1374         if (data->cancelled == 0)
1375                 goto out_free;
1376         /* In case of error, no cleanup! */
1377         if (!data->rpc_done)
1378                 goto out_free;
1379         state = nfs4_opendata_to_nfs4_state(data);
1380         if (!IS_ERR(state))
1381                 nfs4_close_state(state, data->o_arg.fmode);
1382 out_free:
1383         nfs4_opendata_put(data);
1384 }
1385
1386 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1387         .rpc_call_done = nfs4_open_confirm_done,
1388         .rpc_release = nfs4_open_confirm_release,
1389 };
1390
1391 /*
1392  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1393  */
1394 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1395 {
1396         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1397         struct rpc_task *task;
1398         struct  rpc_message msg = {
1399                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1400                 .rpc_argp = &data->c_arg,
1401                 .rpc_resp = &data->c_res,
1402                 .rpc_cred = data->owner->so_cred,
1403         };
1404         struct rpc_task_setup task_setup_data = {
1405                 .rpc_client = server->client,
1406                 .rpc_message = &msg,
1407                 .callback_ops = &nfs4_open_confirm_ops,
1408                 .callback_data = data,
1409                 .workqueue = nfsiod_workqueue,
1410                 .flags = RPC_TASK_ASYNC,
1411         };
1412         int status;
1413
1414         kref_get(&data->kref);
1415         data->rpc_done = 0;
1416         data->rpc_status = 0;
1417         data->timestamp = jiffies;
1418         task = rpc_run_task(&task_setup_data);
1419         if (IS_ERR(task))
1420                 return PTR_ERR(task);
1421         status = nfs4_wait_for_completion_rpc_task(task);
1422         if (status != 0) {
1423                 data->cancelled = 1;
1424                 smp_wmb();
1425         } else
1426                 status = data->rpc_status;
1427         rpc_put_task(task);
1428         return status;
1429 }
1430
1431 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1432 {
1433         struct nfs4_opendata *data = calldata;
1434         struct nfs4_state_owner *sp = data->owner;
1435
1436         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1437                 return;
1438         /*
1439          * Check if we still need to send an OPEN call, or if we can use
1440          * a delegation instead.
1441          */
1442         if (data->state != NULL) {
1443                 struct nfs_delegation *delegation;
1444
1445                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1446                         goto out_no_action;
1447                 rcu_read_lock();
1448                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1449                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1450                     can_open_delegated(delegation, data->o_arg.fmode))
1451                         goto unlock_no_action;
1452                 rcu_read_unlock();
1453         }
1454         /* Update sequence id. */
1455         data->o_arg.id = sp->so_seqid.owner_id;
1456         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1457         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1458                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1459                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1460         }
1461         data->timestamp = jiffies;
1462         if (nfs4_setup_sequence(data->o_arg.server,
1463                                 &data->o_arg.seq_args,
1464                                 &data->o_res.seq_res, task))
1465                 return;
1466         rpc_call_start(task);
1467         return;
1468 unlock_no_action:
1469         rcu_read_unlock();
1470 out_no_action:
1471         task->tk_action = NULL;
1472
1473 }
1474
1475 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1476 {
1477         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1478         nfs4_open_prepare(task, calldata);
1479 }
1480
1481 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1482 {
1483         struct nfs4_opendata *data = calldata;
1484
1485         data->rpc_status = task->tk_status;
1486
1487         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1488                 return;
1489
1490         if (task->tk_status == 0) {
1491                 switch (data->o_res.f_attr->mode & S_IFMT) {
1492                         case S_IFREG:
1493                                 break;
1494                         case S_IFLNK:
1495                                 data->rpc_status = -ELOOP;
1496                                 break;
1497                         case S_IFDIR:
1498                                 data->rpc_status = -EISDIR;
1499                                 break;
1500                         default:
1501                                 data->rpc_status = -ENOTDIR;
1502                 }
1503                 renew_lease(data->o_res.server, data->timestamp);
1504                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1505                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1506         }
1507         data->rpc_done = 1;
1508 }
1509
1510 static void nfs4_open_release(void *calldata)
1511 {
1512         struct nfs4_opendata *data = calldata;
1513         struct nfs4_state *state = NULL;
1514
1515         /* If this request hasn't been cancelled, do nothing */
1516         if (data->cancelled == 0)
1517                 goto out_free;
1518         /* In case of error, no cleanup! */
1519         if (data->rpc_status != 0 || !data->rpc_done)
1520                 goto out_free;
1521         /* In case we need an open_confirm, no cleanup! */
1522         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1523                 goto out_free;
1524         state = nfs4_opendata_to_nfs4_state(data);
1525         if (!IS_ERR(state))
1526                 nfs4_close_state(state, data->o_arg.fmode);
1527 out_free:
1528         nfs4_opendata_put(data);
1529 }
1530
1531 static const struct rpc_call_ops nfs4_open_ops = {
1532         .rpc_call_prepare = nfs4_open_prepare,
1533         .rpc_call_done = nfs4_open_done,
1534         .rpc_release = nfs4_open_release,
1535 };
1536
1537 static const struct rpc_call_ops nfs4_recover_open_ops = {
1538         .rpc_call_prepare = nfs4_recover_open_prepare,
1539         .rpc_call_done = nfs4_open_done,
1540         .rpc_release = nfs4_open_release,
1541 };
1542
1543 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1544 {
1545         struct inode *dir = data->dir->d_inode;
1546         struct nfs_server *server = NFS_SERVER(dir);
1547         struct nfs_openargs *o_arg = &data->o_arg;
1548         struct nfs_openres *o_res = &data->o_res;
1549         struct rpc_task *task;
1550         struct rpc_message msg = {
1551                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1552                 .rpc_argp = o_arg,
1553                 .rpc_resp = o_res,
1554                 .rpc_cred = data->owner->so_cred,
1555         };
1556         struct rpc_task_setup task_setup_data = {
1557                 .rpc_client = server->client,
1558                 .rpc_message = &msg,
1559                 .callback_ops = &nfs4_open_ops,
1560                 .callback_data = data,
1561                 .workqueue = nfsiod_workqueue,
1562                 .flags = RPC_TASK_ASYNC,
1563         };
1564         int status;
1565
1566         nfs41_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1567         kref_get(&data->kref);
1568         data->rpc_done = 0;
1569         data->rpc_status = 0;
1570         data->cancelled = 0;
1571         if (isrecover)
1572                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1573         task = rpc_run_task(&task_setup_data);
1574         if (IS_ERR(task))
1575                 return PTR_ERR(task);
1576         status = nfs4_wait_for_completion_rpc_task(task);
1577         if (status != 0) {
1578                 data->cancelled = 1;
1579                 smp_wmb();
1580         } else
1581                 status = data->rpc_status;
1582         rpc_put_task(task);
1583
1584         return status;
1585 }
1586
1587 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1588 {
1589         struct inode *dir = data->dir->d_inode;
1590         struct nfs_openres *o_res = &data->o_res;
1591         int status;
1592
1593         status = nfs4_run_open_task(data, 1);
1594         if (status != 0 || !data->rpc_done)
1595                 return status;
1596
1597         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
1598
1599         nfs_refresh_inode(dir, o_res->dir_attr);
1600
1601         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1602                 status = _nfs4_proc_open_confirm(data);
1603                 if (status != 0)
1604                         return status;
1605         }
1606
1607         return status;
1608 }
1609
1610 /*
1611  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1612  */
1613 static int _nfs4_proc_open(struct nfs4_opendata *data)
1614 {
1615         struct inode *dir = data->dir->d_inode;
1616         struct nfs_server *server = NFS_SERVER(dir);
1617         struct nfs_openargs *o_arg = &data->o_arg;
1618         struct nfs_openres *o_res = &data->o_res;
1619         int status;
1620
1621         status = nfs4_run_open_task(data, 0);
1622         if (!data->rpc_done)
1623                 return status;
1624         if (status != 0) {
1625                 if (status == -NFS4ERR_BADNAME &&
1626                                 !(o_arg->open_flags & O_CREAT))
1627                         return -ENOENT;
1628                 return status;
1629         }
1630
1631         nfs_fattr_map_and_free_names(server, &data->f_attr);
1632
1633         if (o_arg->open_flags & O_CREAT) {
1634                 update_changeattr(dir, &o_res->cinfo);
1635                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1636         } else
1637                 nfs_refresh_inode(dir, o_res->dir_attr);
1638         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1639                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1640         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1641                 status = _nfs4_proc_open_confirm(data);
1642                 if (status != 0)
1643                         return status;
1644         }
1645         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1646                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1647         return 0;
1648 }
1649
1650 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1651 {
1652         unsigned int loop;
1653         int ret;
1654
1655         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1656                 ret = nfs4_wait_clnt_recover(clp);
1657                 if (ret != 0)
1658                         break;
1659                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1660                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1661                         break;
1662                 nfs4_schedule_state_manager(clp);
1663                 ret = -EIO;
1664         }
1665         return ret;
1666 }
1667
1668 static int nfs4_recover_expired_lease(struct nfs_server *server)
1669 {
1670         return nfs4_client_recover_expired_lease(server->nfs_client);
1671 }
1672
1673 /*
1674  * OPEN_EXPIRED:
1675  *      reclaim state on the server after a network partition.
1676  *      Assumes caller holds the appropriate lock
1677  */
1678 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1679 {
1680         struct nfs4_opendata *opendata;
1681         int ret;
1682
1683         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1684         if (IS_ERR(opendata))
1685                 return PTR_ERR(opendata);
1686         ret = nfs4_open_recover(opendata, state);
1687         if (ret == -ESTALE)
1688                 d_drop(ctx->dentry);
1689         nfs4_opendata_put(opendata);
1690         return ret;
1691 }
1692
1693 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1694 {
1695         struct nfs_server *server = NFS_SERVER(state->inode);
1696         struct nfs4_exception exception = { };
1697         int err;
1698
1699         do {
1700                 err = _nfs4_open_expired(ctx, state);
1701                 switch (err) {
1702                 default:
1703                         goto out;
1704                 case -NFS4ERR_GRACE:
1705                 case -NFS4ERR_DELAY:
1706                         nfs4_handle_exception(server, err, &exception);
1707                         err = 0;
1708                 }
1709         } while (exception.retry);
1710 out:
1711         return err;
1712 }
1713
1714 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1715 {
1716         struct nfs_open_context *ctx;
1717         int ret;
1718
1719         ctx = nfs4_state_find_open_context(state);
1720         if (IS_ERR(ctx))
1721                 return PTR_ERR(ctx);
1722         ret = nfs4_do_open_expired(ctx, state);
1723         put_nfs_open_context(ctx);
1724         return ret;
1725 }
1726
1727 #if defined(CONFIG_NFS_V4_1)
1728 static int nfs41_check_expired_stateid(struct nfs4_state *state, nfs4_stateid *stateid, unsigned int flags)
1729 {
1730         int status = NFS_OK;
1731         struct nfs_server *server = NFS_SERVER(state->inode);
1732
1733         if (state->flags & flags) {
1734                 status = nfs41_test_stateid(server, stateid);
1735                 if (status != NFS_OK) {
1736                         nfs41_free_stateid(server, stateid);
1737                         state->flags &= ~flags;
1738                 }
1739         }
1740         return status;
1741 }
1742
1743 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1744 {
1745         int deleg_status, open_status;
1746         int deleg_flags = 1 << NFS_DELEGATED_STATE;
1747         int open_flags = (1 << NFS_O_RDONLY_STATE) | (1 << NFS_O_WRONLY_STATE) | (1 << NFS_O_RDWR_STATE);
1748
1749         deleg_status = nfs41_check_expired_stateid(state, &state->stateid, deleg_flags);
1750         open_status = nfs41_check_expired_stateid(state,  &state->open_stateid, open_flags);
1751
1752         if ((deleg_status == NFS_OK) && (open_status == NFS_OK))
1753                 return NFS_OK;
1754         return nfs4_open_expired(sp, state);
1755 }
1756 #endif
1757
1758 /*
1759  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1760  * fields corresponding to attributes that were used to store the verifier.
1761  * Make sure we clobber those fields in the later setattr call
1762  */
1763 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1764 {
1765         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1766             !(sattr->ia_valid & ATTR_ATIME_SET))
1767                 sattr->ia_valid |= ATTR_ATIME;
1768
1769         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1770             !(sattr->ia_valid & ATTR_MTIME_SET))
1771                 sattr->ia_valid |= ATTR_MTIME;
1772 }
1773
1774 /*
1775  * Returns a referenced nfs4_state
1776  */
1777 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)
1778 {
1779         struct nfs4_state_owner  *sp;
1780         struct nfs4_state     *state = NULL;
1781         struct nfs_server       *server = NFS_SERVER(dir);
1782         struct nfs4_opendata *opendata;
1783         int status;
1784
1785         /* Protect against reboot recovery conflicts */
1786         status = -ENOMEM;
1787         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
1788         if (sp == NULL) {
1789                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1790                 goto out_err;
1791         }
1792         status = nfs4_recover_expired_lease(server);
1793         if (status != 0)
1794                 goto err_put_state_owner;
1795         if (dentry->d_inode != NULL)
1796                 nfs4_return_incompatible_delegation(dentry->d_inode, fmode);
1797         status = -ENOMEM;
1798         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr, GFP_KERNEL);
1799         if (opendata == NULL)
1800                 goto err_put_state_owner;
1801
1802         if (dentry->d_inode != NULL)
1803                 opendata->state = nfs4_get_open_state(dentry->d_inode, sp);
1804
1805         status = _nfs4_proc_open(opendata);
1806         if (status != 0)
1807                 goto err_opendata_put;
1808
1809         state = nfs4_opendata_to_nfs4_state(opendata);
1810         status = PTR_ERR(state);
1811         if (IS_ERR(state))
1812                 goto err_opendata_put;
1813         if (server->caps & NFS_CAP_POSIX_LOCK)
1814                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1815
1816         if (opendata->o_arg.open_flags & O_EXCL) {
1817                 nfs4_exclusive_attrset(opendata, sattr);
1818
1819                 nfs_fattr_init(opendata->o_res.f_attr);
1820                 status = nfs4_do_setattr(state->inode, cred,
1821                                 opendata->o_res.f_attr, sattr,
1822                                 state);
1823                 if (status == 0)
1824                         nfs_setattr_update_inode(state->inode, sattr);
1825                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1826         }
1827         nfs4_opendata_put(opendata);
1828         nfs4_put_state_owner(sp);
1829         *res = state;
1830         return 0;
1831 err_opendata_put:
1832         nfs4_opendata_put(opendata);
1833 err_put_state_owner:
1834         nfs4_put_state_owner(sp);
1835 out_err:
1836         *res = NULL;
1837         return status;
1838 }
1839
1840
1841 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)
1842 {
1843         struct nfs4_exception exception = { };
1844         struct nfs4_state *res;
1845         int status;
1846
1847         do {
1848                 status = _nfs4_do_open(dir, dentry, fmode, flags, sattr, cred, &res);
1849                 if (status == 0)
1850                         break;
1851                 /* NOTE: BAD_SEQID means the server and client disagree about the
1852                  * book-keeping w.r.t. state-changing operations
1853                  * (OPEN/CLOSE/LOCK/LOCKU...)
1854                  * It is actually a sign of a bug on the client or on the server.
1855                  *
1856                  * If we receive a BAD_SEQID error in the particular case of
1857                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1858                  * have unhashed the old state_owner for us, and that we can
1859                  * therefore safely retry using a new one. We should still warn
1860                  * the user though...
1861                  */
1862                 if (status == -NFS4ERR_BAD_SEQID) {
1863                         printk(KERN_WARNING "NFS: v4 server %s "
1864                                         " returned a bad sequence-id error!\n",
1865                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1866                         exception.retry = 1;
1867                         continue;
1868                 }
1869                 /*
1870                  * BAD_STATEID on OPEN means that the server cancelled our
1871                  * state before it received the OPEN_CONFIRM.
1872                  * Recover by retrying the request as per the discussion
1873                  * on Page 181 of RFC3530.
