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/sunrpc/clnt.h>
  43 #include <linux/nfs.h>
  44 #include <linux/nfs4.h>
  45 #include <linux/nfs_fs.h>
  46 #include <linux/nfs_page.h>
  47 #include <linux/namei.h>
  48 #include <linux/mount.h>
  49 #include <linux/module.h>
  50 #include <linux/sunrpc/bc_xprt.h>
  51
  52 #include "nfs4_fs.h"
  53 #include "delegation.h"
  54 #include "internal.h"
  55 #include "iostat.h"
  56 #include "callback.h"
  57
  58 #define NFSDBG_FACILITY         NFSDBG_PROC
  59
  60 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  61 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  62
  63 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  64
  65 struct nfs4_opendata;
  66 static int _nfs4_proc_open(struct nfs4_opendata *data);
  67 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  68 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  69 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  70 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  71 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  72
  73 /* Prevent leaks of NFSv4 errors into userland */
  74 static int nfs4_map_errors(int err)
  75 {
  76         if (err >= -1000)
  77                 return err;
  78         switch (err) {
  79         case -NFS4ERR_RESOURCE:
  80                 return -EREMOTEIO;
  81         default:
  82                 dprintk("%s could not handle NFSv4 error %d\n",
  83                                 __func__, -err);
  84                 break;
  85         }
  86         return -EIO;
  87 }
  88
  89 /*
  90  * This is our standard bitmap for GETATTR requests.
  91  */
  92 const u32 nfs4_fattr_bitmap[2] = {
  93         FATTR4_WORD0_TYPE
  94         | FATTR4_WORD0_CHANGE
  95         | FATTR4_WORD0_SIZE
  96         | FATTR4_WORD0_FSID
  97         | FATTR4_WORD0_FILEID,
  98         FATTR4_WORD1_MODE
  99         | FATTR4_WORD1_NUMLINKS
 100         | FATTR4_WORD1_OWNER
 101         | FATTR4_WORD1_OWNER_GROUP
 102         | FATTR4_WORD1_RAWDEV
 103         | FATTR4_WORD1_SPACE_USED
 104         | FATTR4_WORD1_TIME_ACCESS
 105         | FATTR4_WORD1_TIME_METADATA
 106         | FATTR4_WORD1_TIME_MODIFY
 107 };
 108
 109 const u32 nfs4_statfs_bitmap[2] = {
 110         FATTR4_WORD0_FILES_AVAIL
 111         | FATTR4_WORD0_FILES_FREE
 112         | FATTR4_WORD0_FILES_TOTAL,
 113         FATTR4_WORD1_SPACE_AVAIL
 114         | FATTR4_WORD1_SPACE_FREE
 115         | FATTR4_WORD1_SPACE_TOTAL
 116 };
 117
 118 const u32 nfs4_pathconf_bitmap[2] = {
 119         FATTR4_WORD0_MAXLINK
 120         | FATTR4_WORD0_MAXNAME,
 121         0
 122 };
 123
 124 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 125                         | FATTR4_WORD0_MAXREAD
 126                         | FATTR4_WORD0_MAXWRITE
 127                         | FATTR4_WORD0_LEASE_TIME,
 128                         0
 129 };
 130
 131 const u32 nfs4_fs_locations_bitmap[2] = {
 132         FATTR4_WORD0_TYPE
 133         | FATTR4_WORD0_CHANGE
 134         | FATTR4_WORD0_SIZE
 135         | FATTR4_WORD0_FSID
 136         | FATTR4_WORD0_FILEID
 137         | FATTR4_WORD0_FS_LOCATIONS,
 138         FATTR4_WORD1_MODE
 139         | FATTR4_WORD1_NUMLINKS
 140         | FATTR4_WORD1_OWNER
 141         | FATTR4_WORD1_OWNER_GROUP
 142         | FATTR4_WORD1_RAWDEV
 143         | FATTR4_WORD1_SPACE_USED
 144         | FATTR4_WORD1_TIME_ACCESS
 145         | FATTR4_WORD1_TIME_METADATA
 146         | FATTR4_WORD1_TIME_MODIFY
 147         | FATTR4_WORD1_MOUNTED_ON_FILEID
 148 };
 149
 150 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 151                 struct nfs4_readdir_arg *readdir)
 152 {
 153         __be32 *start, *p;
 154
 155         BUG_ON(readdir->count < 80);
 156         if (cookie > 2) {
 157                 readdir->cookie = cookie;
 158                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 159                 return;
 160         }
 161
 162         readdir->cookie = 0;
 163         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 164         if (cookie == 2)
 165                 return;
 166
 167         /*
 168          * NFSv4 servers do not return entries for '.' and '..'
 169          * Therefore, we fake these entries here.  We let '.'
 170          * have cookie 0 and '..' have cookie 1.  Note that
 171          * when talking to the server, we always send cookie 0
 172          * instead of 1 or 2.
 173          */
 174         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 175
 176         if (cookie == 0) {
 177                 *p++ = xdr_one;                                  /* next */
 178                 *p++ = xdr_zero;                   /* cookie, first word */
 179                 *p++ = xdr_one;                   /* cookie, second word */
 180                 *p++ = xdr_one;                             /* entry len */
 181                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 182                 p++;
 183                 *p++ = xdr_one;                         /* bitmap length */
 184                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 185                 *p++ = htonl(8);              /* attribute buffer length */
 186                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 187         }
 188
 189         *p++ = xdr_one;                                  /* next */
 190         *p++ = xdr_zero;                   /* cookie, first word */
 191         *p++ = xdr_two;                   /* cookie, second word */
 192         *p++ = xdr_two;                             /* entry len */
 193         memcpy(p, "..\0\0", 4);                         /* entry */
 194         p++;
 195         *p++ = xdr_one;                         /* bitmap length */
 196         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 197         *p++ = htonl(8);              /* attribute buffer length */
 198         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 199
 200         readdir->pgbase = (char *)p - (char *)start;
 201         readdir->count -= readdir->pgbase;
 202         kunmap_atomic(start, KM_USER0);
 203 }
 204
 205 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 206 {
 207         int res;
 208
 209         might_sleep();
 210
 211         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 212                         nfs_wait_bit_killable, TASK_KILLABLE);
 213         return res;
 214 }
 215
 216 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 217 {
 218         int res = 0;
 219
 220         might_sleep();
 221
 222         if (*timeout <= 0)
 223                 *timeout = NFS4_POLL_RETRY_MIN;
 224         if (*timeout > NFS4_POLL_RETRY_MAX)
 225                 *timeout = NFS4_POLL_RETRY_MAX;
 226         schedule_timeout_killable(*timeout);
 227         if (fatal_signal_pending(current))
 228                 res = -ERESTARTSYS;
 229         *timeout <<= 1;
 230         return res;
 231 }
 232
 233 /* This is the error handling routine for processes that are allowed
 234  * to sleep.
 235  */
 236 static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 237 {
 238         struct nfs_client *clp = server->nfs_client;
 239         struct nfs4_state *state = exception->state;
 240         int ret = errorcode;
 241
 242         exception->retry = 0;
 243         switch(errorcode) {
 244                 case 0:
 245                         return 0;
 246                 case -NFS4ERR_ADMIN_REVOKED:
 247                 case -NFS4ERR_BAD_STATEID:
 248                 case -NFS4ERR_OPENMODE:
 249                         if (state == NULL)
 250                                 break;
 251                         nfs4_state_mark_reclaim_nograce(clp, state);
 252                         goto do_state_recovery;
 253                 case -NFS4ERR_STALE_STATEID:
 254                         if (state == NULL)
 255                                 break;
 256                         nfs4_state_mark_reclaim_reboot(clp, state);
 257                 case -NFS4ERR_STALE_CLIENTID:
 258                 case -NFS4ERR_EXPIRED:
 259                         goto do_state_recovery;
 260 #if defined(CONFIG_NFS_V4_1)
 261                 case -NFS4ERR_BADSESSION:
 262                 case -NFS4ERR_BADSLOT:
 263                 case -NFS4ERR_BAD_HIGH_SLOT:
 264                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 265                 case -NFS4ERR_DEADSESSION:
 266                 case -NFS4ERR_SEQ_FALSE_RETRY:
 267                 case -NFS4ERR_SEQ_MISORDERED:
 268                         dprintk("%s ERROR: %d Reset session\n", __func__,
 269                                 errorcode);
 270                         nfs4_schedule_state_recovery(clp);
 271                         exception->retry = 1;
 272                         break;
 273 #endif /* defined(CONFIG_NFS_V4_1) */
 274                 case -NFS4ERR_FILE_OPEN:
 275                         if (exception->timeout > HZ) {
 276                                 /* We have retried a decent amount, time to
 277                                  * fail
 278                                  */
 279                                 ret = -EBUSY;
 280                                 break;
 281                         }
 282                 case -NFS4ERR_GRACE:
 283                 case -NFS4ERR_DELAY:
 284                 case -EKEYEXPIRED:
 285                         ret = nfs4_delay(server->client, &exception->timeout);
 286                         if (ret != 0)
 287                                 break;
 288                 case -NFS4ERR_OLD_STATEID:
 289                         exception->retry = 1;
 290         }
 291         /* We failed to handle the error */
 292         return nfs4_map_errors(ret);
 293 do_state_recovery:
 294         nfs4_schedule_state_recovery(clp);
 295         ret = nfs4_wait_clnt_recover(clp);
 296         if (ret == 0)
 297                 exception->retry = 1;
 298         return ret;
 299 }
 300
 301
 302 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 303 {
 304         struct nfs_client *clp = server->nfs_client;
 305         spin_lock(&clp->cl_lock);
 306         if (time_before(clp->cl_last_renewal,timestamp))
 307                 clp->cl_last_renewal = timestamp;
 308         spin_unlock(&clp->cl_lock);
 309 }
 310
 311 #if defined(CONFIG_NFS_V4_1)
 312
 313 /*
 314  * nfs4_free_slot - free a slot and efficiently update slot table.
 315  *
 316  * freeing a slot is trivially done by clearing its respective bit
 317  * in the bitmap.
 318  * If the freed slotid equals highest_used_slotid we want to update it
 319  * so that the server would be able to size down the slot table if needed,
 320  * otherwise we know that the highest_used_slotid is still in use.
 321  * When updating highest_used_slotid there may be "holes" in the bitmap
 322  * so we need to scan down from highest_used_slotid to 0 looking for the now
 323  * highest slotid in use.
 324  * If none found, highest_used_slotid is set to -1.
 325  *
 326  * Must be called while holding tbl->slot_tbl_lock
 327  */
 328 static void
 329 nfs4_free_slot(struct nfs4_slot_table *tbl, u8 free_slotid)
 330 {
 331         int slotid = free_slotid;
 332
 333         /* clear used bit in bitmap */
 334         __clear_bit(slotid, tbl->used_slots);
 335
 336         /* update highest_used_slotid when it is freed */
 337         if (slotid == tbl->highest_used_slotid) {
 338                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 339                 if (slotid < tbl->max_slots)
 340                         tbl->highest_used_slotid = slotid;
 341                 else
 342                         tbl->highest_used_slotid = -1;
 343         }
 344         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 345                 free_slotid, tbl->highest_used_slotid);
 346 }
 347
 348 /*
 349  * Signal state manager thread if session is drained
 350  */
 351 static void nfs41_check_drain_session_complete(struct nfs4_session *ses)
 352 {
 353         struct rpc_task *task;
 354
 355         if (!test_bit(NFS4CLNT_SESSION_DRAINING, &ses->clp->cl_state)) {
 356                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 357                 if (task)
 358                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 359                 return;
 360         }
 361
 362         if (ses->fc_slot_table.highest_used_slotid != -1)
 363                 return;
 364
 365         dprintk("%s COMPLETE: Session Drained\n", __func__);
 366         complete(&ses->complete);
 367 }
 368
 369 static void nfs41_sequence_free_slot(const struct nfs_client *clp,
 370                               struct nfs4_sequence_res *res)
 371 {
 372         struct nfs4_slot_table *tbl;
 373
 374         tbl = &clp->cl_session->fc_slot_table;
 375         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) {
 376                 /* just wake up the next guy waiting since
 377                  * we may have not consumed a slot after all */
 378                 dprintk("%s: No slot\n", __func__);
 379                 return;
 380         }
 381
 382         spin_lock(&tbl->slot_tbl_lock);
 383         nfs4_free_slot(tbl, res->sr_slotid);
 384         nfs41_check_drain_session_complete(clp->cl_session);
 385         spin_unlock(&tbl->slot_tbl_lock);
 386         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
 387 }
 388
 389 static void nfs41_sequence_done(struct nfs_client *clp,
 390                                 struct nfs4_sequence_res *res,
 391                                 int rpc_status)
 392 {
 393         unsigned long timestamp;
 394         struct nfs4_slot_table *tbl;
 395         struct nfs4_slot *slot;
 396
 397         /*
 398          * sr_status remains 1 if an RPC level error occurred. The server
 399          * may or may not have processed the sequence operation..
 400          * Proceed as if the server received and processed the sequence
 401          * operation.
 402          */
 403         if (res->sr_status == 1)
 404                 res->sr_status = NFS_OK;
 405
 406         /* -ERESTARTSYS can result in skipping nfs41_sequence_setup */
 407         if (res->sr_slotid == NFS4_MAX_SLOT_TABLE)
 408                 goto out;
 409
 410         /* Check the SEQUENCE operation status */
 411         if (res->sr_status == 0) {
 412                 tbl = &clp->cl_session->fc_slot_table;
 413                 slot = tbl->slots + res->sr_slotid;
 414                 /* Update the slot's sequence and clientid lease timer */
 415                 ++slot->seq_nr;
 416                 timestamp = res->sr_renewal_time;
 417                 spin_lock(&clp->cl_lock);
 418                 if (time_before(clp->cl_last_renewal, timestamp))
 419                         clp->cl_last_renewal = timestamp;
 420                 spin_unlock(&clp->cl_lock);
 421                 /* Check sequence flags */
 422                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 423         }
 424 out:
 425         /* The session may be reset by one of the error handlers. */
 426         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 427         nfs41_sequence_free_slot(clp, res);
 428 }
 429
 430 /*
 431  * nfs4_find_slot - efficiently look for a free slot
 432  *
 433  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 434  * If found, we mark the slot as used, update the highest_used_slotid,
 435  * and respectively set up the sequence operation args.
 436  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 437  *
 438  * Note: must be called with under the slot_tbl_lock.
 439  */
 440 static u8
 441 nfs4_find_slot(struct nfs4_slot_table *tbl)
 442 {
 443         int slotid;
 444         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 445         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 446
 447         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 448                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 449                 tbl->max_slots);
 450         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 451         if (slotid >= tbl->max_slots)
 452                 goto out;
 453         __set_bit(slotid, tbl->used_slots);
 454         if (slotid > tbl->highest_used_slotid)
 455                 tbl->highest_used_slotid = slotid;
 456         ret_id = slotid;
 457 out:
 458         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 459                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 460         return ret_id;
 461 }
 462
 463 static int nfs41_setup_sequence(struct nfs4_session *session,
 464                                 struct nfs4_sequence_args *args,
 465                                 struct nfs4_sequence_res *res,
 466                                 int cache_reply,
 467                                 struct rpc_task *task)
 468 {
 469         struct nfs4_slot *slot;
 470         struct nfs4_slot_table *tbl;
 471         u8 slotid;
 472
 473         dprintk("--> %s\n", __func__);
 474         /* slot already allocated? */
 475         if (res->sr_slotid != NFS4_MAX_SLOT_TABLE)
 476                 return 0;
 477
 478         memset(res, 0, sizeof(*res));
 479         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
 480         tbl = &session->fc_slot_table;
 481
 482         spin_lock(&tbl->slot_tbl_lock);
 483         if (test_bit(NFS4CLNT_SESSION_DRAINING, &session->clp->cl_state) &&
 484             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 485                 /*
 486                  * The state manager will wait until the slot table is empty.
 487                  * Schedule the reset thread
 488                  */
 489                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 490                 spin_unlock(&tbl->slot_tbl_lock);
 491                 dprintk("%s Schedule Session Reset\n", __func__);
 492                 return -EAGAIN;
 493         }
 494
 495         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 496             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 497                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 498                 spin_unlock(&tbl->slot_tbl_lock);
 499                 dprintk("%s enforce FIFO order\n", __func__);
 500                 return -EAGAIN;
 501         }
 502
 503         slotid = nfs4_find_slot(tbl);
 504         if (slotid == NFS4_MAX_SLOT_TABLE) {
 505                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 506                 spin_unlock(&tbl->slot_tbl_lock);
 507                 dprintk("<-- %s: no free slots\n", __func__);
 508                 return -EAGAIN;
 509         }
 510         spin_unlock(&tbl->slot_tbl_lock);
 511
 512         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 513         slot = tbl->slots + slotid;
 514         args->sa_session = session;
 515         args->sa_slotid = slotid;
 516         args->sa_cache_this = cache_reply;
 517
 518         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 519
 520         res->sr_session = session;
 521         res->sr_slotid = slotid;
 522         res->sr_renewal_time = jiffies;
 523         /*
 524          * sr_status is only set in decode_sequence, and so will remain
 525          * set to 1 if an rpc level failure occurs.
