Pileus Git - ~andy/linux/blob - fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/sunrpc/gss_api.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/nfs_idmap.h>
  56 #include <linux/sunrpc/bc_xprt.h>
  57 #include <linux/xattr.h>
  58 #include <linux/utsname.h>
  59 #include <linux/freezer.h>
  60
  61 #include "nfs4_fs.h"
  62 #include "delegation.h"
  63 #include "internal.h"
  64 #include "iostat.h"
  65 #include "callback.h"
  66 #include "pnfs.h"
  67
  68 #define NFSDBG_FACILITY         NFSDBG_PROC
  69
  70 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  71 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  72
  73 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  74
  75 static unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  82 static 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                 server->fh_expire_type = res.fh_expire_type;
2224         }
2225
2226         return status;
2227 }
2228
2229 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2230 {
2231         struct nfs4_exception exception = { };
2232         int err;
2233         do {
2234                 err = nfs4_handle_exception(server,
2235                                 _nfs4_server_capabilities(server, fhandle),
2236                                 &exception);
2237         } while (exception.retry);
2238         return err;
2239 }
2240
2241 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2242                 struct nfs_fsinfo *info)
2243 {
2244         struct nfs4_lookup_root_arg args = {
2245                 .bitmask = nfs4_fattr_bitmap,
2246         };
2247         struct nfs4_lookup_res res = {
2248                 .server = server,
2249                 .fattr = info->fattr,
2250                 .fh = fhandle,
2251         };
2252         struct rpc_message msg = {
2253                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2254                 .rpc_argp = &args,
2255                 .rpc_resp = &res,
2256         };
2257
2258         nfs_fattr_init(info->fattr);
2259         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2260 }
2261
2262 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2263                 struct nfs_fsinfo *info)
2264 {
2265         struct nfs4_exception exception = { };
2266         int err;
2267         do {
2268                 err = _nfs4_lookup_root(server, fhandle, info);
2269                 switch (err) {
2270                 case 0:
2271                 case -NFS4ERR_WRONGSEC:
2272                         break;
2273                 default:
2274                         err = nfs4_handle_exception(server, err, &exception);
2275                 }
2276         } while (exception.retry);
2277         return err;
2278 }
2279
2280 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2281                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2282 {
2283         struct rpc_auth *auth;
2284         int ret;
2285
2286         auth = rpcauth_create(flavor, server->client);
2287         if (!auth) {
2288                 ret = -EIO;
2289                 goto out;
2290         }
2291         ret = nfs4_lookup_root(server, fhandle, info);
2292 out:
2293         return ret;
2294 }
2295
2296 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2297                               struct nfs_fsinfo *info)
2298 {
2299         int i, len, status = 0;
2300         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2301
2302         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2303         flav_array[len] = RPC_AUTH_NULL;
2304         len += 1;
2305
2306         for (i = 0; i < len; i++) {
2307                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2308                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2309                         continue;
2310                 break;
2311         }
2312         /*
2313          * -EACCESS could mean that the user doesn't have correct permissions
2314          * to access the mount.  It could also mean that we tried to mount
2315          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2316          * existing mount programs don't handle -EACCES very well so it should
2317          * be mapped to -EPERM instead.
2318          */
2319         if (status == -EACCES)
2320                 status = -EPERM;
2321         return status;
2322 }
2323
2324 /*
2325  * get the file handle for the "/" directory on the server
2326  */
2327 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2328                               struct nfs_fsinfo *info)
2329 {
2330         int minor_version = server->nfs_client->cl_minorversion;
2331         int status = nfs4_lookup_root(server, fhandle, info);
2332         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2333                 /*
2334                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2335                  * by nfs4_map_errors() as this function exits.
2336                  */
2337                 status = nfs_v4_minor_ops[minor_version]->find_root_sec(server, fhandle, info);
2338         if (status == 0)
2339                 status = nfs4_server_capabilities(server, fhandle);
2340         if (status == 0)
2341                 status = nfs4_do_fsinfo(server, fhandle, info);
2342         return nfs4_map_errors(status);
2343 }
2344
2345 /*
2346  * Get locations and (maybe) other attributes of a referral.
2347  * Note that we'll actually follow the referral later when
2348  * we detect fsid mismatch in inode revalidation
2349  */
2350 static int nfs4_get_referral(struct inode *dir, const struct qstr *name,
2351                              struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2352 {
2353         int status = -ENOMEM;
2354         struct page *page = NULL;
2355         struct nfs4_fs_locations *locations = NULL;
2356
2357         page = alloc_page(GFP_KERNEL);
2358         if (page == NULL)
2359                 goto out;
2360         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2361         if (locations == NULL)
2362                 goto out;
2363
2364         status = nfs4_proc_fs_locations(dir, name, locations, page);
2365         if (status != 0)
2366                 goto out;
2367         /* Make sure server returned a different fsid for the referral */
2368         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2369                 dprintk("%s: server did not return a different fsid for"
2370                         " a referral at %s\n", __func__, name->name);
2371                 status = -EIO;
2372                 goto out;
2373         }
2374         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
2375         nfs_fixup_referral_attributes(&locations->fattr);
2376
2377         /* replace the lookup nfs_fattr with the locations nfs_fattr */
2378         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2379         memset(fhandle, 0, sizeof(struct nfs_fh));
2380 out:
2381         if (page)
2382                 __free_page(page);
2383         kfree(locations);
2384         return status;
2385 }
2386
2387 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2388 {
2389         struct nfs4_getattr_arg args = {
2390                 .fh = fhandle,
2391                 .bitmask = server->attr_bitmask,
2392         };
2393         struct nfs4_getattr_res res = {
2394                 .fattr = fattr,
2395                 .server = server,
2396         };
2397         struct rpc_message msg = {
2398                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2399                 .rpc_argp = &args,
2400                 .rpc_resp = &res,
2401         };
2402
2403         nfs_fattr_init(fattr);
2404         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2405 }
2406
2407 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2408 {
2409         struct nfs4_exception exception = { };
2410         int err;
2411         do {
2412                 err = nfs4_handle_exception(server,
2413                                 _nfs4_proc_getattr(server, fhandle, fattr),
2414                                 &exception);
2415         } while (exception.retry);
2416         return err;
2417 }
2418
2419 /*
2420  * The file is not closed if it is opened due to the a request to change
2421  * the size of the file. The open call will not be needed once the
2422  * VFS layer lookup-intents are implemented.
2423  *
2424  * Close is called when the inode is destroyed.
2425  * If we haven't opened the file for O_WRONLY, we
2426  * need to in the size_change case to obtain a stateid.
2427  *
2428  * Got race?
2429  * Because OPEN is always done by name in nfsv4, it is
2430  * possible that we opened a different file by the same
2431  * name.  We can recognize this race condition, but we
2432  * can't do anything about it besides returning an error.
2433  *
2434  * This will be fixed with VFS changes (lookup-intent).
2435  */
2436 static int
2437 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2438                   struct iattr *sattr)
2439 {
2440         struct inode *inode = dentry->d_inode;
2441         struct rpc_cred *cred = NULL;
2442         struct nfs4_state *state = NULL;
2443         int status;
2444
2445         if (pnfs_ld_layoutret_on_setattr(inode))
2446                 pnfs_return_layout(inode);
2447
2448         nfs_fattr_init(fattr);
2449
2450         /* Search for an existing open(O_WRITE) file */
2451         if (sattr->ia_valid & ATTR_FILE) {
2452                 struct nfs_open_context *ctx;
2453
2454                 ctx = nfs_file_open_context(sattr->ia_file);
2455                 if (ctx) {
2456                         cred = ctx->cred;
2457                         state = ctx->state;
2458                 }
2459         }
2460
2461         /* Deal with open(O_TRUNC) */
2462         if (sattr->ia_valid & ATTR_OPEN)
2463                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME|ATTR_OPEN);
2464
2465         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2466         if (status == 0)
2467                 nfs_setattr_update_inode(inode, sattr);
2468         return status;
2469 }
2470
2471 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2472                 const struct qstr *name, struct nfs_fh *fhandle,
2473                 struct nfs_fattr *fattr)
2474 {
2475         struct nfs_server *server = NFS_SERVER(dir);
2476         int                    status;
2477         struct nfs4_lookup_arg args = {
2478                 .bitmask = server->attr_bitmask,
2479                 .dir_fh = NFS_FH(dir),
2480                 .name = name,
2481         };
2482         struct nfs4_lookup_res res = {
2483                 .server = server,
2484                 .fattr = fattr,
2485                 .fh = fhandle,
2486         };
2487         struct rpc_message msg = {
2488                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2489                 .rpc_argp = &args,
2490                 .rpc_resp = &res,
2491         };
2492
2493         nfs_fattr_init(fattr);
2494
2495         dprintk("NFS call  lookup %s\n", name->name);
2496         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2497         dprintk("NFS reply lookup: %d\n", status);
2498         return status;
2499 }
2500
2501 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2502 {
2503         memset(fh, 0, sizeof(struct nfs_fh));
2504         fattr->fsid.major = 1;
2505         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2506                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2507         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2508         fattr->nlink = 2;
2509 }
2510
2511 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2512                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2513 {
2514         struct nfs4_exception exception = { };
2515         int err;
2516         do {
2517                 int status;
2518
2519                 status = _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr);
2520                 switch (status) {
2521                 case -NFS4ERR_BADNAME:
2522                         return -ENOENT;
2523                 case -NFS4ERR_MOVED:
2524                         return nfs4_get_referral(dir, name, fattr, fhandle);
2525                 case -NFS4ERR_WRONGSEC:
2526                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2527                 }
2528                 err = nfs4_handle_exception(NFS_SERVER(dir),
2529                                 status, &exception);
2530         } while (exception.retry);
2531         return err;
2532 }
2533
2534 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2535 {
2536         struct nfs_server *server = NFS_SERVER(inode);
2537         struct nfs4_accessargs args = {
2538                 .fh = NFS_FH(inode),
2539                 .bitmask = server->cache_consistency_bitmask,
2540         };
2541         struct nfs4_accessres res = {
2542                 .server = server,
2543         };
2544         struct rpc_message msg = {
2545                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2546                 .rpc_argp = &args,
2547                 .rpc_resp = &res,
2548                 .rpc_cred = entry->cred,
2549         };
2550         int mode = entry->mask;
2551         int status;
2552
2553         /*
2554          * Determine which access bits we want to ask for...
2555          */
2556         if (mode & MAY_READ)
2557                 args.access |= NFS4_ACCESS_READ;
2558         if (S_ISDIR(inode->i_mode)) {
2559                 if (mode & MAY_WRITE)
2560                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2561                 if (mode & MAY_EXEC)
2562                         args.access |= NFS4_ACCESS_LOOKUP;
2563         } else {
2564                 if (mode & MAY_WRITE)
2565                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2566                 if (mode & MAY_EXEC)
2567                         args.access |= NFS4_ACCESS_EXECUTE;
2568         }
2569
2570         res.fattr = nfs_alloc_fattr();
2571         if (res.fattr == NULL)
2572                 return -ENOMEM;
2573
2574         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2575         if (!status) {
2576                 entry->mask = 0;
2577                 if (res.access & NFS4_ACCESS_READ)
2578                         entry->mask |= MAY_READ;
2579                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2580                         entry->mask |= MAY_WRITE;
2581                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2582                         entry->mask |= MAY_EXEC;
2583                 nfs_refresh_inode(inode, res.fattr);
2584         }
2585         nfs_free_fattr(res.fattr);
2586         return status;
2587 }
2588
2589 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2590 {
2591         struct nfs4_exception exception = { };
2592         int err;
2593         do {
2594                 err = nfs4_handle_exception(NFS_SERVER(inode),
2595                                 _nfs4_proc_access(inode, entry),
2596                                 &exception);
2597         } while (exception.retry);
2598         return err;
2599 }
2600
2601 /*
2602  * TODO: For the time being, we don't try to get any attributes
2603  * along with any of the zero-copy operations READ, READDIR,
2604  * READLINK, WRITE.
2605  *
2606  * In the case of the first three, we want to put the GETATTR
2607  * after the read-type operation -- this is because it is hard
2608  * to predict the length of a GETATTR response in v4, and thus
2609  * align the READ data correctly.  This means that the GETATTR
2610  * may end up partially falling into the page cache, and we should
2611  * shift it into the 'tail' of the xdr_buf before processing.
2612  * To do this efficiently, we need to know the total length
2613  * of data received, which doesn't seem to be available outside
2614  * of the RPC layer.
2615  *
2616  * In the case of WRITE, we also want to put the GETATTR after
2617  * the operation -- in this case because we want to make sure
2618  * we get the post-operation mtime and size.  This means that
2619  * we can't use xdr_encode_pages() as written: we need a variant
2620  * of it which would leave room in the 'tail' iovec.
2621  *
2622  * Both of these changes to the XDR layer would in fact be quite
2623  * minor, but I decided to leave them for a subsequent patch.
2624  */
2625 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2626                 unsigned int pgbase, unsigned int pglen)
2627 {
2628         struct nfs4_readlink args = {
2629                 .fh       = NFS_FH(inode),
2630                 .pgbase   = pgbase,
2631                 .pglen    = pglen,
2632                 .pages    = &page,
2633         };
2634         struct nfs4_readlink_res res;
2635         struct rpc_message msg = {
2636                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2637                 .rpc_argp = &args,
2638                 .rpc_resp = &res,
2639         };
2640
2641         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2642 }
2643
2644 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2645                 unsigned int pgbase, unsigned int pglen)
2646 {
2647         struct nfs4_exception exception = { };
2648         int err;
2649         do {
2650                 err = nfs4_handle_exception(NFS_SERVER(inode),
2651                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2652                                 &exception);
2653         } while (exception.retry);
2654         return err;
2655 }
2656
2657 /*
2658  * Got race?
2659  * We will need to arrange for the VFS layer to provide an atomic open.
2660  * Until then, this create/open method is prone to inefficiency and race
2661  * conditions due to the lookup, create, and open VFS calls from sys_open()
2662  * placed on the wire.
2663  *
2664  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2665  * The file will be opened again in the subsequent VFS open call
2666  * (nfs4_proc_file_open).