1874                  */
1875                 if (status == -NFS4ERR_BAD_STATEID) {
1876                         exception.retry = 1;
1877                         continue;
1878                 }
1879                 if (status == -EAGAIN) {
1880                         /* We must have found a delegation */
1881                         exception.retry = 1;
1882                         continue;
1883                 }
1884                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1885                                         status, &exception));
1886         } while (exception.retry);
1887         return res;
1888 }
1889
1890 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1891                             struct nfs_fattr *fattr, struct iattr *sattr,
1892                             struct nfs4_state *state)
1893 {
1894         struct nfs_server *server = NFS_SERVER(inode);
1895         struct nfs_setattrargs  arg = {
1896                 .fh             = NFS_FH(inode),
1897                 .iap            = sattr,
1898                 .server         = server,
1899                 .bitmask = server->attr_bitmask,
1900         };
1901         struct nfs_setattrres  res = {
1902                 .fattr          = fattr,
1903                 .server         = server,
1904         };
1905         struct rpc_message msg = {
1906                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1907                 .rpc_argp       = &arg,
1908                 .rpc_resp       = &res,
1909                 .rpc_cred       = cred,
1910         };
1911         unsigned long timestamp = jiffies;
1912         int status;
1913
1914         nfs_fattr_init(fattr);
1915
1916         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1917                 /* Use that stateid */
1918         } else if (state != NULL) {
1919                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1920         } else
1921                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1922
1923         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1924         if (status == 0 && state != NULL)
1925                 renew_lease(server, timestamp);
1926         return status;
1927 }
1928
1929 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1930                            struct nfs_fattr *fattr, struct iattr *sattr,
1931                            struct nfs4_state *state)
1932 {
1933         struct nfs_server *server = NFS_SERVER(inode);
1934         struct nfs4_exception exception = { };
1935         int err;
1936         do {
1937                 err = nfs4_handle_exception(server,
1938                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1939                                 &exception);
1940         } while (exception.retry);
1941         return err;
1942 }
1943
1944 struct nfs4_closedata {
1945         struct inode *inode;
1946         struct nfs4_state *state;
1947         struct nfs_closeargs arg;
1948         struct nfs_closeres res;
1949         struct nfs_fattr fattr;
1950         unsigned long timestamp;
1951         bool roc;
1952         u32 roc_barrier;
1953 };
1954
1955 static void nfs4_free_closedata(void *data)
1956 {
1957         struct nfs4_closedata *calldata = data;
1958         struct nfs4_state_owner *sp = calldata->state->owner;
1959         struct super_block *sb = calldata->state->inode->i_sb;
1960
1961         if (calldata->roc)
1962                 pnfs_roc_release(calldata->state->inode);
1963         nfs4_put_open_state(calldata->state);
1964         nfs_free_seqid(calldata->arg.seqid);
1965         nfs4_put_state_owner(sp);
1966         nfs_sb_deactive(sb);
1967         kfree(calldata);
1968 }
1969
1970 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1971                 fmode_t fmode)
1972 {
1973         spin_lock(&state->owner->so_lock);
1974         if (!(fmode & FMODE_READ))
1975                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1976         if (!(fmode & FMODE_WRITE))
1977                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1978         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1979         spin_unlock(&state->owner->so_lock);
1980 }
1981
1982 static void nfs4_close_done(struct rpc_task *task, void *data)
1983 {
1984         struct nfs4_closedata *calldata = data;
1985         struct nfs4_state *state = calldata->state;
1986         struct nfs_server *server = NFS_SERVER(calldata->inode);
1987
1988         dprintk("%s: begin!\n", __func__);
1989         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1990                 return;
1991         /* hmm. we are done with the inode, and in the process of freeing
1992          * the state_owner. we keep this around to process errors
1993          */
1994         switch (task->tk_status) {
1995                 case 0:
1996                         if (calldata->roc)
1997                                 pnfs_roc_set_barrier(state->inode,
1998                                                      calldata->roc_barrier);
1999                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
2000                         renew_lease(server, calldata->timestamp);
2001                         nfs4_close_clear_stateid_flags(state,
2002                                         calldata->arg.fmode);
2003                         break;
2004                 case -NFS4ERR_STALE_STATEID:
2005                 case -NFS4ERR_OLD_STATEID:
2006                 case -NFS4ERR_BAD_STATEID:
2007                 case -NFS4ERR_EXPIRED:
2008                         if (calldata->arg.fmode == 0)
2009                                 break;
2010                 default:
2011                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
2012                                 rpc_restart_call_prepare(task);
2013         }
2014         nfs_release_seqid(calldata->arg.seqid);
2015         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2016         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2017 }
2018
2019 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2020 {
2021         struct nfs4_closedata *calldata = data;
2022         struct nfs4_state *state = calldata->state;
2023         int call_close = 0;
2024
2025         dprintk("%s: begin!\n", __func__);
2026         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2027                 return;
2028
2029         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2030         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
2031         spin_lock(&state->owner->so_lock);
2032         /* Calculate the change in open mode */
2033         if (state->n_rdwr == 0) {
2034                 if (state->n_rdonly == 0) {
2035                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
2036                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2037                         calldata->arg.fmode &= ~FMODE_READ;
2038                 }
2039                 if (state->n_wronly == 0) {
2040                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
2041                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
2042                         calldata->arg.fmode &= ~FMODE_WRITE;
2043                 }
2044         }
2045         spin_unlock(&state->owner->so_lock);
2046
2047         if (!call_close) {
2048                 /* Note: exit _without_ calling nfs4_close_done */
2049                 task->tk_action = NULL;
2050                 goto out;
2051         }
2052
2053         if (calldata->arg.fmode == 0) {
2054                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2055                 if (calldata->roc &&
2056                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
2057                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
2058                                      task, NULL);
2059                         goto out;
2060                 }
2061         }
2062
2063         nfs_fattr_init(calldata->res.fattr);
2064         calldata->timestamp = jiffies;
2065         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
2066                                 &calldata->arg.seq_args,
2067                                 &calldata->res.seq_res,
2068                                 task))
2069                 goto out;
2070         rpc_call_start(task);
2071 out:
2072         dprintk("%s: done!\n", __func__);
2073 }
2074
2075 static const struct rpc_call_ops nfs4_close_ops = {
2076         .rpc_call_prepare = nfs4_close_prepare,
2077         .rpc_call_done = nfs4_close_done,
2078         .rpc_release = nfs4_free_closedata,
2079 };
2080
2081 /*
2082  * It is possible for data to be read/written from a mem-mapped file
2083  * after the sys_close call (which hits the vfs layer as a flush).
2084  * This means that we can't safely call nfsv4 close on a file until
2085  * the inode is cleared. This in turn means that we are not good
2086  * NFSv4 citizens - we do not indicate to the server to update the file's
2087  * share state even when we are done with one of the three share
2088  * stateid's in the inode.
2089  *
2090  * NOTE: Caller must be holding the sp->so_owner semaphore!
2091  */
2092 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2093 {
2094         struct nfs_server *server = NFS_SERVER(state->inode);
2095         struct nfs4_closedata *calldata;
2096         struct nfs4_state_owner *sp = state->owner;
2097         struct rpc_task *task;
2098         struct rpc_message msg = {
2099                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2100                 .rpc_cred = state->owner->so_cred,
2101         };
2102         struct rpc_task_setup task_setup_data = {
2103                 .rpc_client = server->client,
2104                 .rpc_message = &msg,
2105                 .callback_ops = &nfs4_close_ops,
2106                 .workqueue = nfsiod_workqueue,
2107                 .flags = RPC_TASK_ASYNC,
2108         };
2109         int status = -ENOMEM;
2110
2111         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2112         if (calldata == NULL)
2113                 goto out;
2114         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2115         calldata->inode = state->inode;
2116         calldata->state = state;
2117         calldata->arg.fh = NFS_FH(state->inode);
2118         calldata->arg.stateid = &state->open_stateid;
2119         /* Serialization for the sequence id */
2120         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2121         if (calldata->arg.seqid == NULL)
2122                 goto out_free_calldata;
2123         calldata->arg.fmode = 0;
2124         calldata->arg.bitmask = server->cache_consistency_bitmask;
2125         calldata->res.fattr = &calldata->fattr;
2126         calldata->res.seqid = calldata->arg.seqid;
2127         calldata->res.server = server;
2128         calldata->roc = roc;
2129         nfs_sb_active(calldata->inode->i_sb);
2130
2131         msg.rpc_argp = &calldata->arg;
2132         msg.rpc_resp = &calldata->res;
2133         task_setup_data.callback_data = calldata;
2134         task = rpc_run_task(&task_setup_data);
2135         if (IS_ERR(task))
2136                 return PTR_ERR(task);
2137         status = 0;
2138         if (wait)
2139                 status = rpc_wait_for_completion_task(task);
2140         rpc_put_task(task);
2141         return status;
2142 out_free_calldata:
2143         kfree(calldata);
2144 out:
2145         if (roc)
2146                 pnfs_roc_release(state->inode);
2147         nfs4_put_open_state(state);
2148         nfs4_put_state_owner(sp);
2149         return status;
2150 }
2151
2152 static struct inode *
2153 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2154 {
2155         struct nfs4_state *state;
2156
2157         /* Protect against concurrent sillydeletes */
2158         state = nfs4_do_open(dir, ctx->dentry, ctx->mode, open_flags, attr, ctx->cred);
2159         if (IS_ERR(state))
2160                 return ERR_CAST(state);
2161         ctx->state = state;
2162         return igrab(state->inode);
2163 }
2164
2165 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2166 {
2167         if (ctx->state == NULL)
2168                 return;
2169         if (is_sync)
2170                 nfs4_close_sync(ctx->state, ctx->mode);
2171         else
2172                 nfs4_close_state(ctx->state, ctx->mode);
2173 }
2174
2175 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2176 {
2177         struct nfs4_server_caps_arg args = {
2178                 .fhandle = fhandle,
2179         };
2180         struct nfs4_server_caps_res res = {};
2181         struct rpc_message msg = {
2182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2183                 .rpc_argp = &args,
2184                 .rpc_resp = &res,
2185         };
2186         int status;
2187
2188         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2189         if (status == 0) {
2190                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2191                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2192                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2193                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2194                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2195                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2196                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2197                         server->caps |= NFS_CAP_ACLS;
2198                 if (res.has_links != 0)
2199                         server->caps |= NFS_CAP_HARDLINKS;
2200                 if (res.has_symlinks != 0)
2201                         server->caps |= NFS_CAP_SYMLINKS;
2202                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2203                         server->caps |= NFS_CAP_FILEID;
2204                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2205                         server->caps |= NFS_CAP_MODE;
2206                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2207                         server->caps |= NFS_CAP_NLINK;
2208                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2209                         server->caps |= NFS_CAP_OWNER;
2210                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2211                         server->caps |= NFS_CAP_OWNER_GROUP;
2212                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2213                         server->caps |= NFS_CAP_ATIME;
2214                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2215                         server->caps |= NFS_CAP_CTIME;
2216                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2217                         server->caps |= NFS_CAP_MTIME;
2218
2219                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2220                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2221                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2222                 server->acl_bitmask = res.acl_bitmask;
2223         }
2224
2225         return status;
2226 }
2227
2228 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2229 {
2230         struct nfs4_exception exception = { };
2231         int err;
2232         do {
2233                 err = nfs4_handle_exception(server,
2234                                 _nfs4_server_capabilities(server, fhandle),
2235                                 &exception);
2236         } while (exception.retry);
2237         return err;
2238 }
2239
2240 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2241                 struct nfs_fsinfo *info)
2242 {
2243         struct nfs4_lookup_root_arg args = {
2244                 .bitmask = nfs4_fattr_bitmap,
2245         };
2246         struct nfs4_lookup_res res = {
2247                 .server = server,
2248                 .fattr = info->fattr,
2249                 .fh = fhandle,
2250         };
2251         struct rpc_message msg = {
2252                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2253                 .rpc_argp = &args,
2254                 .rpc_resp = &res,
2255         };
2256
2257         nfs_fattr_init(info->fattr);
2258         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2259 }
2260
2261 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2262                 struct nfs_fsinfo *info)
2263 {
2264         struct nfs4_exception exception = { };
2265         int err;
2266         do {
2267                 err = _nfs4_lookup_root(server, fhandle, info);
2268                 switch (err) {
2269                 case 0:
2270                 case -NFS4ERR_WRONGSEC:
2271                         break;
2272                 default:
2273                         err = nfs4_handle_exception(server, err, &exception);
2274                 }
2275         } while (exception.retry);
2276         return err;
2277 }
2278
2279 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2280                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2281 {
2282         struct rpc_auth *auth;
2283         int ret;
2284
2285         auth = rpcauth_create(flavor, server->client);
2286         if (!auth) {
2287                 ret = -EIO;
2288                 goto out;
2289         }
2290         ret = nfs4_lookup_root(server, fhandle, info);
2291 out:
2292         return ret;
2293 }
2294
2295 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2296                               struct nfs_fsinfo *info)
2297 {
2298         int i, len, status = 0;
2299         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2300
2301         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2302         flav_array[len] = RPC_AUTH_NULL;
2303         len += 1;
2304
2305         for (i = 0; i < len; i++) {
2306                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2307                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2308                         continue;
2309                 break;
2310         }
2311         /*
2312          * -EACCESS could mean that the user doesn't have correct permissions
2313          * to access the mount.  It could also mean that we tried to mount
2314          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2315          * existing mount programs don't handle -EACCES very well so it should
2316          * be mapped to -EPERM instead.
2317          */
2318         if (status == -EACCES)
2319                 status = -EPERM;
2320         return status;
2321 }
2322
2323 /*
2324  * get the file handle for the "/" directory on the server
2325  */
2326 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2327                               struct nfs_fsinfo *info)
2328 {
2329         int minor_version = server->nfs_client->cl_minorversion;
2330         int status = nfs4_lookup_root(server, fhandle, info);
2331         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2332                 /*
2333                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2334                  * by nfs4_map_errors() as this function exits.
2335                  */
2336                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2337         if (status == 0)
2338                 status = nfs4_server_capabilities(server, fhandle);
2339         if (status == 0)
2340                 status = nfs4_do_fsinfo(server, fhandle, info);
2341         return nfs4_map_errors(status);
2342 }
2343
2344 /*
2345  * Get locations and (maybe) other attributes of a referral.
2346  * Note that we'll actually follow the referral later when
2347  * we detect fsid mismatch in inode revalidation
2348  */
2349 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2350                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2351 {
2352         int status = -ENOMEM;
2353         struct page *page = NULL;
2354         struct nfs4_fs_locations *locations = NULL;
2355
2356         page = alloc_page(GFP_KERNEL);
2357         if (page == NULL)
2358                 goto out;
2359         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2360         if (locations == NULL)
2361                 goto out;
2362
2363         status = nfs4_proc_fs_locations(dir, name, locations, page);
2364         if (status != 0)
2365                 goto out;
2366         /* Make sure server returned a different fsid for the referral */
2367         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2368                 dprintk("%s: server did not return a different fsid for"
2369                         " a referral at %s\n", __func__, name->name);
2370                 status = -EIO;
2371                 goto out;
2372         }
2373         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2374         nfs_fixup_referral_attributes(&locations->fattr);
2375
2376         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2377         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2378         memset(fhandle, 0, sizeof(struct nfs_fh));
2379 out:
2380         if (page)
2381                 __free_page(page);
2382         kfree(locations);
2383         return status;
2384 }
2385
2386 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2387 {
2388         struct nfs4_getattr_arg args = {
2389                 .fh = fhandle,
2390                 .bitmask = server->attr_bitmask,
2391         };
2392         struct nfs4_getattr_res res = {
2393                 .fattr = fattr,
2394                 .server = server,
2395         };
2396         struct rpc_message msg = {
2397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2398                 .rpc_argp = &args,
2399                 .rpc_resp = &res,
2400         };
2401
2402         nfs_fattr_init(fattr);
2403         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2404 }
2405
2406 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2407 {
2408         struct nfs4_exception exception = { };
2409         int err;
2410         do {
2411                 err = nfs4_handle_exception(server,
2412                                 _nfs4_proc_getattr(server, fhandle, fattr),
2413                                 &exception);
2414         } while (exception.retry);
2415         return err;
2416 }
2417
2418 /*
2419  * The file is not closed if it is opened due to the a request to change
2420  * the size of the file. The open call will not be needed once the
2421  * VFS layer lookup-intents are implemented.
2422  *
2423  * Close is called when the inode is destroyed.
2424  * If we haven't opened the file for O_WRONLY, we
2425  * need to in the size_change case to obtain a stateid.
2426  *
2427  * Got race?
2428  * Because OPEN is always done by name in nfsv4, it is
2429  * possible that we opened a different file by the same
2430  * name.  We can recognize this race condition, but we
2431  * can't do anything about it besides returning an error.
2432  *
2433  * This will be fixed with VFS changes (lookup-intent).
2434  */
2435 static int
2436 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2437                   struct iattr *sattr)
2438 {
2439         struct inode *inode = dentry->d_inode;
2440         struct rpc_cred *cred = NULL;
2441         struct nfs4_state *state = NULL;
2442         int status;
2443
2444         if (pnfs_ld_layoutret_on_setattr(inode))
2445                 pnfs_return_layout(inode);
2446
2447         nfs_fattr_init(fattr);
2448
2449         /* Search for an existing open(O_WRITE) file */
2450         if (sattr->ia_valid & ATTR_FILE) {
2451                 struct nfs_open_context *ctx;
2452
2453                 ctx = nfs_file_open_context(sattr->ia_file);
2454                 if (ctx) {
2455                         cred = ctx->cred;
2456                         state = ctx->state;
2457                 }
2458         }
2459
2460         /* Deal with open(O_TRUNC) */
2461         if (sattr->ia_valid & ATTR_OPEN)
2462                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2463
2464         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2465         if (status == 0)
2466                 nfs_setattr_update_inode(inode, sattr);
2467         return status;
2468 }
2469
2470 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2471                 const struct qstr *name, struct nfs_fh *fhandle,
2472                 struct nfs_fattr *fattr)
2473 {
2474         struct nfs_server *server = NFS_SERVER(dir);
2475         int                    status;
2476         struct nfs4_lookup_arg args = {
2477                 .bitmask = server->attr_bitmask,
2478                 .dir_fh = NFS_FH(dir),
2479                 .name = name,
2480         };
2481         struct nfs4_lookup_res res = {
2482                 .server = server,
2483                 .fattr = fattr,
2484                 .fh = fhandle,
2485         };
2486         struct rpc_message msg = {
2487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2488                 .rpc_argp = &args,
2489                 .rpc_resp = &res,
2490         };
2491
2492         nfs_fattr_init(fattr);
2493
2494         dprintk("NFS call  lookup %s\n", name->name);
2495         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2496         dprintk("NFS reply lookup: %d\n", status);
2497         return status;
2498 }
2499
2500 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2501 {
2502         memset(fh, 0, sizeof(struct nfs_fh));
2503         fattr->fsid.major = 1;
2504         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2505                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2506         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2507         fattr->nlink = 2;
2508 }
2509
2510 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2511                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2512 {
2513         struct nfs4_exception exception = { };
2514         int err;
2515         do {
2516                 int status;
2517
2518                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2519                 switch (status) {
2520                 case -NFS4ERR_BADNAME:
2521                         return -ENOENT;
2522                 case -NFS4ERR_MOVED:
2523                         return nfs4_get_referral(dir, name, fattr, fhandle);
2524                 case -NFS4ERR_WRONGSEC:
2525                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2526                 }
2527                 err = nfs4_handle_exception(NFS_SERVER(dir),
2528                                 status, &exception);
2529         } while (exception.retry);
2530         return err;
2531 }
2532
2533 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2534 {
2535         struct nfs_server *server = NFS_SERVER(inode);
2536         struct nfs4_accessargs args = {
2537                 .fh = NFS_FH(inode),
2538                 .bitmask = server->cache_consistency_bitmask,
2539         };
2540         struct nfs4_accessres res = {
2541                 .server = server,
2542         };
2543         struct rpc_message msg = {
2544                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2545                 .rpc_argp = &args,
2546                 .rpc_resp = &res,
2547                 .rpc_cred = entry->cred,
2548         };
2549         int mode = entry->mask;
2550         int status;
2551
2552         /*
2553          * Determine which access bits we want to ask for...
2554          */
2555         if (mode & MAY_READ)
2556                 args.access |= NFS4_ACCESS_READ;
2557         if (S_ISDIR(inode->i_mode)) {
2558                 if (mode & MAY_WRITE)
2559                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2560                 if (mode & MAY_EXEC)
2561                         args.access |= NFS4_ACCESS_LOOKUP;
2562         } else {
2563                 if (mode & MAY_WRITE)
2564                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2565                 if (mode & MAY_EXEC)
2566                         args.access |= NFS4_ACCESS_EXECUTE;
2567         }
2568
2569         res.fattr = nfs_alloc_fattr();
2570         if (res.fattr == NULL)
2571                 return -ENOMEM;
2572
2573         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2574         if (!status) {
2575                 entry->mask = 0;
2576                 if (res.access & NFS4_ACCESS_READ)
2577                         entry->mask |= MAY_READ;
2578                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2579                         entry->mask |= MAY_WRITE;
2580                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2581                         entry->mask |= MAY_EXEC;
2582                 nfs_refresh_inode(inode, res.fattr);
2583         }
2584         nfs_free_fattr(res.fattr);
2585         return status;
2586 }
2587
2588 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2589 {
2590         struct nfs4_exception exception = { };
2591         int err;
2592         do {
2593                 err = nfs4_handle_exception(NFS_SERVER(inode),
2594                                 _nfs4_proc_access(inode, entry),
2595                                 &exception);
2596         } while (exception.retry);
2597         return err;
2598 }
2599
2600 /*
2601  * TODO: For the time being, we don't try to get any attributes
2602  * along with any of the zero-copy operations READ, READDIR,
2603  * READLINK, WRITE.
2604  *
2605  * In the case of the first three, we want to put the GETATTR
2606  * after the read-type operation -- this is because it is hard
2607  * to predict the length of a GETATTR response in v4, and thus
2608  * align the READ data correctly.  This means that the GETATTR
2609  * may end up partially falling into the page cache, and we should
2610  * shift it into the 'tail' of the xdr_buf before processing.