 526          */
 527         res->sr_status = 1;
 528         return 0;
 529 }
 530
 531 int nfs4_setup_sequence(struct nfs_client *clp,
 532                         struct nfs4_sequence_args *args,
 533                         struct nfs4_sequence_res *res,
 534                         int cache_reply,
 535                         struct rpc_task *task)
 536 {
 537         int ret = 0;
 538
 539         dprintk("--> %s clp %p session %p sr_slotid %d\n",
 540                 __func__, clp, clp->cl_session, res->sr_slotid);
 541
 542         if (!nfs4_has_session(clp))
 543                 goto out;
 544         ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply,
 545                                    task);
 546         if (ret && ret != -EAGAIN) {
 547                 /* terminate rpc task */
 548                 task->tk_status = ret;
 549                 task->tk_action = NULL;
 550         }
 551 out:
 552         dprintk("<-- %s status=%d\n", __func__, ret);
 553         return ret;
 554 }
 555
 556 struct nfs41_call_sync_data {
 557         struct nfs_client *clp;
 558         struct nfs4_sequence_args *seq_args;
 559         struct nfs4_sequence_res *seq_res;
 560         int cache_reply;
 561 };
 562
 563 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 564 {
 565         struct nfs41_call_sync_data *data = calldata;
 566
 567         dprintk("--> %s data->clp->cl_session %p\n", __func__,
 568                 data->clp->cl_session);
 569         if (nfs4_setup_sequence(data->clp, data->seq_args,
 570                                 data->seq_res, data->cache_reply, task))
 571                 return;
 572         rpc_call_start(task);
 573 }
 574
 575 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 576 {
 577         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 578         nfs41_call_sync_prepare(task, calldata);
 579 }
 580
 581 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 582 {
 583         struct nfs41_call_sync_data *data = calldata;
 584
 585         nfs41_sequence_done(data->clp, data->seq_res, task->tk_status);
 586 }
 587
 588 struct rpc_call_ops nfs41_call_sync_ops = {
 589         .rpc_call_prepare = nfs41_call_sync_prepare,
 590         .rpc_call_done = nfs41_call_sync_done,
 591 };
 592
 593 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 594         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 595         .rpc_call_done = nfs41_call_sync_done,
 596 };
 597
 598 static int nfs4_call_sync_sequence(struct nfs_client *clp,
 599                                    struct rpc_clnt *clnt,
 600                                    struct rpc_message *msg,
 601                                    struct nfs4_sequence_args *args,
 602                                    struct nfs4_sequence_res *res,
 603                                    int cache_reply,
 604                                    int privileged)
 605 {
 606         int ret;
 607         struct rpc_task *task;
 608         struct nfs41_call_sync_data data = {
 609                 .clp = clp,
 610                 .seq_args = args,
 611                 .seq_res = res,
 612                 .cache_reply = cache_reply,
 613         };
 614         struct rpc_task_setup task_setup = {
 615                 .rpc_client = clnt,
 616                 .rpc_message = msg,
 617                 .callback_ops = &nfs41_call_sync_ops,
 618                 .callback_data = &data
 619         };
 620
 621         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
 622         if (privileged)
 623                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 624         task = rpc_run_task(&task_setup);
 625         if (IS_ERR(task))
 626                 ret = PTR_ERR(task);
 627         else {
 628                 ret = task->tk_status;
 629                 rpc_put_task(task);
 630         }
 631         return ret;
 632 }
 633
 634 int _nfs4_call_sync_session(struct nfs_server *server,
 635                             struct rpc_message *msg,
 636                             struct nfs4_sequence_args *args,
 637                             struct nfs4_sequence_res *res,
 638                             int cache_reply)
 639 {
 640         return nfs4_call_sync_sequence(server->nfs_client, server->client,
 641                                        msg, args, res, cache_reply, 0);
 642 }
 643
 644 #endif /* CONFIG_NFS_V4_1 */
 645
 646 int _nfs4_call_sync(struct nfs_server *server,
 647                     struct rpc_message *msg,
 648                     struct nfs4_sequence_args *args,
 649                     struct nfs4_sequence_res *res,
 650                     int cache_reply)
 651 {
 652         args->sa_session = res->sr_session = NULL;
 653         return rpc_call_sync(server->client, msg, 0);
 654 }
 655
 656 #define nfs4_call_sync(server, msg, args, res, cache_reply) \
 657         (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \
 658                         &(res)->seq_res, (cache_reply))
 659
 660 static void nfs4_sequence_done(const struct nfs_server *server,
 661                                struct nfs4_sequence_res *res, int rpc_status)
 662 {
 663 #ifdef CONFIG_NFS_V4_1
 664         if (nfs4_has_session(server->nfs_client))
 665                 nfs41_sequence_done(server->nfs_client, res, rpc_status);
 666 #endif /* CONFIG_NFS_V4_1 */
 667 }
 668
 669 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 670 {
 671         struct nfs_inode *nfsi = NFS_I(dir);
 672
 673         spin_lock(&dir->i_lock);
 674         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 675         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 676                 nfs_force_lookup_revalidate(dir);
 677         nfsi->change_attr = cinfo->after;
 678         spin_unlock(&dir->i_lock);
 679 }
 680
 681 struct nfs4_opendata {
 682         struct kref kref;
 683         struct nfs_openargs o_arg;
 684         struct nfs_openres o_res;
 685         struct nfs_open_confirmargs c_arg;
 686         struct nfs_open_confirmres c_res;
 687         struct nfs_fattr f_attr;
 688         struct nfs_fattr dir_attr;
 689         struct path path;
 690         struct dentry *dir;
 691         struct nfs4_state_owner *owner;
 692         struct nfs4_state *state;
 693         struct iattr attrs;
 694         unsigned long timestamp;
 695         unsigned int rpc_done : 1;
 696         int rpc_status;
 697         int cancelled;
 698 };
 699
 700
 701 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 702 {
 703         p->o_res.f_attr = &p->f_attr;
 704         p->o_res.dir_attr = &p->dir_attr;
 705         p->o_res.seqid = p->o_arg.seqid;
 706         p->c_res.seqid = p->c_arg.seqid;
 707         p->o_res.server = p->o_arg.server;
 708         nfs_fattr_init(&p->f_attr);
 709         nfs_fattr_init(&p->dir_attr);
 710         p->o_res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
 711 }
 712
 713 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 714                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 715                 const struct iattr *attrs)
 716 {
 717         struct dentry *parent = dget_parent(path->dentry);
 718         struct inode *dir = parent->d_inode;
 719         struct nfs_server *server = NFS_SERVER(dir);
 720         struct nfs4_opendata *p;
 721
 722         p = kzalloc(sizeof(*p), GFP_KERNEL);
 723         if (p == NULL)
 724                 goto err;
 725         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
 726         if (p->o_arg.seqid == NULL)
 727                 goto err_free;
 728         p->path.mnt = mntget(path->mnt);
 729         p->path.dentry = dget(path->dentry);
 730         p->dir = parent;
 731         p->owner = sp;
 732         atomic_inc(&sp->so_count);
 733         p->o_arg.fh = NFS_FH(dir);
 734         p->o_arg.open_flags = flags;
 735         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 736         p->o_arg.clientid = server->nfs_client->cl_clientid;
 737         p->o_arg.id = sp->so_owner_id.id;
 738         p->o_arg.name = &p->path.dentry->d_name;
 739         p->o_arg.server = server;
 740         p->o_arg.bitmask = server->attr_bitmask;
 741         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 742         if (flags & O_EXCL) {
 743                 if (nfs4_has_persistent_session(server->nfs_client)) {
 744                         /* GUARDED */
 745                         p->o_arg.u.attrs = &p->attrs;
 746                         memcpy(&p->attrs, attrs, sizeof(p->attrs));
 747                 } else { /* EXCLUSIVE4_1 */
 748                         u32 *s = (u32 *) p->o_arg.u.verifier.data;
 749                         s[0] = jiffies;
 750                         s[1] = current->pid;
 751                 }
 752         } else if (flags & O_CREAT) {
 753                 p->o_arg.u.attrs = &p->attrs;
 754                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 755         }
 756         p->c_arg.fh = &p->o_res.fh;
 757         p->c_arg.stateid = &p->o_res.stateid;
 758         p->c_arg.seqid = p->o_arg.seqid;
 759         nfs4_init_opendata_res(p);
 760         kref_init(&p->kref);
 761         return p;
 762 err_free:
 763         kfree(p);
 764 err:
 765         dput(parent);
 766         return NULL;
 767 }
 768
 769 static void nfs4_opendata_free(struct kref *kref)
 770 {
 771         struct nfs4_opendata *p = container_of(kref,
 772                         struct nfs4_opendata, kref);
 773
 774         nfs_free_seqid(p->o_arg.seqid);
 775         if (p->state != NULL)
 776                 nfs4_put_open_state(p->state);
 777         nfs4_put_state_owner(p->owner);
 778         dput(p->dir);
 779         path_put(&p->path);
 780         kfree(p);
 781 }
 782
 783 static void nfs4_opendata_put(struct nfs4_opendata *p)
 784 {
 785         if (p != NULL)
 786                 kref_put(&p->kref, nfs4_opendata_free);
 787 }
 788
 789 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 790 {
 791         int ret;
 792
 793         ret = rpc_wait_for_completion_task(task);
 794         return ret;
 795 }
 796
 797 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 798 {
 799         int ret = 0;
 800
 801         if (open_mode & O_EXCL)
 802                 goto out;
 803         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 804                 case FMODE_READ:
 805                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 806                                 && state->n_rdonly != 0;
 807                         break;
 808                 case FMODE_WRITE:
 809                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 810                                 && state->n_wronly != 0;
 811                         break;
 812                 case FMODE_READ|FMODE_WRITE:
 813                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 814                                 && state->n_rdwr != 0;
 815         }
 816 out:
 817         return ret;
 818 }
 819
 820 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 821 {
 822         if ((delegation->type & fmode) != fmode)
 823                 return 0;
 824         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 825                 return 0;
 826         nfs_mark_delegation_referenced(delegation);
 827         return 1;
 828 }
 829
 830 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 831 {
 832         switch (fmode) {
 833                 case FMODE_WRITE:
 834                         state->n_wronly++;
 835                         break;
 836                 case FMODE_READ:
 837                         state->n_rdonly++;
 838                         break;
 839                 case FMODE_READ|FMODE_WRITE:
 840                         state->n_rdwr++;
 841         }
 842         nfs4_state_set_mode_locked(state, state->state | fmode);
 843 }
 844
 845 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 846 {
 847         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 848                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 849         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 850         switch (fmode) {
 851                 case FMODE_READ:
 852                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 853                         break;
 854                 case FMODE_WRITE:
 855                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 856                         break;
 857                 case FMODE_READ|FMODE_WRITE:
 858                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 859         }
 860 }
 861
 862 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 863 {
 864         write_seqlock(&state->seqlock);
 865         nfs_set_open_stateid_locked(state, stateid, fmode);
 866         write_sequnlock(&state->seqlock);
 867 }
 868
 869 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 870 {
 871         /*
 872          * Protect the call to nfs4_state_set_mode_locked and
 873          * serialise the stateid update
 874          */
 875         write_seqlock(&state->seqlock);
 876         if (deleg_stateid != NULL) {
 877                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 878                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 879         }
 880         if (open_stateid != NULL)
 881                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 882         write_sequnlock(&state->seqlock);
 883         spin_lock(&state->owner->so_lock);
 884         update_open_stateflags(state, fmode);
 885         spin_unlock(&state->owner->so_lock);
 886 }
 887
 888 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 889 {
 890         struct nfs_inode *nfsi = NFS_I(state->inode);
 891         struct nfs_delegation *deleg_cur;
 892         int ret = 0;
 893
 894         fmode &= (FMODE_READ|FMODE_WRITE);
 895
 896         rcu_read_lock();
 897         deleg_cur = rcu_dereference(nfsi->delegation);
 898         if (deleg_cur == NULL)
 899                 goto no_delegation;
 900
 901         spin_lock(&deleg_cur->lock);
 902         if (nfsi->delegation != deleg_cur ||
 903             (deleg_cur->type & fmode) != fmode)
 904                 goto no_delegation_unlock;
 905
 906         if (delegation == NULL)
 907                 delegation = &deleg_cur->stateid;
 908         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 909                 goto no_delegation_unlock;
 910
 911         nfs_mark_delegation_referenced(deleg_cur);
 912         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 913         ret = 1;
 914 no_delegation_unlock:
 915         spin_unlock(&deleg_cur->lock);
 916 no_delegation:
 917         rcu_read_unlock();
 918
 919         if (!ret && open_stateid != NULL) {
 920                 __update_open_stateid(state, open_stateid, NULL, fmode);
 921                 ret = 1;
 922         }
 923
 924         return ret;
 925 }
 926
 927
 928 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 929 {
 930         struct nfs_delegation *delegation;
 931
 932         rcu_read_lock();
 933         delegation = rcu_dereference(NFS_I(inode)->delegation);
 934         if (delegation == NULL || (delegation->type & fmode) == fmode) {
 935                 rcu_read_unlock();
 936                 return;
 937         }
 938         rcu_read_unlock();
 939         nfs_inode_return_delegation(inode);
 940 }
 941
 942 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
 943 {
 944         struct nfs4_state *state = opendata->state;
 945         struct nfs_inode *nfsi = NFS_I(state->inode);
 946         struct nfs_delegation *delegation;
 947         int open_mode = opendata->o_arg.open_flags & O_EXCL;
 948         fmode_t fmode = opendata->o_arg.fmode;
 949         nfs4_stateid stateid;
 950         int ret = -EAGAIN;
 951
 952         for (;;) {
 953                 if (can_open_cached(state, fmode, open_mode)) {
 954                         spin_lock(&state->owner->so_lock);
 955                         if (can_open_cached(state, fmode, open_mode)) {
 956                                 update_open_stateflags(state, fmode);
 957                                 spin_unlock(&state->owner->so_lock);
 958                                 goto out_return_state;
 959                         }
 960                         spin_unlock(&state->owner->so_lock);
 961                 }
 962                 rcu_read_lock();
 963                 delegation = rcu_dereference(nfsi->delegation);
 964                 if (delegation == NULL ||
 965                     !can_open_delegated(delegation, fmode)) {
 966                         rcu_read_unlock();
 967                         break;
 968                 }
 969                 /* Save the delegation */
 970                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
 971                 rcu_read_unlock();
 972                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
 973                 if (ret != 0)
 974                         goto out;
 975                 ret = -EAGAIN;
 976
 977                 /* Try to update the stateid using the delegation */
 978                 if (update_open_stateid(state, NULL, &stateid, fmode))
 979                         goto out_return_state;
 980         }
 981 out:
 982         return ERR_PTR(ret);
 983 out_return_state:
 984         atomic_inc(&state->count);
 985         return state;
 986 }
 987
 988 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
 989 {
 990         struct inode *inode;
 991         struct nfs4_state *state = NULL;
 992         struct nfs_delegation *delegation;
 993         int ret;
 994
 995         if (!data->rpc_done) {
 996                 state = nfs4_try_open_cached(data);
 997                 goto out;
 998         }
 999
1000         ret = -EAGAIN;
1001         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1002                 goto err;
1003         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1004         ret = PTR_ERR(inode);
1005         if (IS_ERR(inode))
1006                 goto err;
1007         ret = -ENOMEM;
1008         state = nfs4_get_open_state(inode, data->owner);
1009         if (state == NULL)
1010                 goto err_put_inode;
1011         if (data->o_res.delegation_type != 0) {
1012                 int delegation_flags = 0;
1013
1014                 rcu_read_lock();
1015                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1016                 if (delegation)
1017                         delegation_flags = delegation->flags;
1018                 rcu_read_unlock();
1019                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1020                         nfs_inode_set_delegation(state->inode,
1021                                         data->owner->so_cred,
1022                                         &data->o_res);
1023                 else
1024                         nfs_inode_reclaim_delegation(state->inode,
1025                                         data->owner->so_cred,
1026                                         &data->o_res);
1027         }
1028
1029         update_open_stateid(state, &data->o_res.stateid, NULL,
1030                         data->o_arg.fmode);
1031         iput(inode);
1032 out:
1033         return state;
1034 err_put_inode:
1035         iput(inode);
1036 err:
1037         return ERR_PTR(ret);
1038 }
1039
1040 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1041 {
1042         struct nfs_inode *nfsi = NFS_I(state->inode);
1043         struct nfs_open_context *ctx;
1044
1045         spin_lock(&state->inode->i_lock);
1046         list_for_each_entry(ctx, &nfsi->open_files, list) {
1047                 if (ctx->state != state)
1048                         continue;
1049                 get_nfs_open_context(ctx);
1050                 spin_unlock(&state->inode->i_lock);
1051                 return ctx;
1052         }
1053         spin_unlock(&state->inode->i_lock);
1054         return ERR_PTR(-ENOENT);
1055 }
1056
1057 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1058 {
1059         struct nfs4_opendata *opendata;
1060
1061         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL);
1062         if (opendata == NULL)
1063                 return ERR_PTR(-ENOMEM);
1064         opendata->state = state;
1065         atomic_inc(&state->count);
1066         return opendata;
1067 }
1068
1069 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1070 {
1071         struct nfs4_state *newstate;
1072         int ret;
1073
1074         opendata->o_arg.open_flags = 0;
1075         opendata->o_arg.fmode = fmode;
1076         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1077         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1078         nfs4_init_opendata_res(opendata);
1079         ret = _nfs4_recover_proc_open(opendata);
1080         if (ret != 0)
1081                 return ret;
1082         newstate = nfs4_opendata_to_nfs4_state(opendata);
1083         if (IS_ERR(newstate))
1084                 return PTR_ERR(newstate);
1085         nfs4_close_state(&opendata->path, newstate, fmode);
1086         *res = newstate;
1087         return 0;
1088 }
1089
1090 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1091 {
1092         struct nfs4_state *newstate;
1093         int ret;
1094
1095         /* memory barrier prior to reading state->n_* */
1096         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1097         smp_rmb();
1098         if (state->n_rdwr != 0) {
1099                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1100                 if (ret != 0)
1101                         return ret;
1102                 if (newstate != state)
1103                         return -ESTALE;
1104         }
1105         if (state->n_wronly != 0) {
1106                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1107                 if (ret != 0)
1108                         return ret;
1109                 if (newstate != state)
1110                         return -ESTALE;
1111         }
1112         if (state->n_rdonly != 0) {
1113                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1114                 if (ret != 0)
1115                         return ret;
1116                 if (newstate != state)
1117                         return -ESTALE;
1118         }
1119         /*
1120          * We may have performed cached opens for all three recoveries.
1121          * Check if we need to update the current stateid.
1122          */
1123         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1124             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1125                 write_seqlock(&state->seqlock);
1126                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1127                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1128                 write_sequnlock(&state->seqlock);
1129         }
1130         return 0;
1131 }
1132
1133 /*
1134  * OPEN_RECLAIM:
1135  *      reclaim state on the server after a reboot.
1136  */
1137 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1138 {
1139         struct nfs_delegation *delegation;
1140         struct nfs4_opendata *opendata;
1141         fmode_t delegation_type = 0;
1142         int status;
1143
1144         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1145         if (IS_ERR(opendata))
1146                 return PTR_ERR(opendata);
1147         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1148         opendata->o_arg.fh = NFS_FH(state->inode);
1149         rcu_read_lock();
1150         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1151         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1152                 delegation_type = delegation->type;
1153         rcu_read_unlock();
1154         opendata->o_arg.u.delegation_type = delegation_type;
1155         status = nfs4_open_recover(opendata, state);
1156         nfs4_opendata_put(opendata);
1157         return status;
1158 }
1159
1160 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1161 {
1162         struct nfs_server *server = NFS_SERVER(state->inode);
1163         struct nfs4_exception exception = { };
1164         int err;
1165         do {
1166                 err = _nfs4_do_open_reclaim(ctx, state);
1167                 if (err != -NFS4ERR_DELAY && err != -EKEYEXPIRED)
1168                         break;
1169                 nfs4_handle_exception(server, err, &exception);
1170         } while (exception.retry);
1171         return err;
1172 }
1173
1174 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1175 {
1176         struct nfs_open_context *ctx;
1177         int ret;
1178
1179         ctx = nfs4_state_find_open_context(state);
1180         if (IS_ERR(ctx))
1181                 return PTR_ERR(ctx);
1182         ret = nfs4_do_open_reclaim(ctx, state);
1183         put_nfs_open_context(ctx);
1184         return ret;
1185 }
1186
1187 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1188 {
1189         struct nfs4_opendata *opendata;
1190         int ret;
1191
1192         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1193         if (IS_ERR(opendata))
1194                 return PTR_ERR(opendata);
1195         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1196         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1197                         sizeof(opendata->o_arg.u.delegation.data));
1198         ret = nfs4_open_recover(opendata, state);
1199         nfs4_opendata_put(opendata);
1200         return ret;
1201 }
1202
1203 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1204 {
1205         struct nfs4_exception exception = { };
1206         struct nfs_server *server = NFS_SERVER(state->inode);
1207         int err;
1208         do {
1209                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1210                 switch (err) {
1211                         case 0:
1212                         case -ENOENT:
1213                         case -ESTALE:
1214                                 goto out;
1215                         case -NFS4ERR_BADSESSION:
1216                         case -NFS4ERR_BADSLOT:
1217                         case -NFS4ERR_BAD_HIGH_SLOT:
1218                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1219                         case -NFS4ERR_DEADSESSION:
1220                                 nfs4_schedule_state_recovery(
1221                                         server->nfs_client);
1222                                 goto out;
1223                         case -NFS4ERR_STALE_CLIENTID:
1224                         case -NFS4ERR_STALE_STATEID:
1225                         case -NFS4ERR_EXPIRED:
1226                                 /* Don't recall a delegation if it was lost */
1227                                 nfs4_schedule_state_recovery(server->nfs_client);
1228                                 goto out;
1229                         case -ERESTARTSYS:
1230                                 /*
1231                                  * The show must go on: exit, but mark the
1232                                  * stateid as needing recovery.
1233                                  */
1234                         case -NFS4ERR_ADMIN_REVOKED:
1235                         case -NFS4ERR_BAD_STATEID:
1236                                 nfs4_state_mark_reclaim_nograce(server->nfs_client, state);
1237                         case -ENOMEM:
1238                                 err = 0;
1239                                 goto out;
1240                 }
1241                 err = nfs4_handle_exception(server, err, &exception);
1242         } while (exception.retry);
1243 out:
1244         return err;
1245 }
1246
1247 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1248 {
1249         struct nfs4_opendata *data = calldata;
1250
1251         data->rpc_status = task->tk_status;
1252         if (RPC_ASSASSINATED(task))
1253                 return;
1254         if (data->rpc_status == 0) {
1255                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1256                                 sizeof(data->o_res.stateid.data));
1257                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1258                 renew_lease(data->o_res.server, data->timestamp);
1259                 data->rpc_done = 1;
1260         }
1261 }
1262
1263 static void nfs4_open_confirm_release(void *calldata)
1264 {
1265         struct nfs4_opendata *data = calldata;
1266         struct nfs4_state *state = NULL;
1267
1268         /* If this request hasn't been cancelled, do nothing */
1269         if (data->cancelled == 0)
1270                 goto out_free;
1271         /* In case of error, no cleanup! */
1272         if (!data->rpc_done)
1273                 goto out_free;
1274         state = nfs4_opendata_to_nfs4_state(data);
1275         if (!IS_ERR(state))
1276                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1277 out_free:
1278         nfs4_opendata_put(data);
1279 }
1280
1281 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1282         .rpc_call_done = nfs4_open_confirm_done,
1283         .rpc_release = nfs4_open_confirm_release,
1284 };
1285
1286 /*
1287  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1288  */
1289 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1290 {
1291         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1292         struct rpc_task *task;
1293         struct  rpc_message msg = {
1294                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1295                 .rpc_argp = &data->c_arg,
1296                 .rpc_resp = &data->c_res,
1297                 .rpc_cred = data->owner->so_cred,
1298         };
1299         struct rpc_task_setup task_setup_data = {
1300                 .rpc_client = server->client,
1301                 .rpc_message = &msg,
1302                 .callback_ops = &nfs4_open_confirm_ops,
1303                 .callback_data = data,
1304                 .workqueue = nfsiod_workqueue,
1305                 .flags = RPC_TASK_ASYNC,
1306         };
1307         int status;
1308
1309         kref_get(&data->kref);
1310         data->rpc_done = 0;
1311         data->rpc_status = 0;
1312         data->timestamp = jiffies;
1313         task = rpc_run_task(&task_setup_data);
1314         if (IS_ERR(task))
1315                 return PTR_ERR(task);
1316         status = nfs4_wait_for_completion_rpc_task(task);
1317         if (status != 0) {
1318                 data->cancelled = 1;
1319                 smp_wmb();
1320         } else
1321                 status = data->rpc_status;
1322         rpc_put_task(task);
1323         return status;
1324 }
1325
1326 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1327 {
1328         struct nfs4_opendata *data = calldata;
1329         struct nfs4_state_owner *sp = data->owner;
1330
1331         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1332                 return;
1333         /*
1334          * Check if we still need to send an OPEN call, or if we can use
1335          * a delegation instead.