2667  *
2668  * The open for read will just hang around to be used by any process that
2669  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2670  */
2671
2672 static int
2673 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2674                  int flags, struct nfs_open_context *ctx)
2675 {
2676         struct dentry *de = dentry;
2677         struct nfs4_state *state;
2678         struct rpc_cred *cred = NULL;
2679         fmode_t fmode = 0;
2680         int status = 0;
2681
2682         if (ctx != NULL) {
2683                 cred = ctx->cred;
2684                 de = ctx->dentry;
2685                 fmode = ctx->mode;
2686         }
2687         sattr->ia_mode &= ~current_umask();
2688         state = nfs4_do_open(dir, de, fmode, flags, sattr, cred);
2689         d_drop(dentry);
2690         if (IS_ERR(state)) {
2691                 status = PTR_ERR(state);
2692                 goto out;
2693         }
2694         d_add(dentry, igrab(state->inode));
2695         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2696         if (ctx != NULL)
2697                 ctx->state = state;
2698         else
2699                 nfs4_close_sync(state, fmode);
2700 out:
2701         return status;
2702 }
2703
2704 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2705 {
2706         struct nfs_server *server = NFS_SERVER(dir);
2707         struct nfs_removeargs args = {
2708                 .fh = NFS_FH(dir),
2709                 .name.len = name->len,
2710                 .name.name = name->name,
2711                 .bitmask = server->attr_bitmask,
2712         };
2713         struct nfs_removeres res = {
2714                 .server = server,
2715         };
2716         struct rpc_message msg = {
2717                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2718                 .rpc_argp = &args,
2719                 .rpc_resp = &res,
2720         };
2721         int status = -ENOMEM;
2722
2723         res.dir_attr = nfs_alloc_fattr();
2724         if (res.dir_attr == NULL)
2725                 goto out;
2726
2727         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2728         if (status == 0) {
2729                 update_changeattr(dir, &res.cinfo);
2730                 nfs_post_op_update_inode(dir, res.dir_attr);
2731         }
2732         nfs_free_fattr(res.dir_attr);
2733 out:
2734         return status;
2735 }
2736
2737 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2738 {
2739         struct nfs4_exception exception = { };
2740         int err;
2741         do {
2742                 err = nfs4_handle_exception(NFS_SERVER(dir),
2743                                 _nfs4_proc_remove(dir, name),
2744                                 &exception);
2745         } while (exception.retry);
2746         return err;
2747 }
2748
2749 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2750 {
2751         struct nfs_server *server = NFS_SERVER(dir);
2752         struct nfs_removeargs *args = msg->rpc_argp;
2753         struct nfs_removeres *res = msg->rpc_resp;
2754
2755         args->bitmask = server->cache_consistency_bitmask;
2756         res->server = server;
2757         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2758         nfs41_init_sequence(&args->seq_args, &res->seq_res, 1);
2759 }
2760
2761 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2762 {
2763         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2764
2765         if (!nfs4_sequence_done(task, &res->seq_res))
2766                 return 0;
2767         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2768                 return 0;
2769         update_changeattr(dir, &res->cinfo);
2770         nfs_post_op_update_inode(dir, res->dir_attr);
2771         return 1;
2772 }
2773
2774 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2775 {
2776         struct nfs_server *server = NFS_SERVER(dir);
2777         struct nfs_renameargs *arg = msg->rpc_argp;
2778         struct nfs_renameres *res = msg->rpc_resp;
2779
2780         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2781         arg->bitmask = server->attr_bitmask;
2782         res->server = server;
2783         nfs41_init_sequence(&arg->seq_args, &res->seq_res, 1);
2784 }
2785
2786 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2787                                  struct inode *new_dir)
2788 {
2789         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2790
2791         if (!nfs4_sequence_done(task, &res->seq_res))
2792                 return 0;
2793         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2794                 return 0;
2795
2796         update_changeattr(old_dir, &res->old_cinfo);
2797         nfs_post_op_update_inode(old_dir, res->old_fattr);
2798         update_changeattr(new_dir, &res->new_cinfo);
2799         nfs_post_op_update_inode(new_dir, res->new_fattr);
2800         return 1;
2801 }
2802
2803 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2804                 struct inode *new_dir, struct qstr *new_name)
2805 {
2806         struct nfs_server *server = NFS_SERVER(old_dir);
2807         struct nfs_renameargs arg = {
2808                 .old_dir = NFS_FH(old_dir),
2809                 .new_dir = NFS_FH(new_dir),
2810                 .old_name = old_name,
2811                 .new_name = new_name,
2812                 .bitmask = server->attr_bitmask,
2813         };
2814         struct nfs_renameres res = {
2815                 .server = server,
2816         };
2817         struct rpc_message msg = {
2818                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2819                 .rpc_argp = &arg,
2820                 .rpc_resp = &res,
2821         };
2822         int status = -ENOMEM;
2823
2824         res.old_fattr = nfs_alloc_fattr();
2825         res.new_fattr = nfs_alloc_fattr();
2826         if (res.old_fattr == NULL || res.new_fattr == NULL)
2827                 goto out;
2828
2829         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2830         if (!status) {
2831                 update_changeattr(old_dir, &res.old_cinfo);
2832                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2833                 update_changeattr(new_dir, &res.new_cinfo);
2834                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2835         }
2836 out:
2837         nfs_free_fattr(res.new_fattr);
2838         nfs_free_fattr(res.old_fattr);
2839         return status;
2840 }
2841
2842 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2843                 struct inode *new_dir, struct qstr *new_name)
2844 {
2845         struct nfs4_exception exception = { };
2846         int err;
2847         do {
2848                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2849                                 _nfs4_proc_rename(old_dir, old_name,
2850                                         new_dir, new_name),
2851                                 &exception);
2852         } while (exception.retry);
2853         return err;
2854 }
2855
2856 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2857 {
2858         struct nfs_server *server = NFS_SERVER(inode);
2859         struct nfs4_link_arg arg = {
2860                 .fh     = NFS_FH(inode),
2861                 .dir_fh = NFS_FH(dir),
2862                 .name   = name,
2863                 .bitmask = server->attr_bitmask,
2864         };
2865         struct nfs4_link_res res = {
2866                 .server = server,
2867         };
2868         struct rpc_message msg = {
2869                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2870                 .rpc_argp = &arg,
2871                 .rpc_resp = &res,
2872         };
2873         int status = -ENOMEM;
2874
2875         res.fattr = nfs_alloc_fattr();
2876         res.dir_attr = nfs_alloc_fattr();
2877         if (res.fattr == NULL || res.dir_attr == NULL)
2878                 goto out;
2879
2880         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2881         if (!status) {
2882                 update_changeattr(dir, &res.cinfo);
2883                 nfs_post_op_update_inode(dir, res.dir_attr);
2884                 nfs_post_op_update_inode(inode, res.fattr);
2885         }
2886 out:
2887         nfs_free_fattr(res.dir_attr);
2888         nfs_free_fattr(res.fattr);
2889         return status;
2890 }
2891
2892 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2893 {
2894         struct nfs4_exception exception = { };
2895         int err;
2896         do {
2897                 err = nfs4_handle_exception(NFS_SERVER(inode),
2898                                 _nfs4_proc_link(inode, dir, name),
2899                                 &exception);
2900         } while (exception.retry);
2901         return err;
2902 }
2903
2904 struct nfs4_createdata {
2905         struct rpc_message msg;
2906         struct nfs4_create_arg arg;
2907         struct nfs4_create_res res;
2908         struct nfs_fh fh;
2909         struct nfs_fattr fattr;
2910         struct nfs_fattr dir_fattr;
2911 };
2912
2913 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2914                 struct qstr *name, struct iattr *sattr, u32 ftype)
2915 {
2916         struct nfs4_createdata *data;
2917
2918         data = kzalloc(sizeof(*data), GFP_KERNEL);
2919         if (data != NULL) {
2920                 struct nfs_server *server = NFS_SERVER(dir);
2921
2922                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2923                 data->msg.rpc_argp = &data->arg;
2924                 data->msg.rpc_resp = &data->res;
2925                 data->arg.dir_fh = NFS_FH(dir);
2926                 data->arg.server = server;
2927                 data->arg.name = name;
2928                 data->arg.attrs = sattr;
2929                 data->arg.ftype = ftype;
2930                 data->arg.bitmask = server->attr_bitmask;
2931                 data->res.server = server;
2932                 data->res.fh = &data->fh;
2933                 data->res.fattr = &data->fattr;
2934                 data->res.dir_fattr = &data->dir_fattr;
2935                 nfs_fattr_init(data->res.fattr);
2936                 nfs_fattr_init(data->res.dir_fattr);
2937         }
2938         return data;
2939 }
2940
2941 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2942 {
2943         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2944                                     &data->arg.seq_args, &data->res.seq_res, 1);
2945         if (status == 0) {
2946                 update_changeattr(dir, &data->res.dir_cinfo);
2947                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2948                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2949         }
2950         return status;
2951 }
2952
2953 static void nfs4_free_createdata(struct nfs4_createdata *data)
2954 {
2955         kfree(data);
2956 }
2957
2958 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2959                 struct page *page, unsigned int len, struct iattr *sattr)
2960 {
2961         struct nfs4_createdata *data;
2962         int status = -ENAMETOOLONG;
2963
2964         if (len > NFS4_MAXPATHLEN)
2965                 goto out;
2966
2967         status = -ENOMEM;
2968         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2969         if (data == NULL)
2970                 goto out;
2971
2972         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2973         data->arg.u.symlink.pages = &page;
2974         data->arg.u.symlink.len = len;
2975
2976         status = nfs4_do_create(dir, dentry, data);
2977
2978         nfs4_free_createdata(data);
2979 out:
2980         return status;
2981 }
2982
2983 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2984                 struct page *page, unsigned int len, struct iattr *sattr)
2985 {
2986         struct nfs4_exception exception = { };
2987         int err;
2988         do {
2989                 err = nfs4_handle_exception(NFS_SERVER(dir),
2990                                 _nfs4_proc_symlink(dir, dentry, page,
2991                                                         len, sattr),
2992                                 &exception);
2993         } while (exception.retry);
2994         return err;
2995 }
2996
2997 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2998                 struct iattr *sattr)
2999 {
3000         struct nfs4_createdata *data;
3001         int status = -ENOMEM;
3002
3003         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3004         if (data == NULL)
3005                 goto out;
3006
3007         status = nfs4_do_create(dir, dentry, data);
3008
3009         nfs4_free_createdata(data);
3010 out:
3011         return status;
3012 }
3013
3014 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3015                 struct iattr *sattr)
3016 {
3017         struct nfs4_exception exception = { };
3018         int err;
3019
3020         sattr->ia_mode &= ~current_umask();
3021         do {
3022                 err = nfs4_handle_exception(NFS_SERVER(dir),
3023                                 _nfs4_proc_mkdir(dir, dentry, sattr),
3024                                 &exception);
3025         } while (exception.retry);
3026         return err;
3027 }
3028
3029 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3030                 u64 cookie, struct page **pages, unsigned int count, int plus)
3031 {
3032         struct inode            *dir = dentry->d_inode;
3033         struct nfs4_readdir_arg args = {
3034                 .fh = NFS_FH(dir),
3035                 .pages = pages,
3036                 .pgbase = 0,
3037                 .count = count,
3038                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
3039                 .plus = plus,
3040         };
3041         struct nfs4_readdir_res res;
3042         struct rpc_message msg = {
3043                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3044                 .rpc_argp = &args,
3045                 .rpc_resp = &res,
3046                 .rpc_cred = cred,
3047         };
3048         int                     status;
3049
3050         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
3051                         dentry->d_parent->d_name.name,
3052                         dentry->d_name.name,
3053                         (unsigned long long)cookie);
3054         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
3055         res.pgbase = args.pgbase;
3056         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3057         if (status >= 0) {
3058                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
3059                 status += args.pgbase;
3060         }
3061
3062         nfs_invalidate_atime(dir);
3063
3064         dprintk("%s: returns %d\n", __func__, status);
3065         return status;
3066 }
3067
3068 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3069                 u64 cookie, struct page **pages, unsigned int count, int plus)
3070 {
3071         struct nfs4_exception exception = { };
3072         int err;
3073         do {
3074                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3075                                 _nfs4_proc_readdir(dentry, cred, cookie,
3076                                         pages, count, plus),
3077                                 &exception);
3078         } while (exception.retry);
3079         return err;
3080 }
3081
3082 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3083                 struct iattr *sattr, dev_t rdev)
3084 {
3085         struct nfs4_createdata *data;
3086         int mode = sattr->ia_mode;
3087         int status = -ENOMEM;
3088
3089         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3090         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3091
3092         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3093         if (data == NULL)
3094                 goto out;
3095
3096         if (S_ISFIFO(mode))
3097                 data->arg.ftype = NF4FIFO;
3098         else if (S_ISBLK(mode)) {
3099                 data->arg.ftype = NF4BLK;
3100                 data->arg.u.device.specdata1 = MAJOR(rdev);
3101                 data->arg.u.device.specdata2 = MINOR(rdev);
3102         }
3103         else if (S_ISCHR(mode)) {
3104                 data->arg.ftype = NF4CHR;
3105                 data->arg.u.device.specdata1 = MAJOR(rdev);
3106                 data->arg.u.device.specdata2 = MINOR(rdev);
3107         }
3108
3109         status = nfs4_do_create(dir, dentry, data);
3110
3111         nfs4_free_createdata(data);
3112 out:
3113         return status;
3114 }
3115
3116 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3117                 struct iattr *sattr, dev_t rdev)
3118 {
3119         struct nfs4_exception exception = { };
3120         int err;
3121
3122         sattr->ia_mode &= ~current_umask();
3123         do {
3124                 err = nfs4_handle_exception(NFS_SERVER(dir),
3125                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3126                                 &exception);
3127         } while (exception.retry);
3128         return err;
3129 }
3130
3131 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3132                  struct nfs_fsstat *fsstat)
3133 {
3134         struct nfs4_statfs_arg args = {
3135                 .fh = fhandle,
3136                 .bitmask = server->attr_bitmask,
3137         };
3138         struct nfs4_statfs_res res = {
3139                 .fsstat = fsstat,
3140         };
3141         struct rpc_message msg = {
3142                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3143                 .rpc_argp = &args,
3144                 .rpc_resp = &res,
3145         };
3146
3147         nfs_fattr_init(fsstat->fattr);
3148         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3149 }
3150
3151 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3152 {
3153         struct nfs4_exception exception = { };
3154         int err;
3155         do {
3156                 err = nfs4_handle_exception(server,
3157                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3158                                 &exception);
3159         } while (exception.retry);
3160         return err;
3161 }
3162
3163 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3164                 struct nfs_fsinfo *fsinfo)
3165 {
3166         struct nfs4_fsinfo_arg args = {
3167                 .fh = fhandle,
3168                 .bitmask = server->attr_bitmask,
3169         };
3170         struct nfs4_fsinfo_res res = {
3171                 .fsinfo = fsinfo,
3172         };
3173         struct rpc_message msg = {
3174                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3175                 .rpc_argp = &args,
3176                 .rpc_resp = &res,
3177         };
3178
3179         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3180 }
3181
3182 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3183 {
3184         struct nfs4_exception exception = { };
3185         int err;
3186
3187         do {
3188                 err = nfs4_handle_exception(server,
3189                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3190                                 &exception);
3191         } while (exception.retry);
3192         return err;
3193 }
3194
3195 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3196 {
3197         nfs_fattr_init(fsinfo->fattr);
3198         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3199 }
3200
3201 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3202                 struct nfs_pathconf *pathconf)
3203 {
3204         struct nfs4_pathconf_arg args = {
3205                 .fh = fhandle,
3206                 .bitmask = server->attr_bitmask,
3207         };
3208         struct nfs4_pathconf_res res = {
3209                 .pathconf = pathconf,
3210         };
3211         struct rpc_message msg = {
3212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3213                 .rpc_argp = &args,
3214                 .rpc_resp = &res,
3215         };
3216
3217         /* None of the pathconf attributes are mandatory to implement */
3218         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3219                 memset(pathconf, 0, sizeof(*pathconf));
3220                 return 0;
3221         }
3222
3223         nfs_fattr_init(pathconf->fattr);
3224         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3225 }
3226
3227 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3228                 struct nfs_pathconf *pathconf)
3229 {
3230         struct nfs4_exception exception = { };
3231         int err;
3232
3233         do {
3234                 err = nfs4_handle_exception(server,
3235                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3236                                 &exception);
3237         } while (exception.retry);
3238         return err;
3239 }
3240
3241 void __nfs4_read_done_cb(struct nfs_read_data *data)
3242 {
3243         nfs_invalidate_atime(data->inode);
3244 }
3245
3246 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3247 {
3248         struct nfs_server *server = NFS_SERVER(data->inode);
3249
3250         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3251                 rpc_restart_call_prepare(task);
3252                 return -EAGAIN;
3253         }
3254
3255         __nfs4_read_done_cb(data);
3256         if (task->tk_status > 0)
3257                 renew_lease(server, data->timestamp);
3258         return 0;
3259 }
3260
3261 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3262 {
3263
3264         dprintk("--> %s\n", __func__);
3265
3266         if (!nfs4_sequence_done(task, &data->res.seq_res))
3267                 return -EAGAIN;
3268
3269         return data->read_done_cb ? data->read_done_cb(task, data) :
3270                                     nfs4_read_done_cb(task, data);
3271 }
3272
3273 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3274 {
3275         data->timestamp   = jiffies;
3276         data->read_done_cb = nfs4_read_done_cb;
3277         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3278         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
3279 }
3280
3281 /* Reset the the nfs_read_data to send the read to the MDS. */
3282 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3283 {
3284         dprintk("%s Reset task for i/o through\n", __func__);
3285         put_lseg(data->lseg);
3286         data->lseg = NULL;
3287         /* offsets will differ in the dense stripe case */
3288         data->args.offset = data->mds_offset;
3289         data->ds_clp = NULL;
3290         data->args.fh     = NFS_FH(data->inode);
3291         data->read_done_cb = nfs4_read_done_cb;
3292         task->tk_ops = data->mds_ops;
3293         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3294 }
3295 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3296
3297 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3298 {
3299         struct inode *inode = data->inode;
3300
3301         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3302                 rpc_restart_call_prepare(task);
3303                 return -EAGAIN;
3304         }
3305         if (task->tk_status >= 0) {
3306                 renew_lease(NFS_SERVER(inode), data->timestamp);
3307                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3308         }
3309         return 0;
3310 }
3311
3312 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3313 {
3314         if (!nfs4_sequence_done(task, &data->res.seq_res))
3315                 return -EAGAIN;
3316         return data->write_done_cb ? data->write_done_cb(task, data) :
3317                 nfs4_write_done_cb(task, data);
3318 }
3319
3320 /* Reset the the nfs_write_data to send the write to the MDS. */
3321 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3322 {
3323         dprintk("%s Reset task for i/o through\n", __func__);
3324         put_lseg(data->lseg);
3325         data->lseg          = NULL;
3326         data->ds_clp        = NULL;
3327         data->write_done_cb = nfs4_write_done_cb;
3328         data->args.fh       = NFS_FH(data->inode);
3329         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3330         data->args.offset   = data->mds_offset;
3331         data->res.fattr     = &data->fattr;
3332         task->tk_ops        = data->mds_ops;
3333         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3334 }
3335 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3336
3337 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3338 {
3339         struct nfs_server *server = NFS_SERVER(data->inode);
3340
3341         if (data->lseg) {
3342                 data->args.bitmask = NULL;
3343                 data->res.fattr = NULL;
3344         } else
3345                 data->args.bitmask = server->cache_consistency_bitmask;
3346         if (!data->write_done_cb)
3347                 data->write_done_cb = nfs4_write_done_cb;
3348         data->res.server = server;
3349         data->timestamp   = jiffies;
3350
3351         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3352         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3353 }
3354
3355 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3356 {
3357         struct inode *inode = data->inode;
3358
3359         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3360                 rpc_restart_call_prepare(task);
3361                 return -EAGAIN;
3362         }
3363         nfs_refresh_inode(inode, data->res.fattr);
3364         return 0;
3365 }
3366
3367 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3368 {
3369         if (!nfs4_sequence_done(task, &data->res.seq_res))
3370                 return -EAGAIN;
3371         return data->write_done_cb(task, data);
3372 }
3373
3374 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3375 {
3376         struct nfs_server *server = NFS_SERVER(data->inode);
3377
3378         if (data->lseg) {
3379                 data->args.bitmask = NULL;
3380                 data->res.fattr = NULL;
3381         } else
3382                 data->args.bitmask = server->cache_consistency_bitmask;
3383         if (!data->write_done_cb)
3384                 data->write_done_cb = nfs4_commit_done_cb;
3385         data->res.server = server;
3386         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3387         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3388 }
3389
3390 struct nfs4_renewdata {
3391         struct nfs_client       *client;
3392         unsigned long           timestamp;
3393 };
3394
3395 /*
3396  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3397  * standalone procedure for queueing an asynchronous RENEW.