2611  * To do this efficiently, we need to know the total length
2612  * of data received, which doesn't seem to be available outside
2613  * of the RPC layer.
2614  *
2615  * In the case of WRITE, we also want to put the GETATTR after
2616  * the operation -- in this case because we want to make sure
2617  * we get the post-operation mtime and size.  This means that
2618  * we can't use xdr_encode_pages() as written: we need a variant
2619  * of it which would leave room in the 'tail' iovec.
2620  *
2621  * Both of these changes to the XDR layer would in fact be quite
2622  * minor, but I decided to leave them for a subsequent patch.
2623  */
2624 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2625                 unsigned int pgbase, unsigned int pglen)
2626 {
2627         struct nfs4_readlink args = {
2628                 .fh       = NFS_FH(inode),
2629                 .pgbase   = pgbase,
2630                 .pglen    = pglen,
2631                 .pages    = &page,
2632         };
2633         struct nfs4_readlink_res res;
2634         struct rpc_message msg = {
2635                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2636                 .rpc_argp = &args,
2637                 .rpc_resp = &res,
2638         };
2639
2640         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2641 }
2642
2643 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2644                 unsigned int pgbase, unsigned int pglen)
2645 {
2646         struct nfs4_exception exception = { };
2647         int err;
2648         do {
2649                 err = nfs4_handle_exception(NFS_SERVER(inode),
2650                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2651                                 &exception);
2652         } while (exception.retry);
2653         return err;
2654 }
2655
2656 /*
2657  * Got race?
2658  * We will need to arrange for the VFS layer to provide an atomic open.
2659  * Until then, this create/open method is prone to inefficiency and race
2660  * conditions due to the lookup, create, and open VFS calls from sys_open()
2661  * placed on the wire.
2662  *
2663  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2664  * The file will be opened again in the subsequent VFS open call
2665  * (nfs4_proc_file_open).
2666  *
2667  * The open for read will just hang around to be used by any process that
2668  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2669  */
2670
2671 static int
2672 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2673                  int flags, struct nfs_open_context *ctx)
2674 {
2675         struct dentry *de = dentry;
2676         struct nfs4_state *state;
2677         struct rpc_cred *cred = NULL;
2678         fmode_t fmode = 0;
2679         int status = 0;
2680
2681         if (ctx != NULL) {
2682                 cred = ctx->cred;
2683                 de = ctx->dentry;
2684                 fmode = ctx->mode;
2685         }
2686         sattr->ia_mode &= ~current_umask();
2687         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2688         d_drop(dentry);
2689         if (IS_ERR(state)) {
2690                 status = PTR_ERR(state);
2691                 goto out;
2692         }
2693         d_add(dentry, igrab(state->inode));
2694         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2695         if (ctx != NULL)
2696                 ctx->state = state;
2697         else
2698                 nfs4_close_sync(state, fmode);
2699 out:
2700         return status;
2701 }
2702
2703 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2704 {
2705         struct nfs_server *server = NFS_SERVER(dir);
2706         struct nfs_removeargs args = {
2707                 .fh = NFS_FH(dir),
2708                 .name.len = name->len,
2709                 .name.name = name->name,
2710                 .bitmask = server->attr_bitmask,
2711         };
2712         struct nfs_removeres res = {
2713                 .server = server,
2714         };
2715         struct rpc_message msg = {
2716                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2717                 .rpc_argp = &args,
2718                 .rpc_resp = &res,
2719         };
2720         int status = -ENOMEM;
2721
2722         res.dir_attr = nfs_alloc_fattr();
2723         if (res.dir_attr == NULL)
2724                 goto out;
2725
2726         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2727         if (status == 0) {
2728                 update_changeattr(dir, &res.cinfo);
2729                 nfs_post_op_update_inode(dir, res.dir_attr);
2730         }
2731         nfs_free_fattr(res.dir_attr);
2732 out:
2733         return status;
2734 }
2735
2736 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2737 {
2738         struct nfs4_exception exception = { };
2739         int err;
2740         do {
2741                 err = nfs4_handle_exception(NFS_SERVER(dir),
2742                                 _nfs4_proc_remove(dir, name),
2743                                 &exception);
2744         } while (exception.retry);
2745         return err;
2746 }
2747
2748 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2749 {
2750         struct nfs_server *server = NFS_SERVER(dir);
2751         struct nfs_removeargs *args = msg->rpc_argp;
2752         struct nfs_removeres *res = msg->rpc_resp;
2753
2754         args->bitmask = server->cache_consistency_bitmask;
2755         res->server = server;
2756         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2757         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2758 }
2759
2760 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2761 {
2762         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2763
2764         if (!nfs4_sequence_done(task, &res->seq_res))
2765                 return 0;
2766         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2767                 return 0;
2768         update_changeattr(dir, &res->cinfo);
2769         nfs_post_op_update_inode(dir, res->dir_attr);
2770         return 1;
2771 }
2772
2773 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2774 {
2775         struct nfs_server *server = NFS_SERVER(dir);
2776         struct nfs_renameargs *arg = msg->rpc_argp;
2777         struct nfs_renameres *res = msg->rpc_resp;
2778
2779         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2780         arg->bitmask = server->attr_bitmask;
2781         res->server = server;
2782         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2783 }
2784
2785 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2786                                  struct inode *new_dir)
2787 {
2788         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2789
2790         if (!nfs4_sequence_done(task, &res->seq_res))
2791                 return 0;
2792         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2793                 return 0;
2794
2795         update_changeattr(old_dir, &res->old_cinfo);
2796         nfs_post_op_update_inode(old_dir, res->old_fattr);
2797         update_changeattr(new_dir, &res->new_cinfo);
2798         nfs_post_op_update_inode(new_dir, res->new_fattr);
2799         return 1;
2800 }
2801
2802 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2803                 struct inode *new_dir, struct qstr *new_name)
2804 {
2805         struct nfs_server *server = NFS_SERVER(old_dir);
2806         struct nfs_renameargs arg = {
2807                 .old_dir = NFS_FH(old_dir),
2808                 .new_dir = NFS_FH(new_dir),
2809                 .old_name = old_name,
2810                 .new_name = new_name,
2811                 .bitmask = server->attr_bitmask,
2812         };
2813         struct nfs_renameres res = {
2814                 .server = server,
2815         };
2816         struct rpc_message msg = {
2817                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2818                 .rpc_argp = &arg,
2819                 .rpc_resp = &res,
2820         };
2821         int status = -ENOMEM;
2822
2823         res.old_fattr = nfs_alloc_fattr();
2824         res.new_fattr = nfs_alloc_fattr();
2825         if (res.old_fattr == NULL || res.new_fattr == NULL)
2826                 goto out;
2827
2828         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2829         if (!status) {
2830                 update_changeattr(old_dir, &res.old_cinfo);
2831                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2832                 update_changeattr(new_dir, &res.new_cinfo);
2833                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2834         }
2835 out:
2836         nfs_free_fattr(res.new_fattr);
2837         nfs_free_fattr(res.old_fattr);
2838         return status;
2839 }
2840
2841 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2842                 struct inode *new_dir, struct qstr *new_name)
2843 {
2844         struct nfs4_exception exception = { };
2845         int err;
2846         do {
2847                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2848                                 _nfs4_proc_rename(old_dir, old_name,
2849                                         new_dir, new_name),
2850                                 &exception);
2851         } while (exception.retry);
2852         return err;
2853 }
2854
2855 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2856 {
2857         struct nfs_server *server = NFS_SERVER(inode);
2858         struct nfs4_link_arg arg = {
2859                 .fh     = NFS_FH(inode),
2860                 .dir_fh = NFS_FH(dir),
2861                 .name   = name,
2862                 .bitmask = server->attr_bitmask,
2863         };
2864         struct nfs4_link_res res = {
2865                 .server = server,
2866         };
2867         struct rpc_message msg = {
2868                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2869                 .rpc_argp = &arg,
2870                 .rpc_resp = &res,
2871         };
2872         int status = -ENOMEM;
2873
2874         res.fattr = nfs_alloc_fattr();
2875         res.dir_attr = nfs_alloc_fattr();
2876         if (res.fattr == NULL || res.dir_attr == NULL)
2877                 goto out;
2878
2879         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2880         if (!status) {
2881                 update_changeattr(dir, &res.cinfo);
2882                 nfs_post_op_update_inode(dir, res.dir_attr);
2883                 nfs_post_op_update_inode(inode, res.fattr);
2884         }
2885 out:
2886         nfs_free_fattr(res.dir_attr);
2887         nfs_free_fattr(res.fattr);
2888         return status;
2889 }
2890
2891 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2892 {
2893         struct nfs4_exception exception = { };
2894         int err;
2895         do {
2896                 err = nfs4_handle_exception(NFS_SERVER(inode),
2897                                 _nfs4_proc_link(inode, dir, name),
2898                                 &exception);
2899         } while (exception.retry);
2900         return err;
2901 }
2902
2903 struct nfs4_createdata {
2904         struct rpc_message msg;
2905         struct nfs4_create_arg arg;
2906         struct nfs4_create_res res;
2907         struct nfs_fh fh;
2908         struct nfs_fattr fattr;
2909         struct nfs_fattr dir_fattr;
2910 };
2911
2912 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2913                 struct qstr *name, struct iattr *sattr, u32 ftype)
2914 {
2915         struct nfs4_createdata *data;
2916
2917         data = kzalloc(sizeof(*data), GFP_KERNEL);
2918         if (data != NULL) {
2919                 struct nfs_server *server = NFS_SERVER(dir);
2920
2921                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2922                 data->msg.rpc_argp = &data->arg;
2923                 data->msg.rpc_resp = &data->res;
2924                 data->arg.dir_fh = NFS_FH(dir);
2925                 data->arg.server = server;
2926                 data->arg.name = name;
2927                 data->arg.attrs = sattr;
2928                 data->arg.ftype = ftype;
2929                 data->arg.bitmask = server->attr_bitmask;
2930                 data->res.server = server;
2931                 data->res.fh = &data->fh;
2932                 data->res.fattr = &data->fattr;
2933                 data->res.dir_fattr = &data->dir_fattr;
2934                 nfs_fattr_init(data->res.fattr);
2935                 nfs_fattr_init(data->res.dir_fattr);
2936         }
2937         return data;
2938 }
2939
2940 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2941 {
2942         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2943                                     &data->arg.seq_args, &data->res.seq_res, 1);
2944         if (status == 0) {
2945                 update_changeattr(dir, &data->res.dir_cinfo);
2946                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2947                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2948         }
2949         return status;
2950 }
2951
2952 static void nfs4_free_createdata(struct nfs4_createdata *data)
2953 {
2954         kfree(data);
2955 }
2956
2957 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2958                 struct page *page, unsigned int len, struct iattr *sattr)
2959 {
2960         struct nfs4_createdata *data;
2961         int status = -ENAMETOOLONG;
2962
2963         if (len > NFS4_MAXPATHLEN)
2964                 goto out;
2965
2966         status = -ENOMEM;
2967         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2968         if (data == NULL)
2969                 goto out;
2970
2971         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2972         data->arg.u.symlink.pages = &page;
2973         data->arg.u.symlink.len = len;
2974
2975         status = nfs4_do_create(dir, dentry, data);
2976
2977         nfs4_free_createdata(data);
2978 out:
2979         return status;
2980 }
2981
2982 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2983                 struct page *page, unsigned int len, struct iattr *sattr)
2984 {
2985         struct nfs4_exception exception = { };
2986         int err;
2987         do {
2988                 err = nfs4_handle_exception(NFS_SERVER(dir),
2989                                 _nfs4_proc_symlink(dir, dentry, page,
2990                                                         len, sattr),
2991                                 &exception);
2992         } while (exception.retry);
2993         return err;
2994 }
2995
2996 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2997                 struct iattr *sattr)
2998 {
2999         struct nfs4_createdata *data;
3000         int status = -ENOMEM;
3001
3002         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3003         if (data == NULL)
3004                 goto out;
3005
3006         status = nfs4_do_create(dir, dentry, data);
3007
3008         nfs4_free_createdata(data);
3009 out:
3010         return status;
3011 }
3012
3013 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3014                 struct iattr *sattr)
3015 {
3016         struct nfs4_exception exception = { };
3017         int err;
3018
3019         sattr->ia_mode &= ~current_umask();
3020         do {
3021                 err = nfs4_handle_exception(NFS_SERVER(dir),
3022                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3023                                 &exception);
3024         } while (exception.retry);
3025         return err;
3026 }
3027
3028 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3029                 u64 cookie, struct page **pages, unsigned int count, int plus)
3030 {
3031         struct inode            *dir = dentry->d_inode;
3032         struct nfs4_readdir_arg args = {
3033                 .fh = NFS_FH(dir),
3034                 .pages = pages,
3035                 .pgbase = 0,
3036                 .count = count,
3037                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3038                 .plus = plus,
3039         };
3040         struct nfs4_readdir_res res;
3041         struct rpc_message msg = {
3042                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3043                 .rpc_argp = &args,
3044                 .rpc_resp = &res,
3045                 .rpc_cred = cred,
3046         };
3047         int                     status;
3048
3049         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3050                         dentry->d_parent->d_name.name,
3051                         dentry->d_name.name,
3052                         (unsigned long long)cookie);
3053         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3054         res.pgbase = args.pgbase;
3055         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3056         if (status >= 0) {
3057                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3058                 status += args.pgbase;
3059         }
3060
3061         nfs_invalidate_atime(dir);
3062
3063         dprintk("%s: returns %d\n", __func__, status);
3064         return status;
3065 }
3066
3067 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3068                 u64 cookie, struct page **pages, unsigned int count, int plus)
3069 {
3070         struct nfs4_exception exception = { };
3071         int err;
3072         do {
3073                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3074                                 _nfs4_proc_readdir(dentry, cred, cookie,
3075                                         pages, count, plus),
3076                                 &exception);
3077         } while (exception.retry);
3078         return err;
3079 }
3080
3081 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3082                 struct iattr *sattr, dev_t rdev)
3083 {
3084         struct nfs4_createdata *data;
3085         int mode = sattr->ia_mode;
3086         int status = -ENOMEM;
3087
3088         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3089         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3090
3091         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3092         if (data == NULL)
3093                 goto out;
3094
3095         if (S_ISFIFO(mode))
3096                 data->arg.ftype = NF4FIFO;
3097         else if (S_ISBLK(mode)) {
3098                 data->arg.ftype = NF4BLK;
3099                 data->arg.u.device.specdata1 = MAJOR(rdev);
3100                 data->arg.u.device.specdata2 = MINOR(rdev);
3101         }
3102         else if (S_ISCHR(mode)) {
3103                 data->arg.ftype = NF4CHR;
3104                 data->arg.u.device.specdata1 = MAJOR(rdev);
3105                 data->arg.u.device.specdata2 = MINOR(rdev);
3106         }
3107
3108         status = nfs4_do_create(dir, dentry, data);
3109
3110         nfs4_free_createdata(data);
3111 out:
3112         return status;
3113 }
3114
3115 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3116                 struct iattr *sattr, dev_t rdev)
3117 {
3118         struct nfs4_exception exception = { };
3119         int err;
3120
3121         sattr->ia_mode &= ~current_umask();
3122         do {
3123                 err = nfs4_handle_exception(NFS_SERVER(dir),
3124                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3125                                 &exception);
3126         } while (exception.retry);
3127         return err;
3128 }
3129
3130 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3131                  struct nfs_fsstat *fsstat)
3132 {
3133         struct nfs4_statfs_arg args = {
3134                 .fh = fhandle,
3135                 .bitmask = server->attr_bitmask,
3136         };
3137         struct nfs4_statfs_res res = {
3138                 .fsstat = fsstat,
3139         };
3140         struct rpc_message msg = {
3141                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3142                 .rpc_argp = &args,
3143                 .rpc_resp = &res,
3144         };
3145
3146         nfs_fattr_init(fsstat->fattr);
3147         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3148 }
3149
3150 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3151 {
3152         struct nfs4_exception exception = { };
3153         int err;
3154         do {
3155                 err = nfs4_handle_exception(server,
3156                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3157                                 &exception);
3158         } while (exception.retry);
3159         return err;
3160 }
3161
3162 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3163                 struct nfs_fsinfo *fsinfo)
3164 {
3165         struct nfs4_fsinfo_arg args = {
3166                 .fh = fhandle,
3167                 .bitmask = server->attr_bitmask,
3168         };
3169         struct nfs4_fsinfo_res res = {
3170                 .fsinfo = fsinfo,
3171         };
3172         struct rpc_message msg = {
3173                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3174                 .rpc_argp = &args,
3175                 .rpc_resp = &res,
3176         };
3177
3178         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3179 }
3180
3181 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3182 {
3183         struct nfs4_exception exception = { };
3184         int err;
3185
3186         do {
3187                 err = nfs4_handle_exception(server,
3188                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3189                                 &exception);
3190         } while (exception.retry);
3191         return err;
3192 }
3193
3194 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3195 {
3196         nfs_fattr_init(fsinfo->fattr);
3197         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3198 }
3199
3200 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3201                 struct nfs_pathconf *pathconf)
3202 {
3203         struct nfs4_pathconf_arg args = {
3204                 .fh = fhandle,
3205                 .bitmask = server->attr_bitmask,
3206         };
3207         struct nfs4_pathconf_res res = {
3208                 .pathconf = pathconf,
3209         };
3210         struct rpc_message msg = {
3211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3212                 .rpc_argp = &args,
3213                 .rpc_resp = &res,
3214         };
3215
3216         /* None of the pathconf attributes are mandatory to implement */
3217         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3218                 memset(pathconf, 0, sizeof(*pathconf));
3219                 return 0;
3220         }
3221
3222         nfs_fattr_init(pathconf->fattr);
3223         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3224 }
3225
3226 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3227                 struct nfs_pathconf *pathconf)
3228 {
3229         struct nfs4_exception exception = { };
3230         int err;
3231
3232         do {
3233                 err = nfs4_handle_exception(server,
3234                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3235                                 &exception);
3236         } while (exception.retry);
3237         return err;
3238 }
3239
3240 void __nfs4_read_done_cb(struct nfs_read_data *data)
3241 {
3242         nfs_invalidate_atime(data->inode);
3243 }
3244
3245 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3246 {
3247         struct nfs_server *server = NFS_SERVER(data->inode);
3248
3249         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3250                 rpc_restart_call_prepare(task);
3251                 return -EAGAIN;
3252         }
3253
3254         __nfs4_read_done_cb(data);
3255         if (task->tk_status > 0)
3256                 renew_lease(server, data->timestamp);
3257         return 0;
3258 }
3259
3260 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3261 {
3262
3263         dprintk("--> %s\n", __func__);
3264
3265         if (!nfs4_sequence_done(task, &data->res.seq_res))
3266                 return -EAGAIN;
3267
3268         return data->read_done_cb ? data->read_done_cb(task, data) :
3269                                     nfs4_read_done_cb(task, data);
3270 }
3271
3272 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3273 {
3274         data->timestamp   = jiffies;
3275         data->read_done_cb = nfs4_read_done_cb;
3276         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3277         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3278 }
3279
3280 /* Reset the the nfs_read_data to send the read to the MDS. */
3281 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3282 {
3283         dprintk("%s Reset task for i/o through\n", __func__);
3284         put_lseg(data->lseg);
3285         data->lseg = NULL;
3286         /* offsets will differ in the dense stripe case */
3287         data->args.offset = data->mds_offset;
3288         data->ds_clp = NULL;
3289         data->args.fh     = NFS_FH(data->inode);
3290         data->read_done_cb = nfs4_read_done_cb;
3291         task->tk_ops = data->mds_ops;
3292         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3293 }
3294 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3295
3296 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3297 {
3298         struct inode *inode = data->inode;
3299
3300         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3301                 rpc_restart_call_prepare(task);
3302                 return -EAGAIN;
3303         }
3304         if (task->tk_status >= 0) {
3305                 renew_lease(NFS_SERVER(inode), data->timestamp);
3306                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3307         }
3308         return 0;
3309 }
3310
3311 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3312 {
3313         if (!nfs4_sequence_done(task, &data->res.seq_res))
3314                 return -EAGAIN;
3315         return data->write_done_cb ? data->write_done_cb(task, data) :
3316                 nfs4_write_done_cb(task, data);
3317 }
3318
3319 /* Reset the the nfs_write_data to send the write to the MDS. */
3320 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3321 {
3322         dprintk("%s Reset task for i/o through\n", __func__);
3323         put_lseg(data->lseg);
3324         data->lseg          = NULL;
3325         data->ds_clp        = NULL;
3326         data->write_done_cb = nfs4_write_done_cb;
3327         data->args.fh       = NFS_FH(data->inode);
3328         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3329         data->args.offset   = data->mds_offset;
3330         data->res.fattr     = &data->fattr;
3331         task->tk_ops        = data->mds_ops;
3332         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3333 }
3334 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3335
3336 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3337 {
3338         struct nfs_server *server = NFS_SERVER(data->inode);
3339
3340         if (data->lseg) {
3341                 data->args.bitmask = NULL;
3342                 data->res.fattr = NULL;
3343         } else
3344                 data->args.bitmask = server->cache_consistency_bitmask;
3345         if (!data->write_done_cb)
3346                 data->write_done_cb = nfs4_write_done_cb;
3347         data->res.server = server;
3348         data->timestamp   = jiffies;
3349
3350         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3351         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3352 }
3353
3354 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3355 {
3356         struct inode *inode = data->inode;
3357
3358         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3359                 rpc_restart_call_prepare(task);
3360                 return -EAGAIN;
3361         }
3362         nfs_refresh_inode(inode, data->res.fattr);
3363         return 0;
3364 }
3365
3366 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3367 {
3368         if (!nfs4_sequence_done(task, &data->res.seq_res))
3369                 return -EAGAIN;
3370         return data->write_done_cb(task, data);
3371 }
3372
3373 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3374 {
3375         struct nfs_server *server = NFS_SERVER(data->inode);
3376
3377         if (data->lseg) {
3378                 data->args.bitmask = NULL;
3379                 data->res.fattr = NULL;
3380         } else
3381                 data->args.bitmask = server->cache_consistency_bitmask;
3382         if (!data->write_done_cb)
3383                 data->write_done_cb = nfs4_commit_done_cb;
3384         data->res.server = server;
3385         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3386         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3387 }
3388
3389 struct nfs4_renewdata {
3390         struct nfs_client       *client;
3391         unsigned long           timestamp;
3392 };
3393
3394 /*
3395  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3396  * standalone procedure for queueing an asynchronous RENEW.