1336          */
1337         if (data->state != NULL) {
1338                 struct nfs_delegation *delegation;
1339
1340                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1341                         goto out_no_action;
1342                 rcu_read_lock();
1343                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1344                 if (delegation != NULL &&
1345                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1346                         rcu_read_unlock();
1347                         goto out_no_action;
1348                 }
1349                 rcu_read_unlock();
1350         }
1351         /* Update sequence id. */
1352         data->o_arg.id = sp->so_owner_id.id;
1353         data->o_arg.clientid = sp->so_client->cl_clientid;
1354         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1355                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1356                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1357         }
1358         data->timestamp = jiffies;
1359         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
1360                                 &data->o_arg.seq_args,
1361                                 &data->o_res.seq_res, 1, task))
1362                 return;
1363         rpc_call_start(task);
1364         return;
1365 out_no_action:
1366         task->tk_action = NULL;
1367
1368 }
1369
1370 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1371 {
1372         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1373         nfs4_open_prepare(task, calldata);
1374 }
1375
1376 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1377 {
1378         struct nfs4_opendata *data = calldata;
1379
1380         data->rpc_status = task->tk_status;
1381
1382         nfs4_sequence_done(data->o_arg.server, &data->o_res.seq_res,
1383                         task->tk_status);
1384
1385         if (RPC_ASSASSINATED(task))
1386                 return;
1387         if (task->tk_status == 0) {
1388                 switch (data->o_res.f_attr->mode & S_IFMT) {
1389                         case S_IFREG:
1390                                 break;
1391                         case S_IFLNK:
1392                                 data->rpc_status = -ELOOP;
1393                                 break;
1394                         case S_IFDIR:
1395                                 data->rpc_status = -EISDIR;
1396                                 break;
1397                         default:
1398                                 data->rpc_status = -ENOTDIR;
1399                 }
1400                 renew_lease(data->o_res.server, data->timestamp);
1401                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1402                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1403         }
1404         data->rpc_done = 1;
1405 }
1406
1407 static void nfs4_open_release(void *calldata)
1408 {
1409         struct nfs4_opendata *data = calldata;
1410         struct nfs4_state *state = NULL;
1411
1412         /* If this request hasn't been cancelled, do nothing */
1413         if (data->cancelled == 0)
1414                 goto out_free;
1415         /* In case of error, no cleanup! */
1416         if (data->rpc_status != 0 || !data->rpc_done)
1417                 goto out_free;
1418         /* In case we need an open_confirm, no cleanup! */
1419         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1420                 goto out_free;
1421         state = nfs4_opendata_to_nfs4_state(data);
1422         if (!IS_ERR(state))
1423                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1424 out_free:
1425         nfs4_opendata_put(data);
1426 }
1427
1428 static const struct rpc_call_ops nfs4_open_ops = {
1429         .rpc_call_prepare = nfs4_open_prepare,
1430         .rpc_call_done = nfs4_open_done,
1431         .rpc_release = nfs4_open_release,
1432 };
1433
1434 static const struct rpc_call_ops nfs4_recover_open_ops = {
1435         .rpc_call_prepare = nfs4_recover_open_prepare,
1436         .rpc_call_done = nfs4_open_done,
1437         .rpc_release = nfs4_open_release,
1438 };
1439
1440 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1441 {
1442         struct inode *dir = data->dir->d_inode;
1443         struct nfs_server *server = NFS_SERVER(dir);
1444         struct nfs_openargs *o_arg = &data->o_arg;
1445         struct nfs_openres *o_res = &data->o_res;
1446         struct rpc_task *task;
1447         struct rpc_message msg = {
1448                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1449                 .rpc_argp = o_arg,
1450                 .rpc_resp = o_res,
1451                 .rpc_cred = data->owner->so_cred,
1452         };
1453         struct rpc_task_setup task_setup_data = {
1454                 .rpc_client = server->client,
1455                 .rpc_message = &msg,
1456                 .callback_ops = &nfs4_open_ops,
1457                 .callback_data = data,
1458                 .workqueue = nfsiod_workqueue,
1459                 .flags = RPC_TASK_ASYNC,
1460         };
1461         int status;
1462
1463         kref_get(&data->kref);
1464         data->rpc_done = 0;
1465         data->rpc_status = 0;
1466         data->cancelled = 0;
1467         if (isrecover)
1468                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1469         task = rpc_run_task(&task_setup_data);
1470         if (IS_ERR(task))
1471                 return PTR_ERR(task);
1472         status = nfs4_wait_for_completion_rpc_task(task);
1473         if (status != 0) {
1474                 data->cancelled = 1;
1475                 smp_wmb();
1476         } else
1477                 status = data->rpc_status;
1478         rpc_put_task(task);
1479
1480         return status;
1481 }
1482
1483 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1484 {
1485         struct inode *dir = data->dir->d_inode;
1486         struct nfs_openres *o_res = &data->o_res;
1487         int status;
1488
1489         status = nfs4_run_open_task(data, 1);
1490         if (status != 0 || !data->rpc_done)
1491                 return status;
1492
1493         nfs_refresh_inode(dir, o_res->dir_attr);
1494
1495         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1496                 status = _nfs4_proc_open_confirm(data);
1497                 if (status != 0)
1498                         return status;
1499         }
1500
1501         return status;
1502 }
1503
1504 /*
1505  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1506  */
1507 static int _nfs4_proc_open(struct nfs4_opendata *data)
1508 {
1509         struct inode *dir = data->dir->d_inode;
1510         struct nfs_server *server = NFS_SERVER(dir);
1511         struct nfs_openargs *o_arg = &data->o_arg;
1512         struct nfs_openres *o_res = &data->o_res;
1513         int status;
1514
1515         status = nfs4_run_open_task(data, 0);
1516         if (status != 0 || !data->rpc_done)
1517                 return status;
1518
1519         if (o_arg->open_flags & O_CREAT) {
1520                 update_changeattr(dir, &o_res->cinfo);
1521                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1522         } else
1523                 nfs_refresh_inode(dir, o_res->dir_attr);
1524         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1525                 status = _nfs4_proc_open_confirm(data);
1526                 if (status != 0)
1527                         return status;
1528         }
1529         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1530                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1531         return 0;
1532 }
1533
1534 static int nfs4_recover_expired_lease(struct nfs_server *server)
1535 {
1536         struct nfs_client *clp = server->nfs_client;
1537         unsigned int loop;
1538         int ret;
1539
1540         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1541                 ret = nfs4_wait_clnt_recover(clp);
1542                 if (ret != 0)
1543                         break;
1544                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1545                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1546                         break;
1547                 nfs4_schedule_state_recovery(clp);
1548                 ret = -EIO;
1549         }
1550         return ret;
1551 }
1552
1553 /*
1554  * OPEN_EXPIRED:
1555  *      reclaim state on the server after a network partition.
1556  *      Assumes caller holds the appropriate lock
1557  */
1558 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1559 {
1560         struct nfs4_opendata *opendata;
1561         int ret;
1562
1563         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1564         if (IS_ERR(opendata))
1565                 return PTR_ERR(opendata);
1566         ret = nfs4_open_recover(opendata, state);
1567         if (ret == -ESTALE)
1568                 d_drop(ctx->path.dentry);
1569         nfs4_opendata_put(opendata);
1570         return ret;
1571 }
1572
1573 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1574 {
1575         struct nfs_server *server = NFS_SERVER(state->inode);
1576         struct nfs4_exception exception = { };
1577         int err;
1578
1579         do {
1580                 err = _nfs4_open_expired(ctx, state);
1581                 switch (err) {
1582                 default:
1583                         goto out;
1584                 case -NFS4ERR_GRACE:
1585                 case -NFS4ERR_DELAY:
1586                 case -EKEYEXPIRED:
1587                         nfs4_handle_exception(server, err, &exception);
1588                         err = 0;
1589                 }
1590         } while (exception.retry);
1591 out:
1592         return err;
1593 }
1594
1595 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1596 {
1597         struct nfs_open_context *ctx;
1598         int ret;
1599
1600         ctx = nfs4_state_find_open_context(state);
1601         if (IS_ERR(ctx))
1602                 return PTR_ERR(ctx);
1603         ret = nfs4_do_open_expired(ctx, state);
1604         put_nfs_open_context(ctx);
1605         return ret;
1606 }
1607
1608 /*
1609  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1610  * fields corresponding to attributes that were used to store the verifier.
1611  * Make sure we clobber those fields in the later setattr call
1612  */
1613 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1614 {
1615         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1616             !(sattr->ia_valid & ATTR_ATIME_SET))
1617                 sattr->ia_valid |= ATTR_ATIME;
1618
1619         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1620             !(sattr->ia_valid & ATTR_MTIME_SET))
1621                 sattr->ia_valid |= ATTR_MTIME;
1622 }
1623
1624 /*
1625  * Returns a referenced nfs4_state
1626  */
1627 static int _nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
1628 {
1629         struct nfs4_state_owner  *sp;
1630         struct nfs4_state     *state = NULL;
1631         struct nfs_server       *server = NFS_SERVER(dir);
1632         struct nfs4_opendata *opendata;
1633         int status;
1634
1635         /* Protect against reboot recovery conflicts */
1636         status = -ENOMEM;
1637         if (!(sp = nfs4_get_state_owner(server, cred))) {
1638                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1639                 goto out_err;
1640         }
1641         status = nfs4_recover_expired_lease(server);
1642         if (status != 0)
1643                 goto err_put_state_owner;
1644         if (path->dentry->d_inode != NULL)
1645                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1646         status = -ENOMEM;
1647         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr);
1648         if (opendata == NULL)
1649                 goto err_put_state_owner;
1650
1651         if (path->dentry->d_inode != NULL)
1652                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1653
1654         status = _nfs4_proc_open(opendata);
1655         if (status != 0)
1656                 goto err_opendata_put;
1657
1658         if (opendata->o_arg.open_flags & O_EXCL)
1659                 nfs4_exclusive_attrset(opendata, sattr);
1660
1661         state = nfs4_opendata_to_nfs4_state(opendata);
1662         status = PTR_ERR(state);
1663         if (IS_ERR(state))
1664                 goto err_opendata_put;
1665         if ((opendata->o_res.rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) != 0)
1666                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1667         nfs4_opendata_put(opendata);
1668         nfs4_put_state_owner(sp);
1669         *res = state;
1670         return 0;
1671 err_opendata_put:
1672         nfs4_opendata_put(opendata);
1673 err_put_state_owner:
1674         nfs4_put_state_owner(sp);
1675 out_err:
1676         *res = NULL;
1677         return status;
1678 }
1679
1680
1681 static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, fmode_t fmode, int flags, struct iattr *sattr, struct rpc_cred *cred)
1682 {
1683         struct nfs4_exception exception = { };
1684         struct nfs4_state *res;
1685         int status;
1686
1687         do {
1688                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1689                 if (status == 0)
1690                         break;
1691                 /* NOTE: BAD_SEQID means the server and client disagree about the
1692                  * book-keeping w.r.t. state-changing operations
1693                  * (OPEN/CLOSE/LOCK/LOCKU...)
1694                  * It is actually a sign of a bug on the client or on the server.
1695                  *
1696                  * If we receive a BAD_SEQID error in the particular case of
1697                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1698                  * have unhashed the old state_owner for us, and that we can
1699                  * therefore safely retry using a new one. We should still warn
1700                  * the user though...
1701                  */
1702                 if (status == -NFS4ERR_BAD_SEQID) {
1703                         printk(KERN_WARNING "NFS: v4 server %s "
1704                                         " returned a bad sequence-id error!\n",
1705                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1706                         exception.retry = 1;
1707                         continue;
1708                 }
1709                 /*
1710                  * BAD_STATEID on OPEN means that the server cancelled our
1711                  * state before it received the OPEN_CONFIRM.
1712                  * Recover by retrying the request as per the discussion
1713                  * on Page 181 of RFC3530.
1714                  */
1715                 if (status == -NFS4ERR_BAD_STATEID) {
1716                         exception.retry = 1;
1717                         continue;
1718                 }
1719                 if (status == -EAGAIN) {
1720                         /* We must have found a delegation */
1721                         exception.retry = 1;
1722                         continue;
1723                 }
1724                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1725                                         status, &exception));
1726         } while (exception.retry);
1727         return res;
1728 }
1729
1730 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1731                             struct nfs_fattr *fattr, struct iattr *sattr,
1732                             struct nfs4_state *state)
1733 {
1734         struct nfs_server *server = NFS_SERVER(inode);
1735         struct nfs_setattrargs  arg = {
1736                 .fh             = NFS_FH(inode),
1737                 .iap            = sattr,
1738                 .server         = server,
1739                 .bitmask = server->attr_bitmask,
1740         };
1741         struct nfs_setattrres  res = {
1742                 .fattr          = fattr,
1743                 .server         = server,
1744         };
1745         struct rpc_message msg = {
1746                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1747                 .rpc_argp       = &arg,
1748                 .rpc_resp       = &res,
1749                 .rpc_cred       = cred,
1750         };
1751         unsigned long timestamp = jiffies;
1752         int status;
1753
1754         nfs_fattr_init(fattr);
1755
1756         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1757                 /* Use that stateid */
1758         } else if (state != NULL) {
1759                 nfs4_copy_stateid(&arg.stateid, state, current->files);
1760         } else
1761                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1762
1763         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
1764         if (status == 0 && state != NULL)
1765                 renew_lease(server, timestamp);
1766         return status;
1767 }
1768
1769 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1770                            struct nfs_fattr *fattr, struct iattr *sattr,
1771                            struct nfs4_state *state)
1772 {
1773         struct nfs_server *server = NFS_SERVER(inode);
1774         struct nfs4_exception exception = { };
1775         int err;
1776         do {
1777                 err = nfs4_handle_exception(server,
1778                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1779                                 &exception);
1780         } while (exception.retry);
1781         return err;
1782 }
1783
1784 struct nfs4_closedata {
1785         struct path path;
1786         struct inode *inode;
1787         struct nfs4_state *state;
1788         struct nfs_closeargs arg;
1789         struct nfs_closeres res;
1790         struct nfs_fattr fattr;
1791         unsigned long timestamp;
1792 };
1793
1794 static void nfs4_free_closedata(void *data)
1795 {
1796         struct nfs4_closedata *calldata = data;
1797         struct nfs4_state_owner *sp = calldata->state->owner;
1798
1799         nfs4_put_open_state(calldata->state);
1800         nfs_free_seqid(calldata->arg.seqid);
1801         nfs4_put_state_owner(sp);
1802         path_put(&calldata->path);
1803         kfree(calldata);
1804 }
1805
1806 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1807                 fmode_t fmode)
1808 {
1809         spin_lock(&state->owner->so_lock);
1810         if (!(fmode & FMODE_READ))
1811                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1812         if (!(fmode & FMODE_WRITE))
1813                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1814         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1815         spin_unlock(&state->owner->so_lock);
1816 }
1817
1818 static void nfs4_close_done(struct rpc_task *task, void *data)
1819 {
1820         struct nfs4_closedata *calldata = data;
1821         struct nfs4_state *state = calldata->state;
1822         struct nfs_server *server = NFS_SERVER(calldata->inode);
1823
1824         nfs4_sequence_done(server, &calldata->res.seq_res, task->tk_status);
1825         if (RPC_ASSASSINATED(task))
1826                 return;
1827         /* hmm. we are done with the inode, and in the process of freeing
1828          * the state_owner. we keep this around to process errors
1829          */
1830         switch (task->tk_status) {
1831                 case 0:
1832                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1833                         renew_lease(server, calldata->timestamp);
1834                         nfs4_close_clear_stateid_flags(state,
1835                                         calldata->arg.fmode);
1836                         break;
1837                 case -NFS4ERR_STALE_STATEID:
1838                 case -NFS4ERR_OLD_STATEID:
1839                 case -NFS4ERR_BAD_STATEID:
1840                 case -NFS4ERR_EXPIRED:
1841                         if (calldata->arg.fmode == 0)
1842                                 break;
1843                 default:
1844                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1845                                 rpc_restart_call_prepare(task);
1846         }
1847         nfs_release_seqid(calldata->arg.seqid);
1848         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1849 }
1850
1851 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1852 {
1853         struct nfs4_closedata *calldata = data;
1854         struct nfs4_state *state = calldata->state;
1855         int call_close = 0;
1856
1857         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1858                 return;
1859
1860         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1861         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1862         spin_lock(&state->owner->so_lock);
1863         /* Calculate the change in open mode */
1864         if (state->n_rdwr == 0) {
1865                 if (state->n_rdonly == 0) {
1866                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1867                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1868                         calldata->arg.fmode &= ~FMODE_READ;
1869                 }
1870                 if (state->n_wronly == 0) {
1871                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1872                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1873                         calldata->arg.fmode &= ~FMODE_WRITE;
1874                 }
1875         }
1876         spin_unlock(&state->owner->so_lock);
1877
1878         if (!call_close) {
1879                 /* Note: exit _without_ calling nfs4_close_done */
1880                 task->tk_action = NULL;
1881                 return;
1882         }
1883
1884         if (calldata->arg.fmode == 0)
1885                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1886
1887         nfs_fattr_init(calldata->res.fattr);
1888         calldata->timestamp = jiffies;
1889         if (nfs4_setup_sequence((NFS_SERVER(calldata->inode))->nfs_client,
1890                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1891                                 1, task))
1892                 return;
1893         rpc_call_start(task);
1894 }
1895
1896 static const struct rpc_call_ops nfs4_close_ops = {
1897         .rpc_call_prepare = nfs4_close_prepare,
1898         .rpc_call_done = nfs4_close_done,
1899         .rpc_release = nfs4_free_closedata,
1900 };
1901
1902 /*
1903  * It is possible for data to be read/written from a mem-mapped file
1904  * after the sys_close call (which hits the vfs layer as a flush).
1905  * This means that we can't safely call nfsv4 close on a file until
1906  * the inode is cleared. This in turn means that we are not good
1907  * NFSv4 citizens - we do not indicate to the server to update the file's
1908  * share state even when we are done with one of the three share
1909  * stateid's in the inode.
1910  *
1911  * NOTE: Caller must be holding the sp->so_owner semaphore!