3398  */
3399 static void nfs4_renew_release(void *calldata)
3400 {
3401         struct nfs4_renewdata *data = calldata;
3402         struct nfs_client *clp = data->client;
3403
3404         if (atomic_read(&clp->cl_count) > 1)
3405                 nfs4_schedule_state_renewal(clp);
3406         nfs_put_client(clp);
3407         kfree(data);
3408 }
3409
3410 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3411 {
3412         struct nfs4_renewdata *data = calldata;
3413         struct nfs_client *clp = data->client;
3414         unsigned long timestamp = data->timestamp;
3415
3416         if (task->tk_status < 0) {
3417                 /* Unless we're shutting down, schedule state recovery! */
3418                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
3419                         return;
3420                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
3421                         nfs4_schedule_lease_recovery(clp);
3422                         return;
3423                 }
3424                 nfs4_schedule_path_down_recovery(clp);
3425         }
3426         do_renew_lease(clp, timestamp);
3427 }
3428
3429 static const struct rpc_call_ops nfs4_renew_ops = {
3430         .rpc_call_done = nfs4_renew_done,
3431         .rpc_release = nfs4_renew_release,
3432 };
3433
3434 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
3435 {
3436         struct rpc_message msg = {
3437                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3438                 .rpc_argp       = clp,
3439                 .rpc_cred       = cred,
3440         };
3441         struct nfs4_renewdata *data;
3442
3443         if (renew_flags == 0)
3444                 return 0;
3445         if (!atomic_inc_not_zero(&clp->cl_count))
3446                 return -EIO;
3447         data = kmalloc(sizeof(*data), GFP_NOFS);
3448         if (data == NULL)
3449                 return -ENOMEM;
3450         data->client = clp;
3451         data->timestamp = jiffies;
3452         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3453                         &nfs4_renew_ops, data);
3454 }
3455
3456 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3457 {
3458         struct rpc_message msg = {
3459                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3460                 .rpc_argp       = clp,
3461                 .rpc_cred       = cred,
3462         };
3463         unsigned long now = jiffies;
3464         int status;
3465
3466         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3467         if (status < 0)
3468                 return status;
3469         do_renew_lease(clp, now);
3470         return 0;
3471 }
3472
3473 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3474 {
3475         return (server->caps & NFS_CAP_ACLS)
3476                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3477                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3478 }
3479
3480 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3481  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3482  * the stack.
3483  */
3484 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3485
3486 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3487                 struct page **pages, unsigned int *pgbase)
3488 {
3489         struct page *newpage, **spages;
3490         int rc = 0;
3491         size_t len;
3492         spages = pages;
3493
3494         do {
3495                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3496                 newpage = alloc_page(GFP_KERNEL);
3497
3498                 if (newpage == NULL)
3499                         goto unwind;
3500                 memcpy(page_address(newpage), buf, len);
3501                 buf += len;
3502                 buflen -= len;
3503                 *pages++ = newpage;
3504                 rc++;
3505         } while (buflen != 0);
3506
3507         return rc;
3508
3509 unwind:
3510         for(; rc > 0; rc--)
3511                 __free_page(spages[rc-1]);
3512         return -ENOMEM;
3513 }
3514
3515 struct nfs4_cached_acl {
3516         int cached;
3517         size_t len;
3518         char data[0];
3519 };
3520
3521 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3522 {
3523         struct nfs_inode *nfsi = NFS_I(inode);
3524
3525         spin_lock(&inode->i_lock);
3526         kfree(nfsi->nfs4_acl);
3527         nfsi->nfs4_acl = acl;
3528         spin_unlock(&inode->i_lock);
3529 }
3530
3531 static void nfs4_zap_acl_attr(struct inode *inode)
3532 {
3533         nfs4_set_cached_acl(inode, NULL);
3534 }
3535
3536 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3537 {
3538         struct nfs_inode *nfsi = NFS_I(inode);
3539         struct nfs4_cached_acl *acl;
3540         int ret = -ENOENT;
3541
3542         spin_lock(&inode->i_lock);
3543         acl = nfsi->nfs4_acl;
3544         if (acl == NULL)
3545                 goto out;
3546         if (buf == NULL) /* user is just asking for length */
3547                 goto out_len;
3548         if (acl->cached == 0)
3549                 goto out;
3550         ret = -ERANGE; /* see getxattr(2) man page */
3551         if (acl->len > buflen)
3552                 goto out;
3553         memcpy(buf, acl->data, acl->len);
3554 out_len:
3555         ret = acl->len;
3556 out:
3557         spin_unlock(&inode->i_lock);
3558         return ret;
3559 }
3560
3561 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3562 {
3563         struct nfs4_cached_acl *acl;
3564
3565         if (buf && acl_len <= PAGE_SIZE) {
3566                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3567                 if (acl == NULL)
3568                         goto out;
3569                 acl->cached = 1;
3570                 memcpy(acl->data, buf, acl_len);
3571         } else {
3572                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3573                 if (acl == NULL)
3574                         goto out;
3575                 acl->cached = 0;
3576         }
3577         acl->len = acl_len;
3578 out:
3579         nfs4_set_cached_acl(inode, acl);
3580 }
3581
3582 /*
3583  * The getxattr API returns the required buffer length when called with a
3584  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
3585  * the required buf.  On a NULL buf, we send a page of data to the server
3586  * guessing that the ACL request can be serviced by a page. If so, we cache
3587  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
3588  * the cache. If not so, we throw away the page, and cache the required
3589  * length. The next getxattr call will then produce another round trip to
3590  * the server, this time with the input buf of the required size.
3591  */
3592 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3593 {
3594         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
3595         struct nfs_getaclargs args = {
3596                 .fh = NFS_FH(inode),
3597                 .acl_pages = pages,
3598                 .acl_len = buflen,
3599         };
3600         struct nfs_getaclres res = {
3601                 .acl_len = buflen,
3602         };
3603         void *resp_buf;
3604         struct rpc_message msg = {
3605                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3606                 .rpc_argp = &args,
3607                 .rpc_resp = &res,
3608         };
3609         int ret = -ENOMEM, npages, i, acl_len = 0;
3610
3611         npages = (buflen + PAGE_SIZE - 1) >> PAGE_SHIFT;
3612         /* As long as we're doing a round trip to the server anyway,
3613          * let's be prepared for a page of acl data. */
3614         if (npages == 0)
3615                 npages = 1;
3616
3617         for (i = 0; i < npages; i++) {
3618                 pages[i] = alloc_page(GFP_KERNEL);
3619                 if (!pages[i])
3620                         goto out_free;
3621         }
3622         if (npages > 1) {
3623                 /* for decoding across pages */
3624                 res.acl_scratch = alloc_page(GFP_KERNEL);
3625                 if (!res.acl_scratch)
3626                         goto out_free;
3627         }
3628         args.acl_len = npages * PAGE_SIZE;
3629         args.acl_pgbase = 0;
3630         /* Let decode_getfacl know not to fail if the ACL data is larger than
3631          * the page we send as a guess */
3632         if (buf == NULL)
3633                 res.acl_flags |= NFS4_ACL_LEN_REQUEST;
3634         resp_buf = page_address(pages[0]);
3635
3636         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
3637                 __func__, buf, buflen, npages, args.acl_len);
3638         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
3639                              &msg, &args.seq_args, &res.seq_res, 0);
3640         if (ret)
3641                 goto out_free;
3642
3643         acl_len = res.acl_len - res.acl_data_offset;
3644         if (acl_len > args.acl_len)
3645                 nfs4_write_cached_acl(inode, NULL, acl_len);
3646         else
3647                 nfs4_write_cached_acl(inode, resp_buf + res.acl_data_offset,
3648                                       acl_len);
3649         if (buf) {
3650                 ret = -ERANGE;
3651                 if (acl_len > buflen)
3652                         goto out_free;
3653                 _copy_from_pages(buf, pages, res.acl_data_offset,
3654                                 res.acl_len);
3655         }
3656         ret = acl_len;
3657 out_free:
3658         for (i = 0; i < npages; i++)
3659                 if (pages[i])
3660                         __free_page(pages[i]);
3661         if (res.acl_scratch)
3662                 __free_page(res.acl_scratch);
3663         return ret;
3664 }
3665
3666 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3667 {
3668         struct nfs4_exception exception = { };
3669         ssize_t ret;
3670         do {
3671                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3672                 if (ret >= 0)
3673                         break;
3674                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3675         } while (exception.retry);
3676         return ret;
3677 }
3678
3679 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3680 {
3681         struct nfs_server *server = NFS_SERVER(inode);
3682         int ret;
3683
3684         if (!nfs4_server_supports_acls(server))
3685                 return -EOPNOTSUPP;
3686         ret = nfs_revalidate_inode(server, inode);
3687         if (ret < 0)
3688                 return ret;
3689         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3690                 nfs_zap_acl_cache(inode);
3691         ret = nfs4_read_cached_acl(inode, buf, buflen);
3692         if (ret != -ENOENT)
3693                 /* -ENOENT is returned if there is no ACL or if there is an ACL
3694                  * but no cached acl data, just the acl length */
3695                 return ret;
3696         return nfs4_get_acl_uncached(inode, buf, buflen);
3697 }
3698
3699 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3700 {
3701         struct nfs_server *server = NFS_SERVER(inode);
3702         struct page *pages[NFS4ACL_MAXPAGES];
3703         struct nfs_setaclargs arg = {
3704                 .fh             = NFS_FH(inode),
3705                 .acl_pages      = pages,
3706                 .acl_len        = buflen,
3707         };
3708         struct nfs_setaclres res;
3709         struct rpc_message msg = {
3710                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3711                 .rpc_argp       = &arg,
3712                 .rpc_resp       = &res,
3713         };
3714         int ret, i;
3715
3716         if (!nfs4_server_supports_acls(server))
3717                 return -EOPNOTSUPP;
3718         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3719         if (i < 0)
3720                 return i;
3721         nfs_inode_return_delegation(inode);
3722         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3723
3724         /*
3725          * Free each page after tx, so the only ref left is
3726          * held by the network stack
3727          */
3728         for (; i > 0; i--)
3729                 put_page(pages[i-1]);
3730
3731         /*
3732          * Acl update can result in inode attribute update.
3733          * so mark the attribute cache invalid.