3397  */
3398 static void nfs4_renew_release(void *calldata)
3399 {
3400         struct nfs4_renewdata *data = calldata;
3401         struct nfs_client *clp = data->client;
3402
3403         if (atomic_read(&clp->cl_count) > 1)
3404                 nfs4_schedule_state_renewal(clp);
3405         nfs_put_client(clp);
3406         kfree(data);
3407 }
3408
3409 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3410 {
3411         struct nfs4_renewdata *data = calldata;
3412         struct nfs_client *clp = data->client;
3413         unsigned long timestamp = data->timestamp;
3414
3415         if (task->tk_status < 0) {
3416                 /* Unless we're shutting down, schedule state recovery! */
3417                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3418                         return;
3419                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3420                         nfs4_schedule_lease_recovery(clp);
3421                         return;
3422                 }
3423                 nfs4_schedule_path_down_recovery(clp);
3424         }
3425         do_renew_lease(clp, timestamp);
3426 }
3427
3428 static const struct rpc_call_ops nfs4_renew_ops = {
3429         .rpc_call_done = nfs4_renew_done,
3430         .rpc_release = nfs4_renew_release,
3431 };
3432
3433 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3434 {
3435         struct rpc_message msg = {
3436                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3437                 .rpc_argp       = clp,
3438                 .rpc_cred       = cred,
3439         };
3440         struct nfs4_renewdata *data;
3441
3442         if (renew_flags == 0)
3443                 return 0;
3444         if (!atomic_inc_not_zero(&clp->cl_count))
3445                 return -EIO;
3446         data = kmalloc(sizeof(*data), GFP_NOFS);
3447         if (data == NULL)
3448                 return -ENOMEM;
3449         data->client = clp;
3450         data->timestamp = jiffies;
3451         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3452                         &nfs4_renew_ops, data);
3453 }
3454
3455 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3456 {
3457         struct rpc_message msg = {
3458                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3459                 .rpc_argp       = clp,
3460                 .rpc_cred       = cred,
3461         };
3462         unsigned long now = jiffies;
3463         int status;
3464
3465         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3466         if (status < 0)
3467                 return status;
3468         do_renew_lease(clp, now);
3469         return 0;
3470 }
3471
3472 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3473 {
3474         return (server->caps & NFS_CAP_ACLS)
3475                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3476                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3477 }
3478
3479 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3480  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3481  * the stack.
3482  */
3483 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3484
3485 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3486                 struct page **pages, unsigned int *pgbase)
3487 {
3488         struct page *newpage, **spages;
3489         int rc = 0;
3490         size_t len;
3491         spages = pages;
3492
3493         do {
3494                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3495                 newpage = alloc_page(GFP_KERNEL);
3496
3497                 if (newpage == NULL)
3498                         goto unwind;
3499                 memcpy(page_address(newpage), buf, len);
3500                 buf += len;
3501                 buflen -= len;
3502                 *pages++ = newpage;
3503                 rc++;
3504         } while (buflen != 0);
3505
3506         return rc;
3507
3508 unwind:
3509         for(; rc > 0; rc--)
3510                 __free_page(spages[rc-1]);
3511         return -ENOMEM;
3512 }
3513
3514 struct nfs4_cached_acl {
3515         int cached;
3516         size_t len;
3517         char data[0];
3518 };
3519
3520 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3521 {
3522         struct nfs_inode *nfsi = NFS_I(inode);
3523
3524         spin_lock(&inode->i_lock);
3525         kfree(nfsi->nfs4_acl);
3526         nfsi->nfs4_acl = acl;
3527         spin_unlock(&inode->i_lock);
3528 }
3529
3530 static void nfs4_zap_acl_attr(struct inode *inode)
3531 {
3532         nfs4_set_cached_acl(inode, NULL);
3533 }
3534
3535 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3536 {
3537         struct nfs_inode *nfsi = NFS_I(inode);
3538         struct nfs4_cached_acl *acl;
3539         int ret = -ENOENT;
3540
3541         spin_lock(&inode->i_lock);
3542         acl = nfsi->nfs4_acl;
3543         if (acl == NULL)
3544                 goto out;
3545         if (buf == NULL) /* user is just asking for length */
3546                 goto out_len;
3547         if (acl->cached == 0)
3548                 goto out;
3549         ret = -ERANGE; /* see getxattr(2) man page */
3550         if (acl->len > buflen)
3551                 goto out;
3552         memcpy(buf, acl->data, acl->len);
3553 out_len:
3554         ret = acl->len;
3555 out:
3556         spin_unlock(&inode->i_lock);
3557         return ret;
3558 }
3559
3560 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3561 {
3562         struct nfs4_cached_acl *acl;
3563
3564         if (buf && acl_len <= PAGE_SIZE) {
3565                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3566                 if (acl == NULL)
3567                         goto out;
3568                 acl->cached = 1;
3569                 memcpy(acl->data, buf, acl_len);
3570         } else {
3571                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3572                 if (acl == NULL)
3573                         goto out;
3574                 acl->cached = 0;
3575         }
3576         acl->len = acl_len;
3577 out:
3578         nfs4_set_cached_acl(inode, acl);
3579 }
3580
3581 /*
3582  * The getxattr API returns the required buffer length when called with a
3583  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3584  * the required buf.  On a NULL buf, we send a page of data to the server
3585  * guessing that the ACL request can be serviced by a page. If so, we cache
3586  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3587  * the cache. If not so, we throw away the page, and cache the required
3588  * length. The next getxattr call will then produce another round trip to
3589  * the server, this time with the input buf of the required size.
3590  */
3591 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3592 {
3593         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3594         struct nfs_getaclargs args = {
3595                 .fh = NFS_FH(inode),
3596                 .acl_pages = pages,
3597                 .acl_len = buflen,
3598         };
3599         struct nfs_getaclres res = {
3600                 .acl_len = buflen,
3601         };
3602         void *resp_buf;
3603         struct rpc_message msg = {
3604                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3605                 .rpc_argp = &args,
3606                 .rpc_resp = &res,
3607         };
3608         int ret = -ENOMEM, npages, i, acl_len = 0;
3609
3610         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3611         /* As long as we're doing a round trip to the server anyway,
3612          * let's be prepared for a page of acl data. */
3613         if (npages == 0)
3614                 npages = 1;
3615
3616         for (i = 0; i < npages; i++) {
3617                 pages[i] = alloc_page(GFP_KERNEL);
3618                 if (!pages[i])
3619                         goto out_free;
3620         }
3621         if (npages > 1) {
3622                 /* for decoding across pages */
3623                 args.acl_scratch = alloc_page(GFP_KERNEL);
3624                 if (!args.acl_scratch)
3625                         goto out_free;
3626         }
3627         args.acl_len = npages * PAGE_SIZE;
3628         args.acl_pgbase = 0;
3629         /* Let decode_getfacl know not to fail if the ACL data is larger than
3630          * the page we send as a guess */
3631         if (buf == NULL)
3632                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3633         resp_buf = page_address(pages[0]);
3634
3635         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3636                 __func__, buf, buflen, npages, args.acl_len);
3637         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3638                              &msg, &args.seq_args, &res.seq_res, 0);
3639         if (ret)
3640                 goto out_free;
3641
3642         acl_len = res.acl_len - res.acl_data_offset;
3643         if (acl_len > args.acl_len)
3644                 nfs4_write_cached_acl(inode, NULL, acl_len);
3645         else
3646                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3647                                       acl_len);
3648         if (buf) {
3649                 ret = -ERANGE;
3650                 if (acl_len > buflen)
3651                         goto out_free;
3652                 _copy_from_pages(buf, pages, res.acl_data_offset,
3653                                 res.acl_len);
3654         }
3655         ret = acl_len;
3656 out_free:
3657         for (i = 0; i < npages; i++)
3658                 if (pages[i])
3659                         __free_page(pages[i]);
3660         if (args.acl_scratch)
3661                 __free_page(args.acl_scratch);
3662         return ret;
3663 }
3664
3665 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3666 {
3667         struct nfs4_exception exception = { };
3668         ssize_t ret;
3669         do {
3670                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3671                 if (ret >= 0)
3672                         break;
3673                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3674         } while (exception.retry);
3675         return ret;
3676 }
3677
3678 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3679 {
3680         struct nfs_server *server = NFS_SERVER(inode);
3681         int ret;
3682
3683         if (!nfs4_server_supports_acls(server))
3684                 return -EOPNOTSUPP;
3685         ret = nfs_revalidate_inode(server, inode);
3686         if (ret < 0)
3687                 return ret;
3688         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3689                 nfs_zap_acl_cache(inode);
3690         ret = nfs4_read_cached_acl(inode, buf, buflen);
3691         if (ret != -ENOENT)
3692                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3693                  * but no cached acl data, just the acl length */
3694                 return ret;
3695         return nfs4_get_acl_uncached(inode, buf, buflen);
3696 }
3697
3698 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3699 {
3700         struct nfs_server *server = NFS_SERVER(inode);
3701         struct page *pages[NFS4ACL_MAXPAGES];
3702         struct nfs_setaclargs arg = {
3703                 .fh             = NFS_FH(inode),
3704                 .acl_pages      = pages,
3705                 .acl_len        = buflen,
3706         };
3707         struct nfs_setaclres res;
3708         struct rpc_message msg = {
3709                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3710                 .rpc_argp       = &arg,
3711                 .rpc_resp       = &res,
3712         };
3713         int ret, i;
3714
3715         if (!nfs4_server_supports_acls(server))
3716                 return -EOPNOTSUPP;
3717         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3718         if (i < 0)
3719                 return i;
3720         nfs_inode_return_delegation(inode);
3721         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3722
3723         /*
3724          * Free each page after tx, so the only ref left is
3725          * held by the network stack
3726          */
3727         for (; i > 0; i--)
3728                 put_page(pages[i-1]);
3729
3730         /*
3731          * Acl update can result in inode attribute update.
3732          * so mark the attribute cache invalid.
3733          */
3734         spin_lock(&inode->i_lock);
3735         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3736         spin_unlock(&inode->i_lock);
3737         nfs_access_zap_cache(inode);
3738         nfs_zap_acl_cache(inode);
3739         return ret;
3740 }
3741
3742 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3743 {
3744         struct nfs4_exception exception = { };
3745         int err;
3746         do {
3747                 err = nfs4_handle_exception(NFS_SERVER(inode),
3748                                 __nfs4_proc_set_acl(inode, buf, buflen),
3749                                 &exception);
3750         } while (exception.retry);
3751         return err;
3752 }
3753
3754 static int
3755 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3756 {
3757         struct nfs_client *clp = server->nfs_client;
3758
3759         if (task->tk_status >= 0)
3760                 return 0;
3761         switch(task->tk_status) {
3762                 case -NFS4ERR_ADMIN_REVOKED:
3763                 case -NFS4ERR_BAD_STATEID:
3764                 case -NFS4ERR_OPENMODE:
3765                         if (state == NULL)
3766                                 break;
3767                         nfs4_schedule_stateid_recovery(server, state);
3768                         goto wait_on_recovery;
3769                 case -NFS4ERR_EXPIRED:
3770                         if (state != NULL)
3771                                 nfs4_schedule_stateid_recovery(server, state);
3772                 case -NFS4ERR_STALE_STATEID:
3773                 case -NFS4ERR_STALE_CLIENTID:
3774                         nfs4_schedule_lease_recovery(clp);
3775                         goto wait_on_recovery;
3776 #if defined(CONFIG_NFS_V4_1)
3777                 case -NFS4ERR_BADSESSION:
3778                 case -NFS4ERR_BADSLOT:
3779                 case -NFS4ERR_BAD_HIGH_SLOT:
3780                 case -NFS4ERR_DEADSESSION:
3781                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3782                 case -NFS4ERR_SEQ_FALSE_RETRY:
3783                 case -NFS4ERR_SEQ_MISORDERED:
3784                         dprintk("%s ERROR %d, Reset session\n", __func__,
3785                                 task->tk_status);
3786                         nfs4_schedule_session_recovery(clp->cl_session);
3787                         task->tk_status = 0;
3788                         return -EAGAIN;
3789 #endif /* CONFIG_NFS_V4_1 */
3790                 case -NFS4ERR_DELAY:
3791                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3792                 case -NFS4ERR_GRACE:
3793                 case -EKEYEXPIRED:
3794                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3795                         task->tk_status = 0;
3796                         return -EAGAIN;
3797                 case -NFS4ERR_RETRY_UNCACHED_REP:
3798                 case -NFS4ERR_OLD_STATEID:
3799                         task->tk_status = 0;
3800                         return -EAGAIN;
3801         }
3802         task->tk_status = nfs4_map_errors(task->tk_status);
3803         return 0;
3804 wait_on_recovery:
3805         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3806         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3807                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3808         task->tk_status = 0;
3809         return -EAGAIN;
3810 }
3811
3812 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3813                 unsigned short port, struct rpc_cred *cred,
3814                 struct nfs4_setclientid_res *res)
3815 {
3816         nfs4_verifier sc_verifier;
3817         struct nfs4_setclientid setclientid = {
3818                 .sc_verifier = &sc_verifier,
3819                 .sc_prog = program,
3820                 .sc_cb_ident = clp->cl_cb_ident,
3821         };
3822         struct rpc_message msg = {
3823                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3824                 .rpc_argp = &setclientid,
3825                 .rpc_resp = res,
3826                 .rpc_cred = cred,
3827         };
3828         __be32 *p;
3829         int loop = 0;
3830         int status;
3831
3832         p = (__be32*)sc_verifier.data;
3833         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3834         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3835
3836         for(;;) {
3837                 rcu_read_lock();
3838                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3839                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3840                                 clp->cl_ipaddr,
3841                                 rpc_peeraddr2str(clp->cl_rpcclient,
3842                                                         RPC_DISPLAY_ADDR),
3843                                 rpc_peeraddr2str(clp->cl_rpcclient,
3844                                                         RPC_DISPLAY_PROTO),
3845                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3846                                 clp->cl_id_uniquifier);
3847                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3848                                 sizeof(setclientid.sc_netid),
3849                                 rpc_peeraddr2str(clp->cl_rpcclient,
3850                                                         RPC_DISPLAY_NETID));
3851                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3852                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3853                                 clp->cl_ipaddr, port >> 8, port & 255);
3854                 rcu_read_unlock();
3855
3856                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3857                 if (status != -NFS4ERR_CLID_INUSE)
3858                         break;
3859                 if (loop != 0) {
3860                         ++clp->cl_id_uniquifier;
3861                         break;
3862                 }
3863                 ++loop;
3864                 ssleep(clp->cl_lease_time / HZ + 1);
3865         }
3866         return status;
3867 }
3868
3869 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3870                 struct nfs4_setclientid_res *arg,
3871                 struct rpc_cred *cred)
3872 {
3873         struct nfs_fsinfo fsinfo;
3874         struct rpc_message msg = {
3875                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3876                 .rpc_argp = arg,
3877                 .rpc_resp = &fsinfo,
3878                 .rpc_cred = cred,
3879         };
3880         unsigned long now;
3881         int status;
3882
3883         now = jiffies;
3884         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3885         if (status == 0) {
3886                 spin_lock(&clp->cl_lock);
3887                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3888                 clp->cl_last_renewal = now;
3889                 spin_unlock(&clp->cl_lock);
3890         }
3891         return status;
3892 }
3893
3894 struct nfs4_delegreturndata {
3895         struct nfs4_delegreturnargs args;
3896         struct nfs4_delegreturnres res;
3897         struct nfs_fh fh;
3898         nfs4_stateid stateid;
3899         unsigned long timestamp;
3900         struct nfs_fattr fattr;
3901         int rpc_status;
3902 };
3903
3904 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3905 {
3906         struct nfs4_delegreturndata *data = calldata;
3907
3908         if (!nfs4_sequence_done(task, &data->res.seq_res))
3909                 return;
3910
3911         switch (task->tk_status) {
3912         case -NFS4ERR_STALE_STATEID:
3913         case -NFS4ERR_EXPIRED:
3914         case 0:
3915                 renew_lease(data->res.server, data->timestamp);
3916                 break;
3917         default:
3918                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3919                                 -EAGAIN) {
3920                         rpc_restart_call_prepare(task);
3921                         return;
3922                 }
3923         }
3924         data->rpc_status = task->tk_status;
3925 }
3926
3927 static void nfs4_delegreturn_release(void *calldata)
3928 {
3929         kfree(calldata);
3930 }
3931
3932 #if defined(CONFIG_NFS_V4_1)
3933 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3934 {
3935         struct nfs4_delegreturndata *d_data;
3936
3937         d_data = (struct nfs4_delegreturndata *)data;
3938
3939         if (nfs4_setup_sequence(d_data->res.server,
3940                                 &d_data->args.seq_args,
3941                                 &d_data->res.seq_res, task))
3942                 return;
3943         rpc_call_start(task);
3944 }
3945 #endif /* CONFIG_NFS_V4_1 */
3946
3947 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3948 #if defined(CONFIG_NFS_V4_1)
3949         .rpc_call_prepare = nfs4_delegreturn_prepare,
3950 #endif /* CONFIG_NFS_V4_1 */
3951         .rpc_call_done = nfs4_delegreturn_done,
3952         .rpc_release = nfs4_delegreturn_release,
3953 };
3954
3955 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3956 {
3957         struct nfs4_delegreturndata *data;
3958         struct nfs_server *server = NFS_SERVER(inode);
3959         struct rpc_task *task;
3960         struct rpc_message msg = {
3961                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3962                 .rpc_cred = cred,
3963         };
3964         struct rpc_task_setup task_setup_data = {
3965                 .rpc_client = server->client,
3966                 .rpc_message = &msg,
3967                 .callback_ops = &nfs4_delegreturn_ops,
3968                 .flags = RPC_TASK_ASYNC,
3969         };
3970         int status = 0;
3971
3972         data = kzalloc(sizeof(*data), GFP_NOFS);
3973         if (data == NULL)
3974                 return -ENOMEM;
3975         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3976         data->args.fhandle = &data->fh;
3977         data->args.stateid = &data->stateid;
3978         data->args.bitmask = server->attr_bitmask;
3979         nfs_copy_fh(&data->fh, NFS_FH(inode));
3980         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3981         data->res.fattr = &data->fattr;
3982         data->res.server = server;
3983         nfs_fattr_init(data->res.fattr);
3984         data->timestamp = jiffies;
3985         data->rpc_status = 0;
3986
3987         task_setup_data.callback_data = data;
3988         msg.rpc_argp = &data->args;
3989         msg.rpc_resp = &data->res;
3990         task = rpc_run_task(&task_setup_data);
3991         if (IS_ERR(task))
3992                 return PTR_ERR(task);
3993         if (!issync)
3994                 goto out;
3995         status = nfs4_wait_for_completion_rpc_task(task);
3996         if (status != 0)
3997                 goto out;
3998         status = data->rpc_status;
3999         if (status != 0)
4000                 goto out;
4001         nfs_refresh_inode(inode, &data->fattr);
4002 out:
4003         rpc_put_task(task);
4004         return status;
4005 }
4006
4007 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4008 {
4009         struct nfs_server *server = NFS_SERVER(inode);
4010         struct nfs4_exception exception = { };
4011         int err;
4012         do {
4013                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4014                 switch (err) {
4015                         case -NFS4ERR_STALE_STATEID:
4016                         case -NFS4ERR_EXPIRED:
4017                         case 0:
4018                                 return 0;
4019                 }
4020                 err = nfs4_handle_exception(server, err, &exception);
4021         } while (exception.retry);
4022         return err;
4023 }
4024
4025 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4026 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4027
4028 /*
4029  * sleep, with exponential backoff, and retry the LOCK operation.