1912  */
1913 int nfs4_do_close(struct path *path, struct nfs4_state *state, int wait)
1914 {
1915         struct nfs_server *server = NFS_SERVER(state->inode);
1916         struct nfs4_closedata *calldata;
1917         struct nfs4_state_owner *sp = state->owner;
1918         struct rpc_task *task;
1919         struct rpc_message msg = {
1920                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
1921                 .rpc_cred = state->owner->so_cred,
1922         };
1923         struct rpc_task_setup task_setup_data = {
1924                 .rpc_client = server->client,
1925                 .rpc_message = &msg,
1926                 .callback_ops = &nfs4_close_ops,
1927                 .workqueue = nfsiod_workqueue,
1928                 .flags = RPC_TASK_ASYNC,
1929         };
1930         int status = -ENOMEM;
1931
1932         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
1933         if (calldata == NULL)
1934                 goto out;
1935         calldata->inode = state->inode;
1936         calldata->state = state;
1937         calldata->arg.fh = NFS_FH(state->inode);
1938         calldata->arg.stateid = &state->open_stateid;
1939         /* Serialization for the sequence id */
1940         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
1941         if (calldata->arg.seqid == NULL)
1942                 goto out_free_calldata;
1943         calldata->arg.fmode = 0;
1944         calldata->arg.bitmask = server->cache_consistency_bitmask;
1945         calldata->res.fattr = &calldata->fattr;
1946         calldata->res.seqid = calldata->arg.seqid;
1947         calldata->res.server = server;
1948         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
1949         calldata->path.mnt = mntget(path->mnt);
1950         calldata->path.dentry = dget(path->dentry);
1951
1952         msg.rpc_argp = &calldata->arg,
1953         msg.rpc_resp = &calldata->res,
1954         task_setup_data.callback_data = calldata;
1955         task = rpc_run_task(&task_setup_data);
1956         if (IS_ERR(task))
1957                 return PTR_ERR(task);
1958         status = 0;
1959         if (wait)
1960                 status = rpc_wait_for_completion_task(task);
1961         rpc_put_task(task);
1962         return status;
1963 out_free_calldata:
1964         kfree(calldata);
1965 out:
1966         nfs4_put_open_state(state);
1967         nfs4_put_state_owner(sp);
1968         return status;
1969 }
1970
1971 static int nfs4_intent_set_file(struct nameidata *nd, struct path *path, struct nfs4_state *state, fmode_t fmode)
1972 {
1973         struct file *filp;
1974         int ret;
1975
1976         /* If the open_intent is for execute, we have an extra check to make */
1977         if (fmode & FMODE_EXEC) {
1978                 ret = nfs_may_open(state->inode,
1979                                 state->owner->so_cred,
1980                                 nd->intent.open.flags);
1981                 if (ret < 0)
1982                         goto out_close;
1983         }
1984         filp = lookup_instantiate_filp(nd, path->dentry, NULL);
1985         if (!IS_ERR(filp)) {
1986                 struct nfs_open_context *ctx;
1987                 ctx = nfs_file_open_context(filp);
1988                 ctx->state = state;
1989                 return 0;
1990         }
1991         ret = PTR_ERR(filp);
1992 out_close:
1993         nfs4_close_sync(path, state, fmode & (FMODE_READ|FMODE_WRITE));
1994         return ret;
1995 }
1996
1997 struct dentry *
1998 nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1999 {
2000         struct path path = {
2001                 .mnt = nd->path.mnt,
2002                 .dentry = dentry,
2003         };
2004         struct dentry *parent;
2005         struct iattr attr;
2006         struct rpc_cred *cred;
2007         struct nfs4_state *state;
2008         struct dentry *res;
2009         fmode_t fmode = nd->intent.open.flags & (FMODE_READ | FMODE_WRITE | FMODE_EXEC);
2010
2011         if (nd->flags & LOOKUP_CREATE) {
2012                 attr.ia_mode = nd->intent.open.create_mode;
2013                 attr.ia_valid = ATTR_MODE;
2014                 if (!IS_POSIXACL(dir))
2015                         attr.ia_mode &= ~current_umask();
2016         } else {
2017                 attr.ia_valid = 0;
2018                 BUG_ON(nd->intent.open.flags & O_CREAT);
2019         }
2020
2021         cred = rpc_lookup_cred();
2022         if (IS_ERR(cred))
2023                 return (struct dentry *)cred;
2024         parent = dentry->d_parent;
2025         /* Protect against concurrent sillydeletes */
2026         nfs_block_sillyrename(parent);
2027         state = nfs4_do_open(dir, &path, fmode, nd->intent.open.flags, &attr, cred);
2028         put_rpccred(cred);
2029         if (IS_ERR(state)) {
2030                 if (PTR_ERR(state) == -ENOENT) {
2031                         d_add(dentry, NULL);
2032                         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2033                 }
2034                 nfs_unblock_sillyrename(parent);
2035                 return (struct dentry *)state;
2036         }
2037         res = d_add_unique(dentry, igrab(state->inode));
2038         if (res != NULL)
2039                 path.dentry = res;
2040         nfs_set_verifier(path.dentry, nfs_save_change_attribute(dir));
2041         nfs_unblock_sillyrename(parent);
2042         nfs4_intent_set_file(nd, &path, state, fmode);
2043         return res;
2044 }
2045
2046 int
2047 nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
2048 {
2049         struct path path = {
2050                 .mnt = nd->path.mnt,
2051                 .dentry = dentry,
2052         };
2053         struct rpc_cred *cred;
2054         struct nfs4_state *state;
2055         fmode_t fmode = openflags & (FMODE_READ | FMODE_WRITE);
2056
2057         cred = rpc_lookup_cred();
2058         if (IS_ERR(cred))
2059                 return PTR_ERR(cred);
2060         state = nfs4_do_open(dir, &path, fmode, openflags, NULL, cred);
2061         put_rpccred(cred);
2062         if (IS_ERR(state)) {
2063                 switch (PTR_ERR(state)) {
2064                         case -EPERM:
2065                         case -EACCES:
2066                         case -EDQUOT:
2067                         case -ENOSPC:
2068                         case -EROFS:
2069                                 lookup_instantiate_filp(nd, (struct dentry *)state, NULL);
2070                                 return 1;
2071                         default:
2072                                 goto out_drop;
2073                 }
2074         }
2075         if (state->inode == dentry->d_inode) {
2076                 nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2077                 nfs4_intent_set_file(nd, &path, state, fmode);
2078                 return 1;
2079         }
2080         nfs4_close_sync(&path, state, fmode);
2081 out_drop:
2082         d_drop(dentry);
2083         return 0;
2084 }
2085
2086 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2087 {
2088         if (ctx->state == NULL)
2089                 return;
2090         if (is_sync)
2091                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2092         else
2093                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2094 }
2095
2096 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2097 {
2098         struct nfs4_server_caps_arg args = {
2099                 .fhandle = fhandle,
2100         };
2101         struct nfs4_server_caps_res res = {};
2102         struct rpc_message msg = {
2103                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2104                 .rpc_argp = &args,
2105                 .rpc_resp = &res,
2106         };
2107         int status;
2108
2109         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2110         if (status == 0) {
2111                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2112                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2113                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2114                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2115                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2116                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2117                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2118                         server->caps |= NFS_CAP_ACLS;
2119                 if (res.has_links != 0)
2120                         server->caps |= NFS_CAP_HARDLINKS;
2121                 if (res.has_symlinks != 0)
2122                         server->caps |= NFS_CAP_SYMLINKS;
2123                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2124                         server->caps |= NFS_CAP_FILEID;
2125                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2126                         server->caps |= NFS_CAP_MODE;
2127                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2128                         server->caps |= NFS_CAP_NLINK;
2129                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2130                         server->caps |= NFS_CAP_OWNER;
2131                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2132                         server->caps |= NFS_CAP_OWNER_GROUP;
2133                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2134                         server->caps |= NFS_CAP_ATIME;
2135                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2136                         server->caps |= NFS_CAP_CTIME;
2137                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2138                         server->caps |= NFS_CAP_MTIME;
2139
2140                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2141                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2142                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2143                 server->acl_bitmask = res.acl_bitmask;
2144         }
2145
2146         return status;
2147 }
2148
2149 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2150 {
2151         struct nfs4_exception exception = { };
2152         int err;
2153         do {
2154                 err = nfs4_handle_exception(server,
2155                                 _nfs4_server_capabilities(server, fhandle),
2156                                 &exception);
2157         } while (exception.retry);
2158         return err;
2159 }
2160
2161 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2162                 struct nfs_fsinfo *info)
2163 {
2164         struct nfs4_lookup_root_arg args = {
2165                 .bitmask = nfs4_fattr_bitmap,
2166         };
2167         struct nfs4_lookup_res res = {
2168                 .server = server,
2169                 .fattr = info->fattr,
2170                 .fh = fhandle,
2171         };
2172         struct rpc_message msg = {
2173                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2174                 .rpc_argp = &args,
2175                 .rpc_resp = &res,
2176         };
2177
2178         nfs_fattr_init(info->fattr);
2179         return nfs4_call_sync(server, &msg, &args, &res, 0);
2180 }
2181
2182 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2183                 struct nfs_fsinfo *info)
2184 {
2185         struct nfs4_exception exception = { };
2186         int err;
2187         do {
2188                 err = nfs4_handle_exception(server,
2189                                 _nfs4_lookup_root(server, fhandle, info),
2190                                 &exception);
2191         } while (exception.retry);
2192         return err;
2193 }
2194
2195 /*
2196  * get the file handle for the "/" directory on the server
2197  */
2198 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2199                               struct nfs_fsinfo *info)
2200 {
2201         int status;
2202
2203         status = nfs4_lookup_root(server, fhandle, info);
2204         if (status == 0)
2205                 status = nfs4_server_capabilities(server, fhandle);
2206         if (status == 0)
2207                 status = nfs4_do_fsinfo(server, fhandle, info);
2208         return nfs4_map_errors(status);
2209 }
2210
2211 /*
2212  * Get locations and (maybe) other attributes of a referral.
2213  * Note that we'll actually follow the referral later when
2214  * we detect fsid mismatch in inode revalidation
2215  */
2216 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2217 {
2218         int status = -ENOMEM;
2219         struct page *page = NULL;
2220         struct nfs4_fs_locations *locations = NULL;
2221
2222         page = alloc_page(GFP_KERNEL);
2223         if (page == NULL)
2224                 goto out;
2225         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2226         if (locations == NULL)
2227                 goto out;
2228
2229         status = nfs4_proc_fs_locations(dir, name, locations, page);
2230         if (status != 0)
2231                 goto out;
2232         /* Make sure server returned a different fsid for the referral */
2233         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2234                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2235                 status = -EIO;
2236                 goto out;
2237         }
2238
2239         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2240         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2241         if (!fattr->mode)
2242                 fattr->mode = S_IFDIR;
2243         memset(fhandle, 0, sizeof(struct nfs_fh));
2244 out:
2245         if (page)
2246                 __free_page(page);
2247         if (locations)
2248                 kfree(locations);
2249         return status;
2250 }
2251
2252 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2253 {
2254         struct nfs4_getattr_arg args = {
2255                 .fh = fhandle,
2256                 .bitmask = server->attr_bitmask,
2257         };
2258         struct nfs4_getattr_res res = {
2259                 .fattr = fattr,
2260                 .server = server,
2261         };
2262         struct rpc_message msg = {
2263                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2264                 .rpc_argp = &args,
2265                 .rpc_resp = &res,
2266         };
2267
2268         nfs_fattr_init(fattr);
2269         return nfs4_call_sync(server, &msg, &args, &res, 0);
2270 }
2271
2272 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2273 {
2274         struct nfs4_exception exception = { };
2275         int err;
2276         do {
2277                 err = nfs4_handle_exception(server,
2278                                 _nfs4_proc_getattr(server, fhandle, fattr),
2279                                 &exception);
2280         } while (exception.retry);
2281         return err;
2282 }
2283
2284 /*
2285  * The file is not closed if it is opened due to the a request to change
2286  * the size of the file. The open call will not be needed once the
2287  * VFS layer lookup-intents are implemented.
2288  *
2289  * Close is called when the inode is destroyed.
2290  * If we haven't opened the file for O_WRONLY, we
2291  * need to in the size_change case to obtain a stateid.
2292  *
2293  * Got race?
2294  * Because OPEN is always done by name in nfsv4, it is
2295  * possible that we opened a different file by the same
2296  * name.  We can recognize this race condition, but we
2297  * can't do anything about it besides returning an error.
2298  *
2299  * This will be fixed with VFS changes (lookup-intent).
2300  */
2301 static int
2302 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2303                   struct iattr *sattr)
2304 {
2305         struct inode *inode = dentry->d_inode;
2306         struct rpc_cred *cred = NULL;
2307         struct nfs4_state *state = NULL;
2308         int status;
2309
2310         nfs_fattr_init(fattr);
2311
2312         /* Search for an existing open(O_WRITE) file */
2313         if (sattr->ia_valid & ATTR_FILE) {
2314                 struct nfs_open_context *ctx;
2315
2316                 ctx = nfs_file_open_context(sattr->ia_file);
2317                 if (ctx) {
2318                         cred = ctx->cred;
2319                         state = ctx->state;
2320                 }
2321         }
2322
2323         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2324         if (status == 0)
2325                 nfs_setattr_update_inode(inode, sattr);
2326         return status;
2327 }
2328
2329 static int _nfs4_proc_lookupfh(struct nfs_server *server, const struct nfs_fh *dirfh,
2330                 const struct qstr *name, struct nfs_fh *fhandle,
2331                 struct nfs_fattr *fattr)
2332 {
2333         int                    status;
2334         struct nfs4_lookup_arg args = {
2335                 .bitmask = server->attr_bitmask,
2336                 .dir_fh = dirfh,
2337                 .name = name,
2338         };
2339         struct nfs4_lookup_res res = {
2340                 .server = server,
2341                 .fattr = fattr,
2342                 .fh = fhandle,
2343         };
2344         struct rpc_message msg = {
2345                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2346                 .rpc_argp = &args,
2347                 .rpc_resp = &res,
2348         };
2349
2350         nfs_fattr_init(fattr);
2351
2352         dprintk("NFS call  lookupfh %s\n", name->name);
2353         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2354         dprintk("NFS reply lookupfh: %d\n", status);
2355         return status;
2356 }
2357
2358 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2359                               struct qstr *name, struct nfs_fh *fhandle,
2360                               struct nfs_fattr *fattr)
2361 {
2362         struct nfs4_exception exception = { };
2363         int err;
2364         do {
2365                 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
2366                 /* FIXME: !!!! */
2367                 if (err == -NFS4ERR_MOVED) {
2368                         err = -EREMOTE;
2369                         break;
2370                 }
2371                 err = nfs4_handle_exception(server, err, &exception);
2372         } while (exception.retry);
2373         return err;
2374 }
2375
2376 static int _nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
2377                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2378 {
2379         int status;
2380
2381         dprintk("NFS call  lookup %s\n", name->name);
2382         status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2383         if (status == -NFS4ERR_MOVED)
2384                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2385         dprintk("NFS reply lookup: %d\n", status);
2386         return status;
2387 }
2388
2389 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2390 {
2391         struct nfs4_exception exception = { };
2392         int err;
2393         do {
2394                 err = nfs4_handle_exception(NFS_SERVER(dir),
2395                                 _nfs4_proc_lookup(dir, name, fhandle, fattr),
2396                                 &exception);
2397         } while (exception.retry);
2398         return err;
2399 }
2400
2401 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2402 {
2403         struct nfs_server *server = NFS_SERVER(inode);
2404         struct nfs_fattr fattr;
2405         struct nfs4_accessargs args = {
2406                 .fh = NFS_FH(inode),
2407                 .bitmask = server->attr_bitmask,
2408         };
2409         struct nfs4_accessres res = {
2410                 .server = server,
2411                 .fattr = &fattr,
2412         };
2413         struct rpc_message msg = {
2414                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2415                 .rpc_argp = &args,
2416                 .rpc_resp = &res,
2417                 .rpc_cred = entry->cred,
2418         };
2419         int mode = entry->mask;
2420         int status;
2421
2422         /*
2423          * Determine which access bits we want to ask for...
2424          */
2425         if (mode & MAY_READ)
2426                 args.access |= NFS4_ACCESS_READ;
2427         if (S_ISDIR(inode->i_mode)) {
2428                 if (mode & MAY_WRITE)
2429                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2430                 if (mode & MAY_EXEC)
2431                         args.access |= NFS4_ACCESS_LOOKUP;
2432         } else {
2433                 if (mode & MAY_WRITE)
2434                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2435                 if (mode & MAY_EXEC)
2436                         args.access |= NFS4_ACCESS_EXECUTE;
2437         }
2438         nfs_fattr_init(&fattr);
2439         status = nfs4_call_sync(server, &msg, &args, &res, 0);
2440         if (!status) {
2441                 entry->mask = 0;
2442                 if (res.access & NFS4_ACCESS_READ)
2443                         entry->mask |= MAY_READ;
2444                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2445                         entry->mask |= MAY_WRITE;
2446                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2447                         entry->mask |= MAY_EXEC;
2448                 nfs_refresh_inode(inode, &fattr);
2449         }
2450         return status;
2451 }
2452
2453 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2454 {
2455         struct nfs4_exception exception = { };
2456         int err;
2457         do {
2458                 err = nfs4_handle_exception(NFS_SERVER(inode),
2459                                 _nfs4_proc_access(inode, entry),
2460                                 &exception);
2461         } while (exception.retry);
2462         return err;
2463 }
2464
2465 /*
2466  * TODO: For the time being, we don't try to get any attributes
2467  * along with any of the zero-copy operations READ, READDIR,
2468  * READLINK, WRITE.
2469  *
2470  * In the case of the first three, we want to put the GETATTR
2471  * after the read-type operation -- this is because it is hard
2472  * to predict the length of a GETATTR response in v4, and thus
2473  * align the READ data correctly.  This means that the GETATTR
2474  * may end up partially falling into the page cache, and we should
2475  * shift it into the 'tail' of the xdr_buf before processing.
2476  * To do this efficiently, we need to know the total length
2477  * of data received, which doesn't seem to be available outside
2478  * of the RPC layer.
2479  *
2480  * In the case of WRITE, we also want to put the GETATTR after
2481  * the operation -- in this case because we want to make sure
2482  * we get the post-operation mtime and size.  This means that
2483  * we can't use xdr_encode_pages() as written: we need a variant
2484  * of it which would leave room in the 'tail' iovec.
2485  *
2486  * Both of these changes to the XDR layer would in fact be quite
2487  * minor, but I decided to leave them for a subsequent patch.
2488  */
2489 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2490                 unsigned int pgbase, unsigned int pglen)
2491 {
2492         struct nfs4_readlink args = {
2493                 .fh       = NFS_FH(inode),
2494                 .pgbase   = pgbase,
2495                 .pglen    = pglen,
2496                 .pages    = &page,
2497         };
2498         struct nfs4_readlink_res res;
2499         struct rpc_message msg = {
2500                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2501                 .rpc_argp = &args,
2502                 .rpc_resp = &res,
2503         };
2504
2505         return nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
2506 }
2507
2508 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2509                 unsigned int pgbase, unsigned int pglen)
2510 {
2511         struct nfs4_exception exception = { };
2512         int err;
2513         do {
2514                 err = nfs4_handle_exception(NFS_SERVER(inode),
2515                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2516                                 &exception);
2517         } while (exception.retry);
2518         return err;
2519 }
2520
2521 /*
2522  * Got race?
2523  * We will need to arrange for the VFS layer to provide an atomic open.
2524  * Until then, this create/open method is prone to inefficiency and race
2525  * conditions due to the lookup, create, and open VFS calls from sys_open()
2526  * placed on the wire.
2527  *
2528  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2529  * The file will be opened again in the subsequent VFS open call
2530  * (nfs4_proc_file_open).