3734          */
3735         spin_lock(&inode->i_lock);
3736         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3737         spin_unlock(&inode->i_lock);
3738         nfs_access_zap_cache(inode);
3739         nfs_zap_acl_cache(inode);
3740         return ret;
3741 }
3742
3743 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3744 {
3745         struct nfs4_exception exception = { };
3746         int err;
3747         do {
3748                 err = nfs4_handle_exception(NFS_SERVER(inode),
3749                                 __nfs4_proc_set_acl(inode, buf, buflen),
3750                                 &exception);
3751         } while (exception.retry);
3752         return err;
3753 }
3754
3755 static int
3756 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3757 {
3758         struct nfs_client *clp = server->nfs_client;
3759
3760         if (task->tk_status >= 0)
3761                 return 0;
3762         switch(task->tk_status) {
3763                 case -NFS4ERR_ADMIN_REVOKED:
3764                 case -NFS4ERR_BAD_STATEID:
3765                 case -NFS4ERR_OPENMODE:
3766                         if (state == NULL)
3767                                 break;
3768                         nfs4_schedule_stateid_recovery(server, state);
3769                         goto wait_on_recovery;
3770                 case -NFS4ERR_EXPIRED:
3771                         if (state != NULL)
3772                                 nfs4_schedule_stateid_recovery(server, state);
3773                 case -NFS4ERR_STALE_STATEID:
3774                 case -NFS4ERR_STALE_CLIENTID:
3775                         nfs4_schedule_lease_recovery(clp);
3776                         goto wait_on_recovery;
3777 #if defined(CONFIG_NFS_V4_1)
3778                 case -NFS4ERR_BADSESSION:
3779                 case -NFS4ERR_BADSLOT:
3780                 case -NFS4ERR_BAD_HIGH_SLOT:
3781                 case -NFS4ERR_DEADSESSION:
3782                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3783                 case -NFS4ERR_SEQ_FALSE_RETRY:
3784                 case -NFS4ERR_SEQ_MISORDERED:
3785                         dprintk("%s ERROR %d, Reset session\n", __func__,
3786                                 task->tk_status);
3787                         nfs4_schedule_session_recovery(clp->cl_session);
3788                         task->tk_status = 0;
3789                         return -EAGAIN;
3790 #endif /* CONFIG_NFS_V4_1 */
3791                 case -NFS4ERR_DELAY:
3792                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3793                 case -NFS4ERR_GRACE:
3794                 case -EKEYEXPIRED:
3795                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3796                         task->tk_status = 0;
3797                         return -EAGAIN;
3798                 case -NFS4ERR_RETRY_UNCACHED_REP:
3799                 case -NFS4ERR_OLD_STATEID:
3800                         task->tk_status = 0;
3801                         return -EAGAIN;
3802         }
3803         task->tk_status = nfs4_map_errors(task->tk_status);
3804         return 0;
3805 wait_on_recovery:
3806         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3807         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3808                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3809         task->tk_status = 0;
3810         return -EAGAIN;
3811 }
3812
3813 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3814                 unsigned short port, struct rpc_cred *cred,
3815                 struct nfs4_setclientid_res *res)
3816 {
3817         nfs4_verifier sc_verifier;
3818         struct nfs4_setclientid setclientid = {
3819                 .sc_verifier = &sc_verifier,
3820                 .sc_prog = program,
3821                 .sc_cb_ident = clp->cl_cb_ident,
3822         };
3823         struct rpc_message msg = {
3824                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3825                 .rpc_argp = &setclientid,
3826                 .rpc_resp = res,
3827                 .rpc_cred = cred,
3828         };
3829         __be32 *p;
3830         int loop = 0;
3831         int status;
3832
3833         p = (__be32*)sc_verifier.data;
3834         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3835         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3836
3837         for(;;) {
3838                 rcu_read_lock();
3839                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3840                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3841                                 clp->cl_ipaddr,
3842                                 rpc_peeraddr2str(clp->cl_rpcclient,
3843                                                         RPC_DISPLAY_ADDR),
3844                                 rpc_peeraddr2str(clp->cl_rpcclient,
3845                                                         RPC_DISPLAY_PROTO),
3846                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3847                                 clp->cl_id_uniquifier);
3848                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3849                                 sizeof(setclientid.sc_netid),
3850                                 rpc_peeraddr2str(clp->cl_rpcclient,
3851                                                         RPC_DISPLAY_NETID));
3852                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3853                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3854                                 clp->cl_ipaddr, port >> 8, port & 255);
3855                 rcu_read_unlock();
3856
3857                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3858                 if (status != -NFS4ERR_CLID_INUSE)
3859                         break;
3860                 if (loop != 0) {
3861                         ++clp->cl_id_uniquifier;
3862                         break;
3863                 }
3864                 ++loop;
3865                 ssleep(clp->cl_lease_time / HZ + 1);
3866         }
3867         return status;
3868 }
3869
3870 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3871                 struct nfs4_setclientid_res *arg,
3872                 struct rpc_cred *cred)
3873 {
3874         struct nfs_fsinfo fsinfo;
3875         struct rpc_message msg = {
3876                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3877                 .rpc_argp = arg,
3878                 .rpc_resp = &fsinfo,
3879                 .rpc_cred = cred,
3880         };
3881         unsigned long now;
3882         int status;
3883
3884         now = jiffies;
3885         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3886         if (status == 0) {
3887                 spin_lock(&clp->cl_lock);
3888                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3889                 clp->cl_last_renewal = now;
3890                 spin_unlock(&clp->cl_lock);
3891         }
3892         return status;
3893 }
3894
3895 struct nfs4_delegreturndata {
3896         struct nfs4_delegreturnargs args;
3897         struct nfs4_delegreturnres res;
3898         struct nfs_fh fh;
3899         nfs4_stateid stateid;
3900         unsigned long timestamp;
3901         struct nfs_fattr fattr;
3902         int rpc_status;
3903 };
3904
3905 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3906 {
3907         struct nfs4_delegreturndata *data = calldata;
3908
3909         if (!nfs4_sequence_done(task, &data->res.seq_res))
3910                 return;
3911
3912         switch (task->tk_status) {
3913         case -NFS4ERR_STALE_STATEID:
3914         case -NFS4ERR_EXPIRED:
3915         case 0:
3916                 renew_lease(data->res.server, data->timestamp);
3917                 break;
3918         default:
3919                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3920                                 -EAGAIN) {
3921                         rpc_restart_call_prepare(task);
3922                         return;
3923                 }
3924         }
3925         data->rpc_status = task->tk_status;
3926 }
3927
3928 static void nfs4_delegreturn_release(void *calldata)
3929 {
3930         kfree(calldata);
3931 }
3932
3933 #if defined(CONFIG_NFS_V4_1)
3934 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3935 {
3936         struct nfs4_delegreturndata *d_data;
3937
3938         d_data = (struct nfs4_delegreturndata *)data;
3939
3940         if (nfs4_setup_sequence(d_data->res.server,
3941                                 &d_data->args.seq_args,
3942                                 &d_data->res.seq_res, task))
3943                 return;
3944         rpc_call_start(task);
3945 }
3946 #endif /* CONFIG_NFS_V4_1 */
3947
3948 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3949 #if defined(CONFIG_NFS_V4_1)
3950         .rpc_call_prepare = nfs4_delegreturn_prepare,
3951 #endif /* CONFIG_NFS_V4_1 */
3952         .rpc_call_done = nfs4_delegreturn_done,
3953         .rpc_release = nfs4_delegreturn_release,
3954 };
3955
3956 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3957 {
3958         struct nfs4_delegreturndata *data;
3959         struct nfs_server *server = NFS_SERVER(inode);
3960         struct rpc_task *task;
3961         struct rpc_message msg = {
3962                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3963                 .rpc_cred = cred,
3964         };
3965         struct rpc_task_setup task_setup_data = {
3966                 .rpc_client = server->client,
3967                 .rpc_message = &msg,
3968                 .callback_ops = &nfs4_delegreturn_ops,
3969                 .flags = RPC_TASK_ASYNC,
3970         };
3971         int status = 0;
3972
3973         data = kzalloc(sizeof(*data), GFP_NOFS);
3974         if (data == NULL)
3975                 return -ENOMEM;
3976         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
3977         data->args.fhandle = &data->fh;
3978         data->args.stateid = &data->stateid;
3979         data->args.bitmask = server->attr_bitmask;
3980         nfs_copy_fh(&data->fh, NFS_FH(inode));
3981         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3982         data->res.fattr = &data->fattr;
3983         data->res.server = server;
3984         nfs_fattr_init(data->res.fattr);
3985         data->timestamp = jiffies;
3986         data->rpc_status = 0;
3987
3988         task_setup_data.callback_data = data;
3989         msg.rpc_argp = &data->args;
3990         msg.rpc_resp = &data->res;
3991         task = rpc_run_task(&task_setup_data);
3992         if (IS_ERR(task))
3993                 return PTR_ERR(task);
3994         if (!issync)
3995                 goto out;
3996         status = nfs4_wait_for_completion_rpc_task(task);
3997         if (status != 0)
3998                 goto out;
3999         status = data->rpc_status;
4000         if (status != 0)
4001                 goto out;
4002         nfs_refresh_inode(inode, &data->fattr);
4003 out:
4004         rpc_put_task(task);
4005         return status;
4006 }
4007
4008 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
4009 {
4010         struct nfs_server *server = NFS_SERVER(inode);
4011         struct nfs4_exception exception = { };
4012         int err;
4013         do {
4014                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
4015                 switch (err) {
4016                         case -NFS4ERR_STALE_STATEID:
4017                         case -NFS4ERR_EXPIRED:
4018                         case 0:
4019                                 return 0;
4020                 }
4021                 err = nfs4_handle_exception(server, err, &exception);
4022         } while (exception.retry);
4023         return err;
4024 }
4025
4026 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
4027 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
4028
4029 /*
4030  * sleep, with exponential backoff, and retry the LOCK operation.
4031  */
4032 static unsigned long
4033 nfs4_set_lock_task_retry(unsigned long timeout)
4034 {
4035         freezable_schedule_timeout_killable(timeout);
4036         timeout <<= 1;
4037         if (timeout > NFS4_LOCK_MAXTIMEOUT)
4038                 return NFS4_LOCK_MAXTIMEOUT;
4039         return timeout;
4040 }
4041
4042 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4043 {
4044         struct inode *inode = state->inode;
4045         struct nfs_server *server = NFS_SERVER(inode);
4046         struct nfs_client *clp = server->nfs_client;
4047         struct nfs_lockt_args arg = {
4048                 .fh = NFS_FH(inode),
4049                 .fl = request,
4050         };
4051         struct nfs_lockt_res res = {
4052                 .denied = request,
4053         };
4054         struct rpc_message msg = {
4055                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
4056                 .rpc_argp       = &arg,
4057                 .rpc_resp       = &res,
4058                 .rpc_cred       = state->owner->so_cred,
4059         };
4060         struct nfs4_lock_state *lsp;
4061         int status;
4062
4063         arg.lock_owner.clientid = clp->cl_clientid;
4064         status = nfs4_set_lock_state(state, request);
4065         if (status != 0)
4066                 goto out;
4067         lsp = request->fl_u.nfs4_fl.owner;
4068         arg.lock_owner.id = lsp->ls_seqid.owner_id;
4069         arg.lock_owner.s_dev = server->s_dev;
4070         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4071         switch (status) {
4072                 case 0:
4073                         request->fl_type = F_UNLCK;
4074                         break;
4075                 case -NFS4ERR_DENIED:
4076                         status = 0;
4077         }
4078         request->fl_ops->fl_release_private(request);
4079 out:
4080         return status;
4081 }
4082
4083 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4084 {
4085         struct nfs4_exception exception = { };
4086         int err;
4087
4088         do {
4089                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4090                                 _nfs4_proc_getlk(state, cmd, request),
4091                                 &exception);
4092         } while (exception.retry);
4093         return err;
4094 }
4095
4096 static int do_vfs_lock(struct file *file, struct file_lock *fl)
4097 {
4098         int res = 0;
4099         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
4100                 case FL_POSIX:
4101                         res = posix_lock_file_wait(file, fl);
4102                         break;
4103                 case FL_FLOCK:
4104                         res = flock_lock_file_wait(file, fl);
4105                         break;
4106                 default:
4107                         BUG();
4108         }
4109         return res;
4110 }
4111
4112 struct nfs4_unlockdata {
4113         struct nfs_locku_args arg;
4114         struct nfs_locku_res res;
4115         struct nfs4_lock_state *lsp;
4116         struct nfs_open_context *ctx;
4117         struct file_lock fl;
4118         const struct nfs_server *server;
4119         unsigned long timestamp;
4120 };
4121
4122 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4123                 struct nfs_open_context *ctx,
4124                 struct nfs4_lock_state *lsp,
4125                 struct nfs_seqid *seqid)
4126 {
4127         struct nfs4_unlockdata *p;
4128         struct inode *inode = lsp->ls_state->inode;
4129
4130         p = kzalloc(sizeof(*p), GFP_NOFS);
4131         if (p == NULL)
4132                 return NULL;
4133         p->arg.fh = NFS_FH(inode);
4134         p->arg.fl = &p->fl;
4135         p->arg.seqid = seqid;
4136         p->res.seqid = seqid;
4137         p->arg.stateid = &lsp->ls_stateid;
4138         p->lsp = lsp;
4139         atomic_inc(&lsp->ls_count);
4140         /* Ensure we don't close file until we're done freeing locks! */
4141         p->ctx = get_nfs_open_context(ctx);
4142         memcpy(&p->fl, fl, sizeof(p->fl));
4143         p->server = NFS_SERVER(inode);
4144         return p;
4145 }
4146
4147 static void nfs4_locku_release_calldata(void *data)
4148 {
4149         struct nfs4_unlockdata *calldata = data;
4150         nfs_free_seqid(calldata->arg.seqid);
4151         nfs4_put_lock_state(calldata->lsp);
4152         put_nfs_open_context(calldata->ctx);
4153         kfree(calldata);
4154 }
4155
4156 static void nfs4_locku_done(struct rpc_task *task, void *data)
4157 {
4158         struct nfs4_unlockdata *calldata = data;
4159
4160         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4161                 return;
4162         switch (task->tk_status) {
4163                 case 0:
4164                         memcpy(calldata->lsp->ls_stateid.data,
4165                                         calldata->res.stateid.data,
4166                                         sizeof(calldata->lsp->ls_stateid.data));
4167                         renew_lease(calldata->server, calldata->timestamp);
4168                         break;
4169                 case -NFS4ERR_BAD_STATEID:
4170                 case -NFS4ERR_OLD_STATEID:
4171                 case -NFS4ERR_STALE_STATEID:
4172                 case -NFS4ERR_EXPIRED:
4173                         break;
4174                 default:
4175                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4176                                 rpc_restart_call_prepare(task);
4177         }
4178 }
4179
4180 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4181 {
4182         struct nfs4_unlockdata *calldata = data;
4183
4184         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4185                 return;
4186         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4187                 /* Note: exit _without_ running nfs4_locku_done */
4188                 task->tk_action = NULL;
4189                 return;
4190         }
4191         calldata->timestamp = jiffies;
4192         if (nfs4_setup_sequence(calldata->server,
4193                                 &calldata->arg.seq_args,
4194                                 &calldata->res.seq_res, task))
4195                 return;
4196         rpc_call_start(task);
4197 }
4198
4199 static const struct rpc_call_ops nfs4_locku_ops = {
4200         .rpc_call_prepare = nfs4_locku_prepare,
4201         .rpc_call_done = nfs4_locku_done,
4202         .rpc_release = nfs4_locku_release_calldata,
4203 };
4204
4205 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4206                 struct nfs_open_context *ctx,
4207                 struct nfs4_lock_state *lsp,
4208                 struct nfs_seqid *seqid)
4209 {
4210         struct nfs4_unlockdata *data;
4211         struct rpc_message msg = {
4212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4213                 .rpc_cred = ctx->cred,
4214         };
4215         struct rpc_task_setup task_setup_data = {
4216                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4217                 .rpc_message = &msg,
4218                 .callback_ops = &nfs4_locku_ops,
4219                 .workqueue = nfsiod_workqueue,
4220                 .flags = RPC_TASK_ASYNC,
4221         };
4222
4223         /* Ensure this is an unlock - when canceling a lock, the
4224          * canceled lock is passed in, and it won't be an unlock.