4030  */
4031 static unsigned long
4032 nfs4_set_lock_task_retry(unsigned long timeout)
4033 {
4034         freezable_schedule_timeout_killable(timeout);
4035         timeout <<= 1;
4036         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4037                 return NFS4_LOCK_MAXTIMEOUT;
4038         return timeout;
4039 }
4040
4041 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4042 {
4043         struct inode *inode = state->inode;
4044         struct nfs_server *server = NFS_SERVER(inode);
4045         struct nfs_client *clp = server->nfs_client;
4046         struct nfs_lockt_args arg = {
4047                 .fh = NFS_FH(inode),
4048                 .fl = request,
4049         };
4050         struct nfs_lockt_res res = {
4051                 .denied = request,
4052         };
4053         struct rpc_message msg = {
4054                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4055                 .rpc_argp       = &arg,
4056                 .rpc_resp       = &res,
4057                 .rpc_cred       = state->owner->so_cred,
4058         };
4059         struct nfs4_lock_state *lsp;
4060         int status;
4061
4062         arg.lock_owner.clientid = clp->cl_clientid;
4063         status = nfs4_set_lock_state(state, request);
4064         if (status != 0)
4065                 goto out;
4066         lsp = request->fl_u.nfs4_fl.owner;
4067         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4068         arg.lock_owner.s_dev = server->s_dev;
4069         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4070         switch (status) {
4071                 case 0:
4072                         request->fl_type = F_UNLCK;
4073                         break;
4074                 case -NFS4ERR_DENIED:
4075                         status = 0;
4076         }
4077         request->fl_ops->fl_release_private(request);
4078 out:
4079         return status;
4080 }
4081
4082 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4083 {
4084         struct nfs4_exception exception = { };
4085         int err;
4086
4087         do {
4088                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4089                                 _nfs4_proc_getlk(state, cmd, request),
4090                                 &exception);
4091         } while (exception.retry);
4092         return err;
4093 }
4094
4095 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4096 {
4097         int res = 0;
4098         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4099                 case FL_POSIX:
4100                         res = posix_lock_file_wait(file, fl);
4101                         break;
4102                 case FL_FLOCK:
4103                         res = flock_lock_file_wait(file, fl);
4104                         break;
4105                 default:
4106                         BUG();
4107         }
4108         return res;
4109 }
4110
4111 struct nfs4_unlockdata {
4112         struct nfs_locku_args arg;
4113         struct nfs_locku_res res;
4114         struct nfs4_lock_state *lsp;
4115         struct nfs_open_context *ctx;
4116         struct file_lock fl;
4117         const struct nfs_server *server;
4118         unsigned long timestamp;
4119 };
4120
4121 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4122                 struct nfs_open_context *ctx,
4123                 struct nfs4_lock_state *lsp,
4124                 struct nfs_seqid *seqid)
4125 {
4126         struct nfs4_unlockdata *p;
4127         struct inode *inode = lsp->ls_state->inode;
4128
4129         p = kzalloc(sizeof(*p), GFP_NOFS);
4130         if (p == NULL)
4131                 return NULL;
4132         p->arg.fh = NFS_FH(inode);
4133         p->arg.fl = &p->fl;
4134         p->arg.seqid = seqid;
4135         p->res.seqid = seqid;
4136         p->arg.stateid = &lsp->ls_stateid;
4137         p->lsp = lsp;
4138         atomic_inc(&lsp->ls_count);
4139         /* Ensure we don't close file until we're done freeing locks! */
4140         p->ctx = get_nfs_open_context(ctx);
4141         memcpy(&p->fl, fl, sizeof(p->fl));
4142         p->server = NFS_SERVER(inode);
4143         return p;
4144 }
4145
4146 static void nfs4_locku_release_calldata(void *data)
4147 {
4148         struct nfs4_unlockdata *calldata = data;
4149         nfs_free_seqid(calldata->arg.seqid);
4150         nfs4_put_lock_state(calldata->lsp);
4151         put_nfs_open_context(calldata->ctx);
4152         kfree(calldata);
4153 }
4154
4155 static void nfs4_locku_done(struct rpc_task *task, void *data)
4156 {
4157         struct nfs4_unlockdata *calldata = data;
4158
4159         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4160                 return;
4161         switch (task->tk_status) {
4162                 case 0:
4163                         memcpy(calldata->lsp->ls_stateid.data,
4164                                         calldata->res.stateid.data,
4165                                         sizeof(calldata->lsp->ls_stateid.data));
4166                         renew_lease(calldata->server, calldata->timestamp);
4167                         break;
4168                 case -NFS4ERR_BAD_STATEID:
4169                 case -NFS4ERR_OLD_STATEID:
4170                 case -NFS4ERR_STALE_STATEID:
4171                 case -NFS4ERR_EXPIRED:
4172                         break;
4173                 default:
4174                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4175                                 rpc_restart_call_prepare(task);
4176         }
4177 }
4178
4179 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4180 {
4181         struct nfs4_unlockdata *calldata = data;
4182
4183         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4184                 return;
4185         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4186                 /* Note: exit _without_ running nfs4_locku_done */
4187                 task->tk_action = NULL;
4188                 return;
4189         }
4190         calldata->timestamp = jiffies;
4191         if (nfs4_setup_sequence(calldata->server,
4192                                 &calldata->arg.seq_args,
4193                                 &calldata->res.seq_res, task))
4194                 return;
4195         rpc_call_start(task);
4196 }
4197
4198 static const struct rpc_call_ops nfs4_locku_ops = {
4199         .rpc_call_prepare = nfs4_locku_prepare,
4200         .rpc_call_done = nfs4_locku_done,
4201         .rpc_release = nfs4_locku_release_calldata,
4202 };
4203
4204 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4205                 struct nfs_open_context *ctx,
4206                 struct nfs4_lock_state *lsp,
4207                 struct nfs_seqid *seqid)
4208 {
4209         struct nfs4_unlockdata *data;
4210         struct rpc_message msg = {
4211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4212                 .rpc_cred = ctx->cred,
4213         };
4214         struct rpc_task_setup task_setup_data = {
4215                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4216                 .rpc_message = &msg,
4217                 .callback_ops = &nfs4_locku_ops,
4218                 .workqueue = nfsiod_workqueue,
4219                 .flags = RPC_TASK_ASYNC,
4220         };
4221
4222         /* Ensure this is an unlock - when canceling a lock, the
4223          * canceled lock is passed in, and it won't be an unlock.
4224          */
4225         fl->fl_type = F_UNLCK;
4226
4227         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4228         if (data == NULL) {
4229                 nfs_free_seqid(seqid);
4230                 return ERR_PTR(-ENOMEM);
4231         }
4232
4233         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4234         msg.rpc_argp = &data->arg;
4235         msg.rpc_resp = &data->res;
4236         task_setup_data.callback_data = data;
4237         return rpc_run_task(&task_setup_data);
4238 }
4239
4240 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4241 {
4242         struct nfs_inode *nfsi = NFS_I(state->inode);
4243         struct nfs_seqid *seqid;
4244         struct nfs4_lock_state *lsp;
4245         struct rpc_task *task;
4246         int status = 0;
4247         unsigned char fl_flags = request->fl_flags;
4248
4249         status = nfs4_set_lock_state(state, request);
4250         /* Unlock _before_ we do the RPC call */
4251         request->fl_flags |= FL_EXISTS;
4252         down_read(&nfsi->rwsem);
4253         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4254                 up_read(&nfsi->rwsem);
4255                 goto out;
4256         }
4257         up_read(&nfsi->rwsem);
4258         if (status != 0)
4259                 goto out;
4260         /* Is this a delegated lock? */
4261         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4262                 goto out;
4263         lsp = request->fl_u.nfs4_fl.owner;
4264         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4265         status = -ENOMEM;
4266         if (seqid == NULL)
4267                 goto out;
4268         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4269         status = PTR_ERR(task);
4270         if (IS_ERR(task))
4271                 goto out;
4272         status = nfs4_wait_for_completion_rpc_task(task);
4273         rpc_put_task(task);
4274 out:
4275         request->fl_flags = fl_flags;
4276         return status;
4277 }
4278
4279 struct nfs4_lockdata {
4280         struct nfs_lock_args arg;
4281         struct nfs_lock_res res;
4282         struct nfs4_lock_state *lsp;
4283         struct nfs_open_context *ctx;
4284         struct file_lock fl;
4285         unsigned long timestamp;
4286         int rpc_status;
4287         int cancelled;
4288         struct nfs_server *server;
4289 };
4290
4291 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4292                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4293                 gfp_t gfp_mask)
4294 {
4295         struct nfs4_lockdata *p;
4296         struct inode *inode = lsp->ls_state->inode;
4297         struct nfs_server *server = NFS_SERVER(inode);
4298
4299         p = kzalloc(sizeof(*p), gfp_mask);
4300         if (p == NULL)
4301                 return NULL;
4302
4303         p->arg.fh = NFS_FH(inode);
4304         p->arg.fl = &p->fl;
4305         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4306         if (p->arg.open_seqid == NULL)
4307                 goto out_free;
4308         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4309         if (p->arg.lock_seqid == NULL)
4310                 goto out_free_seqid;
4311         p->arg.lock_stateid = &lsp->ls_stateid;
4312         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4313         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4314         p->arg.lock_owner.s_dev = server->s_dev;
4315         p->res.lock_seqid = p->arg.lock_seqid;
4316         p->lsp = lsp;
4317         p->server = server;
4318         atomic_inc(&lsp->ls_count);
4319         p->ctx = get_nfs_open_context(ctx);
4320         memcpy(&p->fl, fl, sizeof(p->fl));
4321         return p;
4322 out_free_seqid:
4323         nfs_free_seqid(p->arg.open_seqid);
4324 out_free:
4325         kfree(p);
4326         return NULL;
4327 }
4328
4329 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4330 {
4331         struct nfs4_lockdata *data = calldata;
4332         struct nfs4_state *state = data->lsp->ls_state;
4333
4334         dprintk("%s: begin!\n", __func__);
4335         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4336                 return;
4337         /* Do we need to do an open_to_lock_owner? */
4338         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4339                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4340                         return;
4341                 data->arg.open_stateid = &state->stateid;
4342                 data->arg.new_lock_owner = 1;
4343                 data->res.open_seqid = data->arg.open_seqid;
4344         } else
4345                 data->arg.new_lock_owner = 0;
4346         data->timestamp = jiffies;
4347         if (nfs4_setup_sequence(data->server,
4348                                 &data->arg.seq_args,
4349                                 &data->res.seq_res, task))
4350                 return;
4351         rpc_call_start(task);
4352         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4353 }
4354
4355 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4356 {
4357         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4358         nfs4_lock_prepare(task, calldata);
4359 }
4360
4361 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4362 {
4363         struct nfs4_lockdata *data = calldata;
4364
4365         dprintk("%s: begin!\n", __func__);
4366
4367         if (!nfs4_sequence_done(task, &data->res.seq_res))
4368                 return;
4369
4370         data->rpc_status = task->tk_status;
4371         if (data->arg.new_lock_owner != 0) {
4372                 if (data->rpc_status == 0)
4373                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4374                 else
4375                         goto out;
4376         }
4377         if (data->rpc_status == 0) {
4378                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4379                                         sizeof(data->lsp->ls_stateid.data));
4380                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4381                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4382         }
4383 out:
4384         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4385 }
4386
4387 static void nfs4_lock_release(void *calldata)
4388 {
4389         struct nfs4_lockdata *data = calldata;
4390
4391         dprintk("%s: begin!\n", __func__);
4392         nfs_free_seqid(data->arg.open_seqid);
4393         if (data->cancelled != 0) {
4394                 struct rpc_task *task;
4395                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4396                                 data->arg.lock_seqid);
4397                 if (!IS_ERR(task))
4398                         rpc_put_task_async(task);
4399                 dprintk("%s: cancelling lock!\n", __func__);
4400         } else
4401                 nfs_free_seqid(data->arg.lock_seqid);
4402         nfs4_put_lock_state(data->lsp);
4403         put_nfs_open_context(data->ctx);
4404         kfree(data);
4405         dprintk("%s: done!\n", __func__);
4406 }
4407
4408 static const struct rpc_call_ops nfs4_lock_ops = {
4409         .rpc_call_prepare = nfs4_lock_prepare,
4410         .rpc_call_done = nfs4_lock_done,
4411         .rpc_release = nfs4_lock_release,
4412 };
4413
4414 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4415         .rpc_call_prepare = nfs4_recover_lock_prepare,
4416         .rpc_call_done = nfs4_lock_done,
4417         .rpc_release = nfs4_lock_release,
4418 };
4419
4420 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4421 {
4422         switch (error) {
4423         case -NFS4ERR_ADMIN_REVOKED:
4424         case -NFS4ERR_BAD_STATEID:
4425                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4426                 if (new_lock_owner != 0 ||
4427                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4428                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4429                 break;
4430         case -NFS4ERR_STALE_STATEID:
4431                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4432         case -NFS4ERR_EXPIRED:
4433                 nfs4_schedule_lease_recovery(server->nfs_client);
4434         };
4435 }
4436
4437 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4438 {
4439         struct nfs4_lockdata *data;
4440         struct rpc_task *task;
4441         struct rpc_message msg = {
4442                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4443                 .rpc_cred = state->owner->so_cred,
4444         };
4445         struct rpc_task_setup task_setup_data = {
4446                 .rpc_client = NFS_CLIENT(state->inode),
4447                 .rpc_message = &msg,
4448                 .callback_ops = &nfs4_lock_ops,
4449                 .workqueue = nfsiod_workqueue,
4450                 .flags = RPC_TASK_ASYNC,
4451         };
4452         int ret;
4453
4454         dprintk("%s: begin!\n", __func__);
4455         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4456                         fl->fl_u.nfs4_fl.owner,
4457                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4458         if (data == NULL)
4459                 return -ENOMEM;
4460         if (IS_SETLKW(cmd))
4461                 data->arg.block = 1;
4462         if (recovery_type > NFS_LOCK_NEW) {
4463                 if (recovery_type == NFS_LOCK_RECLAIM)
4464                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4465                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4466         }
4467         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4468         msg.rpc_argp = &data->arg;
4469         msg.rpc_resp = &data->res;
4470         task_setup_data.callback_data = data;
4471         task = rpc_run_task(&task_setup_data);
4472         if (IS_ERR(task))
4473                 return PTR_ERR(task);
4474         ret = nfs4_wait_for_completion_rpc_task(task);
4475         if (ret == 0) {
4476                 ret = data->rpc_status;
4477                 if (ret)
4478                         nfs4_handle_setlk_error(data->server, data->lsp,
4479                                         data->arg.new_lock_owner, ret);
4480         } else
4481                 data->cancelled = 1;
4482         rpc_put_task(task);
4483         dprintk("%s: done, ret = %d!\n", __func__, ret);
4484         return ret;
4485 }
4486
4487 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4488 {
4489         struct nfs_server *server = NFS_SERVER(state->inode);
4490         struct nfs4_exception exception = { };
4491         int err;
4492
4493         do {
4494                 /* Cache the lock if possible... */
4495                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4496                         return 0;
4497                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4498                 if (err != -NFS4ERR_DELAY)
4499                         break;
4500                 nfs4_handle_exception(server, err, &exception);
4501         } while (exception.retry);
4502         return err;
4503 }
4504
4505 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4506 {
4507         struct nfs_server *server = NFS_SERVER(state->inode);
4508         struct nfs4_exception exception = { };
4509         int err;
4510
4511         err = nfs4_set_lock_state(state, request);
4512         if (err != 0)
4513                 return err;
4514         do {
4515                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4516                         return 0;
4517                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4518                 switch (err) {
4519                 default:
4520                         goto out;
4521                 case -NFS4ERR_GRACE:
4522                 case -NFS4ERR_DELAY:
4523                         nfs4_handle_exception(server, err, &exception);
4524                         err = 0;
4525                 }
4526         } while (exception.retry);
4527 out:
4528         return err;
4529 }
4530
4531 #if defined(CONFIG_NFS_V4_1)
4532 static int nfs41_check_expired_locks(struct nfs4_state *state)
4533 {
4534         int status, ret = NFS_OK;
4535         struct nfs4_lock_state *lsp;
4536         struct nfs_server *server = NFS_SERVER(state->inode);
4537
4538         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4539                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4540                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4541                         if (status != NFS_OK) {
4542                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4543                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4544                                 ret = status;
4545                         }
4546                 }
4547         };
4548
4549         return ret;
4550 }
4551
4552 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4553 {
4554         int status = NFS_OK;
4555
4556         if (test_bit(LK_STATE_IN_USE, &state->flags))
4557                 status = nfs41_check_expired_locks(state);
4558         if (status == NFS_OK)
4559                 return status;
4560         return nfs4_lock_expired(state, request);
4561 }
4562 #endif
4563
4564 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4565 {
4566         struct nfs_inode *nfsi = NFS_I(state->inode);
4567         unsigned char fl_flags = request->fl_flags;
4568         int status = -ENOLCK;
4569
4570         if ((fl_flags & FL_POSIX) &&
4571                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4572                 goto out;
4573         /* Is this a delegated open? */
4574         status = nfs4_set_lock_state(state, request);
4575         if (status != 0)
4576                 goto out;
4577         request->fl_flags |= FL_ACCESS;
4578         status = do_vfs_lock(request->fl_file, request);
4579         if (status < 0)
4580                 goto out;
4581         down_read(&nfsi->rwsem);
4582         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4583                 /* Yes: cache locks! */
4584                 /* ...but avoid races with delegation recall... */
4585                 request->fl_flags = fl_flags & ~FL_SLEEP;
4586                 status = do_vfs_lock(request->fl_file, request);
4587                 goto out_unlock;
4588         }
4589         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4590         if (status != 0)
4591                 goto out_unlock;
4592         /* Note: we always want to sleep here! */
4593         request->fl_flags = fl_flags | FL_SLEEP;
4594         if (do_vfs_lock(request->fl_file, request) < 0)
4595                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4596                         "manager!\n", __func__);
4597 out_unlock:
4598         up_read(&nfsi->rwsem);
4599 out:
4600         request->fl_flags = fl_flags;
4601         return status;
4602 }
4603
4604 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4605 {
4606         struct nfs4_exception exception = { };
4607         int err;
4608
4609         do {
4610                 err = _nfs4_proc_setlk(state, cmd, request);
4611                 if (err == -NFS4ERR_DENIED)
4612                         err = -EAGAIN;
4613                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4614                                 err, &exception);
4615         } while (exception.retry);
4616         return err;
4617 }
4618
4619 static int
4620 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4621 {
4622         struct nfs_open_context *ctx;
4623         struct nfs4_state *state;
4624         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4625         int status;
4626
4627         /* verify open state */
4628         ctx = nfs_file_open_context(filp);
4629         state = ctx->state;
4630
4631         if (request->fl_start < 0 || request->fl_end < 0)
4632                 return -EINVAL;
4633
4634         if (IS_GETLK(cmd)) {
4635                 if (state != NULL)
4636                         return nfs4_proc_getlk(state, F_GETLK, request);
4637                 return 0;
4638         }
4639
4640         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4641                 return -EINVAL;
4642
4643         if (request->fl_type == F_UNLCK) {
4644                 if (state != NULL)
4645                         return nfs4_proc_unlck(state, cmd, request);
4646                 return 0;
4647         }
4648
4649         if (state == NULL)