2531  *
2532  * The open for read will just hang around to be used by any process that
2533  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2534  */
2535
2536 static int
2537 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2538                  int flags, struct nameidata *nd)
2539 {
2540         struct path path = {
2541                 .mnt = nd->path.mnt,
2542                 .dentry = dentry,
2543         };
2544         struct nfs4_state *state;
2545         struct rpc_cred *cred;
2546         fmode_t fmode = flags & (FMODE_READ | FMODE_WRITE);
2547         int status = 0;
2548
2549         cred = rpc_lookup_cred();
2550         if (IS_ERR(cred)) {
2551                 status = PTR_ERR(cred);
2552                 goto out;
2553         }
2554         state = nfs4_do_open(dir, &path, fmode, flags, sattr, cred);
2555         d_drop(dentry);
2556         if (IS_ERR(state)) {
2557                 status = PTR_ERR(state);
2558                 goto out_putcred;
2559         }
2560         d_add(dentry, igrab(state->inode));
2561         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2562         if (flags & O_EXCL) {
2563                 struct nfs_fattr fattr;
2564                 status = nfs4_do_setattr(state->inode, cred, &fattr, sattr, state);
2565                 if (status == 0)
2566                         nfs_setattr_update_inode(state->inode, sattr);
2567                 nfs_post_op_update_inode(state->inode, &fattr);
2568         }
2569         if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0)
2570                 status = nfs4_intent_set_file(nd, &path, state, fmode);
2571         else
2572                 nfs4_close_sync(&path, state, fmode);
2573 out_putcred:
2574         put_rpccred(cred);
2575 out:
2576         return status;
2577 }
2578
2579 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2580 {
2581         struct nfs_server *server = NFS_SERVER(dir);
2582         struct nfs_removeargs args = {
2583                 .fh = NFS_FH(dir),
2584                 .name.len = name->len,
2585                 .name.name = name->name,
2586                 .bitmask = server->attr_bitmask,
2587         };
2588         struct nfs_removeres res = {
2589                 .server = server,
2590         };
2591         struct rpc_message msg = {
2592                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2593                 .rpc_argp = &args,
2594                 .rpc_resp = &res,
2595         };
2596         int                     status;
2597
2598         nfs_fattr_init(&res.dir_attr);
2599         status = nfs4_call_sync(server, &msg, &args, &res, 1);
2600         if (status == 0) {
2601                 update_changeattr(dir, &res.cinfo);
2602                 nfs_post_op_update_inode(dir, &res.dir_attr);
2603         }
2604         return status;
2605 }
2606
2607 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2608 {
2609         struct nfs4_exception exception = { };
2610         int err;
2611         do {
2612                 err = nfs4_handle_exception(NFS_SERVER(dir),
2613                                 _nfs4_proc_remove(dir, name),
2614                                 &exception);
2615         } while (exception.retry);
2616         return err;
2617 }
2618
2619 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2620 {
2621         struct nfs_server *server = NFS_SERVER(dir);
2622         struct nfs_removeargs *args = msg->rpc_argp;
2623         struct nfs_removeres *res = msg->rpc_resp;
2624
2625         args->bitmask = server->cache_consistency_bitmask;
2626         res->server = server;
2627         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2628 }
2629
2630 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2631 {
2632         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2633
2634         nfs4_sequence_done(res->server, &res->seq_res, task->tk_status);
2635         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2636                 return 0;
2637         update_changeattr(dir, &res->cinfo);
2638         nfs_post_op_update_inode(dir, &res->dir_attr);
2639         return 1;
2640 }
2641
2642 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2643                 struct inode *new_dir, struct qstr *new_name)
2644 {
2645         struct nfs_server *server = NFS_SERVER(old_dir);
2646         struct nfs4_rename_arg arg = {
2647                 .old_dir = NFS_FH(old_dir),
2648                 .new_dir = NFS_FH(new_dir),
2649                 .old_name = old_name,
2650                 .new_name = new_name,
2651                 .bitmask = server->attr_bitmask,
2652         };
2653         struct nfs_fattr old_fattr, new_fattr;
2654         struct nfs4_rename_res res = {
2655                 .server = server,
2656                 .old_fattr = &old_fattr,
2657                 .new_fattr = &new_fattr,
2658         };
2659         struct rpc_message msg = {
2660                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2661                 .rpc_argp = &arg,
2662                 .rpc_resp = &res,
2663         };
2664         int                     status;
2665
2666         nfs_fattr_init(res.old_fattr);
2667         nfs_fattr_init(res.new_fattr);
2668         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2669
2670         if (!status) {
2671                 update_changeattr(old_dir, &res.old_cinfo);
2672                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2673                 update_changeattr(new_dir, &res.new_cinfo);
2674                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2675         }
2676         return status;
2677 }
2678
2679 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2680                 struct inode *new_dir, struct qstr *new_name)
2681 {
2682         struct nfs4_exception exception = { };
2683         int err;
2684         do {
2685                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2686                                 _nfs4_proc_rename(old_dir, old_name,
2687                                         new_dir, new_name),
2688                                 &exception);
2689         } while (exception.retry);
2690         return err;
2691 }
2692
2693 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2694 {
2695         struct nfs_server *server = NFS_SERVER(inode);
2696         struct nfs4_link_arg arg = {
2697                 .fh     = NFS_FH(inode),
2698                 .dir_fh = NFS_FH(dir),
2699                 .name   = name,
2700                 .bitmask = server->attr_bitmask,
2701         };
2702         struct nfs_fattr fattr, dir_attr;
2703         struct nfs4_link_res res = {
2704                 .server = server,
2705                 .fattr = &fattr,
2706                 .dir_attr = &dir_attr,
2707         };
2708         struct rpc_message msg = {
2709                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2710                 .rpc_argp = &arg,
2711                 .rpc_resp = &res,
2712         };
2713         int                     status;
2714
2715         nfs_fattr_init(res.fattr);
2716         nfs_fattr_init(res.dir_attr);
2717         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
2718         if (!status) {
2719                 update_changeattr(dir, &res.cinfo);
2720                 nfs_post_op_update_inode(dir, res.dir_attr);
2721                 nfs_post_op_update_inode(inode, res.fattr);
2722         }
2723
2724         return status;
2725 }
2726
2727 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2728 {
2729         struct nfs4_exception exception = { };
2730         int err;
2731         do {
2732                 err = nfs4_handle_exception(NFS_SERVER(inode),
2733                                 _nfs4_proc_link(inode, dir, name),
2734                                 &exception);
2735         } while (exception.retry);
2736         return err;
2737 }
2738
2739 struct nfs4_createdata {
2740         struct rpc_message msg;
2741         struct nfs4_create_arg arg;
2742         struct nfs4_create_res res;
2743         struct nfs_fh fh;
2744         struct nfs_fattr fattr;
2745         struct nfs_fattr dir_fattr;
2746 };
2747
2748 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2749                 struct qstr *name, struct iattr *sattr, u32 ftype)
2750 {
2751         struct nfs4_createdata *data;
2752
2753         data = kzalloc(sizeof(*data), GFP_KERNEL);
2754         if (data != NULL) {
2755                 struct nfs_server *server = NFS_SERVER(dir);
2756
2757                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2758                 data->msg.rpc_argp = &data->arg;
2759                 data->msg.rpc_resp = &data->res;
2760                 data->arg.dir_fh = NFS_FH(dir);
2761                 data->arg.server = server;
2762                 data->arg.name = name;
2763                 data->arg.attrs = sattr;
2764                 data->arg.ftype = ftype;
2765                 data->arg.bitmask = server->attr_bitmask;
2766                 data->res.server = server;
2767                 data->res.fh = &data->fh;
2768                 data->res.fattr = &data->fattr;
2769                 data->res.dir_fattr = &data->dir_fattr;
2770                 nfs_fattr_init(data->res.fattr);
2771                 nfs_fattr_init(data->res.dir_fattr);
2772         }
2773         return data;
2774 }
2775
2776 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2777 {
2778         int status = nfs4_call_sync(NFS_SERVER(dir), &data->msg,
2779                                     &data->arg, &data->res, 1);
2780         if (status == 0) {
2781                 update_changeattr(dir, &data->res.dir_cinfo);
2782                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2783                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2784         }
2785         return status;
2786 }
2787
2788 static void nfs4_free_createdata(struct nfs4_createdata *data)
2789 {
2790         kfree(data);
2791 }
2792
2793 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2794                 struct page *page, unsigned int len, struct iattr *sattr)
2795 {
2796         struct nfs4_createdata *data;
2797         int status = -ENAMETOOLONG;
2798
2799         if (len > NFS4_MAXPATHLEN)
2800                 goto out;
2801
2802         status = -ENOMEM;
2803         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2804         if (data == NULL)
2805                 goto out;
2806
2807         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2808         data->arg.u.symlink.pages = &page;
2809         data->arg.u.symlink.len = len;
2810
2811         status = nfs4_do_create(dir, dentry, data);
2812
2813         nfs4_free_createdata(data);
2814 out:
2815         return status;
2816 }
2817
2818 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2819                 struct page *page, unsigned int len, struct iattr *sattr)
2820 {
2821         struct nfs4_exception exception = { };
2822         int err;
2823         do {
2824                 err = nfs4_handle_exception(NFS_SERVER(dir),
2825                                 _nfs4_proc_symlink(dir, dentry, page,
2826                                                         len, sattr),
2827                                 &exception);
2828         } while (exception.retry);
2829         return err;
2830 }
2831
2832 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2833                 struct iattr *sattr)
2834 {
2835         struct nfs4_createdata *data;
2836         int status = -ENOMEM;
2837
2838         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2839         if (data == NULL)
2840                 goto out;
2841
2842         status = nfs4_do_create(dir, dentry, data);
2843
2844         nfs4_free_createdata(data);
2845 out:
2846         return status;
2847 }
2848
2849 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2850                 struct iattr *sattr)
2851 {
2852         struct nfs4_exception exception = { };
2853         int err;
2854         do {
2855                 err = nfs4_handle_exception(NFS_SERVER(dir),
2856                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2857                                 &exception);
2858         } while (exception.retry);
2859         return err;
2860 }
2861
2862 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2863                   u64 cookie, struct page *page, unsigned int count, int plus)
2864 {
2865         struct inode            *dir = dentry->d_inode;
2866         struct nfs4_readdir_arg args = {
2867                 .fh = NFS_FH(dir),
2868                 .pages = &page,
2869                 .pgbase = 0,
2870                 .count = count,
2871                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2872         };
2873         struct nfs4_readdir_res res;
2874         struct rpc_message msg = {
2875                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2876                 .rpc_argp = &args,
2877                 .rpc_resp = &res,
2878                 .rpc_cred = cred,
2879         };
2880         int                     status;
2881
2882         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2883                         dentry->d_parent->d_name.name,
2884                         dentry->d_name.name,
2885                         (unsigned long long)cookie);
2886         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2887         res.pgbase = args.pgbase;
2888         status = nfs4_call_sync(NFS_SERVER(dir), &msg, &args, &res, 0);
2889         if (status == 0)
2890                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2891
2892         nfs_invalidate_atime(dir);
2893
2894         dprintk("%s: returns %d\n", __func__, status);
2895         return status;
2896 }
2897
2898 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2899                   u64 cookie, struct page *page, unsigned int count, int plus)
2900 {
2901         struct nfs4_exception exception = { };
2902         int err;
2903         do {
2904                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
2905                                 _nfs4_proc_readdir(dentry, cred, cookie,
2906                                         page, count, plus),
2907                                 &exception);
2908         } while (exception.retry);
2909         return err;
2910 }
2911
2912 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2913                 struct iattr *sattr, dev_t rdev)
2914 {
2915         struct nfs4_createdata *data;
2916         int mode = sattr->ia_mode;
2917         int status = -ENOMEM;
2918
2919         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
2920         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
2921
2922         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
2923         if (data == NULL)
2924                 goto out;
2925
2926         if (S_ISFIFO(mode))
2927                 data->arg.ftype = NF4FIFO;
2928         else if (S_ISBLK(mode)) {
2929                 data->arg.ftype = NF4BLK;
2930                 data->arg.u.device.specdata1 = MAJOR(rdev);
2931                 data->arg.u.device.specdata2 = MINOR(rdev);
2932         }
2933         else if (S_ISCHR(mode)) {
2934                 data->arg.ftype = NF4CHR;
2935                 data->arg.u.device.specdata1 = MAJOR(rdev);
2936                 data->arg.u.device.specdata2 = MINOR(rdev);
2937         }
2938
2939         status = nfs4_do_create(dir, dentry, data);
2940
2941         nfs4_free_createdata(data);
2942 out:
2943         return status;
2944 }
2945
2946 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
2947                 struct iattr *sattr, dev_t rdev)
2948 {
2949         struct nfs4_exception exception = { };
2950         int err;
2951         do {
2952                 err = nfs4_handle_exception(NFS_SERVER(dir),
2953                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
2954                                 &exception);
2955         } while (exception.retry);
2956         return err;
2957 }
2958
2959 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
2960                  struct nfs_fsstat *fsstat)
2961 {
2962         struct nfs4_statfs_arg args = {
2963                 .fh = fhandle,
2964                 .bitmask = server->attr_bitmask,
2965         };
2966         struct nfs4_statfs_res res = {
2967                 .fsstat = fsstat,
2968         };
2969         struct rpc_message msg = {
2970                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
2971                 .rpc_argp = &args,
2972                 .rpc_resp = &res,
2973         };
2974
2975         nfs_fattr_init(fsstat->fattr);
2976         return  nfs4_call_sync(server, &msg, &args, &res, 0);
2977 }
2978
2979 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
2980 {
2981         struct nfs4_exception exception = { };
2982         int err;
2983         do {
2984                 err = nfs4_handle_exception(server,
2985                                 _nfs4_proc_statfs(server, fhandle, fsstat),
2986                                 &exception);
2987         } while (exception.retry);
2988         return err;
2989 }
2990
2991 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
2992                 struct nfs_fsinfo *fsinfo)
2993 {
2994         struct nfs4_fsinfo_arg args = {
2995                 .fh = fhandle,
2996                 .bitmask = server->attr_bitmask,
2997         };
2998         struct nfs4_fsinfo_res res = {
2999                 .fsinfo = fsinfo,
3000         };
3001         struct rpc_message msg = {
3002                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3003                 .rpc_argp = &args,
3004                 .rpc_resp = &res,
3005         };
3006
3007         return nfs4_call_sync(server, &msg, &args, &res, 0);
3008 }
3009
3010 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3011 {
3012         struct nfs4_exception exception = { };
3013         int err;
3014
3015         do {
3016                 err = nfs4_handle_exception(server,
3017                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3018                                 &exception);
3019         } while (exception.retry);
3020         return err;
3021 }
3022
3023 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3024 {
3025         nfs_fattr_init(fsinfo->fattr);
3026         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3027 }
3028
3029 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3030                 struct nfs_pathconf *pathconf)
3031 {
3032         struct nfs4_pathconf_arg args = {
3033                 .fh = fhandle,
3034                 .bitmask = server->attr_bitmask,
3035         };
3036         struct nfs4_pathconf_res res = {
3037                 .pathconf = pathconf,
3038         };
3039         struct rpc_message msg = {
3040                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3041                 .rpc_argp = &args,
3042                 .rpc_resp = &res,
3043         };
3044
3045         /* None of the pathconf attributes are mandatory to implement */
3046         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3047                 memset(pathconf, 0, sizeof(*pathconf));
3048                 return 0;
3049         }
3050
3051         nfs_fattr_init(pathconf->fattr);
3052         return nfs4_call_sync(server, &msg, &args, &res, 0);
3053 }
3054
3055 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3056                 struct nfs_pathconf *pathconf)
3057 {
3058         struct nfs4_exception exception = { };
3059         int err;
3060
3061         do {
3062                 err = nfs4_handle_exception(server,
3063                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3064                                 &exception);
3065         } while (exception.retry);
3066         return err;
3067 }
3068
3069 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3070 {
3071         struct nfs_server *server = NFS_SERVER(data->inode);
3072
3073         dprintk("--> %s\n", __func__);
3074
3075         nfs4_sequence_done(server, &data->res.seq_res, task->tk_status);
3076
3077         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3078                 nfs_restart_rpc(task, server->nfs_client);
3079                 return -EAGAIN;
3080         }
3081
3082         nfs_invalidate_atime(data->inode);
3083         if (task->tk_status > 0)
3084                 renew_lease(server, data->timestamp);
3085         return 0;
3086 }
3087
3088 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3089 {
3090         data->timestamp   = jiffies;
3091         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3092 }
3093
3094 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3095 {
3096         struct inode *inode = data->inode;
3097
3098         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
3099                            task->tk_status);
3100
3101         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3102                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3103                 return -EAGAIN;
3104         }
3105         if (task->tk_status >= 0) {
3106                 renew_lease(NFS_SERVER(inode), data->timestamp);
3107                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3108         }
3109         return 0;
3110 }
3111
3112 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3113 {
3114         struct nfs_server *server = NFS_SERVER(data->inode);
3115
3116         data->args.bitmask = server->cache_consistency_bitmask;
3117         data->res.server = server;
3118         data->timestamp   = jiffies;
3119
3120         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3121 }
3122
3123 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3124 {
3125         struct inode *inode = data->inode;
3126
3127         nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res,
3128                            task->tk_status);
3129         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3130                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3131                 return -EAGAIN;
3132         }
3133         nfs_refresh_inode(inode, data->res.fattr);
3134         return 0;
3135 }
3136
3137 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3138 {
3139         struct nfs_server *server = NFS_SERVER(data->inode);
3140
3141         data->args.bitmask = server->cache_consistency_bitmask;
3142         data->res.server = server;
3143         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3144 }
3145
3146 /*
3147  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3148  * standalone procedure for queueing an asynchronous RENEW.
3149  */
3150 static void nfs4_renew_release(void *data)
3151 {
3152         struct nfs_client *clp = data;
3153
3154         nfs4_schedule_state_renewal(clp);
3155 }
3156
3157 static void nfs4_renew_done(struct rpc_task *task, void *data)
3158 {
3159         struct nfs_client *clp = data;
3160         unsigned long timestamp = task->tk_start;
3161
3162         if (task->tk_status < 0) {
3163                 /* Unless we're shutting down, schedule state recovery! */
3164                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3165                         nfs4_schedule_state_recovery(clp);
3166                 return;
3167         }
3168         spin_lock(&clp->cl_lock);
3169         if (time_before(clp->cl_last_renewal,timestamp))
3170                 clp->cl_last_renewal = timestamp;
3171         spin_unlock(&clp->cl_lock);
3172 }
3173
3174 static const struct rpc_call_ops nfs4_renew_ops = {
3175         .rpc_call_done = nfs4_renew_done,
3176         .rpc_release = nfs4_renew_release,
3177 };
3178
3179 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3180 {
3181         struct rpc_message msg = {
3182                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3183                 .rpc_argp       = clp,
3184                 .rpc_cred       = cred,
3185         };
3186
3187         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3188                         &nfs4_renew_ops, clp);
3189 }
3190
3191 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3192 {
3193         struct rpc_message msg = {
3194                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3195                 .rpc_argp       = clp,
3196                 .rpc_cred       = cred,
3197         };
3198         unsigned long now = jiffies;
3199         int status;
3200
3201         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3202         if (status < 0)
3203                 return status;
3204         spin_lock(&clp->cl_lock);
3205         if (time_before(clp->cl_last_renewal,now))
3206                 clp->cl_last_renewal = now;
3207         spin_unlock(&clp->cl_lock);
3208         return 0;
3209 }
3210
3211 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3212 {
3213         return (server->caps & NFS_CAP_ACLS)
3214                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3215                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3216 }
3217
3218 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3219  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3220  * the stack.