4225          */
4226         fl->fl_type = F_UNLCK;
4227
4228         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4229         if (data == NULL) {
4230                 nfs_free_seqid(seqid);
4231                 return ERR_PTR(-ENOMEM);
4232         }
4233
4234         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4235         msg.rpc_argp = &data->arg;
4236         msg.rpc_resp = &data->res;
4237         task_setup_data.callback_data = data;
4238         return rpc_run_task(&task_setup_data);
4239 }
4240
4241 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4242 {
4243         struct nfs_inode *nfsi = NFS_I(state->inode);
4244         struct nfs_seqid *seqid;
4245         struct nfs4_lock_state *lsp;
4246         struct rpc_task *task;
4247         int status = 0;
4248         unsigned char fl_flags = request->fl_flags;
4249
4250         status = nfs4_set_lock_state(state, request);
4251         /* Unlock _before_ we do the RPC call */
4252         request->fl_flags |= FL_EXISTS;
4253         down_read(&nfsi->rwsem);
4254         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4255                 up_read(&nfsi->rwsem);
4256                 goto out;
4257         }
4258         up_read(&nfsi->rwsem);
4259         if (status != 0)
4260                 goto out;
4261         /* Is this a delegated lock? */
4262         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4263                 goto out;
4264         lsp = request->fl_u.nfs4_fl.owner;
4265         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4266         status = -ENOMEM;
4267         if (seqid == NULL)
4268                 goto out;
4269         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4270         status = PTR_ERR(task);
4271         if (IS_ERR(task))
4272                 goto out;
4273         status = nfs4_wait_for_completion_rpc_task(task);
4274         rpc_put_task(task);
4275 out:
4276         request->fl_flags = fl_flags;
4277         return status;
4278 }
4279
4280 struct nfs4_lockdata {
4281         struct nfs_lock_args arg;
4282         struct nfs_lock_res res;
4283         struct nfs4_lock_state *lsp;
4284         struct nfs_open_context *ctx;
4285         struct file_lock fl;
4286         unsigned long timestamp;
4287         int rpc_status;
4288         int cancelled;
4289         struct nfs_server *server;
4290 };
4291
4292 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4293                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4294                 gfp_t gfp_mask)
4295 {
4296         struct nfs4_lockdata *p;
4297         struct inode *inode = lsp->ls_state->inode;
4298         struct nfs_server *server = NFS_SERVER(inode);
4299
4300         p = kzalloc(sizeof(*p), gfp_mask);
4301         if (p == NULL)
4302                 return NULL;
4303
4304         p->arg.fh = NFS_FH(inode);
4305         p->arg.fl = &p->fl;
4306         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4307         if (p->arg.open_seqid == NULL)
4308                 goto out_free;
4309         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4310         if (p->arg.lock_seqid == NULL)
4311                 goto out_free_seqid;
4312         p->arg.lock_stateid = &lsp->ls_stateid;
4313         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4314         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
4315         p->arg.lock_owner.s_dev = server->s_dev;
4316         p->res.lock_seqid = p->arg.lock_seqid;
4317         p->lsp = lsp;
4318         p->server = server;
4319         atomic_inc(&lsp->ls_count);
4320         p->ctx = get_nfs_open_context(ctx);
4321         memcpy(&p->fl, fl, sizeof(p->fl));
4322         return p;
4323 out_free_seqid:
4324         nfs_free_seqid(p->arg.open_seqid);
4325 out_free:
4326         kfree(p);
4327         return NULL;
4328 }
4329
4330 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4331 {
4332         struct nfs4_lockdata *data = calldata;
4333         struct nfs4_state *state = data->lsp->ls_state;
4334
4335         dprintk("%s: begin!\n", __func__);
4336         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4337                 return;
4338         /* Do we need to do an open_to_lock_owner? */
4339         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4340                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4341                         return;
4342                 data->arg.open_stateid = &state->stateid;
4343                 data->arg.new_lock_owner = 1;
4344                 data->res.open_seqid = data->arg.open_seqid;
4345         } else
4346                 data->arg.new_lock_owner = 0;
4347         data->timestamp = jiffies;
4348         if (nfs4_setup_sequence(data->server,
4349                                 &data->arg.seq_args,
4350                                 &data->res.seq_res, task))
4351                 return;
4352         rpc_call_start(task);
4353         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4354 }
4355
4356 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4357 {
4358         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4359         nfs4_lock_prepare(task, calldata);
4360 }
4361
4362 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4363 {
4364         struct nfs4_lockdata *data = calldata;
4365
4366         dprintk("%s: begin!\n", __func__);
4367
4368         if (!nfs4_sequence_done(task, &data->res.seq_res))
4369                 return;
4370
4371         data->rpc_status = task->tk_status;
4372         if (data->arg.new_lock_owner != 0) {
4373                 if (data->rpc_status == 0)
4374                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4375                 else
4376                         goto out;
4377         }
4378         if (data->rpc_status == 0) {
4379                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4380                                         sizeof(data->lsp->ls_stateid.data));
4381                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4382                 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp);
4383         }
4384 out:
4385         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4386 }
4387
4388 static void nfs4_lock_release(void *calldata)
4389 {
4390         struct nfs4_lockdata *data = calldata;
4391
4392         dprintk("%s: begin!\n", __func__);
4393         nfs_free_seqid(data->arg.open_seqid);
4394         if (data->cancelled != 0) {
4395                 struct rpc_task *task;
4396                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4397                                 data->arg.lock_seqid);
4398                 if (!IS_ERR(task))
4399                         rpc_put_task_async(task);
4400                 dprintk("%s: cancelling lock!\n", __func__);
4401         } else
4402                 nfs_free_seqid(data->arg.lock_seqid);
4403         nfs4_put_lock_state(data->lsp);
4404         put_nfs_open_context(data->ctx);
4405         kfree(data);
4406         dprintk("%s: done!\n", __func__);
4407 }
4408
4409 static const struct rpc_call_ops nfs4_lock_ops = {
4410         .rpc_call_prepare = nfs4_lock_prepare,
4411         .rpc_call_done = nfs4_lock_done,
4412         .rpc_release = nfs4_lock_release,
4413 };
4414
4415 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4416         .rpc_call_prepare = nfs4_recover_lock_prepare,
4417         .rpc_call_done = nfs4_lock_done,
4418         .rpc_release = nfs4_lock_release,
4419 };
4420
4421 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4422 {
4423         switch (error) {
4424         case -NFS4ERR_ADMIN_REVOKED:
4425         case -NFS4ERR_BAD_STATEID:
4426                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4427                 if (new_lock_owner != 0 ||
4428                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4429                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4430                 break;
4431         case -NFS4ERR_STALE_STATEID:
4432                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4433         case -NFS4ERR_EXPIRED:
4434                 nfs4_schedule_lease_recovery(server->nfs_client);
4435         };
4436 }
4437
4438 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4439 {
4440         struct nfs4_lockdata *data;
4441         struct rpc_task *task;
4442         struct rpc_message msg = {
4443                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4444                 .rpc_cred = state->owner->so_cred,
4445         };
4446         struct rpc_task_setup task_setup_data = {
4447                 .rpc_client = NFS_CLIENT(state->inode),
4448                 .rpc_message = &msg,
4449                 .callback_ops = &nfs4_lock_ops,
4450                 .workqueue = nfsiod_workqueue,
4451                 .flags = RPC_TASK_ASYNC,
4452         };
4453         int ret;
4454
4455         dprintk("%s: begin!\n", __func__);
4456         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4457                         fl->fl_u.nfs4_fl.owner,
4458                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4459         if (data == NULL)
4460                 return -ENOMEM;
4461         if (IS_SETLKW(cmd))
4462                 data->arg.block = 1;
4463         if (recovery_type > NFS_LOCK_NEW) {
4464                 if (recovery_type == NFS_LOCK_RECLAIM)
4465                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4466                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4467         }
4468         nfs41_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
4469         msg.rpc_argp = &data->arg;
4470         msg.rpc_resp = &data->res;
4471         task_setup_data.callback_data = data;
4472         task = rpc_run_task(&task_setup_data);
4473         if (IS_ERR(task))
4474                 return PTR_ERR(task);
4475         ret = nfs4_wait_for_completion_rpc_task(task);
4476         if (ret == 0) {
4477                 ret = data->rpc_status;
4478                 if (ret)
4479                         nfs4_handle_setlk_error(data->server, data->lsp,
4480                                         data->arg.new_lock_owner, ret);
4481         } else
4482                 data->cancelled = 1;
4483         rpc_put_task(task);
4484         dprintk("%s: done, ret = %d!\n", __func__, ret);
4485         return ret;
4486 }
4487
4488 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4489 {
4490         struct nfs_server *server = NFS_SERVER(state->inode);
4491         struct nfs4_exception exception = { };
4492         int err;
4493
4494         do {
4495                 /* Cache the lock if possible... */
4496                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4497                         return 0;
4498                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4499                 if (err != -NFS4ERR_DELAY)
4500                         break;
4501                 nfs4_handle_exception(server, err, &exception);
4502         } while (exception.retry);
4503         return err;
4504 }
4505
4506 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4507 {
4508         struct nfs_server *server = NFS_SERVER(state->inode);
4509         struct nfs4_exception exception = { };
4510         int err;
4511
4512         err = nfs4_set_lock_state(state, request);
4513         if (err != 0)
4514                 return err;
4515         do {
4516                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4517                         return 0;
4518                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4519                 switch (err) {
4520                 default:
4521                         goto out;
4522                 case -NFS4ERR_GRACE:
4523                 case -NFS4ERR_DELAY:
4524                         nfs4_handle_exception(server, err, &exception);
4525                         err = 0;
4526                 }
4527         } while (exception.retry);
4528 out:
4529         return err;
4530 }
4531
4532 #if defined(CONFIG_NFS_V4_1)
4533 static int nfs41_check_expired_locks(struct nfs4_state *state)
4534 {
4535         int status, ret = NFS_OK;
4536         struct nfs4_lock_state *lsp;
4537         struct nfs_server *server = NFS_SERVER(state->inode);
4538
4539         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
4540                 if (lsp->ls_flags & NFS_LOCK_INITIALIZED) {
4541                         status = nfs41_test_stateid(server, &lsp->ls_stateid);
4542                         if (status != NFS_OK) {
4543                                 nfs41_free_stateid(server, &lsp->ls_stateid);
4544                                 lsp->ls_flags &= ~NFS_LOCK_INITIALIZED;
4545                                 ret = status;
4546                         }
4547                 }
4548         };
4549
4550         return ret;
4551 }
4552
4553 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
4554 {
4555         int status = NFS_OK;
4556
4557         if (test_bit(LK_STATE_IN_USE, &state->flags))
4558                 status = nfs41_check_expired_locks(state);
4559         if (status == NFS_OK)
4560                 return status;
4561         return nfs4_lock_expired(state, request);
4562 }
4563 #endif
4564
4565 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4566 {
4567         struct nfs_inode *nfsi = NFS_I(state->inode);
4568         unsigned char fl_flags = request->fl_flags;
4569         int status = -ENOLCK;
4570
4571         if ((fl_flags & FL_POSIX) &&
4572                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4573                 goto out;
4574         /* Is this a delegated open? */
4575         status = nfs4_set_lock_state(state, request);
4576         if (status != 0)
4577                 goto out;
4578         request->fl_flags |= FL_ACCESS;
4579         status = do_vfs_lock(request->fl_file, request);
4580         if (status < 0)
4581                 goto out;
4582         down_read(&nfsi->rwsem);
4583         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4584                 /* Yes: cache locks! */
4585                 /* ...but avoid races with delegation recall... */
4586                 request->fl_flags = fl_flags & ~FL_SLEEP;
4587                 status = do_vfs_lock(request->fl_file, request);
4588                 goto out_unlock;
4589         }
4590         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4591         if (status != 0)
4592                 goto out_unlock;
4593         /* Note: we always want to sleep here! */
4594         request->fl_flags = fl_flags | FL_SLEEP;
4595         if (do_vfs_lock(request->fl_file, request) < 0)
4596                 printk(KERN_WARNING "NFS: %s: VFS is out of sync with lock "
4597                         "manager!\n", __func__);
4598 out_unlock:
4599         up_read(&nfsi->rwsem);
4600 out:
4601         request->fl_flags = fl_flags;
4602         return status;
4603 }
4604
4605 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4606 {
4607         struct nfs4_exception exception = { };
4608         int err;
4609
4610         do {
4611                 err = _nfs4_proc_setlk(state, cmd, request);
4612                 if (err == -NFS4ERR_DENIED)
4613                         err = -EAGAIN;
4614                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4615                                 err, &exception);
4616         } while (exception.retry);
4617         return err;
4618 }
4619
4620 static int
4621 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4622 {
4623         struct nfs_open_context *ctx;
4624         struct nfs4_state *state;
4625         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4626         int status;
4627
4628         /* verify open state */
4629         ctx = nfs_file_open_context(filp);
4630         state = ctx->state;
4631
4632         if (request->fl_start < 0 || request->fl_end < 0)
4633                 return -EINVAL;
4634
4635         if (IS_GETLK(cmd)) {
4636                 if (state != NULL)
4637                         return nfs4_proc_getlk(state, F_GETLK, request);
4638                 return 0;
4639         }
4640
4641         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4642                 return -EINVAL;
4643
4644         if (request->fl_type == F_UNLCK) {
4645                 if (state != NULL)
4646                         return nfs4_proc_unlck(state, cmd, request);
4647                 return 0;
4648         }
4649
4650         if (state == NULL)
4651                 return -ENOLCK;
4652         do {
4653                 status = nfs4_proc_setlk(state, cmd, request);
4654                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4655                         break;
4656                 timeout = nfs4_set_lock_task_retry(timeout);
4657                 status = -ERESTARTSYS;
4658                 if (signalled())
4659                         break;
4660         } while(status < 0);
4661         return status;
4662 }
4663
4664 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4665 {
4666         struct nfs_server *server = NFS_SERVER(state->inode);
4667         struct nfs4_exception exception = { };
4668         int err;
4669
4670         err = nfs4_set_lock_state(state, fl);
4671         if (err != 0)
4672                 goto out;
4673         do {
4674                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4675                 switch (err) {
4676                         default:
4677                                 printk(KERN_ERR "NFS: %s: unhandled error "
4678                                         "%d.\n", __func__, err);
4679                         case 0:
4680                         case -ESTALE:
4681                                 goto out;
4682                         case -NFS4ERR_EXPIRED:
4683                                 nfs4_schedule_stateid_recovery(server, state);
4684                         case -NFS4ERR_STALE_CLIENTID:
4685                         case -NFS4ERR_STALE_STATEID:
4686                                 nfs4_schedule_lease_recovery(server->nfs_client);
4687                                 goto out;
4688                         case -NFS4ERR_BADSESSION:
4689                         case -NFS4ERR_BADSLOT:
4690                         case -NFS4ERR_BAD_HIGH_SLOT:
4691                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4692                         case -NFS4ERR_DEADSESSION:
4693                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4694                                 goto out;
4695                         case -ERESTARTSYS:
4696                                 /*
4697                                  * The show must go on: exit, but mark the
4698                                  * stateid as needing recovery.
4699                                  */
4700                         case -NFS4ERR_ADMIN_REVOKED:
4701                         case -NFS4ERR_BAD_STATEID:
4702                         case -NFS4ERR_OPENMODE:
4703                                 nfs4_schedule_stateid_recovery(server, state);
4704                                 err = 0;
4705                                 goto out;
4706                         case -EKEYEXPIRED:
4707                                 /*
4708                                  * User RPCSEC_GSS context has expired.
4709                                  * We cannot recover this stateid now, so
4710                                  * skip it and allow recovery thread to
4711                                  * proceed.
4712                                  */
4713                                 err = 0;
4714                                 goto out;
4715                         case -ENOMEM:
4716                         case -NFS4ERR_DENIED:
4717                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4718                                 err = 0;
4719                                 goto out;
4720                         case -NFS4ERR_DELAY:
4721                                 break;
4722                 }
4723                 err = nfs4_handle_exception(server, err, &exception);
4724         } while (exception.retry);
4725 out:
4726         return err;
4727 }
4728
4729 static void nfs4_release_lockowner_release(void *calldata)
4730 {
4731         kfree(calldata);
4732 }
4733
4734 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4735         .rpc_release = nfs4_release_lockowner_release,
4736 };
4737
4738 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4739 {
4740         struct nfs_server *server = lsp->ls_state->owner->so_server;
4741         struct nfs_release_lockowner_args *args;
4742         struct rpc_message msg = {
4743                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4744         };
4745
4746         if (server->nfs_client->cl_mvops->minor_version != 0)
4747                 return;
4748         args = kmalloc(sizeof(*args), GFP_NOFS);
4749         if (!args)
4750                 return;
4751         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4752         args->lock_owner.id = lsp->ls_seqid.owner_id;
4753         args->lock_owner.s_dev = server->s_dev;
4754         msg.rpc_argp = args;
4755         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4756 }
4757
4758 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4759
4760 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4761                                    const void *buf, size_t buflen,
4762                                    int flags, int type)
4763 {
4764         if (strcmp(key, "") != 0)
4765                 return -EINVAL;
4766
4767         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4768 }
4769
4770 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4771                                    void *buf, size_t buflen, int type)
4772 {
4773         if (strcmp(key, "") != 0)
4774                 return -EINVAL;
4775
4776         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4777 }
4778
4779 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4780                                        size_t list_len, const char *name,
4781                                        size_t name_len, int type)
4782 {
4783         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4784
4785         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4786                 return 0;
4787
4788         if (list && len <= list_len)
4789                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4790         return len;
4791 }
4792
4793 /*
4794  * nfs_fhget will use either the mounted_on_fileid or the fileid
4795  */
4796 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4797 {
4798         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
4799                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
4800               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4801               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
4802                 return;
4803
4804         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4805                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
4806         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4807         fattr->nlink = 2;
4808 }
4809
4810 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4811                 struct nfs4_fs_locations *fs_locations, struct page *page)
4812 {
4813         struct nfs_server *server = NFS_SERVER(dir);
4814         u32 bitmask[2] = {
4815                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4816         };
4817         struct nfs4_fs_locations_arg args = {
4818                 .dir_fh = NFS_FH(dir),
4819                 .name = name,
4820                 .page = page,
4821                 .bitmask = bitmask,
4822         };
4823         struct nfs4_fs_locations_res res = {
4824                 .fs_locations = fs_locations,
4825         };
4826         struct rpc_message msg = {
4827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4828                 .rpc_argp = &args,
4829                 .rpc_resp = &res,
4830         };
4831         int status;
4832
4833         dprintk("%s: start\n", __func__);
4834
4835         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
4836          * is not supported */
4837         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
4838                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
4839         else
4840                 bitmask[0] |= FATTR4_WORD0_FILEID;
4841
4842         nfs_fattr_init(&fs_locations->fattr);
4843         fs_locations->server = server;
4844         fs_locations->nlocations = 0;
4845         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4846         dprintk("%s: returned status = %d\n", __func__, status);
4847         return status;
4848 }
4849
4850 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4851 {
4852         int status;
4853         struct nfs4_secinfo_arg args = {
4854                 .dir_fh = NFS_FH(dir),
4855                 .name   = name,
4856         };
4857         struct nfs4_secinfo_res res = {
4858                 .flavors     = flavors,
4859         };
4860         struct rpc_message msg = {
4861                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4862                 .rpc_argp = &args,
4863                 .rpc_resp = &res,
4864         };
4865
4866         dprintk("NFS call  secinfo %s\n", name->name);
4867         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4868         dprintk("NFS reply  secinfo: %d\n", status);
4869         return status;
4870 }
4871
4872 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4873 {
4874         struct nfs4_exception exception = { };
4875         int err;
4876         do {
4877                 err = nfs4_handle_exception(NFS_SERVER(dir),
4878                                 _nfs4_proc_secinfo(dir, name, flavors),
4879                                 &exception);
4880         } while (exception.retry);
4881         return err;
4882 }
4883
4884 #ifdef CONFIG_NFS_V4_1
4885 /*
4886  * Check the exchange flags returned by the server for invalid flags, having
4887  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4888  * DS flags set.