4650                 return -ENOLCK;
4651         do {
4652                 status = nfs4_proc_setlk(state, cmd, request);
4653                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4654                         break;
4655                 timeout = nfs4_set_lock_task_retry(timeout);
4656                 status = -ERESTARTSYS;
4657                 if (signalled())
4658                         break;
4659         } while(status < 0);
4660         return status;
4661 }
4662
4663 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4664 {
4665         struct nfs_server *server = NFS_SERVER(state->inode);
4666         struct nfs4_exception exception = { };
4667         int err;
4668
4669         err = nfs4_set_lock_state(state, fl);
4670         if (err != 0)
4671                 goto out;
4672         do {
4673                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4674                 switch (err) {
4675                         default:
4676                                 printk(KERN_ERR "NFS: %s: unhandled error "
4677                                         "%d.\n", __func__, err);
4678                         case 0:
4679                         case -ESTALE:
4680                                 goto out;
4681                         case -NFS4ERR_EXPIRED:
4682                                 nfs4_schedule_stateid_recovery(server, state);
4683                         case -NFS4ERR_STALE_CLIENTID:
4684                         case -NFS4ERR_STALE_STATEID:
4685                                 nfs4_schedule_lease_recovery(server->nfs_client);
4686                                 goto out;
4687                         case -NFS4ERR_BADSESSION:
4688                         case -NFS4ERR_BADSLOT:
4689                         case -NFS4ERR_BAD_HIGH_SLOT:
4690                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4691                         case -NFS4ERR_DEADSESSION:
4692                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4693                                 goto out;
4694                         case -ERESTARTSYS:
4695                                 /*
4696                                  * The show must go on: exit, but mark the
4697                                  * stateid as needing recovery.
4698                                  */
4699                         case -NFS4ERR_ADMIN_REVOKED:
4700                         case -NFS4ERR_BAD_STATEID:
4701                         case -NFS4ERR_OPENMODE:
4702                                 nfs4_schedule_stateid_recovery(server, state);
4703                                 err = 0;
4704                                 goto out;
4705                         case -EKEYEXPIRED:
4706                                 /*
4707                                  * User RPCSEC_GSS context has expired.
4708                                  * We cannot recover this stateid now, so
4709                                  * skip it and allow recovery thread to
4710                                  * proceed.
4711                                  */
4712                                 err = 0;
4713                                 goto out;
4714                         case -ENOMEM:
4715                         case -NFS4ERR_DENIED:
4716                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4717                                 err = 0;
4718                                 goto out;
4719                         case -NFS4ERR_DELAY:
4720                                 break;
4721                 }
4722                 err = nfs4_handle_exception(server, err, &exception);
4723         } while (exception.retry);
4724 out:
4725         return err;
4726 }
4727
4728 static void nfs4_release_lockowner_release(void *calldata)
4729 {
4730         kfree(calldata);
4731 }
4732
4733 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4734         .rpc_release = nfs4_release_lockowner_release,
4735 };
4736
4737 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4738 {
4739         struct nfs_server *server = lsp->ls_state->owner->so_server;
4740         struct nfs_release_lockowner_args *args;
4741         struct rpc_message msg = {
4742                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4743         };
4744
4745         if (server->nfs_client->cl_mvops->minor_version != 0)
4746                 return;
4747         args = kmalloc(sizeof(*args), GFP_NOFS);
4748         if (!args)
4749                 return;
4750         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4751         args->lock_owner.id = lsp->ls_seqid.owner_id;
4752         args->lock_owner.s_dev = server->s_dev;
4753         msg.rpc_argp = args;
4754         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4755 }
4756
4757 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4758
4759 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4760                                    const void *buf, size_t buflen,
4761                                    int flags, int type)
4762 {
4763         if (strcmp(key, "") != 0)
4764                 return -EINVAL;
4765
4766         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4767 }
4768
4769 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4770                                    void *buf, size_t buflen, int type)
4771 {
4772         if (strcmp(key, "") != 0)
4773                 return -EINVAL;
4774
4775         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4776 }
4777
4778 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4779                                        size_t list_len, const char *name,
4780                                        size_t name_len, int type)
4781 {
4782         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4783
4784         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4785                 return 0;
4786
4787         if (list && len <= list_len)
4788                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4789         return len;
4790 }
4791
4792 /*
4793  * nfs_fhget will use either the mounted_on_fileid or the fileid
4794  */
4795 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4796 {
4797         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4798                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4799               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4800               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4801                 return;
4802
4803         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4804                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4805         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4806         fattr->nlink = 2;
4807 }
4808
4809 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4810                 struct nfs4_fs_locations *fs_locations, struct page *page)
4811 {
4812         struct nfs_server *server = NFS_SERVER(dir);
4813         u32 bitmask[2] = {
4814                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4815         };
4816         struct nfs4_fs_locations_arg args = {
4817                 .dir_fh = NFS_FH(dir),
4818                 .name = name,
4819                 .page = page,
4820                 .bitmask = bitmask,
4821         };
4822         struct nfs4_fs_locations_res res = {
4823                 .fs_locations = fs_locations,
4824         };
4825         struct rpc_message msg = {
4826                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4827                 .rpc_argp = &args,
4828                 .rpc_resp = &res,
4829         };
4830         int status;
4831
4832         dprintk("%s: start\n", __func__);
4833
4834         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4835          * is not supported */
4836         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4837                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4838         else
4839                 bitmask[0] |= FATTR4_WORD0_FILEID;
4840
4841         nfs_fattr_init(&fs_locations->fattr);
4842         fs_locations->server = server;
4843         fs_locations->nlocations = 0;
4844         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4845         dprintk("%s: returned status = %d\n", __func__, status);
4846         return status;
4847 }
4848
4849 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4850 {
4851         int status;
4852         struct nfs4_secinfo_arg args = {
4853                 .dir_fh = NFS_FH(dir),
4854                 .name   = name,
4855         };
4856         struct nfs4_secinfo_res res = {
4857                 .flavors     = flavors,
4858         };
4859         struct rpc_message msg = {
4860                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4861                 .rpc_argp = &args,
4862                 .rpc_resp = &res,
4863         };
4864
4865         dprintk("NFS call  secinfo %s\n", name->name);
4866         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4867         dprintk("NFS reply  secinfo: %d\n", status);
4868         return status;
4869 }
4870
4871 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4872 {
4873         struct nfs4_exception exception = { };
4874         int err;
4875         do {
4876                 err = nfs4_handle_exception(NFS_SERVER(dir),
4877                                 _nfs4_proc_secinfo(dir, name, flavors),
4878                                 &exception);
4879         } while (exception.retry);
4880         return err;
4881 }
4882
4883 #ifdef CONFIG_NFS_V4_1
4884 /*
4885  * Check the exchange flags returned by the server for invalid flags, having
4886  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4887  * DS flags set.
4888  */
4889 static int nfs4_check_cl_exchange_flags(u32 flags)
4890 {
4891         if (flags & ~EXCHGID4_FLAG_MASK_R)
4892                 goto out_inval;
4893         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4894             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4895                 goto out_inval;
4896         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4897                 goto out_inval;
4898         return NFS_OK;
4899 out_inval:
4900         return -NFS4ERR_INVAL;
4901 }
4902
4903 static bool
4904 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4905 {
4906         if (a->server_scope_sz == b->server_scope_sz &&
4907             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4908                 return true;
4909
4910         return false;
4911 }
4912
4913 /*
4914  * nfs4_proc_exchange_id()
4915  *
4916  * Since the clientid has expired, all compounds using sessions
4917  * associated with the stale clientid will be returning
4918  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4919  * be in some phase of session reset.
4920  */
4921 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4922 {
4923         nfs4_verifier verifier;
4924         struct nfs41_exchange_id_args args = {
4925                 .client = clp,
4926                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4927         };
4928         struct nfs41_exchange_id_res res = {
4929                 .client = clp,
4930         };
4931         int status;
4932         struct rpc_message msg = {
4933                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4934                 .rpc_argp = &args,
4935                 .rpc_resp = &res,
4936                 .rpc_cred = cred,
4937         };
4938         __be32 *p;
4939
4940         dprintk("--> %s\n", __func__);
4941         BUG_ON(clp == NULL);
4942
4943         p = (u32 *)verifier.data;
4944         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4945         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4946         args.verifier = &verifier;
4947
4948         args.id_len = scnprintf(args.id, sizeof(args.id),
4949                                 "%s/%s.%s/%u",
4950                                 clp->cl_ipaddr,
4951                                 init_utsname()->nodename,
4952                                 init_utsname()->domainname,
4953                                 clp->cl_rpcclient->cl_auth->au_flavor);
4954
4955         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4956         if (unlikely(!res.server_scope)) {
4957                 status = -ENOMEM;
4958                 goto out;
4959         }
4960
4961         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
4962         if (unlikely(!res.impl_id)) {
4963                 status = -ENOMEM;
4964                 goto out_server_scope;
4965         }
4966
4967         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4968         if (!status)
4969                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4970
4971         if (!status) {
4972                 /* use the most recent implementation id */
4973                 kfree(clp->impl_id);
4974                 clp->impl_id = res.impl_id;
4975         } else
4976                 kfree(res.impl_id);
4977
4978         if (!status) {
4979                 if (clp->server_scope &&
4980                     !nfs41_same_server_scope(clp->server_scope,
4981                                              res.server_scope)) {
4982                         dprintk("%s: server_scope mismatch detected\n",
4983                                 __func__);
4984                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4985                         kfree(clp->server_scope);
4986                         clp->server_scope = NULL;
4987                 }
4988
4989                 if (!clp->server_scope) {
4990                         clp->server_scope = res.server_scope;
4991                         goto out;
4992                 }
4993         }
4994
4995 out_server_scope:
4996         kfree(res.server_scope);
4997 out:
4998         if (clp->impl_id)
4999                 dprintk("%s: Server Implementation ID: "
5000                         "domain: %s, name: %s, date: %llu,%u\n",
5001                         __func__, clp->impl_id->domain, clp->impl_id->name,
5002                         clp->impl_id->date.seconds,
5003                         clp->impl_id->date.nseconds);
5004         dprintk("<-- %s status= %d\n", __func__, status);
5005         return status;
5006 }
5007
5008 struct nfs4_get_lease_time_data {
5009         struct nfs4_get_lease_time_args *args;
5010         struct nfs4_get_lease_time_res *res;
5011         struct nfs_client *clp;
5012 };
5013
5014 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5015                                         void *calldata)
5016 {
5017         int ret;
5018         struct nfs4_get_lease_time_data *data =
5019                         (struct nfs4_get_lease_time_data *)calldata;
5020
5021         dprintk("--> %s\n", __func__);
5022         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5023         /* just setup sequence, do not trigger session recovery
5024            since we're invoked within one */
5025         ret = nfs41_setup_sequence(data->clp->cl_session,
5026                                    &data->args->la_seq_args,
5027                                    &data->res->lr_seq_res, task);
5028
5029         BUG_ON(ret == -EAGAIN);
5030         rpc_call_start(task);
5031         dprintk("<-- %s\n", __func__);
5032 }
5033
5034 /*
5035  * Called from nfs4_state_manager thread for session setup, so don't recover
5036  * from sequence operation or clientid errors.
5037  */
5038 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5039 {
5040         struct nfs4_get_lease_time_data *data =
5041                         (struct nfs4_get_lease_time_data *)calldata;
5042
5043         dprintk("--> %s\n", __func__);
5044         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5045                 return;
5046         switch (task->tk_status) {
5047         case -NFS4ERR_DELAY:
5048         case -NFS4ERR_GRACE:
5049                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5050                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5051                 task->tk_status = 0;
5052                 /* fall through */
5053         case -NFS4ERR_RETRY_UNCACHED_REP:
5054                 rpc_restart_call_prepare(task);
5055                 return;
5056         }
5057         dprintk("<-- %s\n", __func__);
5058 }
5059
5060 struct rpc_call_ops nfs4_get_lease_time_ops = {
5061         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5062         .rpc_call_done = nfs4_get_lease_time_done,
5063 };
5064
5065 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5066 {
5067         struct rpc_task *task;
5068         struct nfs4_get_lease_time_args args;
5069         struct nfs4_get_lease_time_res res = {
5070                 .lr_fsinfo = fsinfo,
5071         };
5072         struct nfs4_get_lease_time_data data = {
5073                 .args = &args,
5074                 .res = &res,
5075                 .clp = clp,
5076         };
5077         struct rpc_message msg = {
5078                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5079                 .rpc_argp = &args,
5080                 .rpc_resp = &res,
5081         };
5082         struct rpc_task_setup task_setup = {
5083                 .rpc_client = clp->cl_rpcclient,
5084                 .rpc_message = &msg,
5085                 .callback_ops = &nfs4_get_lease_time_ops,
5086                 .callback_data = &data,
5087                 .flags = RPC_TASK_TIMEOUT,
5088         };
5089         int status;
5090
5091         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5092         dprintk("--> %s\n", __func__);
5093         task = rpc_run_task(&task_setup);
5094
5095         if (IS_ERR(task))
5096                 status = PTR_ERR(task);
5097         else {
5098                 status = task->tk_status;
5099                 rpc_put_task(task);
5100         }
5101         dprintk("<-- %s return %d\n", __func__, status);
5102
5103         return status;
5104 }
5105
5106 /*
5107  * Reset a slot table
5108  */
5109 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5110                                  int ivalue)
5111 {
5112         struct nfs4_slot *new = NULL;
5113         int i;
5114         int ret = 0;
5115
5116         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5117                 max_reqs, tbl->max_slots);
5118
5119         /* Does the newly negotiated max_reqs match the existing slot table? */
5120         if (max_reqs != tbl->max_slots) {
5121                 ret = -ENOMEM;
5122                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
5123                               GFP_NOFS);
5124                 if (!new)
5125                         goto out;
5126                 ret = 0;
5127                 kfree(tbl->slots);
5128         }
5129         spin_lock(&tbl->slot_tbl_lock);
5130         if (new) {
5131                 tbl->slots = new;
5132                 tbl->max_slots = max_reqs;
5133         }
5134         for (i = 0; i < tbl->max_slots; ++i)
5135                 tbl->slots[i].seq_nr = ivalue;
5136         spin_unlock(&tbl->slot_tbl_lock);
5137         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5138                 tbl, tbl->slots, tbl->max_slots);
5139 out:
5140         dprintk("<-- %s: return %d\n", __func__, ret);
5141         return ret;
5142 }
5143
5144 /* Destroy the slot table */
5145 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5146 {
5147         if (session->fc_slot_table.slots != NULL) {
5148                 kfree(session->fc_slot_table.slots);
5149                 session->fc_slot_table.slots = NULL;
5150         }
5151         if (session->bc_slot_table.slots != NULL) {
5152                 kfree(session->bc_slot_table.slots);
5153                 session->bc_slot_table.slots = NULL;
5154         }
5155         return;
5156 }
5157
5158 /*
5159  * Initialize slot table
5160  */
5161 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
5162                 int max_slots, int ivalue)
5163 {
5164         struct nfs4_slot *slot;
5165         int ret = -ENOMEM;
5166
5167         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
5168
5169         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
5170
5171         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
5172         if (!slot)
5173                 goto out;
5174         ret = 0;
5175
5176         spin_lock(&tbl->slot_tbl_lock);
5177         tbl->max_slots = max_slots;
5178         tbl->slots = slot;
5179         tbl->highest_used_slotid = NFS4_NO_SLOT;  /* no slot is currently used */
5180         spin_unlock(&tbl->slot_tbl_lock);
5181         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5182                 tbl, tbl->slots, tbl->max_slots);
5183 out:
5184         dprintk("<-- %s: return %d\n", __func__, ret);
5185         return ret;
5186 }
5187
5188 /*
5189  * Initialize or reset the forechannel and backchannel tables
5190  */
5191 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5192 {
5193         struct nfs4_slot_table *tbl;
5194         int status;
5195
5196         dprintk("--> %s\n", __func__);
5197         /* Fore channel */
5198         tbl = &ses->fc_slot_table;
5199         if (tbl->slots == NULL) {
5200                 status = nfs4_init_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5201                 if (status) /* -ENOMEM */
5202                         return status;
5203         } else {
5204                 status = nfs4_reset_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5205                 if (status)
5206                         return status;
5207         }
5208         /* Back channel */
5209         tbl = &ses->bc_slot_table;
5210         if (tbl->slots == NULL) {
5211                 status = nfs4_init_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5212                 if (status)
5213                         /* Fore and back channel share a connection so get
5214                          * both slot tables or neither */
5215                         nfs4_destroy_slot_tables(ses);
5216         } else
5217                 status = nfs4_reset_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5218         return status;
5219 }
5220
5221 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5222 {
5223         struct nfs4_session *session;
5224         struct nfs4_slot_table *tbl;
5225
5226         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5227         if (!session)
5228                 return NULL;
5229
5230         tbl = &session->fc_slot_table;
5231         tbl->highest_used_slotid = NFS4_NO_SLOT;
5232         spin_lock_init(&tbl->slot_tbl_lock);
5233         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5234         init_completion(&tbl->complete);
5235
5236         tbl = &session->bc_slot_table;
5237         tbl->highest_used_slotid = NFS4_NO_SLOT;
5238         spin_lock_init(&tbl->slot_tbl_lock);
5239         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5240         init_completion(&tbl->complete);
5241
5242         session->session_state = 1<<NFS4_SESSION_INITING;
5243
5244         session->clp = clp;
5245         return session;
5246 }
5247
5248 void nfs4_destroy_session(struct nfs4_session *session)
5249 {
5250         struct rpc_xprt *xprt;
5251
5252         nfs4_proc_destroy_session(session);
5253
5254         rcu_read_lock();
5255         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5256         rcu_read_unlock();
5257         dprintk("%s Destroy backchannel for xprt %p\n",
5258                 __func__, xprt);
5259         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5260         nfs4_destroy_slot_tables(session);
5261         kfree(session);
5262 }
5263
5264 /*
5265  * Initialize the values to be used by the client in CREATE_SESSION
5266  * If nfs4_init_session set the fore channel request and response sizes,
5267  * use them.