3221  */
3222 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3223
3224 static void buf_to_pages(const void *buf, size_t buflen,
3225                 struct page **pages, unsigned int *pgbase)
3226 {
3227         const void *p = buf;
3228
3229         *pgbase = offset_in_page(buf);
3230         p -= *pgbase;
3231         while (p < buf + buflen) {
3232                 *(pages++) = virt_to_page(p);
3233                 p += PAGE_CACHE_SIZE;
3234         }
3235 }
3236
3237 struct nfs4_cached_acl {
3238         int cached;
3239         size_t len;
3240         char data[0];
3241 };
3242
3243 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3244 {
3245         struct nfs_inode *nfsi = NFS_I(inode);
3246
3247         spin_lock(&inode->i_lock);
3248         kfree(nfsi->nfs4_acl);
3249         nfsi->nfs4_acl = acl;
3250         spin_unlock(&inode->i_lock);
3251 }
3252
3253 static void nfs4_zap_acl_attr(struct inode *inode)
3254 {
3255         nfs4_set_cached_acl(inode, NULL);
3256 }
3257
3258 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3259 {
3260         struct nfs_inode *nfsi = NFS_I(inode);
3261         struct nfs4_cached_acl *acl;
3262         int ret = -ENOENT;
3263
3264         spin_lock(&inode->i_lock);
3265         acl = nfsi->nfs4_acl;
3266         if (acl == NULL)
3267                 goto out;
3268         if (buf == NULL) /* user is just asking for length */
3269                 goto out_len;
3270         if (acl->cached == 0)
3271                 goto out;
3272         ret = -ERANGE; /* see getxattr(2) man page */
3273         if (acl->len > buflen)
3274                 goto out;
3275         memcpy(buf, acl->data, acl->len);
3276 out_len:
3277         ret = acl->len;
3278 out:
3279         spin_unlock(&inode->i_lock);
3280         return ret;
3281 }
3282
3283 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3284 {
3285         struct nfs4_cached_acl *acl;
3286
3287         if (buf && acl_len <= PAGE_SIZE) {
3288                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3289                 if (acl == NULL)
3290                         goto out;
3291                 acl->cached = 1;
3292                 memcpy(acl->data, buf, acl_len);
3293         } else {
3294                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3295                 if (acl == NULL)
3296                         goto out;
3297                 acl->cached = 0;
3298         }
3299         acl->len = acl_len;
3300 out:
3301         nfs4_set_cached_acl(inode, acl);
3302 }
3303
3304 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3305 {
3306         struct page *pages[NFS4ACL_MAXPAGES];
3307         struct nfs_getaclargs args = {
3308                 .fh = NFS_FH(inode),
3309                 .acl_pages = pages,
3310                 .acl_len = buflen,
3311         };
3312         struct nfs_getaclres res = {
3313                 .acl_len = buflen,
3314         };
3315         void *resp_buf;
3316         struct rpc_message msg = {
3317                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3318                 .rpc_argp = &args,
3319                 .rpc_resp = &res,
3320         };
3321         struct page *localpage = NULL;
3322         int ret;
3323
3324         if (buflen < PAGE_SIZE) {
3325                 /* As long as we're doing a round trip to the server anyway,
3326                  * let's be prepared for a page of acl data. */
3327                 localpage = alloc_page(GFP_KERNEL);
3328                 resp_buf = page_address(localpage);
3329                 if (localpage == NULL)
3330                         return -ENOMEM;
3331                 args.acl_pages[0] = localpage;
3332                 args.acl_pgbase = 0;
3333                 args.acl_len = PAGE_SIZE;
3334         } else {
3335                 resp_buf = buf;
3336                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3337         }
3338         ret = nfs4_call_sync(NFS_SERVER(inode), &msg, &args, &res, 0);
3339         if (ret)
3340                 goto out_free;
3341         if (res.acl_len > args.acl_len)
3342                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3343         else
3344                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3345         if (buf) {
3346                 ret = -ERANGE;
3347                 if (res.acl_len > buflen)
3348                         goto out_free;
3349                 if (localpage)
3350                         memcpy(buf, resp_buf, res.acl_len);
3351         }
3352         ret = res.acl_len;
3353 out_free:
3354         if (localpage)
3355                 __free_page(localpage);
3356         return ret;
3357 }
3358
3359 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3360 {
3361         struct nfs4_exception exception = { };
3362         ssize_t ret;
3363         do {
3364                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3365                 if (ret >= 0)
3366                         break;
3367                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3368         } while (exception.retry);
3369         return ret;
3370 }
3371
3372 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3373 {
3374         struct nfs_server *server = NFS_SERVER(inode);
3375         int ret;
3376
3377         if (!nfs4_server_supports_acls(server))
3378                 return -EOPNOTSUPP;
3379         ret = nfs_revalidate_inode(server, inode);
3380         if (ret < 0)
3381                 return ret;
3382         ret = nfs4_read_cached_acl(inode, buf, buflen);
3383         if (ret != -ENOENT)
3384                 return ret;
3385         return nfs4_get_acl_uncached(inode, buf, buflen);
3386 }
3387
3388 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3389 {
3390         struct nfs_server *server = NFS_SERVER(inode);
3391         struct page *pages[NFS4ACL_MAXPAGES];
3392         struct nfs_setaclargs arg = {
3393                 .fh             = NFS_FH(inode),
3394                 .acl_pages      = pages,
3395                 .acl_len        = buflen,
3396         };
3397         struct nfs_setaclres res;
3398         struct rpc_message msg = {
3399                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3400                 .rpc_argp       = &arg,
3401                 .rpc_resp       = &res,
3402         };
3403         int ret;
3404
3405         if (!nfs4_server_supports_acls(server))
3406                 return -EOPNOTSUPP;
3407         nfs_inode_return_delegation(inode);
3408         buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3409         ret = nfs4_call_sync(server, &msg, &arg, &res, 1);
3410         nfs_access_zap_cache(inode);
3411         nfs_zap_acl_cache(inode);
3412         return ret;
3413 }
3414
3415 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3416 {
3417         struct nfs4_exception exception = { };
3418         int err;
3419         do {
3420                 err = nfs4_handle_exception(NFS_SERVER(inode),
3421                                 __nfs4_proc_set_acl(inode, buf, buflen),
3422                                 &exception);
3423         } while (exception.retry);
3424         return err;
3425 }
3426
3427 static int
3428 _nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs_client *clp, struct nfs4_state *state)
3429 {
3430         if (!clp || task->tk_status >= 0)
3431                 return 0;
3432         switch(task->tk_status) {
3433                 case -NFS4ERR_ADMIN_REVOKED:
3434                 case -NFS4ERR_BAD_STATEID:
3435                 case -NFS4ERR_OPENMODE:
3436                         if (state == NULL)
3437                                 break;
3438                         nfs4_state_mark_reclaim_nograce(clp, state);
3439                         goto do_state_recovery;
3440                 case -NFS4ERR_STALE_STATEID:
3441                         if (state == NULL)
3442                                 break;
3443                         nfs4_state_mark_reclaim_reboot(clp, state);
3444                 case -NFS4ERR_STALE_CLIENTID:
3445                 case -NFS4ERR_EXPIRED:
3446                         goto do_state_recovery;
3447 #if defined(CONFIG_NFS_V4_1)
3448                 case -NFS4ERR_BADSESSION:
3449                 case -NFS4ERR_BADSLOT:
3450                 case -NFS4ERR_BAD_HIGH_SLOT:
3451                 case -NFS4ERR_DEADSESSION:
3452                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3453                 case -NFS4ERR_SEQ_FALSE_RETRY:
3454                 case -NFS4ERR_SEQ_MISORDERED:
3455                         dprintk("%s ERROR %d, Reset session\n", __func__,
3456                                 task->tk_status);
3457                         nfs4_schedule_state_recovery(clp);
3458                         task->tk_status = 0;
3459                         return -EAGAIN;
3460 #endif /* CONFIG_NFS_V4_1 */
3461                 case -NFS4ERR_DELAY:
3462                         if (server)
3463                                 nfs_inc_server_stats(server, NFSIOS_DELAY);
3464                 case -NFS4ERR_GRACE:
3465                 case -EKEYEXPIRED:
3466                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3467                         task->tk_status = 0;
3468                         return -EAGAIN;
3469                 case -NFS4ERR_OLD_STATEID:
3470                         task->tk_status = 0;
3471                         return -EAGAIN;
3472         }
3473         task->tk_status = nfs4_map_errors(task->tk_status);
3474         return 0;
3475 do_state_recovery:
3476         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3477         nfs4_schedule_state_recovery(clp);
3478         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3479                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3480         task->tk_status = 0;
3481         return -EAGAIN;
3482 }
3483
3484 static int
3485 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3486 {
3487         return _nfs4_async_handle_error(task, server, server->nfs_client, state);
3488 }
3489
3490 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program, unsigned short port, struct rpc_cred *cred)
3491 {
3492         nfs4_verifier sc_verifier;
3493         struct nfs4_setclientid setclientid = {
3494                 .sc_verifier = &sc_verifier,
3495                 .sc_prog = program,
3496         };
3497         struct rpc_message msg = {
3498                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3499                 .rpc_argp = &setclientid,
3500                 .rpc_resp = clp,
3501                 .rpc_cred = cred,
3502         };
3503         __be32 *p;
3504         int loop = 0;
3505         int status;
3506
3507         p = (__be32*)sc_verifier.data;
3508         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3509         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3510
3511         for(;;) {
3512                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3513                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3514                                 clp->cl_ipaddr,
3515                                 rpc_peeraddr2str(clp->cl_rpcclient,
3516                                                         RPC_DISPLAY_ADDR),
3517                                 rpc_peeraddr2str(clp->cl_rpcclient,
3518                                                         RPC_DISPLAY_PROTO),
3519                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3520                                 clp->cl_id_uniquifier);
3521                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3522                                 sizeof(setclientid.sc_netid),
3523                                 rpc_peeraddr2str(clp->cl_rpcclient,
3524                                                         RPC_DISPLAY_NETID));
3525                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3526                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3527                                 clp->cl_ipaddr, port >> 8, port & 255);
3528
3529                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3530                 if (status != -NFS4ERR_CLID_INUSE)
3531                         break;
3532                 if (signalled())
3533                         break;
3534                 if (loop++ & 1)
3535                         ssleep(clp->cl_lease_time + 1);
3536                 else
3537                         if (++clp->cl_id_uniquifier == 0)
3538                                 break;
3539         }
3540         return status;
3541 }
3542
3543 static int _nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3544 {
3545         struct nfs_fsinfo fsinfo;
3546         struct rpc_message msg = {
3547                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3548                 .rpc_argp = clp,
3549                 .rpc_resp = &fsinfo,
3550                 .rpc_cred = cred,
3551         };
3552         unsigned long now;
3553         int status;
3554
3555         now = jiffies;
3556         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3557         if (status == 0) {
3558                 spin_lock(&clp->cl_lock);
3559                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3560                 clp->cl_last_renewal = now;
3561                 spin_unlock(&clp->cl_lock);
3562         }
3563         return status;
3564 }
3565
3566 int nfs4_proc_setclientid_confirm(struct nfs_client *clp, struct rpc_cred *cred)
3567 {
3568         long timeout = 0;
3569         int err;
3570         do {
3571                 err = _nfs4_proc_setclientid_confirm(clp, cred);
3572                 switch (err) {
3573                         case 0:
3574                                 return err;
3575                         case -NFS4ERR_RESOURCE:
3576                                 /* The IBM lawyers misread another document! */
3577                         case -NFS4ERR_DELAY:
3578                         case -EKEYEXPIRED:
3579                                 err = nfs4_delay(clp->cl_rpcclient, &timeout);
3580                 }
3581         } while (err == 0);
3582         return err;
3583 }
3584
3585 struct nfs4_delegreturndata {
3586         struct nfs4_delegreturnargs args;
3587         struct nfs4_delegreturnres res;
3588         struct nfs_fh fh;
3589         nfs4_stateid stateid;
3590         unsigned long timestamp;
3591         struct nfs_fattr fattr;
3592         int rpc_status;
3593 };
3594
3595 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3596 {
3597         struct nfs4_delegreturndata *data = calldata;
3598
3599         nfs4_sequence_done(data->res.server, &data->res.seq_res,
3600                         task->tk_status);
3601
3602         switch (task->tk_status) {
3603         case -NFS4ERR_STALE_STATEID:
3604         case -NFS4ERR_EXPIRED:
3605         case 0:
3606                 renew_lease(data->res.server, data->timestamp);
3607                 break;
3608         default:
3609                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3610                                 -EAGAIN) {
3611                         nfs_restart_rpc(task, data->res.server->nfs_client);
3612                         return;
3613                 }
3614         }
3615         data->rpc_status = task->tk_status;
3616 }
3617
3618 static void nfs4_delegreturn_release(void *calldata)
3619 {
3620         kfree(calldata);
3621 }
3622
3623 #if defined(CONFIG_NFS_V4_1)
3624 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3625 {
3626         struct nfs4_delegreturndata *d_data;
3627
3628         d_data = (struct nfs4_delegreturndata *)data;
3629
3630         if (nfs4_setup_sequence(d_data->res.server->nfs_client,
3631                                 &d_data->args.seq_args,
3632                                 &d_data->res.seq_res, 1, task))
3633                 return;
3634         rpc_call_start(task);
3635 }
3636 #endif /* CONFIG_NFS_V4_1 */
3637
3638 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3639 #if defined(CONFIG_NFS_V4_1)
3640         .rpc_call_prepare = nfs4_delegreturn_prepare,
3641 #endif /* CONFIG_NFS_V4_1 */
3642         .rpc_call_done = nfs4_delegreturn_done,
3643         .rpc_release = nfs4_delegreturn_release,
3644 };
3645
3646 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3647 {
3648         struct nfs4_delegreturndata *data;
3649         struct nfs_server *server = NFS_SERVER(inode);
3650         struct rpc_task *task;
3651         struct rpc_message msg = {
3652                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3653                 .rpc_cred = cred,
3654         };
3655         struct rpc_task_setup task_setup_data = {
3656                 .rpc_client = server->client,
3657                 .rpc_message = &msg,
3658                 .callback_ops = &nfs4_delegreturn_ops,
3659                 .flags = RPC_TASK_ASYNC,
3660         };
3661         int status = 0;
3662
3663         data = kzalloc(sizeof(*data), GFP_KERNEL);
3664         if (data == NULL)
3665                 return -ENOMEM;
3666         data->args.fhandle = &data->fh;
3667         data->args.stateid = &data->stateid;
3668         data->args.bitmask = server->attr_bitmask;
3669         nfs_copy_fh(&data->fh, NFS_FH(inode));
3670         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3671         data->res.fattr = &data->fattr;
3672         data->res.server = server;
3673         data->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3674         nfs_fattr_init(data->res.fattr);
3675         data->timestamp = jiffies;
3676         data->rpc_status = 0;
3677
3678         task_setup_data.callback_data = data;
3679         msg.rpc_argp = &data->args,
3680         msg.rpc_resp = &data->res,
3681         task = rpc_run_task(&task_setup_data);
3682         if (IS_ERR(task))
3683                 return PTR_ERR(task);
3684         if (!issync)
3685                 goto out;
3686         status = nfs4_wait_for_completion_rpc_task(task);
3687         if (status != 0)
3688                 goto out;
3689         status = data->rpc_status;
3690         if (status != 0)
3691                 goto out;
3692         nfs_refresh_inode(inode, &data->fattr);
3693 out:
3694         rpc_put_task(task);
3695         return status;
3696 }
3697
3698 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3699 {
3700         struct nfs_server *server = NFS_SERVER(inode);
3701         struct nfs4_exception exception = { };
3702         int err;
3703         do {
3704                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3705                 switch (err) {
3706                         case -NFS4ERR_STALE_STATEID:
3707                         case -NFS4ERR_EXPIRED:
3708                         case 0:
3709                                 return 0;
3710                 }
3711                 err = nfs4_handle_exception(server, err, &exception);
3712         } while (exception.retry);
3713         return err;
3714 }
3715
3716 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3717 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3718
3719 /*
3720  * sleep, with exponential backoff, and retry the LOCK operation.
3721  */
3722 static unsigned long
3723 nfs4_set_lock_task_retry(unsigned long timeout)
3724 {
3725         schedule_timeout_killable(timeout);
3726         timeout <<= 1;
3727         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3728                 return NFS4_LOCK_MAXTIMEOUT;
3729         return timeout;
3730 }
3731
3732 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3733 {
3734         struct inode *inode = state->inode;
3735         struct nfs_server *server = NFS_SERVER(inode);
3736         struct nfs_client *clp = server->nfs_client;
3737         struct nfs_lockt_args arg = {
3738                 .fh = NFS_FH(inode),
3739                 .fl = request,
3740         };
3741         struct nfs_lockt_res res = {
3742                 .denied = request,
3743         };
3744         struct rpc_message msg = {
3745                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3746                 .rpc_argp       = &arg,
3747                 .rpc_resp       = &res,
3748                 .rpc_cred       = state->owner->so_cred,
3749         };
3750         struct nfs4_lock_state *lsp;
3751         int status;
3752
3753         arg.lock_owner.clientid = clp->cl_clientid;
3754         status = nfs4_set_lock_state(state, request);
3755         if (status != 0)
3756                 goto out;
3757         lsp = request->fl_u.nfs4_fl.owner;
3758         arg.lock_owner.id = lsp->ls_id.id;
3759         status = nfs4_call_sync(server, &msg, &arg, &res, 1);
3760         switch (status) {
3761                 case 0:
3762                         request->fl_type = F_UNLCK;
3763                         break;
3764                 case -NFS4ERR_DENIED:
3765                         status = 0;
3766         }
3767         request->fl_ops->fl_release_private(request);
3768 out:
3769         return status;
3770 }
3771
3772 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3773 {
3774         struct nfs4_exception exception = { };
3775         int err;
3776
3777         do {
3778                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3779                                 _nfs4_proc_getlk(state, cmd, request),
3780                                 &exception);
3781         } while (exception.retry);
3782         return err;
3783 }
3784
3785 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3786 {
3787         int res = 0;
3788         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3789                 case FL_POSIX:
3790                         res = posix_lock_file_wait(file, fl);
3791                         break;
3792                 case FL_FLOCK:
3793                         res = flock_lock_file_wait(file, fl);
3794                         break;
3795                 default:
3796                         BUG();
3797         }
3798         return res;
3799 }
3800
3801 struct nfs4_unlockdata {
3802         struct nfs_locku_args arg;
3803         struct nfs_locku_res res;
3804         struct nfs4_lock_state *lsp;
3805         struct nfs_open_context *ctx;
3806         struct file_lock fl;
3807         const struct nfs_server *server;
3808         unsigned long timestamp;
3809 };
3810
3811 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
3812                 struct nfs_open_context *ctx,
3813                 struct nfs4_lock_state *lsp,
3814                 struct nfs_seqid *seqid)
3815 {
3816         struct nfs4_unlockdata *p;
3817         struct inode *inode = lsp->ls_state->inode;
3818
3819         p = kzalloc(sizeof(*p), GFP_KERNEL);
3820         if (p == NULL)
3821                 return NULL;
3822         p->arg.fh = NFS_FH(inode);
3823         p->arg.fl = &p->fl;
3824         p->arg.seqid = seqid;
3825         p->res.seqid = seqid;
3826         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
3827         p->arg.stateid = &lsp->ls_stateid;
3828         p->lsp = lsp;
3829         atomic_inc(&lsp->ls_count);
3830         /* Ensure we don't close file until we're done freeing locks! */
3831         p->ctx = get_nfs_open_context(ctx);
3832         memcpy(&p->fl, fl, sizeof(p->fl));
3833         p->server = NFS_SERVER(inode);
3834         return p;
3835 }
3836
3837 static void nfs4_locku_release_calldata(void *data)
3838 {
3839         struct nfs4_unlockdata *calldata = data;
3840         nfs_free_seqid(calldata->arg.seqid);
3841         nfs4_put_lock_state(calldata->lsp);
3842         put_nfs_open_context(calldata->ctx);
3843         kfree(calldata);
3844 }
3845
3846 static void nfs4_locku_done(struct rpc_task *task, void *data)
3847 {
3848         struct nfs4_unlockdata *calldata = data;
3849
3850         nfs4_sequence_done(calldata->server, &calldata->res.seq_res,
3851                            task->tk_status);
3852         if (RPC_ASSASSINATED(task))
3853                 return;
3854         switch (task->tk_status) {
3855                 case 0:
3856                         memcpy(calldata->lsp->ls_stateid.data,
3857                                         calldata->res.stateid.data,
3858                                         sizeof(calldata->lsp->ls_stateid.data));
3859                         renew_lease(calldata->server, calldata->timestamp);
3860                         break;
3861                 case -NFS4ERR_BAD_STATEID:
3862                 case -NFS4ERR_OLD_STATEID:
3863                 case -NFS4ERR_STALE_STATEID:
3864                 case -NFS4ERR_EXPIRED:
3865                         break;
3866                 default:
3867                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
3868                                 nfs_restart_rpc(task,
3869                                                  calldata->server->nfs_client);
3870         }
3871 }
3872
3873 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
3874 {
3875         struct nfs4_unlockdata *calldata = data;
3876
3877         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3878                 return;
3879         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
3880                 /* Note: exit _without_ running nfs4_locku_done */
3881                 task->tk_action = NULL;
3882                 return;
3883         }
3884         calldata->timestamp = jiffies;
3885         if (nfs4_setup_sequence(calldata->server->nfs_client,
3886                                 &calldata->arg.seq_args,
3887                                 &calldata->res.seq_res, 1, task))
3888                 return;
3889         rpc_call_start(task);
3890 }
3891
3892 static const struct rpc_call_ops nfs4_locku_ops = {
3893         .rpc_call_prepare = nfs4_locku_prepare,
3894         .rpc_call_done = nfs4_locku_done,
3895         .rpc_release = nfs4_locku_release_calldata,
3896 };
3897
3898 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
3899                 struct nfs_open_context *ctx,
3900                 struct nfs4_lock_state *lsp,
3901                 struct nfs_seqid *seqid)
3902 {
3903         struct nfs4_unlockdata *data;
3904         struct rpc_message msg = {
3905                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
3906                 .rpc_cred = ctx->cred,
3907         };
3908         struct rpc_task_setup task_setup_data = {
3909                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
3910                 .rpc_message = &msg,
3911                 .callback_ops = &nfs4_locku_ops,
3912                 .workqueue = nfsiod_workqueue,
3913                 .flags = RPC_TASK_ASYNC,
3914         };
3915
3916         /* Ensure this is an unlock - when canceling a lock, the
3917          * canceled lock is passed in, and it won't be an unlock.