4889  */
4890 static int nfs4_check_cl_exchange_flags(u32 flags)
4891 {
4892         if (flags & ~EXCHGID4_FLAG_MASK_R)
4893                 goto out_inval;
4894         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4895             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4896                 goto out_inval;
4897         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4898                 goto out_inval;
4899         return NFS_OK;
4900 out_inval:
4901         return -NFS4ERR_INVAL;
4902 }
4903
4904 static bool
4905 nfs41_same_server_scope(struct server_scope *a, struct server_scope *b)
4906 {
4907         if (a->server_scope_sz == b->server_scope_sz &&
4908             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
4909                 return true;
4910
4911         return false;
4912 }
4913
4914 /*
4915  * nfs4_proc_exchange_id()
4916  *
4917  * Since the clientid has expired, all compounds using sessions
4918  * associated with the stale clientid will be returning
4919  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4920  * be in some phase of session reset.
4921  */
4922 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4923 {
4924         nfs4_verifier verifier;
4925         struct nfs41_exchange_id_args args = {
4926                 .client = clp,
4927                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4928         };
4929         struct nfs41_exchange_id_res res = {
4930                 .client = clp,
4931         };
4932         int status;
4933         struct rpc_message msg = {
4934                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4935                 .rpc_argp = &args,
4936                 .rpc_resp = &res,
4937                 .rpc_cred = cred,
4938         };
4939         __be32 *p;
4940
4941         dprintk("--> %s\n", __func__);
4942         BUG_ON(clp == NULL);
4943
4944         p = (u32 *)verifier.data;
4945         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4946         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4947         args.verifier = &verifier;
4948
4949         args.id_len = scnprintf(args.id, sizeof(args.id),
4950                                 "%s/%s.%s/%u",
4951                                 clp->cl_ipaddr,
4952                                 init_utsname()->nodename,
4953                                 init_utsname()->domainname,
4954                                 clp->cl_rpcclient->cl_auth->au_flavor);
4955
4956         res.server_scope = kzalloc(sizeof(struct server_scope), GFP_KERNEL);
4957         if (unlikely(!res.server_scope)) {
4958                 status = -ENOMEM;
4959                 goto out;
4960         }
4961
4962         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_KERNEL);
4963         if (unlikely(!res.impl_id)) {
4964                 status = -ENOMEM;
4965                 goto out_server_scope;
4966         }
4967
4968         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4969         if (!status)
4970                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4971
4972         if (!status) {
4973                 /* use the most recent implementation id */
4974                 kfree(clp->impl_id);
4975                 clp->impl_id = res.impl_id;
4976         } else
4977                 kfree(res.impl_id);
4978
4979         if (!status) {
4980                 if (clp->server_scope &&
4981                     !nfs41_same_server_scope(clp->server_scope,
4982                                              res.server_scope)) {
4983                         dprintk("%s: server_scope mismatch detected\n",
4984                                 __func__);
4985                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
4986                         kfree(clp->server_scope);
4987                         clp->server_scope = NULL;
4988                 }
4989
4990                 if (!clp->server_scope) {
4991                         clp->server_scope = res.server_scope;
4992                         goto out;
4993                 }
4994         }
4995
4996 out_server_scope:
4997         kfree(res.server_scope);
4998 out:
4999         if (clp->impl_id)
5000                 dprintk("%s: Server Implementation ID: "
5001                         "domain: %s, name: %s, date: %llu,%u\n",
5002                         __func__, clp->impl_id->domain, clp->impl_id->name,
5003                         clp->impl_id->date.seconds,
5004                         clp->impl_id->date.nseconds);
5005         dprintk("<-- %s status= %d\n", __func__, status);
5006         return status;
5007 }
5008
5009 struct nfs4_get_lease_time_data {
5010         struct nfs4_get_lease_time_args *args;
5011         struct nfs4_get_lease_time_res *res;
5012         struct nfs_client *clp;
5013 };
5014
5015 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
5016                                         void *calldata)
5017 {
5018         int ret;
5019         struct nfs4_get_lease_time_data *data =
5020                         (struct nfs4_get_lease_time_data *)calldata;
5021
5022         dprintk("--> %s\n", __func__);
5023         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5024         /* just setup sequence, do not trigger session recovery
5025            since we're invoked within one */
5026         ret = nfs41_setup_sequence(data->clp->cl_session,
5027                                    &data->args->la_seq_args,
5028                                    &data->res->lr_seq_res, task);
5029
5030         BUG_ON(ret == -EAGAIN);
5031         rpc_call_start(task);
5032         dprintk("<-- %s\n", __func__);
5033 }
5034
5035 /*
5036  * Called from nfs4_state_manager thread for session setup, so don't recover
5037  * from sequence operation or clientid errors.
5038  */
5039 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
5040 {
5041         struct nfs4_get_lease_time_data *data =
5042                         (struct nfs4_get_lease_time_data *)calldata;
5043
5044         dprintk("--> %s\n", __func__);
5045         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
5046                 return;
5047         switch (task->tk_status) {
5048         case -NFS4ERR_DELAY:
5049         case -NFS4ERR_GRACE:
5050                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
5051                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
5052                 task->tk_status = 0;
5053                 /* fall through */
5054         case -NFS4ERR_RETRY_UNCACHED_REP:
5055                 rpc_restart_call_prepare(task);
5056                 return;
5057         }
5058         dprintk("<-- %s\n", __func__);
5059 }
5060
5061 struct rpc_call_ops nfs4_get_lease_time_ops = {
5062         .rpc_call_prepare = nfs4_get_lease_time_prepare,
5063         .rpc_call_done = nfs4_get_lease_time_done,
5064 };
5065
5066 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
5067 {
5068         struct rpc_task *task;
5069         struct nfs4_get_lease_time_args args;
5070         struct nfs4_get_lease_time_res res = {
5071                 .lr_fsinfo = fsinfo,
5072         };
5073         struct nfs4_get_lease_time_data data = {
5074                 .args = &args,
5075                 .res = &res,
5076                 .clp = clp,
5077         };
5078         struct rpc_message msg = {
5079                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
5080                 .rpc_argp = &args,
5081                 .rpc_resp = &res,
5082         };
5083         struct rpc_task_setup task_setup = {
5084                 .rpc_client = clp->cl_rpcclient,
5085                 .rpc_message = &msg,
5086                 .callback_ops = &nfs4_get_lease_time_ops,
5087                 .callback_data = &data,
5088                 .flags = RPC_TASK_TIMEOUT,
5089         };
5090         int status;
5091
5092         nfs41_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
5093         dprintk("--> %s\n", __func__);
5094         task = rpc_run_task(&task_setup);
5095
5096         if (IS_ERR(task))
5097                 status = PTR_ERR(task);
5098         else {
5099                 status = task->tk_status;
5100                 rpc_put_task(task);
5101         }
5102         dprintk("<-- %s return %d\n", __func__, status);
5103
5104         return status;
5105 }
5106
5107 static struct nfs4_slot *nfs4_alloc_slots(u32 max_slots, gfp_t gfp_flags)
5108 {
5109         return kcalloc(max_slots, sizeof(struct nfs4_slot), gfp_flags);
5110 }
5111
5112 static void nfs4_add_and_init_slots(struct nfs4_slot_table *tbl,
5113                 struct nfs4_slot *new,
5114                 u32 max_slots,
5115                 u32 ivalue)
5116 {
5117         struct nfs4_slot *old = NULL;
5118         u32 i;
5119
5120         spin_lock(&tbl->slot_tbl_lock);
5121         if (new) {
5122                 old = tbl->slots;
5123                 tbl->slots = new;
5124                 tbl->max_slots = max_slots;
5125         }
5126         tbl->highest_used_slotid = -1;  /* no slot is currently used */
5127         for (i = 0; i < tbl->max_slots; i++)
5128                 tbl->slots[i].seq_nr = ivalue;
5129         spin_unlock(&tbl->slot_tbl_lock);
5130         kfree(old);
5131 }
5132
5133 /*
5134  * (re)Initialise a slot table
5135  */
5136 static int nfs4_realloc_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
5137                                  u32 ivalue)
5138 {
5139         struct nfs4_slot *new = NULL;
5140         int ret = -ENOMEM;
5141
5142         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
5143                 max_reqs, tbl->max_slots);
5144
5145         /* Does the newly negotiated max_reqs match the existing slot table? */
5146         if (max_reqs != tbl->max_slots) {
5147                 new = nfs4_alloc_slots(max_reqs, GFP_NOFS);
5148                 if (!new)
5149                         goto out;
5150         }
5151         ret = 0;
5152
5153         nfs4_add_and_init_slots(tbl, new, max_reqs, ivalue);
5154         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
5155                 tbl, tbl->slots, tbl->max_slots);
5156 out:
5157         dprintk("<-- %s: return %d\n", __func__, ret);
5158         return ret;
5159 }
5160
5161 /* Destroy the slot table */
5162 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
5163 {
5164         if (session->fc_slot_table.slots != NULL) {
5165                 kfree(session->fc_slot_table.slots);
5166                 session->fc_slot_table.slots = NULL;
5167         }
5168         if (session->bc_slot_table.slots != NULL) {
5169                 kfree(session->bc_slot_table.slots);
5170                 session->bc_slot_table.slots = NULL;
5171         }
5172         return;
5173 }
5174
5175 /*
5176  * Initialize or reset the forechannel and backchannel tables
5177  */
5178 static int nfs4_setup_session_slot_tables(struct nfs4_session *ses)
5179 {
5180         struct nfs4_slot_table *tbl;
5181         int status;
5182
5183         dprintk("--> %s\n", __func__);
5184         /* Fore channel */
5185         tbl = &ses->fc_slot_table;
5186         status = nfs4_realloc_slot_table(tbl, ses->fc_attrs.max_reqs, 1);
5187         if (status) /* -ENOMEM */
5188                 return status;
5189         /* Back channel */
5190         tbl = &ses->bc_slot_table;
5191         status = nfs4_realloc_slot_table(tbl, ses->bc_attrs.max_reqs, 0);
5192         if (status && tbl->slots == NULL)
5193                 /* Fore and back channel share a connection so get
5194                  * both slot tables or neither */
5195                 nfs4_destroy_slot_tables(ses);
5196         return status;
5197 }
5198
5199 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5200 {
5201         struct nfs4_session *session;
5202         struct nfs4_slot_table *tbl;
5203
5204         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5205         if (!session)
5206                 return NULL;
5207
5208         tbl = &session->fc_slot_table;
5209         tbl->highest_used_slotid = NFS4_NO_SLOT;
5210         spin_lock_init(&tbl->slot_tbl_lock);
5211         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5212         init_completion(&tbl->complete);
5213
5214         tbl = &session->bc_slot_table;
5215         tbl->highest_used_slotid = NFS4_NO_SLOT;
5216         spin_lock_init(&tbl->slot_tbl_lock);
5217         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5218         init_completion(&tbl->complete);
5219
5220         session->session_state = 1<<NFS4_SESSION_INITING;
5221
5222         session->clp = clp;
5223         return session;
5224 }
5225
5226 void nfs4_destroy_session(struct nfs4_session *session)
5227 {
5228         struct rpc_xprt *xprt;
5229
5230         nfs4_proc_destroy_session(session);
5231
5232         rcu_read_lock();
5233         xprt = rcu_dereference(session->clp->cl_rpcclient->cl_xprt);
5234         rcu_read_unlock();
5235         dprintk("%s Destroy backchannel for xprt %p\n",
5236                 __func__, xprt);
5237         xprt_destroy_backchannel(xprt, NFS41_BC_MIN_CALLBACKS);
5238         nfs4_destroy_slot_tables(session);
5239         kfree(session);
5240 }
5241
5242 /*
5243  * Initialize the values to be used by the client in CREATE_SESSION
5244  * If nfs4_init_session set the fore channel request and response sizes,
5245  * use them.
5246  *
5247  * Set the back channel max_resp_sz_cached to zero to force the client to
5248  * always set csa_cachethis to FALSE because the current implementation
5249  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5250  */
5251 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5252 {
5253         struct nfs4_session *session = args->client->cl_session;
5254         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5255                      mxresp_sz = session->fc_attrs.max_resp_sz;
5256
5257         if (mxrqst_sz == 0)
5258                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5259         if (mxresp_sz == 0)
5260                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5261         /* Fore channel attributes */
5262         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5263         args->fc_attrs.max_resp_sz = mxresp_sz;
5264         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5265         args->fc_attrs.max_reqs = max_session_slots;
5266
5267         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5268                 "max_ops=%u max_reqs=%u\n",
5269                 __func__,
5270                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5271                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5272
5273         /* Back channel attributes */
5274         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5275         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5276         args->bc_attrs.max_resp_sz_cached = 0;
5277         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5278         args->bc_attrs.max_reqs = 1;
5279
5280         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5281                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5282                 __func__,
5283                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5284                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5285                 args->bc_attrs.max_reqs);
5286 }
5287
5288 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5289 {
5290         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5291         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5292
5293         if (rcvd->max_resp_sz > sent->max_resp_sz)
5294                 return -EINVAL;
5295         /*
5296          * Our requested max_ops is the minimum we need; we're not
5297          * prepared to break up compounds into smaller pieces than that.
5298          * So, no point even trying to continue if the server won't
5299          * cooperate:
5300          */
5301         if (rcvd->max_ops < sent->max_ops)
5302                 return -EINVAL;
5303         if (rcvd->max_reqs == 0)
5304                 return -EINVAL;
5305         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
5306                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
5307         return 0;
5308 }
5309
5310 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5311 {
5312         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5313         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5314
5315         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5316                 return -EINVAL;
5317         if (rcvd->max_resp_sz < sent->max_resp_sz)
5318                 return -EINVAL;
5319         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5320                 return -EINVAL;
5321         /* These would render the backchannel useless: */
5322         if (rcvd->max_ops != sent->max_ops)
5323                 return -EINVAL;
5324         if (rcvd->max_reqs != sent->max_reqs)
5325                 return -EINVAL;
5326         return 0;
5327 }
5328
5329 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5330                                      struct nfs4_session *session)
5331 {
5332         int ret;
5333
5334         ret = nfs4_verify_fore_channel_attrs(args, session);
5335         if (ret)
5336                 return ret;
5337         return nfs4_verify_back_channel_attrs(args, session);
5338 }
5339
5340 static int _nfs4_proc_create_session(struct nfs_client *clp)
5341 {
5342         struct nfs4_session *session = clp->cl_session;
5343         struct nfs41_create_session_args args = {
5344                 .client = clp,
5345                 .cb_program = NFS4_CALLBACK,
5346         };
5347         struct nfs41_create_session_res res = {
5348                 .client = clp,
5349         };
5350         struct rpc_message msg = {
5351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5352                 .rpc_argp = &args,
5353                 .rpc_resp = &res,
5354         };
5355         int status;
5356
5357         nfs4_init_channel_attrs(&args);
5358         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5359
5360         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5361
5362         if (!status)
5363                 /* Verify the session's negotiated channel_attrs values */
5364                 status = nfs4_verify_channel_attrs(&args, session);
5365         if (!status) {
5366                 /* Increment the clientid slot sequence id */
5367                 clp->cl_seqid++;
5368         }
5369
5370         return status;
5371 }
5372
5373 /*
5374  * Issues a CREATE_SESSION operation to the server.
5375  * It is the responsibility of the caller to verify the session is
5376  * expired before calling this routine.