5268  *
5269  * Set the back channel max_resp_sz_cached to zero to force the client to
5270  * always set csa_cachethis to FALSE because the current implementation
5271  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5272  */
5273 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5274 {
5275         struct nfs4_session *session = args->client->cl_session;
5276         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5277                      mxresp_sz = session->fc_attrs.max_resp_sz;
5278
5279         if (mxrqst_sz == 0)
5280                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5281         if (mxresp_sz == 0)
5282                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5283         /* Fore channel attributes */
5284         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5285         args->fc_attrs.max_resp_sz = mxresp_sz;
5286         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5287         args->fc_attrs.max_reqs = max_session_slots;
5288
5289         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5290                 "max_ops=%u max_reqs=%u\n",
5291                 __func__,
5292                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5293                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5294
5295         /* Back channel attributes */
5296         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5297         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5298         args->bc_attrs.max_resp_sz_cached = 0;
5299         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5300         args->bc_attrs.max_reqs = 1;
5301
5302         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5303                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5304                 __func__,
5305                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5306                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5307                 args->bc_attrs.max_reqs);
5308 }
5309
5310 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5311 {
5312         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5313         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5314
5315         if (rcvd->max_resp_sz > sent->max_resp_sz)
5316                 return -EINVAL;
5317         /*
5318          * Our requested max_ops is the minimum we need; we're not
5319          * prepared to break up compounds into smaller pieces than that.
5320          * So, no point even trying to continue if the server won't
5321          * cooperate:
5322          */
5323         if (rcvd->max_ops < sent->max_ops)
5324                 return -EINVAL;
5325         if (rcvd->max_reqs == 0)
5326                 return -EINVAL;
5327         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5328                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5329         return 0;
5330 }
5331
5332 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5333 {
5334         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5335         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5336
5337         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5338                 return -EINVAL;
5339         if (rcvd->max_resp_sz < sent->max_resp_sz)
5340                 return -EINVAL;
5341         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5342                 return -EINVAL;
5343         /* These would render the backchannel useless: */
5344         if (rcvd->max_ops != sent->max_ops)
5345                 return -EINVAL;
5346         if (rcvd->max_reqs != sent->max_reqs)
5347                 return -EINVAL;
5348         return 0;
5349 }
5350
5351 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5352                                      struct nfs4_session *session)
5353 {
5354         int ret;
5355
5356         ret = nfs4_verify_fore_channel_attrs(args, session);
5357         if (ret)
5358                 return ret;
5359         return nfs4_verify_back_channel_attrs(args, session);
5360 }
5361
5362 static int _nfs4_proc_create_session(struct nfs_client *clp)
5363 {
5364         struct nfs4_session *session = clp->cl_session;
5365         struct nfs41_create_session_args args = {
5366                 .client = clp,
5367                 .cb_program = NFS4_CALLBACK,
5368         };
5369         struct nfs41_create_session_res res = {
5370                 .client = clp,
5371         };
5372         struct rpc_message msg = {
5373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5374                 .rpc_argp = &args,
5375                 .rpc_resp = &res,
5376         };
5377         int status;
5378
5379         nfs4_init_channel_attrs(&args);
5380         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5381
5382         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5383
5384         if (!status)
5385                 /* Verify the session's negotiated channel_attrs values */
5386                 status = nfs4_verify_channel_attrs(&args, session);
5387         if (!status) {
5388                 /* Increment the clientid slot sequence id */
5389                 clp->cl_seqid++;
5390         }
5391
5392         return status;
5393 }
5394
5395 /*
5396  * Issues a CREATE_SESSION operation to the server.
5397  * It is the responsibility of the caller to verify the session is
5398  * expired before calling this routine.
5399  */
5400 int nfs4_proc_create_session(struct nfs_client *clp)
5401 {
5402         int status;
5403         unsigned *ptr;
5404         struct nfs4_session *session = clp->cl_session;
5405
5406         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5407
5408         status = _nfs4_proc_create_session(clp);
5409         if (status)
5410                 goto out;
5411
5412         /* Init or reset the session slot tables */
5413         status = nfs4_setup_session_slot_tables(session);
5414         dprintk("slot table setup returned %d\n", status);
5415         if (status)
5416                 goto out;
5417
5418         ptr = (unsigned *)&session->sess_id.data[0];
5419         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5420                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5421 out:
5422         dprintk("<-- %s\n", __func__);
5423         return status;
5424 }
5425
5426 /*
5427  * Issue the over-the-wire RPC DESTROY_SESSION.
5428  * The caller must serialize access to this routine.
5429  */
5430 int nfs4_proc_destroy_session(struct nfs4_session *session)
5431 {
5432         int status = 0;
5433         struct rpc_message msg;
5434
5435         dprintk("--> nfs4_proc_destroy_session\n");
5436
5437         /* session is still being setup */
5438         if (session->clp->cl_cons_state != NFS_CS_READY)
5439                 return status;
5440
5441         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5442         msg.rpc_argp = session;
5443         msg.rpc_resp = NULL;
5444         msg.rpc_cred = NULL;
5445         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5446
5447         if (status)
5448                 printk(KERN_WARNING
5449                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5450                         "Session has been destroyed regardless...\n", status);
5451
5452         dprintk("<-- nfs4_proc_destroy_session\n");
5453         return status;
5454 }
5455
5456 int nfs4_init_session(struct nfs_server *server)
5457 {
5458         struct nfs_client *clp = server->nfs_client;
5459         struct nfs4_session *session;
5460         unsigned int rsize, wsize;
5461         int ret;
5462
5463         if (!nfs4_has_session(clp))
5464                 return 0;
5465
5466         session = clp->cl_session;
5467         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5468                 return 0;
5469
5470         rsize = server->rsize;
5471         if (rsize == 0)
5472                 rsize = NFS_MAX_FILE_IO_SIZE;
5473         wsize = server->wsize;
5474         if (wsize == 0)
5475                 wsize = NFS_MAX_FILE_IO_SIZE;
5476
5477         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5478         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5479
5480         ret = nfs4_recover_expired_lease(server);
5481         if (!ret)
5482                 ret = nfs4_check_client_ready(clp);
5483         return ret;
5484 }
5485
5486 int nfs4_init_ds_session(struct nfs_client *clp)
5487 {
5488         struct nfs4_session *session = clp->cl_session;
5489         int ret;
5490
5491         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5492                 return 0;
5493
5494         ret = nfs4_client_recover_expired_lease(clp);
5495         if (!ret)
5496                 /* Test for the DS role */
5497                 if (!is_ds_client(clp))
5498                         ret = -ENODEV;
5499         if (!ret)
5500                 ret = nfs4_check_client_ready(clp);
5501         return ret;
5502
5503 }
5504 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5505
5506
5507 /*
5508  * Renew the cl_session lease.
5509  */
5510 struct nfs4_sequence_data {
5511         struct nfs_client *clp;
5512         struct nfs4_sequence_args args;
5513         struct nfs4_sequence_res res;
5514 };
5515
5516 static void nfs41_sequence_release(void *data)
5517 {
5518         struct nfs4_sequence_data *calldata = data;
5519         struct nfs_client *clp = calldata->clp;
5520
5521         if (atomic_read(&clp->cl_count) > 1)
5522                 nfs4_schedule_state_renewal(clp);
5523         nfs_put_client(clp);
5524         kfree(calldata);
5525 }
5526
5527 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5528 {
5529         switch(task->tk_status) {
5530         case -NFS4ERR_DELAY:
5531                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5532                 return -EAGAIN;
5533         default:
5534                 nfs4_schedule_lease_recovery(clp);
5535         }
5536         return 0;
5537 }
5538
5539 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5540 {
5541         struct nfs4_sequence_data *calldata = data;
5542         struct nfs_client *clp = calldata->clp;
5543
5544         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5545                 return;
5546
5547         if (task->tk_status < 0) {
5548                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5549                 if (atomic_read(&clp->cl_count) == 1)
5550                         goto out;
5551
5552                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5553                         rpc_restart_call_prepare(task);
5554                         return;
5555                 }
5556         }
5557         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5558 out:
5559         dprintk("<-- %s\n", __func__);
5560 }
5561
5562 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5563 {
5564         struct nfs4_sequence_data *calldata = data;
5565         struct nfs_client *clp = calldata->clp;
5566         struct nfs4_sequence_args *args;
5567         struct nfs4_sequence_res *res;
5568
5569         args = task->tk_msg.rpc_argp;
5570         res = task->tk_msg.rpc_resp;
5571
5572         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5573                 return;
5574         rpc_call_start(task);
5575 }
5576
5577 static const struct rpc_call_ops nfs41_sequence_ops = {
5578         .rpc_call_done = nfs41_sequence_call_done,
5579         .rpc_call_prepare = nfs41_sequence_prepare,
5580         .rpc_release = nfs41_sequence_release,
5581 };
5582
5583 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5584 {
5585         struct nfs4_sequence_data *calldata;
5586         struct rpc_message msg = {
5587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5588                 .rpc_cred = cred,
5589         };
5590         struct rpc_task_setup task_setup_data = {
5591                 .rpc_client = clp->cl_rpcclient,
5592                 .rpc_message = &msg,
5593                 .callback_ops = &nfs41_sequence_ops,
5594                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5595         };
5596
5597         if (!atomic_inc_not_zero(&clp->cl_count))
5598                 return ERR_PTR(-EIO);
5599         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5600         if (calldata == NULL) {
5601                 nfs_put_client(clp);
5602                 return ERR_PTR(-ENOMEM);
5603         }
5604         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5605         msg.rpc_argp = &calldata->args;
5606         msg.rpc_resp = &calldata->res;
5607         calldata->clp = clp;
5608         task_setup_data.callback_data = calldata;
5609
5610         return rpc_run_task(&task_setup_data);
5611 }
5612
5613 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5614 {
5615         struct rpc_task *task;
5616         int ret = 0;
5617
5618         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5619                 return 0;
5620         task = _nfs41_proc_sequence(clp, cred);
5621         if (IS_ERR(task))
5622                 ret = PTR_ERR(task);
5623         else
5624                 rpc_put_task_async(task);
5625         dprintk("<-- %s status=%d\n", __func__, ret);
5626         return ret;
5627 }
5628
5629 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5630 {
5631         struct rpc_task *task;
5632         int ret;
5633
5634         task = _nfs41_proc_sequence(clp, cred);
5635         if (IS_ERR(task)) {
5636                 ret = PTR_ERR(task);
5637                 goto out;
5638         }
5639         ret = rpc_wait_for_completion_task(task);
5640         if (!ret) {
5641                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5642
5643                 if (task->tk_status == 0)
5644                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5645                 ret = task->tk_status;
5646         }
5647         rpc_put_task(task);
5648 out:
5649         dprintk("<-- %s status=%d\n", __func__, ret);
5650         return ret;
5651 }
5652
5653 struct nfs4_reclaim_complete_data {
5654         struct nfs_client *clp;
5655         struct nfs41_reclaim_complete_args arg;
5656         struct nfs41_reclaim_complete_res res;
5657 };
5658
5659 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5660 {
5661         struct nfs4_reclaim_complete_data *calldata = data;
5662
5663         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5664         if (nfs41_setup_sequence(calldata->clp->cl_session,
5665                                 &calldata->arg.seq_args,
5666                                 &calldata->res.seq_res, task))
5667                 return;
5668
5669         rpc_call_start(task);
5670 }
5671
5672 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5673 {
5674         switch(task->tk_status) {
5675         case 0:
5676         case -NFS4ERR_COMPLETE_ALREADY:
5677         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5678                 break;
5679         case -NFS4ERR_DELAY:
5680                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5681                 /* fall through */
5682         case -NFS4ERR_RETRY_UNCACHED_REP:
5683                 return -EAGAIN;
5684         default:
5685                 nfs4_schedule_lease_recovery(clp);
5686         }
5687         return 0;
5688 }
5689
5690 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5691 {
5692         struct nfs4_reclaim_complete_data *calldata = data;
5693         struct nfs_client *clp = calldata->clp;
5694         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5695
5696         dprintk("--> %s\n", __func__);
5697         if (!nfs41_sequence_done(task, res))
5698                 return;
5699
5700         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5701                 rpc_restart_call_prepare(task);
5702                 return;
5703         }
5704         dprintk("<-- %s\n", __func__);
5705 }
5706
5707 static void nfs4_free_reclaim_complete_data(void *data)
5708 {
5709         struct nfs4_reclaim_complete_data *calldata = data;
5710
5711         kfree(calldata);
5712 }
5713
5714 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5715         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5716         .rpc_call_done = nfs4_reclaim_complete_done,
5717         .rpc_release = nfs4_free_reclaim_complete_data,
5718 };
5719
5720 /*
5721  * Issue a global reclaim complete.
5722  */
5723 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5724 {
5725         struct nfs4_reclaim_complete_data *calldata;
5726         struct rpc_task *task;
5727         struct rpc_message msg = {
5728                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5729         };
5730         struct rpc_task_setup task_setup_data = {
5731                 .rpc_client = clp->cl_rpcclient,
5732                 .rpc_message = &msg,
5733                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5734                 .flags = RPC_TASK_ASYNC,
5735         };
5736         int status = -ENOMEM;
5737
5738         dprintk("--> %s\n", __func__);
5739         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5740         if (calldata == NULL)
5741                 goto out;
5742         calldata->clp = clp;
5743         calldata->arg.one_fs = 0;
5744
5745         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5746         msg.rpc_argp = &calldata->arg;
5747         msg.rpc_resp = &calldata->res;
5748         task_setup_data.callback_data = calldata;
5749         task = rpc_run_task(&task_setup_data);
5750         if (IS_ERR(task)) {
5751                 status = PTR_ERR(task);
5752                 goto out;
5753         }
5754         status = nfs4_wait_for_completion_rpc_task(task);
5755         if (status == 0)
5756                 status = task->tk_status;
5757         rpc_put_task(task);
5758         return 0;
5759 out:
5760         dprintk("<-- %s status=%d\n", __func__, status);
5761         return status;
5762 }
5763
5764 static void
5765 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5766 {
5767         struct nfs4_layoutget *lgp = calldata;
5768         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5769
5770         dprintk("--> %s\n", __func__);
5771         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5772          * right now covering the LAYOUTGET we are about to send.
5773          * However, that is not so catastrophic, and there seems
5774          * to be no way to prevent it completely.