3918          */
3919         fl->fl_type = F_UNLCK;
3920
3921         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
3922         if (data == NULL) {
3923                 nfs_free_seqid(seqid);
3924                 return ERR_PTR(-ENOMEM);
3925         }
3926
3927         msg.rpc_argp = &data->arg,
3928         msg.rpc_resp = &data->res,
3929         task_setup_data.callback_data = data;
3930         return rpc_run_task(&task_setup_data);
3931 }
3932
3933 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
3934 {
3935         struct nfs_inode *nfsi = NFS_I(state->inode);
3936         struct nfs_seqid *seqid;
3937         struct nfs4_lock_state *lsp;
3938         struct rpc_task *task;
3939         int status = 0;
3940         unsigned char fl_flags = request->fl_flags;
3941
3942         status = nfs4_set_lock_state(state, request);
3943         /* Unlock _before_ we do the RPC call */
3944         request->fl_flags |= FL_EXISTS;
3945         down_read(&nfsi->rwsem);
3946         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
3947                 up_read(&nfsi->rwsem);
3948                 goto out;
3949         }
3950         up_read(&nfsi->rwsem);
3951         if (status != 0)
3952                 goto out;
3953         /* Is this a delegated lock? */
3954         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
3955                 goto out;
3956         lsp = request->fl_u.nfs4_fl.owner;
3957         seqid = nfs_alloc_seqid(&lsp->ls_seqid);
3958         status = -ENOMEM;
3959         if (seqid == NULL)
3960                 goto out;
3961         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
3962         status = PTR_ERR(task);
3963         if (IS_ERR(task))
3964                 goto out;
3965         status = nfs4_wait_for_completion_rpc_task(task);
3966         rpc_put_task(task);
3967 out:
3968         request->fl_flags = fl_flags;
3969         return status;
3970 }
3971
3972 struct nfs4_lockdata {
3973         struct nfs_lock_args arg;
3974         struct nfs_lock_res res;
3975         struct nfs4_lock_state *lsp;
3976         struct nfs_open_context *ctx;
3977         struct file_lock fl;
3978         unsigned long timestamp;
3979         int rpc_status;
3980         int cancelled;
3981         struct nfs_server *server;
3982 };
3983
3984 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
3985                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp)
3986 {
3987         struct nfs4_lockdata *p;
3988         struct inode *inode = lsp->ls_state->inode;
3989         struct nfs_server *server = NFS_SERVER(inode);
3990
3991         p = kzalloc(sizeof(*p), GFP_KERNEL);
3992         if (p == NULL)
3993                 return NULL;
3994
3995         p->arg.fh = NFS_FH(inode);
3996         p->arg.fl = &p->fl;
3997         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid);
3998         if (p->arg.open_seqid == NULL)
3999                 goto out_free;
4000         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
4001         if (p->arg.lock_seqid == NULL)
4002                 goto out_free_seqid;
4003         p->arg.lock_stateid = &lsp->ls_stateid;
4004         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4005         p->arg.lock_owner.id = lsp->ls_id.id;
4006         p->res.lock_seqid = p->arg.lock_seqid;
4007         p->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
4008         p->lsp = lsp;
4009         p->server = server;
4010         atomic_inc(&lsp->ls_count);
4011         p->ctx = get_nfs_open_context(ctx);
4012         memcpy(&p->fl, fl, sizeof(p->fl));
4013         return p;
4014 out_free_seqid:
4015         nfs_free_seqid(p->arg.open_seqid);
4016 out_free:
4017         kfree(p);
4018         return NULL;
4019 }
4020
4021 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4022 {
4023         struct nfs4_lockdata *data = calldata;
4024         struct nfs4_state *state = data->lsp->ls_state;
4025
4026         dprintk("%s: begin!\n", __func__);
4027         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4028                 return;
4029         /* Do we need to do an open_to_lock_owner? */
4030         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4031                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4032                         return;
4033                 data->arg.open_stateid = &state->stateid;
4034                 data->arg.new_lock_owner = 1;
4035                 data->res.open_seqid = data->arg.open_seqid;
4036         } else
4037                 data->arg.new_lock_owner = 0;
4038         data->timestamp = jiffies;
4039         if (nfs4_setup_sequence(data->server->nfs_client, &data->arg.seq_args,
4040                                 &data->res.seq_res, 1, task))
4041                 return;
4042         rpc_call_start(task);
4043         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4044 }
4045
4046 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4047 {
4048         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4049         nfs4_lock_prepare(task, calldata);
4050 }
4051
4052 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4053 {
4054         struct nfs4_lockdata *data = calldata;
4055
4056         dprintk("%s: begin!\n", __func__);
4057
4058         nfs4_sequence_done(data->server, &data->res.seq_res,
4059                         task->tk_status);
4060
4061         data->rpc_status = task->tk_status;
4062         if (RPC_ASSASSINATED(task))
4063                 goto out;
4064         if (data->arg.new_lock_owner != 0) {
4065                 if (data->rpc_status == 0)
4066                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4067                 else
4068                         goto out;
4069         }
4070         if (data->rpc_status == 0) {
4071                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4072                                         sizeof(data->lsp->ls_stateid.data));
4073                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4074                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4075         }
4076 out:
4077         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4078 }
4079
4080 static void nfs4_lock_release(void *calldata)
4081 {
4082         struct nfs4_lockdata *data = calldata;
4083
4084         dprintk("%s: begin!\n", __func__);
4085         nfs_free_seqid(data->arg.open_seqid);
4086         if (data->cancelled != 0) {
4087                 struct rpc_task *task;
4088                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4089                                 data->arg.lock_seqid);
4090                 if (!IS_ERR(task))
4091                         rpc_put_task(task);
4092                 dprintk("%s: cancelling lock!\n", __func__);
4093         } else
4094                 nfs_free_seqid(data->arg.lock_seqid);
4095         nfs4_put_lock_state(data->lsp);
4096         put_nfs_open_context(data->ctx);
4097         kfree(data);
4098         dprintk("%s: done!\n", __func__);
4099 }
4100
4101 static const struct rpc_call_ops nfs4_lock_ops = {
4102         .rpc_call_prepare = nfs4_lock_prepare,
4103         .rpc_call_done = nfs4_lock_done,
4104         .rpc_release = nfs4_lock_release,
4105 };
4106
4107 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4108         .rpc_call_prepare = nfs4_recover_lock_prepare,
4109         .rpc_call_done = nfs4_lock_done,
4110         .rpc_release = nfs4_lock_release,
4111 };
4112
4113 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4114 {
4115         struct nfs_client *clp = server->nfs_client;
4116         struct nfs4_state *state = lsp->ls_state;
4117
4118         switch (error) {
4119         case -NFS4ERR_ADMIN_REVOKED:
4120         case -NFS4ERR_BAD_STATEID:
4121         case -NFS4ERR_EXPIRED:
4122                 if (new_lock_owner != 0 ||
4123                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4124                         nfs4_state_mark_reclaim_nograce(clp, state);
4125                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4126                 break;
4127         case -NFS4ERR_STALE_STATEID:
4128                 if (new_lock_owner != 0 ||
4129                     (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4130                         nfs4_state_mark_reclaim_reboot(clp, state);
4131                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4132         };
4133 }
4134
4135 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4136 {
4137         struct nfs4_lockdata *data;
4138         struct rpc_task *task;
4139         struct rpc_message msg = {
4140                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4141                 .rpc_cred = state->owner->so_cred,
4142         };
4143         struct rpc_task_setup task_setup_data = {
4144                 .rpc_client = NFS_CLIENT(state->inode),
4145                 .rpc_message = &msg,
4146                 .callback_ops = &nfs4_lock_ops,
4147                 .workqueue = nfsiod_workqueue,
4148                 .flags = RPC_TASK_ASYNC,
4149         };
4150         int ret;
4151
4152         dprintk("%s: begin!\n", __func__);
4153         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4154                         fl->fl_u.nfs4_fl.owner);
4155         if (data == NULL)
4156                 return -ENOMEM;
4157         if (IS_SETLKW(cmd))
4158                 data->arg.block = 1;
4159         if (recovery_type > NFS_LOCK_NEW) {
4160                 if (recovery_type == NFS_LOCK_RECLAIM)
4161                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4162                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4163         }
4164         msg.rpc_argp = &data->arg,
4165         msg.rpc_resp = &data->res,
4166         task_setup_data.callback_data = data;
4167         task = rpc_run_task(&task_setup_data);
4168         if (IS_ERR(task))
4169                 return PTR_ERR(task);
4170         ret = nfs4_wait_for_completion_rpc_task(task);
4171         if (ret == 0) {
4172                 ret = data->rpc_status;
4173                 if (ret)
4174                         nfs4_handle_setlk_error(data->server, data->lsp,
4175                                         data->arg.new_lock_owner, ret);
4176         } else
4177                 data->cancelled = 1;
4178         rpc_put_task(task);
4179         dprintk("%s: done, ret = %d!\n", __func__, ret);
4180         return ret;
4181 }
4182
4183 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4184 {
4185         struct nfs_server *server = NFS_SERVER(state->inode);
4186         struct nfs4_exception exception = { };
4187         int err;
4188
4189         do {
4190                 /* Cache the lock if possible... */
4191                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4192                         return 0;
4193                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4194                 if (err != -NFS4ERR_DELAY && err != -EKEYEXPIRED)
4195                         break;
4196                 nfs4_handle_exception(server, err, &exception);
4197         } while (exception.retry);
4198         return err;
4199 }
4200
4201 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4202 {
4203         struct nfs_server *server = NFS_SERVER(state->inode);
4204         struct nfs4_exception exception = { };
4205         int err;
4206
4207         err = nfs4_set_lock_state(state, request);
4208         if (err != 0)
4209                 return err;
4210         do {
4211                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4212                         return 0;
4213                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4214                 switch (err) {
4215                 default:
4216                         goto out;
4217                 case -NFS4ERR_GRACE:
4218                 case -NFS4ERR_DELAY:
4219                 case -EKEYEXPIRED:
4220                         nfs4_handle_exception(server, err, &exception);
4221                         err = 0;
4222                 }
4223         } while (exception.retry);
4224 out:
4225         return err;
4226 }
4227
4228 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4229 {
4230         struct nfs_inode *nfsi = NFS_I(state->inode);
4231         unsigned char fl_flags = request->fl_flags;
4232         int status = -ENOLCK;
4233
4234         if ((fl_flags & FL_POSIX) &&
4235                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4236                 goto out;
4237         /* Is this a delegated open? */
4238         status = nfs4_set_lock_state(state, request);
4239         if (status != 0)
4240                 goto out;
4241         request->fl_flags |= FL_ACCESS;
4242         status = do_vfs_lock(request->fl_file, request);
4243         if (status < 0)
4244                 goto out;
4245         down_read(&nfsi->rwsem);
4246         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4247                 /* Yes: cache locks! */
4248                 /* ...but avoid races with delegation recall... */
4249                 request->fl_flags = fl_flags & ~FL_SLEEP;
4250                 status = do_vfs_lock(request->fl_file, request);
4251                 goto out_unlock;
4252         }
4253         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4254         if (status != 0)
4255                 goto out_unlock;
4256         /* Note: we always want to sleep here! */
4257         request->fl_flags = fl_flags | FL_SLEEP;
4258         if (do_vfs_lock(request->fl_file, request) < 0)
4259                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4260 out_unlock:
4261         up_read(&nfsi->rwsem);
4262 out:
4263         request->fl_flags = fl_flags;
4264         return status;
4265 }
4266
4267 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4268 {
4269         struct nfs4_exception exception = { };
4270         int err;
4271
4272         do {
4273                 err = _nfs4_proc_setlk(state, cmd, request);
4274                 if (err == -NFS4ERR_DENIED)
4275                         err = -EAGAIN;
4276                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4277                                 err, &exception);
4278         } while (exception.retry);
4279         return err;
4280 }
4281
4282 static int
4283 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4284 {
4285         struct nfs_open_context *ctx;
4286         struct nfs4_state *state;
4287         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4288         int status;
4289
4290         /* verify open state */
4291         ctx = nfs_file_open_context(filp);
4292         state = ctx->state;
4293
4294         if (request->fl_start < 0 || request->fl_end < 0)
4295                 return -EINVAL;
4296
4297         if (IS_GETLK(cmd)) {
4298                 if (state != NULL)
4299                         return nfs4_proc_getlk(state, F_GETLK, request);
4300                 return 0;
4301         }
4302
4303         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4304                 return -EINVAL;
4305
4306         if (request->fl_type == F_UNLCK) {
4307                 if (state != NULL)
4308                         return nfs4_proc_unlck(state, cmd, request);
4309                 return 0;
4310         }
4311
4312         if (state == NULL)
4313                 return -ENOLCK;
4314         do {
4315                 status = nfs4_proc_setlk(state, cmd, request);
4316                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4317                         break;
4318                 timeout = nfs4_set_lock_task_retry(timeout);
4319                 status = -ERESTARTSYS;
4320                 if (signalled())
4321                         break;
4322         } while(status < 0);
4323         return status;
4324 }
4325
4326 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4327 {
4328         struct nfs_server *server = NFS_SERVER(state->inode);
4329         struct nfs4_exception exception = { };
4330         int err;
4331
4332         err = nfs4_set_lock_state(state, fl);
4333         if (err != 0)
4334                 goto out;
4335         do {
4336                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4337                 switch (err) {
4338                         default:
4339                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4340                                                 __func__, err);
4341                         case 0:
4342                         case -ESTALE:
4343                                 goto out;
4344                         case -NFS4ERR_EXPIRED:
4345                         case -NFS4ERR_STALE_CLIENTID:
4346                         case -NFS4ERR_STALE_STATEID:
4347                         case -NFS4ERR_BADSESSION:
4348                         case -NFS4ERR_BADSLOT:
4349                         case -NFS4ERR_BAD_HIGH_SLOT:
4350                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4351                         case -NFS4ERR_DEADSESSION:
4352                                 nfs4_schedule_state_recovery(server->nfs_client);
4353                                 goto out;
4354                         case -ERESTARTSYS:
4355                                 /*
4356                                  * The show must go on: exit, but mark the
4357                                  * stateid as needing recovery.
4358                                  */
4359                         case -NFS4ERR_ADMIN_REVOKED:
4360                         case -NFS4ERR_BAD_STATEID:
4361                         case -NFS4ERR_OPENMODE:
4362                                 nfs4_state_mark_reclaim_nograce(server->nfs_client, state);
4363                                 err = 0;
4364                                 goto out;
4365                         case -ENOMEM:
4366                         case -NFS4ERR_DENIED:
4367                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4368                                 err = 0;
4369                                 goto out;
4370                         case -NFS4ERR_DELAY:
4371                         case -EKEYEXPIRED:
4372                                 break;
4373                 }
4374                 err = nfs4_handle_exception(server, err, &exception);
4375         } while (exception.retry);
4376 out:
4377         return err;
4378 }
4379
4380 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4381
4382 int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf,
4383                 size_t buflen, int flags)
4384 {
4385         struct inode *inode = dentry->d_inode;
4386
4387         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4388                 return -EOPNOTSUPP;
4389
4390         return nfs4_proc_set_acl(inode, buf, buflen);
4391 }
4392
4393 /* The getxattr man page suggests returning -ENODATA for unknown attributes,
4394  * and that's what we'll do for e.g. user attributes that haven't been set.
4395  * But we'll follow ext2/ext3's lead by returning -EOPNOTSUPP for unsupported
4396  * attributes in kernel-managed attribute namespaces. */
4397 ssize_t nfs4_getxattr(struct dentry *dentry, const char *key, void *buf,
4398                 size_t buflen)
4399 {
4400         struct inode *inode = dentry->d_inode;
4401
4402         if (strcmp(key, XATTR_NAME_NFSV4_ACL) != 0)
4403                 return -EOPNOTSUPP;
4404
4405         return nfs4_proc_get_acl(inode, buf, buflen);
4406 }
4407
4408 ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
4409 {
4410         size_t len = strlen(XATTR_NAME_NFSV4_ACL) + 1;
4411
4412         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4413                 return 0;
4414         if (buf && buflen < len)
4415                 return -ERANGE;
4416         if (buf)
4417                 memcpy(buf, XATTR_NAME_NFSV4_ACL, len);
4418         return len;
4419 }
4420
4421 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4422 {
4423         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4424                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4425                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4426                 return;
4427
4428         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4429                 NFS_ATTR_FATTR_NLINK;
4430         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4431         fattr->nlink = 2;
4432 }
4433
4434 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4435                 struct nfs4_fs_locations *fs_locations, struct page *page)
4436 {
4437         struct nfs_server *server = NFS_SERVER(dir);
4438         u32 bitmask[2] = {
4439                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4440                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4441         };
4442         struct nfs4_fs_locations_arg args = {
4443                 .dir_fh = NFS_FH(dir),
4444                 .name = name,
4445                 .page = page,
4446                 .bitmask = bitmask,
4447         };
4448         struct nfs4_fs_locations_res res = {
4449                 .fs_locations = fs_locations,
4450         };
4451         struct rpc_message msg = {
4452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4453                 .rpc_argp = &args,
4454                 .rpc_resp = &res,
4455         };
4456         int status;
4457
4458         dprintk("%s: start\n", __func__);
4459         nfs_fattr_init(&fs_locations->fattr);
4460         fs_locations->server = server;
4461         fs_locations->nlocations = 0;
4462         status = nfs4_call_sync(server, &msg, &args, &res, 0);
4463         nfs_fixup_referral_attributes(&fs_locations->fattr);
4464         dprintk("%s: returned status = %d\n", __func__, status);
4465         return status;
4466 }
4467
4468 #ifdef CONFIG_NFS_V4_1
4469 /*
4470  * nfs4_proc_exchange_id()
4471  *
4472  * Since the clientid has expired, all compounds using sessions
4473  * associated with the stale clientid will be returning
4474  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4475  * be in some phase of session reset.
4476  */
4477 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4478 {
4479         nfs4_verifier verifier;
4480         struct nfs41_exchange_id_args args = {
4481                 .client = clp,
4482                 .flags = clp->cl_exchange_flags,
4483         };
4484         struct nfs41_exchange_id_res res = {
4485                 .client = clp,
4486         };
4487         int status;
4488         struct rpc_message msg = {
4489                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4490                 .rpc_argp = &args,
4491                 .rpc_resp = &res,
4492                 .rpc_cred = cred,
4493         };
4494         __be32 *p;
4495
4496         dprintk("--> %s\n", __func__);
4497         BUG_ON(clp == NULL);
4498
4499         /* Remove server-only flags */
4500         args.flags &= ~EXCHGID4_FLAG_CONFIRMED_R;
4501
4502         p = (u32 *)verifier.data;
4503         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4504         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4505         args.verifier = &verifier;
4506
4507         while (1) {
4508                 args.id_len = scnprintf(args.id, sizeof(args.id),
4509                                         "%s/%s %u",
4510                                         clp->cl_ipaddr,
4511                                         rpc_peeraddr2str(clp->cl_rpcclient,
4512                                                          RPC_DISPLAY_ADDR),
4513                                         clp->cl_id_uniquifier);
4514
4515                 status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
4516
4517                 if (status != NFS4ERR_CLID_INUSE)
4518                         break;
4519
4520                 if (signalled())
4521                         break;
4522
4523                 if (++clp->cl_id_uniquifier == 0)
4524                         break;
4525         }
4526
4527         dprintk("<-- %s status= %d\n", __func__, status);
4528         return status;
4529 }
4530
4531 struct nfs4_get_lease_time_data {
4532         struct nfs4_get_lease_time_args *args;
4533         struct nfs4_get_lease_time_res *res;
4534         struct nfs_client *clp;
4535 };
4536
4537 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4538                                         void *calldata)
4539 {
4540         int ret;
4541         struct nfs4_get_lease_time_data *data =
4542                         (struct nfs4_get_lease_time_data *)calldata;
4543
4544         dprintk("--> %s\n", __func__);
4545         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4546         /* just setup sequence, do not trigger session recovery
4547            since we're invoked within one */
4548         ret = nfs41_setup_sequence(data->clp->cl_session,
4549                                    &data->args->la_seq_args,
4550                                    &data->res->lr_seq_res, 0, task);
4551
4552         BUG_ON(ret == -EAGAIN);
4553         rpc_call_start(task);
4554         dprintk("<-- %s\n", __func__);
4555 }
4556
4557 /*
4558  * Called from nfs4_state_manager thread for session setup, so don't recover
4559  * from sequence operation or clientid errors.