5377  */
5378 int nfs4_proc_create_session(struct nfs_client *clp)
5379 {
5380         int status;
5381         unsigned *ptr;
5382         struct nfs4_session *session = clp->cl_session;
5383
5384         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5385
5386         status = _nfs4_proc_create_session(clp);
5387         if (status)
5388                 goto out;
5389
5390         /* Init or reset the session slot tables */
5391         status = nfs4_setup_session_slot_tables(session);
5392         dprintk("slot table setup returned %d\n", status);
5393         if (status)
5394                 goto out;
5395
5396         ptr = (unsigned *)&session->sess_id.data[0];
5397         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5398                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5399 out:
5400         dprintk("<-- %s\n", __func__);
5401         return status;
5402 }
5403
5404 /*
5405  * Issue the over-the-wire RPC DESTROY_SESSION.
5406  * The caller must serialize access to this routine.
5407  */
5408 int nfs4_proc_destroy_session(struct nfs4_session *session)
5409 {
5410         int status = 0;
5411         struct rpc_message msg;
5412
5413         dprintk("--> nfs4_proc_destroy_session\n");
5414
5415         /* session is still being setup */
5416         if (session->clp->cl_cons_state != NFS_CS_READY)
5417                 return status;
5418
5419         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5420         msg.rpc_argp = session;
5421         msg.rpc_resp = NULL;
5422         msg.rpc_cred = NULL;
5423         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5424
5425         if (status)
5426                 printk(KERN_WARNING
5427                         "NFS: Got error %d from the server on DESTROY_SESSION. "
5428                         "Session has been destroyed regardless...\n", status);
5429
5430         dprintk("<-- nfs4_proc_destroy_session\n");
5431         return status;
5432 }
5433
5434 int nfs4_init_session(struct nfs_server *server)
5435 {
5436         struct nfs_client *clp = server->nfs_client;
5437         struct nfs4_session *session;
5438         unsigned int rsize, wsize;
5439         int ret;
5440
5441         if (!nfs4_has_session(clp))
5442                 return 0;
5443
5444         session = clp->cl_session;
5445         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5446                 return 0;
5447
5448         rsize = server->rsize;
5449         if (rsize == 0)
5450                 rsize = NFS_MAX_FILE_IO_SIZE;
5451         wsize = server->wsize;
5452         if (wsize == 0)
5453                 wsize = NFS_MAX_FILE_IO_SIZE;
5454
5455         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5456         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5457
5458         ret = nfs4_recover_expired_lease(server);
5459         if (!ret)
5460                 ret = nfs4_check_client_ready(clp);
5461         return ret;
5462 }
5463
5464 int nfs4_init_ds_session(struct nfs_client *clp)
5465 {
5466         struct nfs4_session *session = clp->cl_session;
5467         int ret;
5468
5469         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5470                 return 0;
5471
5472         ret = nfs4_client_recover_expired_lease(clp);
5473         if (!ret)
5474                 /* Test for the DS role */
5475                 if (!is_ds_client(clp))
5476                         ret = -ENODEV;
5477         if (!ret)
5478                 ret = nfs4_check_client_ready(clp);
5479         return ret;
5480
5481 }
5482 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5483
5484
5485 /*
5486  * Renew the cl_session lease.
5487  */
5488 struct nfs4_sequence_data {
5489         struct nfs_client *clp;
5490         struct nfs4_sequence_args args;
5491         struct nfs4_sequence_res res;
5492 };
5493
5494 static void nfs41_sequence_release(void *data)
5495 {
5496         struct nfs4_sequence_data *calldata = data;
5497         struct nfs_client *clp = calldata->clp;
5498
5499         if (atomic_read(&clp->cl_count) > 1)
5500                 nfs4_schedule_state_renewal(clp);
5501         nfs_put_client(clp);
5502         kfree(calldata);
5503 }
5504
5505 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5506 {
5507         switch(task->tk_status) {
5508         case -NFS4ERR_DELAY:
5509                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5510                 return -EAGAIN;
5511         default:
5512                 nfs4_schedule_lease_recovery(clp);
5513         }
5514         return 0;
5515 }
5516
5517 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5518 {
5519         struct nfs4_sequence_data *calldata = data;
5520         struct nfs_client *clp = calldata->clp;
5521
5522         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5523                 return;
5524
5525         if (task->tk_status < 0) {
5526                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5527                 if (atomic_read(&clp->cl_count) == 1)
5528                         goto out;
5529
5530                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5531                         rpc_restart_call_prepare(task);
5532                         return;
5533                 }
5534         }
5535         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5536 out:
5537         dprintk("<-- %s\n", __func__);
5538 }
5539
5540 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5541 {
5542         struct nfs4_sequence_data *calldata = data;
5543         struct nfs_client *clp = calldata->clp;
5544         struct nfs4_sequence_args *args;
5545         struct nfs4_sequence_res *res;
5546
5547         args = task->tk_msg.rpc_argp;
5548         res = task->tk_msg.rpc_resp;
5549
5550         if (nfs41_setup_sequence(clp->cl_session, args, res, task))
5551                 return;
5552         rpc_call_start(task);
5553 }
5554
5555 static const struct rpc_call_ops nfs41_sequence_ops = {
5556         .rpc_call_done = nfs41_sequence_call_done,
5557         .rpc_call_prepare = nfs41_sequence_prepare,
5558         .rpc_release = nfs41_sequence_release,
5559 };
5560
5561 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5562 {
5563         struct nfs4_sequence_data *calldata;
5564         struct rpc_message msg = {
5565                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5566                 .rpc_cred = cred,
5567         };
5568         struct rpc_task_setup task_setup_data = {
5569                 .rpc_client = clp->cl_rpcclient,
5570                 .rpc_message = &msg,
5571                 .callback_ops = &nfs41_sequence_ops,
5572                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5573         };
5574
5575         if (!atomic_inc_not_zero(&clp->cl_count))
5576                 return ERR_PTR(-EIO);
5577         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5578         if (calldata == NULL) {
5579                 nfs_put_client(clp);
5580                 return ERR_PTR(-ENOMEM);
5581         }
5582         nfs41_init_sequence(&calldata->args, &calldata->res, 0);
5583         msg.rpc_argp = &calldata->args;
5584         msg.rpc_resp = &calldata->res;
5585         calldata->clp = clp;
5586         task_setup_data.callback_data = calldata;
5587
5588         return rpc_run_task(&task_setup_data);
5589 }
5590
5591 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
5592 {
5593         struct rpc_task *task;
5594         int ret = 0;
5595
5596         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
5597                 return 0;
5598         task = _nfs41_proc_sequence(clp, cred);
5599         if (IS_ERR(task))
5600                 ret = PTR_ERR(task);
5601         else
5602                 rpc_put_task_async(task);
5603         dprintk("<-- %s status=%d\n", __func__, ret);
5604         return ret;
5605 }
5606
5607 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5608 {
5609         struct rpc_task *task;
5610         int ret;
5611
5612         task = _nfs41_proc_sequence(clp, cred);
5613         if (IS_ERR(task)) {
5614                 ret = PTR_ERR(task);
5615                 goto out;
5616         }
5617         ret = rpc_wait_for_completion_task(task);
5618         if (!ret) {
5619                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5620
5621                 if (task->tk_status == 0)
5622                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5623                 ret = task->tk_status;
5624         }
5625         rpc_put_task(task);
5626 out:
5627         dprintk("<-- %s status=%d\n", __func__, ret);
5628         return ret;
5629 }
5630
5631 struct nfs4_reclaim_complete_data {
5632         struct nfs_client *clp;
5633         struct nfs41_reclaim_complete_args arg;
5634         struct nfs41_reclaim_complete_res res;
5635 };
5636
5637 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5638 {
5639         struct nfs4_reclaim_complete_data *calldata = data;
5640
5641         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5642         if (nfs41_setup_sequence(calldata->clp->cl_session,
5643                                 &calldata->arg.seq_args,
5644                                 &calldata->res.seq_res, task))
5645                 return;
5646
5647         rpc_call_start(task);
5648 }
5649
5650 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5651 {
5652         switch(task->tk_status) {
5653         case 0:
5654         case -NFS4ERR_COMPLETE_ALREADY:
5655         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5656                 break;
5657         case -NFS4ERR_DELAY:
5658                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5659                 /* fall through */
5660         case -NFS4ERR_RETRY_UNCACHED_REP:
5661                 return -EAGAIN;
5662         default:
5663                 nfs4_schedule_lease_recovery(clp);
5664         }
5665         return 0;
5666 }
5667
5668 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5669 {
5670         struct nfs4_reclaim_complete_data *calldata = data;
5671         struct nfs_client *clp = calldata->clp;
5672         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5673
5674         dprintk("--> %s\n", __func__);
5675         if (!nfs41_sequence_done(task, res))
5676                 return;
5677
5678         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5679                 rpc_restart_call_prepare(task);
5680                 return;
5681         }
5682         dprintk("<-- %s\n", __func__);
5683 }
5684
5685 static void nfs4_free_reclaim_complete_data(void *data)
5686 {
5687         struct nfs4_reclaim_complete_data *calldata = data;
5688
5689         kfree(calldata);
5690 }
5691
5692 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5693         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5694         .rpc_call_done = nfs4_reclaim_complete_done,
5695         .rpc_release = nfs4_free_reclaim_complete_data,
5696 };
5697
5698 /*
5699  * Issue a global reclaim complete.
5700  */
5701 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5702 {
5703         struct nfs4_reclaim_complete_data *calldata;
5704         struct rpc_task *task;
5705         struct rpc_message msg = {
5706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5707         };
5708         struct rpc_task_setup task_setup_data = {
5709                 .rpc_client = clp->cl_rpcclient,
5710                 .rpc_message = &msg,
5711                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5712                 .flags = RPC_TASK_ASYNC,
5713         };
5714         int status = -ENOMEM;
5715
5716         dprintk("--> %s\n", __func__);
5717         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5718         if (calldata == NULL)
5719                 goto out;
5720         calldata->clp = clp;
5721         calldata->arg.one_fs = 0;
5722
5723         nfs41_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
5724         msg.rpc_argp = &calldata->arg;
5725         msg.rpc_resp = &calldata->res;
5726         task_setup_data.callback_data = calldata;
5727         task = rpc_run_task(&task_setup_data);
5728         if (IS_ERR(task)) {
5729                 status = PTR_ERR(task);
5730                 goto out;
5731         }
5732         status = nfs4_wait_for_completion_rpc_task(task);
5733         if (status == 0)
5734                 status = task->tk_status;
5735         rpc_put_task(task);
5736         return 0;
5737 out:
5738         dprintk("<-- %s status=%d\n", __func__, status);
5739         return status;
5740 }
5741
5742 static void
5743 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5744 {
5745         struct nfs4_layoutget *lgp = calldata;
5746         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5747
5748         dprintk("--> %s\n", __func__);
5749         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5750          * right now covering the LAYOUTGET we are about to send.
5751          * However, that is not so catastrophic, and there seems
5752          * to be no way to prevent it completely.
5753          */
5754         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5755                                 &lgp->res.seq_res, task))
5756                 return;
5757         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5758                                           NFS_I(lgp->args.inode)->layout,
5759                                           lgp->args.ctx->state)) {
5760                 rpc_exit(task, NFS4_OK);
5761                 return;
5762         }
5763         rpc_call_start(task);
5764 }
5765
5766 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5767 {
5768         struct nfs4_layoutget *lgp = calldata;
5769         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5770
5771         dprintk("--> %s\n", __func__);
5772
5773         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5774                 return;
5775
5776         switch (task->tk_status) {
5777         case 0:
5778                 break;
5779         case -NFS4ERR_LAYOUTTRYLATER:
5780         case -NFS4ERR_RECALLCONFLICT:
5781                 task->tk_status = -NFS4ERR_DELAY;
5782                 /* Fall through */
5783         default:
5784                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5785                         rpc_restart_call_prepare(task);
5786                         return;
5787                 }
5788         }
5789         dprintk("<-- %s\n", __func__);
5790 }
5791
5792 static void nfs4_layoutget_release(void *calldata)
5793 {
5794         struct nfs4_layoutget *lgp = calldata;
5795
5796         dprintk("--> %s\n", __func__);
5797         put_nfs_open_context(lgp->args.ctx);
5798         kfree(calldata);
5799         dprintk("<-- %s\n", __func__);
5800 }
5801
5802 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5803         .rpc_call_prepare = nfs4_layoutget_prepare,
5804         .rpc_call_done = nfs4_layoutget_done,
5805         .rpc_release = nfs4_layoutget_release,
5806 };
5807
5808 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5809 {
5810         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5811         struct rpc_task *task;
5812         struct rpc_message msg = {
5813                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5814                 .rpc_argp = &lgp->args,
5815                 .rpc_resp = &lgp->res,
5816         };
5817         struct rpc_task_setup task_setup_data = {
5818                 .rpc_client = server->client,
5819                 .rpc_message = &msg,
5820                 .callback_ops = &nfs4_layoutget_call_ops,
5821                 .callback_data = lgp,
5822                 .flags = RPC_TASK_ASYNC,
5823         };
5824         int status = 0;
5825
5826         dprintk("--> %s\n", __func__);
5827
5828         lgp->res.layoutp = &lgp->args.layout;
5829         lgp->res.seq_res.sr_slot = NULL;
5830         nfs41_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
5831         task = rpc_run_task(&task_setup_data);
5832         if (IS_ERR(task))
5833                 return PTR_ERR(task);
5834         status = nfs4_wait_for_completion_rpc_task(task);
5835         if (status == 0)
5836                 status = task->tk_status;
5837         if (status == 0)
5838                 status = pnfs_layout_process(lgp);
5839         rpc_put_task(task);
5840         dprintk("<-- %s status=%d\n", __func__, status);
5841         return status;
5842 }
5843
5844 static void
5845 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
5846 {
5847         struct nfs4_layoutreturn *lrp = calldata;
5848
5849         dprintk("--> %s\n", __func__);
5850         if (nfs41_setup_sequence(lrp->clp->cl_session, &lrp->args.seq_args,
5851                                 &lrp->res.seq_res, task))
5852                 return;
5853         rpc_call_start(task);
5854 }
5855
5856 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
5857 {
5858         struct nfs4_layoutreturn *lrp = calldata;
5859         struct nfs_server *server;
5860         struct pnfs_layout_hdr *lo = lrp->args.layout;
5861
5862         dprintk("--> %s\n", __func__);
5863
5864         if (!nfs4_sequence_done(task, &lrp->res.seq_res))
5865                 return;
5866
5867         server = NFS_SERVER(lrp->args.inode);
5868         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5869                 rpc_restart_call_prepare(task);
5870                 return;
5871         }
5872         spin_lock(&lo->plh_inode->i_lock);
5873         if (task->tk_status == 0) {
5874                 if (lrp->res.lrs_present) {
5875                         pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
5876                 } else
5877                         BUG_ON(!list_empty(&lo->plh_segs));
5878         }
5879         lo->plh_block_lgets--;
5880         spin_unlock(&lo->plh_inode->i_lock);
5881         dprintk("<-- %s\n", __func__);
5882 }
5883
5884 static void nfs4_layoutreturn_release(void *calldata)
5885 {
5886         struct nfs4_layoutreturn *lrp = calldata;
5887
5888         dprintk("--> %s\n", __func__);
5889         put_layout_hdr(lrp->args.layout);
5890         kfree(calldata);
5891         dprintk("<-- %s\n", __func__);
5892 }
5893
5894 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
5895         .rpc_call_prepare = nfs4_layoutreturn_prepare,
5896         .rpc_call_done = nfs4_layoutreturn_done,
5897         .rpc_release = nfs4_layoutreturn_release,
5898 };
5899
5900 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp)
5901 {
5902         struct rpc_task *task;
5903         struct rpc_message msg = {
5904                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
5905                 .rpc_argp = &lrp->args,
5906                 .rpc_resp = &lrp->res,
5907         };
5908         struct rpc_task_setup task_setup_data = {
5909                 .rpc_client = lrp->clp->cl_rpcclient,
5910                 .rpc_message = &msg,
5911                 .callback_ops = &nfs4_layoutreturn_call_ops,
5912                 .callback_data = lrp,
5913         };
5914         int status;
5915
5916         dprintk("--> %s\n", __func__);
5917         nfs41_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
5918         task = rpc_run_task(&task_setup_data);
5919         if (IS_ERR(task))
5920                 return PTR_ERR(task);
5921         status = task->tk_status;
5922         dprintk("<-- %s status=%d\n", __func__, status);
5923         rpc_put_task(task);
5924         return status;
5925 }
5926
5927 /*
5928  * Retrieve the list of Data Server devices from the MDS.