5775          */
5776         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5777                                 &lgp->res.seq_res, task))
5778                 return;
5779         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5780                                           NFS_I(lgp->args.inode)->layout,
5781                                           lgp->args.ctx->state)) {
5782                 rpc_exit(task, NFS4_OK);
5783                 return;
5784         }
5785         rpc_call_start(task);
5786 }
5787
5788 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5789 {
5790         struct nfs4_layoutget *lgp = calldata;
5791         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5792
5793         dprintk("--> %s\n", __func__);
5794
5795         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5796                 return;
5797
5798         switch (task->tk_status) {
5799         case 0:
5800                 break;
5801         case -NFS4ERR_LAYOUTTRYLATER:
5802         case -NFS4ERR_RECALLCONFLICT:
5803                 task->tk_status = -NFS4ERR_DELAY;
5804                 /* Fall through */
5805         default:
5806                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5807                         rpc_restart_call_prepare(task);
5808                         return;
5809                 }
5810         }
5811         dprintk("<-- %s\n", __func__);
5812 }
5813
5814 static void nfs4_layoutget_release(void *calldata)
5815 {
5816         struct nfs4_layoutget *lgp = calldata;
5817
5818         dprintk("--> %s\n", __func__);
5819         put_nfs_open_context(lgp->args.ctx);
5820         kfree(calldata);
5821         dprintk("<-- %s\n", __func__);
5822 }
5823
5824 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5825         .rpc_call_prepare = nfs4_layoutget_prepare,
5826         .rpc_call_done = nfs4_layoutget_done,
5827         .rpc_release = nfs4_layoutget_release,
5828 };
5829
5830 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5831 {
5832         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5833         struct rpc_task *task;
5834         struct rpc_message msg = {
5835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5836                 .rpc_argp = &lgp->args,
5837                 .rpc_resp = &lgp->res,
5838         };
5839         struct rpc_task_setup task_setup_data = {
5840                 .rpc_client = server->client,
5841                 .rpc_message = &msg,
5842                 .callback_ops = &nfs4_layoutget_call_ops,
5843                 .callback_data = lgp,
5844                 .flags = RPC_TASK_ASYNC,
5845         };
5846         int status = 0;
5847
5848         dprintk("--> %s\n", __func__);
5849
5850         lgp->res.layoutp = &lgp->args.layout;
5851         lgp->res.seq_res.sr_slot = NULL;
5852         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5853         task = rpc_run_task(&task_setup_data);
5854         if (IS_ERR(task))
5855                 return PTR_ERR(task);
5856         status = nfs4_wait_for_completion_rpc_task(task);
5857         if (status == 0)
5858                 status = task->tk_status;
5859         if (status == 0)
5860                 status = pnfs_layout_process(lgp);
5861         rpc_put_task(task);
5862         dprintk("<-- %s status=%d\n", __func__, status);
5863         return status;
5864 }
5865
5866 static void
5867 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5868 {
5869         struct nfs4_layoutreturn *lrp = calldata;
5870
5871         dprintk("--> %s\n", __func__);
5872         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5873                                 &lrp->res.seq_res, task))
5874                 return;
5875         rpc_call_start(task);
5876 }
5877
5878 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5879 {
5880         struct nfs4_layoutreturn *lrp = calldata;
5881         struct nfs_server *server;
5882         struct pnfs_layout_hdr *lo = lrp->args.layout;
5883
5884         dprintk("--> %s\n", __func__);
5885
5886         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5887                 return;
5888
5889         server = NFS_SERVER(lrp->args.inode);
5890         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5891                 rpc_restart_call_prepare(task);
5892                 return;
5893         }
5894         spin_lock(&lo->plh_inode->i_lock);
5895         if (task->tk_status == 0) {
5896                 if (lrp->res.lrs_present) {
5897                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5898                 } else
5899                         BUG_ON(!list_empty(&lo->plh_segs));
5900         }
5901         lo->plh_block_lgets--;
5902         spin_unlock(&lo->plh_inode->i_lock);
5903         dprintk("<-- %s\n", __func__);
5904 }
5905
5906 static void nfs4_layoutreturn_release(void *calldata)
5907 {
5908         struct nfs4_layoutreturn *lrp = calldata;
5909
5910         dprintk("--> %s\n", __func__);
5911         put_layout_hdr(lrp->args.layout);
5912         kfree(calldata);
5913         dprintk("<-- %s\n", __func__);
5914 }
5915
5916 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5917         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5918         .rpc_call_done = nfs4_layoutreturn_done,
5919         .rpc_release = nfs4_layoutreturn_release,
5920 };
5921
5922 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5923 {
5924         struct rpc_task *task;
5925         struct rpc_message msg = {
5926                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5927                 .rpc_argp = &lrp->args,
5928                 .rpc_resp = &lrp->res,
5929         };
5930         struct rpc_task_setup task_setup_data = {
5931                 .rpc_client = lrp->clp->cl_rpcclient,
5932                 .rpc_message = &msg,
5933                 .callback_ops = &nfs4_layoutreturn_call_ops,
5934                 .callback_data = lrp,
5935         };
5936         int status;
5937
5938         dprintk("--> %s\n", __func__);
5939         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5940         task = rpc_run_task(&task_setup_data);
5941         if (IS_ERR(task))
5942                 return PTR_ERR(task);
5943         status = task->tk_status;
5944         dprintk("<-- %s status=%d\n", __func__, status);
5945         rpc_put_task(task);
5946         return status;
5947 }
5948
5949 /*
5950  * Retrieve the list of Data Server devices from the MDS.
5951  */
5952 static int _nfs4_getdevicelist(struct nfs_server *server,
5953                                     const struct nfs_fh *fh,
5954                                     struct pnfs_devicelist *devlist)
5955 {
5956         struct nfs4_getdevicelist_args args = {
5957                 .fh = fh,
5958                 .layoutclass = server->pnfs_curr_ld->id,
5959         };
5960         struct nfs4_getdevicelist_res res = {
5961                 .devlist = devlist,
5962         };
5963         struct rpc_message msg = {
5964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5965                 .rpc_argp = &args,
5966                 .rpc_resp = &res,
5967         };
5968         int status;
5969
5970         dprintk("--> %s\n", __func__);
5971         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5972                                 &res.seq_res, 0);
5973         dprintk("<-- %s status=%d\n", __func__, status);
5974         return status;
5975 }
5976
5977 int nfs4_proc_getdevicelist(struct nfs_server *server,
5978                             const struct nfs_fh *fh,
5979                             struct pnfs_devicelist *devlist)
5980 {
5981         struct nfs4_exception exception = { };
5982         int err;
5983
5984         do {
5985                 err = nfs4_handle_exception(server,
5986                                 _nfs4_getdevicelist(server, fh, devlist),
5987                                 &exception);
5988         } while (exception.retry);
5989
5990         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5991                 err, devlist->num_devs);
5992
5993         return err;
5994 }
5995 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5996
5997 static int
5998 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5999 {
6000         struct nfs4_getdeviceinfo_args args = {
6001                 .pdev = pdev,
6002         };
6003         struct nfs4_getdeviceinfo_res res = {
6004                 .pdev = pdev,
6005         };
6006         struct rpc_message msg = {
6007                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
6008                 .rpc_argp = &args,
6009                 .rpc_resp = &res,
6010         };
6011         int status;
6012
6013         dprintk("--> %s\n", __func__);
6014         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6015         dprintk("<-- %s status=%d\n", __func__, status);
6016
6017         return status;
6018 }
6019
6020 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
6021 {
6022         struct nfs4_exception exception = { };
6023         int err;
6024
6025         do {
6026                 err = nfs4_handle_exception(server,
6027                                         _nfs4_proc_getdeviceinfo(server, pdev),
6028                                         &exception);
6029         } while (exception.retry);
6030         return err;
6031 }
6032 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6033
6034 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6035 {
6036         struct nfs4_layoutcommit_data *data = calldata;
6037         struct nfs_server *server = NFS_SERVER(data->args.inode);
6038
6039         if (nfs4_setup_sequence(server, &data->args.seq_args,
6040                                 &data->res.seq_res, task))
6041                 return;
6042         rpc_call_start(task);
6043 }
6044
6045 static void
6046 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6047 {
6048         struct nfs4_layoutcommit_data *data = calldata;
6049         struct nfs_server *server = NFS_SERVER(data->args.inode);
6050
6051         if (!nfs4_sequence_done(task, &data->res.seq_res))
6052                 return;
6053
6054         switch (task->tk_status) { /* Just ignore these failures */
6055         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6056         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6057         case NFS4ERR_BADLAYOUT:     /* no layout */
6058         case NFS4ERR_GRACE:         /* loca_recalim always false */
6059                 task->tk_status = 0;
6060         }
6061
6062         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6063                 rpc_restart_call_prepare(task);
6064                 return;
6065         }
6066
6067         if (task->tk_status == 0)
6068                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6069                                                    data->res.fattr);
6070 }
6071
6072 static void nfs4_layoutcommit_release(void *calldata)
6073 {
6074         struct nfs4_layoutcommit_data *data = calldata;
6075         struct pnfs_layout_segment *lseg, *tmp;
6076         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6077
6078         pnfs_cleanup_layoutcommit(data);
6079         /* Matched by references in pnfs_set_layoutcommit */
6080         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6081                 list_del_init(&lseg->pls_lc_list);
6082                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6083                                        &lseg->pls_flags))
6084                         put_lseg(lseg);
6085         }
6086
6087         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6088         smp_mb__after_clear_bit();
6089         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6090
6091         put_rpccred(data->cred);
6092         kfree(data);
6093 }
6094
6095 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6096         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6097         .rpc_call_done = nfs4_layoutcommit_done,
6098         .rpc_release = nfs4_layoutcommit_release,
6099 };
6100
6101 int
6102 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6103 {
6104         struct rpc_message msg = {
6105                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6106                 .rpc_argp = &data->args,
6107                 .rpc_resp = &data->res,
6108                 .rpc_cred = data->cred,
6109         };
6110         struct rpc_task_setup task_setup_data = {
6111                 .task = &data->task,
6112                 .rpc_client = NFS_CLIENT(data->args.inode),
6113                 .rpc_message = &msg,
6114                 .callback_ops = &nfs4_layoutcommit_ops,
6115                 .callback_data = data,
6116                 .flags = RPC_TASK_ASYNC,
6117         };
6118         struct rpc_task *task;
6119         int status = 0;
6120
6121         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6122                 "lbw: %llu inode %lu\n",
6123                 data->task.tk_pid, sync,
6124                 data->args.lastbytewritten,
6125                 data->args.inode->i_ino);
6126
6127         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6128         task = rpc_run_task(&task_setup_data);
6129         if (IS_ERR(task))
6130                 return PTR_ERR(task);
6131         if (sync == false)
6132                 goto out;
6133         status = nfs4_wait_for_completion_rpc_task(task);
6134         if (status != 0)
6135                 goto out;
6136         status = task->tk_status;
6137 out:
6138         dprintk("%s: status %d\n", __func__, status);
6139         rpc_put_task(task);
6140         return status;
6141 }
6142
6143 static int
6144 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6145                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6146 {
6147         struct nfs41_secinfo_no_name_args args = {
6148                 .style = SECINFO_STYLE_CURRENT_FH,
6149         };
6150         struct nfs4_secinfo_res res = {
6151                 .flavors = flavors,
6152         };
6153         struct rpc_message msg = {
6154                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6155                 .rpc_argp = &args,
6156                 .rpc_resp = &res,
6157         };
6158         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6159 }
6160
6161 static int
6162 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6163                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6164 {
6165         struct nfs4_exception exception = { };
6166         int err;
6167         do {
6168                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6169                 switch (err) {
6170                 case 0:
6171                 case -NFS4ERR_WRONGSEC:
6172                 case -NFS4ERR_NOTSUPP:
6173                         break;
6174                 default:
6175                         err = nfs4_handle_exception(server, err, &exception);
6176                 }
6177         } while (exception.retry);
6178         return err;
6179 }
6180
6181 static int
6182 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6183                     struct nfs_fsinfo *info)
6184 {
6185         int err;
6186         struct page *page;
6187         rpc_authflavor_t flavor;
6188         struct nfs4_secinfo_flavors *flavors;
6189
6190         page = alloc_page(GFP_KERNEL);
6191         if (!page) {
6192                 err = -ENOMEM;
6193                 goto out;
6194         }
6195
6196         flavors = page_address(page);
6197         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6198
6199         /*
6200          * Fall back on "guess and check" method if
6201          * the server doesn't support SECINFO_NO_NAME
6202          */
6203         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6204                 err = nfs4_find_root_sec(server, fhandle, info);
6205                 goto out_freepage;
6206         }
6207         if (err)
6208                 goto out_freepage;
6209
6210         flavor = nfs_find_best_sec(flavors);
6211         if (err == 0)
6212                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6213
6214 out_freepage:
6215         put_page(page);
6216         if (err == -EACCES)
6217                 return -EPERM;
6218 out:
6219         return err;
6220 }
6221
6222 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6223 {
6224         int status;
6225         struct nfs41_test_stateid_args args = {
6226                 .stateid = stateid,
6227         };
6228         struct nfs41_test_stateid_res res;
6229         struct rpc_message msg = {
6230                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6231                 .rpc_argp = &args,
6232                 .rpc_resp = &res,
6233         };
6234
6235         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6236         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6237
6238         if (status == NFS_OK)
6239                 return res.status;
6240         return status;
6241 }
6242
6243 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6244 {
6245         struct nfs4_exception exception = { };
6246         int err;
6247         do {
6248                 err = nfs4_handle_exception(server,
6249                                 _nfs41_test_stateid(server, stateid),
6250                                 &exception);
6251         } while (exception.retry);
6252         return err;
6253 }
6254
6255 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6256 {
6257         struct nfs41_free_stateid_args args = {
6258                 .stateid = stateid,
6259         };
6260         struct nfs41_free_stateid_res res;
6261         struct rpc_message msg = {
6262                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6263                 .rpc_argp = &args,
6264                 .rpc_resp = &res,
6265         };
6266
6267         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6268         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6269 }
6270
6271 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6272 {
6273         struct nfs4_exception exception = { };
6274         int err;
6275         do {
6276                 err = nfs4_handle_exception(server,
6277                                 _nfs4_free_stateid(server, stateid),
6278                                 &exception);
6279         } while (exception.retry);
6280         return err;
6281 }
6282 #endif /* CONFIG_NFS_V4_1 */
6283
6284 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6285         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6286         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6287         .recover_open   = nfs4_open_reclaim,
6288         .recover_lock   = nfs4_lock_reclaim,
6289         .establish_clid = nfs4_init_clientid,
6290         .get_clid_cred  = nfs4_get_setclientid_cred,
6291 };
6292
6293 #if defined(CONFIG_NFS_V4_1)
6294 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6295         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6296         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6297         .recover_open   = nfs4_open_reclaim,
6298         .recover_lock   = nfs4_lock_reclaim,
6299         .establish_clid = nfs41_init_clientid,
6300         .get_clid_cred  = nfs4_get_exchange_id_cred,
6301         .reclaim_complete = nfs41_proc_reclaim_complete,
6302 };
6303 #endif /* CONFIG_NFS_V4_1 */
6304
6305 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6306         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6307         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6308         .recover_open   = nfs4_open_expired,
6309         .recover_lock   = nfs4_lock_expired,
6310         .establish_clid = nfs4_init_clientid,
6311         .get_clid_cred  = nfs4_get_setclientid_cred,
6312 };
6313
6314 #if defined(CONFIG_NFS_V4_1)
6315 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6316         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6317         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6318         .recover_open   = nfs41_open_expired,
6319         .recover_lock   = nfs41_lock_expired,
6320         .establish_clid = nfs41_init_clientid,
6321         .get_clid_cred  = nfs4_get_exchange_id_cred,
6322 };
6323 #endif /* CONFIG_NFS_V4_1 */
6324
6325 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6326         .sched_state_renewal = nfs4_proc_async_renew,
6327         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6328         .renew_lease = nfs4_proc_renew,
6329 };
6330
6331 #if defined(CONFIG_NFS_V4_1)
6332 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6333         .sched_state_renewal = nfs41_proc_async_sequence,
6334         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6335         .renew_lease = nfs4_proc_sequence,
6336 };
6337 #endif
6338
6339 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6340         .minor_version = 0,
6341         .call_sync = _nfs4_call_sync,
6342         .validate_stateid = nfs4_validate_delegation_stateid,
6343         .find_root_sec = nfs4_find_root_sec,
6344         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6345         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6346         .state_renewal_ops = &nfs40_state_renewal_ops,
6347 };
6348
6349 #if defined(CONFIG_NFS_V4_1)
6350 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6351         .minor_version = 1,
6352         .call_sync = _nfs4_call_sync_session,
6353         .validate_stateid = nfs41_validate_delegation_stateid,
6354         .find_root_sec = nfs41_find_root_sec,
6355         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6356         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6357         .state_renewal_ops = &nfs41_state_renewal_ops,
6358 };
6359 #endif
6360
6361 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6362         [0] = &nfs_v4_0_minor_ops,
6363 #if defined(CONFIG_NFS_V4_1)
6364         [1] = &nfs_v4_1_minor_ops,
6365 #endif
6366 };
6367
6368 static const struct inode_operations nfs4_file_inode_operations = {
6369         .permission     = nfs_permission,
6370         .getattr        = nfs_getattr,
6371         .setattr        = nfs_setattr,
6372         .getxattr       = generic_getxattr,
6373         .setxattr       = generic_setxattr,
6374         .listxattr      = generic_listxattr,
6375         .removexattr    = generic_removexattr,
6376 };
6377
6378 const struct nfs_rpc_ops nfs_v4_clientops = {
6379         .version        = 4,                    /* protocol version */
6380         .dentry_ops     = &nfs4_dentry_operations,
6381         .dir_inode_ops  = &nfs4_dir_inode_operations,
6382         .file_inode_ops = &nfs4_file_inode_operations,
6383         .file_ops       = &nfs4_file_operations,
6384         .getroot        = nfs4_proc_get_root,
6385         .getattr        = nfs4_proc_getattr,
6386         .setattr        = nfs4_proc_setattr,
6387         .lookup         = nfs4_proc_lookup,
6388         .access         = nfs4_proc_access,
6389         .readlink       = nfs4_proc_readlink,
6390         .create         = nfs4_proc_create,
6391         .remove         = nfs4_proc_remove,
6392         .unlink_setup   = nfs4_proc_unlink_setup,
6393         .unlink_done    = nfs4_proc_unlink_done,
6394         .rename         = nfs4_proc_rename,
6395         .rename_setup   = nfs4_proc_rename_setup,
6396         .rename_done    = nfs4_proc_rename_done,
6397         .link           = nfs4_proc_link,
6398         .symlink        = nfs4_proc_symlink,
6399         .mkdir          = nfs4_proc_mkdir,
6400         .rmdir          = nfs4_proc_remove,
6401         .readdir        = nfs4_proc_readdir,
6402         .mknod          = nfs4_proc_mknod,
6403         .statfs         = nfs4_proc_statfs,
6404         .fsinfo         = nfs4_proc_fsinfo,
6405         .pathconf       = nfs4_proc_pathconf,
6406         .set_capabilities = nfs4_server_capabilities,
6407         .decode_dirent  = nfs4_decode_dirent,
6408         .read_setup     = nfs4_proc_read_setup,
6409         .read_done      = nfs4_read_done,
6410         .write_setup    = nfs4_proc_write_setup,
6411         .write_done     = nfs4_write_done,
6412         .commit_setup   = nfs4_proc_commit_setup,
6413         .commit_done    = nfs4_commit_done,
6414         .lock           = nfs4_proc_lock,
6415         .clear_acl_cache = nfs4_zap_acl_attr,
6416         .close_context  = nfs4_close_context,
6417         .open_context   = nfs4_atomic_open,
6418         .init_client    = nfs4_init_client,
6419         .secinfo        = nfs4_proc_secinfo,
6420 };
6421
6422 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6423         .prefix = XATTR_NAME_NFSV4_ACL,
6424         .list   = nfs4_xattr_list_nfs4_acl,
6425         .get    = nfs4_xattr_get_nfs4_acl,
6426         .set    = nfs4_xattr_set_nfs4_acl,
6427 };
6428
6429 const struct xattr_handler *nfs4_xattr_handlers[] = {
6430         &nfs4_xattr_nfs4_acl_handler,
6431         NULL
6432 };
6433
6434 module_param(max_session_slots, ushort, 0644);
6435 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6436                 "requests the client will negotiate");
6437
6438 /*
6439  * Local variables:
6440  *  c-basic-offset: 8
6441  * End:
6442  */