5929  */
5930 static int _nfs4_getdevicelist(struct nfs_server *server,
5931                                     const struct nfs_fh *fh,
5932                                     struct pnfs_devicelist *devlist)
5933 {
5934         struct nfs4_getdevicelist_args args = {
5935                 .fh = fh,
5936                 .layoutclass = server->pnfs_curr_ld->id,
5937         };
5938         struct nfs4_getdevicelist_res res = {
5939                 .devlist = devlist,
5940         };
5941         struct rpc_message msg = {
5942                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
5943                 .rpc_argp = &args,
5944                 .rpc_resp = &res,
5945         };
5946         int status;
5947
5948         dprintk("--> %s\n", __func__);
5949         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5950                                 &res.seq_res, 0);
5951         dprintk("<-- %s status=%d\n", __func__, status);
5952         return status;
5953 }
5954
5955 int nfs4_proc_getdevicelist(struct nfs_server *server,
5956                             const struct nfs_fh *fh,
5957                             struct pnfs_devicelist *devlist)
5958 {
5959         struct nfs4_exception exception = { };
5960         int err;
5961
5962         do {
5963                 err = nfs4_handle_exception(server,
5964                                 _nfs4_getdevicelist(server, fh, devlist),
5965                                 &exception);
5966         } while (exception.retry);
5967
5968         dprintk("%s: err=%d, num_devs=%u\n", __func__,
5969                 err, devlist->num_devs);
5970
5971         return err;
5972 }
5973 EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
5974
5975 static int
5976 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5977 {
5978         struct nfs4_getdeviceinfo_args args = {
5979                 .pdev = pdev,
5980         };
5981         struct nfs4_getdeviceinfo_res res = {
5982                 .pdev = pdev,
5983         };
5984         struct rpc_message msg = {
5985                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5986                 .rpc_argp = &args,
5987                 .rpc_resp = &res,
5988         };
5989         int status;
5990
5991         dprintk("--> %s\n", __func__);
5992         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5993         dprintk("<-- %s status=%d\n", __func__, status);
5994
5995         return status;
5996 }
5997
5998 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5999 {
6000         struct nfs4_exception exception = { };
6001         int err;
6002
6003         do {
6004                 err = nfs4_handle_exception(server,
6005                                         _nfs4_proc_getdeviceinfo(server, pdev),
6006                                         &exception);
6007         } while (exception.retry);
6008         return err;
6009 }
6010 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
6011
6012 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
6013 {
6014         struct nfs4_layoutcommit_data *data = calldata;
6015         struct nfs_server *server = NFS_SERVER(data->args.inode);
6016
6017         if (nfs4_setup_sequence(server, &data->args.seq_args,
6018                                 &data->res.seq_res, task))
6019                 return;
6020         rpc_call_start(task);
6021 }
6022
6023 static void
6024 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
6025 {
6026         struct nfs4_layoutcommit_data *data = calldata;
6027         struct nfs_server *server = NFS_SERVER(data->args.inode);
6028
6029         if (!nfs4_sequence_done(task, &data->res.seq_res))
6030                 return;
6031
6032         switch (task->tk_status) { /* Just ignore these failures */
6033         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
6034         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
6035         case NFS4ERR_BADLAYOUT:     /* no layout */
6036         case NFS4ERR_GRACE:         /* loca_recalim always false */
6037                 task->tk_status = 0;
6038         }
6039
6040         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
6041                 rpc_restart_call_prepare(task);
6042                 return;
6043         }
6044
6045         if (task->tk_status == 0)
6046                 nfs_post_op_update_inode_force_wcc(data->args.inode,
6047                                                    data->res.fattr);
6048 }
6049
6050 static void nfs4_layoutcommit_release(void *calldata)
6051 {
6052         struct nfs4_layoutcommit_data *data = calldata;
6053         struct pnfs_layout_segment *lseg, *tmp;
6054         unsigned long *bitlock = &NFS_I(data->args.inode)->flags;
6055
6056         pnfs_cleanup_layoutcommit(data);
6057         /* Matched by references in pnfs_set_layoutcommit */
6058         list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
6059                 list_del_init(&lseg->pls_lc_list);
6060                 if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
6061                                        &lseg->pls_flags))
6062                         put_lseg(lseg);
6063         }
6064
6065         clear_bit_unlock(NFS_INO_LAYOUTCOMMITTING, bitlock);
6066         smp_mb__after_clear_bit();
6067         wake_up_bit(bitlock, NFS_INO_LAYOUTCOMMITTING);
6068
6069         put_rpccred(data->cred);
6070         kfree(data);
6071 }
6072
6073 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
6074         .rpc_call_prepare = nfs4_layoutcommit_prepare,
6075         .rpc_call_done = nfs4_layoutcommit_done,
6076         .rpc_release = nfs4_layoutcommit_release,
6077 };
6078
6079 int
6080 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
6081 {
6082         struct rpc_message msg = {
6083                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
6084                 .rpc_argp = &data->args,
6085                 .rpc_resp = &data->res,
6086                 .rpc_cred = data->cred,
6087         };
6088         struct rpc_task_setup task_setup_data = {
6089                 .task = &data->task,
6090                 .rpc_client = NFS_CLIENT(data->args.inode),
6091                 .rpc_message = &msg,
6092                 .callback_ops = &nfs4_layoutcommit_ops,
6093                 .callback_data = data,
6094                 .flags = RPC_TASK_ASYNC,
6095         };
6096         struct rpc_task *task;
6097         int status = 0;
6098
6099         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
6100                 "lbw: %llu inode %lu\n",
6101                 data->task.tk_pid, sync,
6102                 data->args.lastbytewritten,
6103                 data->args.inode->i_ino);
6104
6105         nfs41_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
6106         task = rpc_run_task(&task_setup_data);
6107         if (IS_ERR(task))
6108                 return PTR_ERR(task);
6109         if (sync == false)
6110                 goto out;
6111         status = nfs4_wait_for_completion_rpc_task(task);
6112         if (status != 0)
6113                 goto out;
6114         status = task->tk_status;
6115 out:
6116         dprintk("%s: status %d\n", __func__, status);
6117         rpc_put_task(task);
6118         return status;
6119 }
6120
6121 static int
6122 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6123                     struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6124 {
6125         struct nfs41_secinfo_no_name_args args = {
6126                 .style = SECINFO_STYLE_CURRENT_FH,
6127         };
6128         struct nfs4_secinfo_res res = {
6129                 .flavors = flavors,
6130         };
6131         struct rpc_message msg = {
6132                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
6133                 .rpc_argp = &args,
6134                 .rpc_resp = &res,
6135         };
6136         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
6137 }
6138
6139 static int
6140 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
6141                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
6142 {
6143         struct nfs4_exception exception = { };
6144         int err;
6145         do {
6146                 err = _nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6147                 switch (err) {
6148                 case 0:
6149                 case -NFS4ERR_WRONGSEC:
6150                 case -NFS4ERR_NOTSUPP:
6151                         break;
6152                 default:
6153                         err = nfs4_handle_exception(server, err, &exception);
6154                 }
6155         } while (exception.retry);
6156         return err;
6157 }
6158
6159 static int
6160 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
6161                     struct nfs_fsinfo *info)
6162 {
6163         int err;
6164         struct page *page;
6165         rpc_authflavor_t flavor;
6166         struct nfs4_secinfo_flavors *flavors;
6167
6168         page = alloc_page(GFP_KERNEL);
6169         if (!page) {
6170                 err = -ENOMEM;
6171                 goto out;
6172         }
6173
6174         flavors = page_address(page);
6175         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
6176
6177         /*
6178          * Fall back on "guess and check" method if
6179          * the server doesn't support SECINFO_NO_NAME
6180          */
6181         if (err == -NFS4ERR_WRONGSEC || err == -NFS4ERR_NOTSUPP) {
6182                 err = nfs4_find_root_sec(server, fhandle, info);
6183                 goto out_freepage;
6184         }
6185         if (err)
6186                 goto out_freepage;
6187
6188         flavor = nfs_find_best_sec(flavors);
6189         if (err == 0)
6190                 err = nfs4_lookup_root_sec(server, fhandle, info, flavor);
6191
6192 out_freepage:
6193         put_page(page);
6194         if (err == -EACCES)
6195                 return -EPERM;
6196 out:
6197         return err;
6198 }
6199
6200 static int _nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6201 {
6202         int status;
6203         struct nfs41_test_stateid_args args = {
6204                 .stateid = stateid,
6205         };
6206         struct nfs41_test_stateid_res res;
6207         struct rpc_message msg = {
6208                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
6209                 .rpc_argp = &args,
6210                 .rpc_resp = &res,
6211         };
6212
6213         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6214         status = nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6215
6216         if (status == NFS_OK)
6217                 return res.status;
6218         return status;
6219 }
6220
6221 static int nfs41_test_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6222 {
6223         struct nfs4_exception exception = { };
6224         int err;
6225         do {
6226                 err = nfs4_handle_exception(server,
6227                                 _nfs41_test_stateid(server, stateid),
6228                                 &exception);
6229         } while (exception.retry);
6230         return err;
6231 }
6232
6233 static int _nfs4_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6234 {
6235         struct nfs41_free_stateid_args args = {
6236                 .stateid = stateid,
6237         };
6238         struct nfs41_free_stateid_res res;
6239         struct rpc_message msg = {
6240                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
6241                 .rpc_argp = &args,
6242                 .rpc_resp = &res,
6243         };
6244
6245         nfs41_init_sequence(&args.seq_args, &res.seq_res, 0);
6246         return nfs4_call_sync_sequence(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
6247 }
6248
6249 static int nfs41_free_stateid(struct nfs_server *server, nfs4_stateid *stateid)
6250 {
6251         struct nfs4_exception exception = { };
6252         int err;
6253         do {
6254                 err = nfs4_handle_exception(server,
6255                                 _nfs4_free_stateid(server, stateid),
6256                                 &exception);
6257         } while (exception.retry);
6258         return err;
6259 }
6260 #endif /* CONFIG_NFS_V4_1 */
6261
6262 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
6263         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6264         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6265         .recover_open   = nfs4_open_reclaim,
6266         .recover_lock   = nfs4_lock_reclaim,
6267         .establish_clid = nfs4_init_clientid,
6268         .get_clid_cred  = nfs4_get_setclientid_cred,
6269 };
6270
6271 #if defined(CONFIG_NFS_V4_1)
6272 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
6273         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
6274         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
6275         .recover_open   = nfs4_open_reclaim,
6276         .recover_lock   = nfs4_lock_reclaim,
6277         .establish_clid = nfs41_init_clientid,
6278         .get_clid_cred  = nfs4_get_exchange_id_cred,
6279         .reclaim_complete = nfs41_proc_reclaim_complete,
6280 };
6281 #endif /* CONFIG_NFS_V4_1 */
6282
6283 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
6284         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6285         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6286         .recover_open   = nfs4_open_expired,
6287         .recover_lock   = nfs4_lock_expired,
6288         .establish_clid = nfs4_init_clientid,
6289         .get_clid_cred  = nfs4_get_setclientid_cred,
6290 };
6291
6292 #if defined(CONFIG_NFS_V4_1)
6293 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
6294         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
6295         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
6296         .recover_open   = nfs41_open_expired,
6297         .recover_lock   = nfs41_lock_expired,
6298         .establish_clid = nfs41_init_clientid,
6299         .get_clid_cred  = nfs4_get_exchange_id_cred,
6300 };
6301 #endif /* CONFIG_NFS_V4_1 */
6302
6303 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
6304         .sched_state_renewal = nfs4_proc_async_renew,
6305         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
6306         .renew_lease = nfs4_proc_renew,
6307 };
6308
6309 #if defined(CONFIG_NFS_V4_1)
6310 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
6311         .sched_state_renewal = nfs41_proc_async_sequence,
6312         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
6313         .renew_lease = nfs4_proc_sequence,
6314 };
6315 #endif
6316
6317 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
6318         .minor_version = 0,
6319         .call_sync = _nfs4_call_sync,
6320         .validate_stateid = nfs4_validate_delegation_stateid,
6321         .find_root_sec = nfs4_find_root_sec,
6322         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
6323         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
6324         .state_renewal_ops = &nfs40_state_renewal_ops,
6325 };
6326
6327 #if defined(CONFIG_NFS_V4_1)
6328 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
6329         .minor_version = 1,
6330         .call_sync = _nfs4_call_sync_session,
6331         .validate_stateid = nfs41_validate_delegation_stateid,
6332         .find_root_sec = nfs41_find_root_sec,
6333         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
6334         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
6335         .state_renewal_ops = &nfs41_state_renewal_ops,
6336 };
6337 #endif
6338
6339 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
6340         [0] = &nfs_v4_0_minor_ops,
6341 #if defined(CONFIG_NFS_V4_1)
6342         [1] = &nfs_v4_1_minor_ops,
6343 #endif
6344 };
6345
6346 static const struct inode_operations nfs4_file_inode_operations = {
6347         .permission     = nfs_permission,
6348         .getattr        = nfs_getattr,
6349         .setattr        = nfs_setattr,
6350         .getxattr       = generic_getxattr,
6351         .setxattr       = generic_setxattr,
6352         .listxattr      = generic_listxattr,
6353         .removexattr    = generic_removexattr,
6354 };
6355
6356 const struct nfs_rpc_ops nfs_v4_clientops = {
6357         .version        = 4,                    /* protocol version */
6358         .dentry_ops     = &nfs4_dentry_operations,
6359         .dir_inode_ops  = &nfs4_dir_inode_operations,
6360         .file_inode_ops = &nfs4_file_inode_operations,
6361         .file_ops       = &nfs4_file_operations,
6362         .getroot        = nfs4_proc_get_root,
6363         .getattr        = nfs4_proc_getattr,
6364         .setattr        = nfs4_proc_setattr,
6365         .lookup         = nfs4_proc_lookup,
6366         .access         = nfs4_proc_access,
6367         .readlink       = nfs4_proc_readlink,
6368         .create         = nfs4_proc_create,
6369         .remove         = nfs4_proc_remove,
6370         .unlink_setup   = nfs4_proc_unlink_setup,
6371         .unlink_done    = nfs4_proc_unlink_done,
6372         .rename         = nfs4_proc_rename,
6373         .rename_setup   = nfs4_proc_rename_setup,
6374         .rename_done    = nfs4_proc_rename_done,
6375         .link           = nfs4_proc_link,
6376         .symlink        = nfs4_proc_symlink,
6377         .mkdir          = nfs4_proc_mkdir,
6378         .rmdir          = nfs4_proc_remove,
6379         .readdir        = nfs4_proc_readdir,
6380         .mknod          = nfs4_proc_mknod,
6381         .statfs         = nfs4_proc_statfs,
6382         .fsinfo         = nfs4_proc_fsinfo,
6383         .pathconf       = nfs4_proc_pathconf,
6384         .set_capabilities = nfs4_server_capabilities,
6385         .decode_dirent  = nfs4_decode_dirent,
6386         .read_setup     = nfs4_proc_read_setup,
6387         .read_done      = nfs4_read_done,
6388         .write_setup    = nfs4_proc_write_setup,
6389         .write_done     = nfs4_write_done,
6390         .commit_setup   = nfs4_proc_commit_setup,
6391         .commit_done    = nfs4_commit_done,
6392         .lock           = nfs4_proc_lock,
6393         .clear_acl_cache = nfs4_zap_acl_attr,
6394         .close_context  = nfs4_close_context,
6395         .open_context   = nfs4_atomic_open,
6396         .init_client    = nfs4_init_client,
6397         .secinfo        = nfs4_proc_secinfo,
6398 };
6399
6400 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
6401         .prefix = XATTR_NAME_NFSV4_ACL,
6402         .list   = nfs4_xattr_list_nfs4_acl,
6403         .get    = nfs4_xattr_get_nfs4_acl,
6404         .set    = nfs4_xattr_set_nfs4_acl,
6405 };
6406
6407 const struct xattr_handler *nfs4_xattr_handlers[] = {
6408         &nfs4_xattr_nfs4_acl_handler,
6409         NULL
6410 };
6411
6412 module_param(max_session_slots, ushort, 0644);
6413 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
6414                 "requests the client will negotiate");
6415
6416 /*
6417  * Local variables:
6418  *  c-basic-offset: 8
6419  * End:
6420  */