4560  */
4561 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4562 {
4563         struct nfs4_get_lease_time_data *data =
4564                         (struct nfs4_get_lease_time_data *)calldata;
4565
4566         dprintk("--> %s\n", __func__);
4567         nfs41_sequence_done(data->clp, &data->res->lr_seq_res, task->tk_status);
4568         switch (task->tk_status) {
4569         case -NFS4ERR_DELAY:
4570         case -NFS4ERR_GRACE:
4571         case -EKEYEXPIRED:
4572                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4573                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4574                 task->tk_status = 0;
4575                 nfs_restart_rpc(task, data->clp);
4576                 return;
4577         }
4578         dprintk("<-- %s\n", __func__);
4579 }
4580
4581 struct rpc_call_ops nfs4_get_lease_time_ops = {
4582         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4583         .rpc_call_done = nfs4_get_lease_time_done,
4584 };
4585
4586 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4587 {
4588         struct rpc_task *task;
4589         struct nfs4_get_lease_time_args args;
4590         struct nfs4_get_lease_time_res res = {
4591                 .lr_fsinfo = fsinfo,
4592         };
4593         struct nfs4_get_lease_time_data data = {
4594                 .args = &args,
4595                 .res = &res,
4596                 .clp = clp,
4597         };
4598         struct rpc_message msg = {
4599                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4600                 .rpc_argp = &args,
4601                 .rpc_resp = &res,
4602         };
4603         struct rpc_task_setup task_setup = {
4604                 .rpc_client = clp->cl_rpcclient,
4605                 .rpc_message = &msg,
4606                 .callback_ops = &nfs4_get_lease_time_ops,
4607                 .callback_data = &data
4608         };
4609         int status;
4610
4611         res.lr_seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
4612         dprintk("--> %s\n", __func__);
4613         task = rpc_run_task(&task_setup);
4614
4615         if (IS_ERR(task))
4616                 status = PTR_ERR(task);
4617         else {
4618                 status = task->tk_status;
4619                 rpc_put_task(task);
4620         }
4621         dprintk("<-- %s return %d\n", __func__, status);
4622
4623         return status;
4624 }
4625
4626 /*
4627  * Reset a slot table
4628  */
4629 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4630                                  int ivalue)
4631 {
4632         struct nfs4_slot *new = NULL;
4633         int i;
4634         int ret = 0;
4635
4636         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4637                 max_reqs, tbl->max_slots);
4638
4639         /* Does the newly negotiated max_reqs match the existing slot table? */
4640         if (max_reqs != tbl->max_slots) {
4641                 ret = -ENOMEM;
4642                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4643                               GFP_KERNEL);
4644                 if (!new)
4645                         goto out;
4646                 ret = 0;
4647                 kfree(tbl->slots);
4648         }
4649         spin_lock(&tbl->slot_tbl_lock);
4650         if (new) {
4651                 tbl->slots = new;
4652                 tbl->max_slots = max_reqs;
4653         }
4654         for (i = 0; i < tbl->max_slots; ++i)
4655                 tbl->slots[i].seq_nr = ivalue;
4656         spin_unlock(&tbl->slot_tbl_lock);
4657         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4658                 tbl, tbl->slots, tbl->max_slots);
4659 out:
4660         dprintk("<-- %s: return %d\n", __func__, ret);
4661         return ret;
4662 }
4663
4664 /*
4665  * Reset the forechannel and backchannel slot tables
4666  */
4667 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4668 {
4669         int status;
4670
4671         status = nfs4_reset_slot_table(&session->fc_slot_table,
4672                         session->fc_attrs.max_reqs, 1);
4673         if (status)
4674                 return status;
4675
4676         status = nfs4_reset_slot_table(&session->bc_slot_table,
4677                         session->bc_attrs.max_reqs, 0);
4678         return status;
4679 }
4680
4681 /* Destroy the slot table */
4682 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4683 {
4684         if (session->fc_slot_table.slots != NULL) {
4685                 kfree(session->fc_slot_table.slots);
4686                 session->fc_slot_table.slots = NULL;
4687         }
4688         if (session->bc_slot_table.slots != NULL) {
4689                 kfree(session->bc_slot_table.slots);
4690                 session->bc_slot_table.slots = NULL;
4691         }
4692         return;
4693 }
4694
4695 /*
4696  * Initialize slot table
4697  */
4698 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4699                 int max_slots, int ivalue)
4700 {
4701         struct nfs4_slot *slot;
4702         int ret = -ENOMEM;
4703
4704         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4705
4706         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4707
4708         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_KERNEL);
4709         if (!slot)
4710                 goto out;
4711         ret = 0;
4712
4713         spin_lock(&tbl->slot_tbl_lock);
4714         tbl->max_slots = max_slots;
4715         tbl->slots = slot;
4716         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4717         spin_unlock(&tbl->slot_tbl_lock);
4718         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4719                 tbl, tbl->slots, tbl->max_slots);
4720 out:
4721         dprintk("<-- %s: return %d\n", __func__, ret);
4722         return ret;
4723 }
4724
4725 /*
4726  * Initialize the forechannel and backchannel tables
4727  */
4728 static int nfs4_init_slot_tables(struct nfs4_session *session)
4729 {
4730         struct nfs4_slot_table *tbl;
4731         int status = 0;
4732
4733         tbl = &session->fc_slot_table;
4734         if (tbl->slots == NULL) {
4735                 status = nfs4_init_slot_table(tbl,
4736                                 session->fc_attrs.max_reqs, 1);
4737                 if (status)
4738                         return status;
4739         }
4740
4741         tbl = &session->bc_slot_table;
4742         if (tbl->slots == NULL) {
4743                 status = nfs4_init_slot_table(tbl,
4744                                 session->bc_attrs.max_reqs, 0);
4745                 if (status)
4746                         nfs4_destroy_slot_tables(session);
4747         }
4748
4749         return status;
4750 }
4751
4752 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
4753 {
4754         struct nfs4_session *session;
4755         struct nfs4_slot_table *tbl;
4756
4757         session = kzalloc(sizeof(struct nfs4_session), GFP_KERNEL);
4758         if (!session)
4759                 return NULL;
4760
4761         /*
4762          * The create session reply races with the server back
4763          * channel probe. Mark the client NFS_CS_SESSION_INITING
4764          * so that the client back channel can find the
4765          * nfs_client struct
4766          */
4767         clp->cl_cons_state = NFS_CS_SESSION_INITING;
4768         init_completion(&session->complete);
4769
4770         tbl = &session->fc_slot_table;
4771         tbl->highest_used_slotid = -1;
4772         spin_lock_init(&tbl->slot_tbl_lock);
4773         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
4774
4775         tbl = &session->bc_slot_table;
4776         tbl->highest_used_slotid = -1;
4777         spin_lock_init(&tbl->slot_tbl_lock);
4778         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
4779
4780         session->clp = clp;
4781         return session;
4782 }
4783
4784 void nfs4_destroy_session(struct nfs4_session *session)
4785 {
4786         nfs4_proc_destroy_session(session);
4787         dprintk("%s Destroy backchannel for xprt %p\n",
4788                 __func__, session->clp->cl_rpcclient->cl_xprt);
4789         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
4790                                 NFS41_BC_MIN_CALLBACKS);
4791         nfs4_destroy_slot_tables(session);
4792         kfree(session);
4793 }
4794
4795 /*
4796  * Initialize the values to be used by the client in CREATE_SESSION
4797  * If nfs4_init_session set the fore channel request and response sizes,
4798  * use them.
4799  *
4800  * Set the back channel max_resp_sz_cached to zero to force the client to
4801  * always set csa_cachethis to FALSE because the current implementation
4802  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
4803  */
4804 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
4805 {
4806         struct nfs4_session *session = args->client->cl_session;
4807         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
4808                      mxresp_sz = session->fc_attrs.max_resp_sz;
4809
4810         if (mxrqst_sz == 0)
4811                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
4812         if (mxresp_sz == 0)
4813                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
4814         /* Fore channel attributes */
4815         args->fc_attrs.headerpadsz = 0;
4816         args->fc_attrs.max_rqst_sz = mxrqst_sz;
4817         args->fc_attrs.max_resp_sz = mxresp_sz;
4818         args->fc_attrs.max_ops = NFS4_MAX_OPS;
4819         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
4820
4821         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
4822                 "max_ops=%u max_reqs=%u\n",
4823                 __func__,
4824                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
4825                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
4826
4827         /* Back channel attributes */
4828         args->bc_attrs.headerpadsz = 0;
4829         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
4830         args->bc_attrs.max_resp_sz = PAGE_SIZE;
4831         args->bc_attrs.max_resp_sz_cached = 0;
4832         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
4833         args->bc_attrs.max_reqs = 1;
4834
4835         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
4836                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
4837                 __func__,
4838                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
4839                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
4840                 args->bc_attrs.max_reqs);
4841 }
4842
4843 static int _verify_channel_attr(char *chan, char *attr_name, u32 sent, u32 rcvd)
4844 {
4845         if (rcvd <= sent)
4846                 return 0;
4847         printk(KERN_WARNING "%s: Session INVALID: %s channel %s increased. "
4848                 "sent=%u rcvd=%u\n", __func__, chan, attr_name, sent, rcvd);
4849         return -EINVAL;
4850 }
4851
4852 #define _verify_fore_channel_attr(_name_) \
4853         _verify_channel_attr("fore", #_name_, \
4854                              args->fc_attrs._name_, \
4855                              session->fc_attrs._name_)
4856
4857 #define _verify_back_channel_attr(_name_) \
4858         _verify_channel_attr("back", #_name_, \
4859                              args->bc_attrs._name_, \
4860                              session->bc_attrs._name_)
4861
4862 /*
4863  * The server is not allowed to increase the fore channel header pad size,
4864  * maximum response size, or maximum number of operations.
4865  *
4866  * The back channel attributes are only negotiatied down: We send what the
4867  * (back channel) server insists upon.
4868  */
4869 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
4870                                      struct nfs4_session *session)
4871 {
4872         int ret = 0;
4873
4874         ret |= _verify_fore_channel_attr(headerpadsz);
4875         ret |= _verify_fore_channel_attr(max_resp_sz);
4876         ret |= _verify_fore_channel_attr(max_ops);
4877
4878         ret |= _verify_back_channel_attr(headerpadsz);
4879         ret |= _verify_back_channel_attr(max_rqst_sz);
4880         ret |= _verify_back_channel_attr(max_resp_sz);
4881         ret |= _verify_back_channel_attr(max_resp_sz_cached);
4882         ret |= _verify_back_channel_attr(max_ops);
4883         ret |= _verify_back_channel_attr(max_reqs);
4884
4885         return ret;
4886 }
4887
4888 static int _nfs4_proc_create_session(struct nfs_client *clp)
4889 {
4890         struct nfs4_session *session = clp->cl_session;
4891         struct nfs41_create_session_args args = {
4892                 .client = clp,
4893                 .cb_program = NFS4_CALLBACK,
4894         };
4895         struct nfs41_create_session_res res = {
4896                 .client = clp,
4897         };
4898         struct rpc_message msg = {
4899                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
4900                 .rpc_argp = &args,
4901                 .rpc_resp = &res,
4902         };
4903         int status;
4904
4905         nfs4_init_channel_attrs(&args);
4906         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
4907
4908         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4909
4910         if (!status)
4911                 /* Verify the session's negotiated channel_attrs values */
4912                 status = nfs4_verify_channel_attrs(&args, session);
4913         if (!status) {
4914                 /* Increment the clientid slot sequence id */
4915                 clp->cl_seqid++;
4916         }
4917
4918         return status;
4919 }
4920
4921 /*
4922  * Issues a CREATE_SESSION operation to the server.
4923  * It is the responsibility of the caller to verify the session is
4924  * expired before calling this routine.
4925  */
4926 int nfs4_proc_create_session(struct nfs_client *clp)
4927 {
4928         int status;
4929         unsigned *ptr;
4930         struct nfs4_session *session = clp->cl_session;
4931
4932         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
4933
4934         status = _nfs4_proc_create_session(clp);
4935         if (status)
4936                 goto out;
4937
4938         /* Init and reset the fore channel */
4939         status = nfs4_init_slot_tables(session);
4940         dprintk("slot table initialization returned %d\n", status);
4941         if (status)
4942                 goto out;
4943         status = nfs4_reset_slot_tables(session);
4944         dprintk("slot table reset returned %d\n", status);
4945         if (status)
4946                 goto out;
4947
4948         ptr = (unsigned *)&session->sess_id.data[0];
4949         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
4950                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
4951 out:
4952         dprintk("<-- %s\n", __func__);
4953         return status;
4954 }
4955
4956 /*
4957  * Issue the over-the-wire RPC DESTROY_SESSION.
4958  * The caller must serialize access to this routine.
4959  */
4960 int nfs4_proc_destroy_session(struct nfs4_session *session)
4961 {
4962         int status = 0;
4963         struct rpc_message msg;
4964
4965         dprintk("--> nfs4_proc_destroy_session\n");
4966
4967         /* session is still being setup */
4968         if (session->clp->cl_cons_state != NFS_CS_READY)
4969                 return status;
4970
4971         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
4972         msg.rpc_argp = session;
4973         msg.rpc_resp = NULL;
4974         msg.rpc_cred = NULL;
4975         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, 0);
4976
4977         if (status)
4978                 printk(KERN_WARNING
4979                         "Got error %d from the server on DESTROY_SESSION. "
4980                         "Session has been destroyed regardless...\n", status);
4981
4982         dprintk("<-- nfs4_proc_destroy_session\n");
4983         return status;
4984 }
4985
4986 int nfs4_init_session(struct nfs_server *server)
4987 {
4988         struct nfs_client *clp = server->nfs_client;
4989         struct nfs4_session *session;
4990         unsigned int rsize, wsize;
4991         int ret;
4992
4993         if (!nfs4_has_session(clp))
4994                 return 0;
4995
4996         rsize = server->rsize;
4997         if (rsize == 0)
4998                 rsize = NFS_MAX_FILE_IO_SIZE;
4999         wsize = server->wsize;
5000         if (wsize == 0)
5001                 wsize = NFS_MAX_FILE_IO_SIZE;
5002
5003         session = clp->cl_session;
5004         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5005         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5006
5007         ret = nfs4_recover_expired_lease(server);
5008         if (!ret)
5009                 ret = nfs4_check_client_ready(clp);
5010         return ret;
5011 }
5012
5013 /*
5014  * Renew the cl_session lease.
5015  */
5016 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5017 {
5018         struct nfs4_sequence_args args;
5019         struct nfs4_sequence_res res;
5020
5021         struct rpc_message msg = {
5022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5023                 .rpc_argp = &args,
5024                 .rpc_resp = &res,
5025                 .rpc_cred = cred,
5026         };
5027
5028         args.sa_cache_this = 0;
5029
5030         return nfs4_call_sync_sequence(clp, clp->cl_rpcclient, &msg, &args,
5031                                        &res, args.sa_cache_this, 1);
5032 }
5033
5034 static void nfs41_sequence_release(void *data)
5035 {
5036         struct nfs_client *clp = (struct nfs_client *)data;
5037
5038         nfs4_schedule_state_renewal(clp);
5039 }
5040
5041 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5042 {
5043         struct nfs_client *clp = (struct nfs_client *)data;
5044
5045         nfs41_sequence_done(clp, task->tk_msg.rpc_resp, task->tk_status);
5046
5047         if (task->tk_status < 0) {
5048                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5049
5050                 if (_nfs4_async_handle_error(task, NULL, clp, NULL)
5051                                                                 == -EAGAIN) {
5052                         nfs_restart_rpc(task, clp);
5053                         return;
5054                 }
5055         }
5056         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5057
5058         kfree(task->tk_msg.rpc_argp);
5059         kfree(task->tk_msg.rpc_resp);
5060
5061         dprintk("<-- %s\n", __func__);
5062 }
5063
5064 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5065 {
5066         struct nfs_client *clp;
5067         struct nfs4_sequence_args *args;
5068         struct nfs4_sequence_res *res;
5069
5070         clp = (struct nfs_client *)data;
5071         args = task->tk_msg.rpc_argp;
5072         res = task->tk_msg.rpc_resp;
5073
5074         if (nfs4_setup_sequence(clp, args, res, 0, task))
5075                 return;
5076         rpc_call_start(task);
5077 }
5078
5079 static const struct rpc_call_ops nfs41_sequence_ops = {
5080         .rpc_call_done = nfs41_sequence_call_done,
5081         .rpc_call_prepare = nfs41_sequence_prepare,
5082         .rpc_release = nfs41_sequence_release,
5083 };
5084
5085 static int nfs41_proc_async_sequence(struct nfs_client *clp,
5086                                      struct rpc_cred *cred)
5087 {
5088         struct nfs4_sequence_args *args;
5089         struct nfs4_sequence_res *res;
5090         struct rpc_message msg = {
5091                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5092                 .rpc_cred = cred,
5093         };
5094
5095         args = kzalloc(sizeof(*args), GFP_KERNEL);
5096         if (!args)
5097                 return -ENOMEM;
5098         res = kzalloc(sizeof(*res), GFP_KERNEL);
5099         if (!res) {
5100                 kfree(args);
5101                 return -ENOMEM;
5102         }
5103         res->sr_slotid = NFS4_MAX_SLOT_TABLE;
5104         msg.rpc_argp = args;
5105         msg.rpc_resp = res;
5106
5107         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
5108                               &nfs41_sequence_ops, (void *)clp);
5109 }
5110
5111 struct nfs4_reclaim_complete_data {
5112         struct nfs_client *clp;
5113         struct nfs41_reclaim_complete_args arg;
5114         struct nfs41_reclaim_complete_res res;
5115 };
5116
5117 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5118 {
5119         struct nfs4_reclaim_complete_data *calldata = data;
5120
5121         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5122         if (nfs4_setup_sequence(calldata->clp, &calldata->arg.seq_args,
5123                                 &calldata->res.seq_res, 0, task))
5124                 return;
5125
5126         rpc_call_start(task);
5127 }
5128
5129 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5130 {
5131         struct nfs4_reclaim_complete_data *calldata = data;
5132         struct nfs_client *clp = calldata->clp;
5133         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5134
5135         dprintk("--> %s\n", __func__);
5136         nfs41_sequence_done(clp, res, task->tk_status);
5137         switch (task->tk_status) {
5138         case 0:
5139         case -NFS4ERR_COMPLETE_ALREADY:
5140                 break;
5141         case -NFS4ERR_BADSESSION:
5142         case -NFS4ERR_DEADSESSION:
5143                 /*
5144                  * Handle the session error, but do not retry the operation, as
5145                  * we have no way of telling whether the clientid had to be
5146                  * reset before we got our reply.  If reset, a new wave of
5147                  * reclaim operations will follow, containing their own reclaim
5148                  * complete.  We don't want our retry to get on the way of
5149                  * recovery by incorrectly indicating to the server that we're
5150                  * done reclaiming state since the process had to be restarted.
5151                  */
5152                 _nfs4_async_handle_error(task, NULL, clp, NULL);
5153                 break;
5154         default:
5155                 if (_nfs4_async_handle_error(
5156                                 task, NULL, clp, NULL) == -EAGAIN) {
5157                         rpc_restart_call_prepare(task);
5158                         return;
5159                 }
5160         }
5161
5162         dprintk("<-- %s\n", __func__);
5163 }
5164
5165 static void nfs4_free_reclaim_complete_data(void *data)
5166 {
5167         struct nfs4_reclaim_complete_data *calldata = data;
5168
5169         kfree(calldata);
5170 }
5171
5172 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5173         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5174         .rpc_call_done = nfs4_reclaim_complete_done,
5175         .rpc_release = nfs4_free_reclaim_complete_data,
5176 };
5177
5178 /*
5179  * Issue a global reclaim complete.
5180  */
5181 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5182 {
5183         struct nfs4_reclaim_complete_data *calldata;
5184         struct rpc_task *task;
5185         struct rpc_message msg = {
5186                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5187         };
5188         struct rpc_task_setup task_setup_data = {
5189                 .rpc_client = clp->cl_rpcclient,
5190                 .rpc_message = &msg,
5191                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5192                 .flags = RPC_TASK_ASYNC,
5193         };
5194         int status = -ENOMEM;
5195
5196         dprintk("--> %s\n", __func__);
5197         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
5198         if (calldata == NULL)
5199                 goto out;
5200         calldata->clp = clp;
5201         calldata->arg.one_fs = 0;
5202         calldata->res.seq_res.sr_slotid = NFS4_MAX_SLOT_TABLE;
5203
5204         msg.rpc_argp = &calldata->arg;
5205         msg.rpc_resp = &calldata->res;
5206         task_setup_data.callback_data = calldata;
5207         task = rpc_run_task(&task_setup_data);
5208         if (IS_ERR(task))
5209                 status = PTR_ERR(task);
5210         rpc_put_task(task);
5211 out:
5212         dprintk("<-- %s status=%d\n", __func__, status);
5213         return status;
5214 }
5215 #endif /* CONFIG_NFS_V4_1 */
5216
5217 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5218         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5219         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5220         .recover_open   = nfs4_open_reclaim,
5221         .recover_lock   = nfs4_lock_reclaim,
5222         .establish_clid = nfs4_init_clientid,
5223         .get_clid_cred  = nfs4_get_setclientid_cred,
5224 };
5225
5226 #if defined(CONFIG_NFS_V4_1)
5227 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5228         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5229         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5230         .recover_open   = nfs4_open_reclaim,
5231         .recover_lock   = nfs4_lock_reclaim,
5232         .establish_clid = nfs41_init_clientid,
5233         .get_clid_cred  = nfs4_get_exchange_id_cred,
5234         .reclaim_complete = nfs41_proc_reclaim_complete,
5235 };
5236 #endif /* CONFIG_NFS_V4_1 */
5237
5238 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5239         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5240         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5241         .recover_open   = nfs4_open_expired,
5242         .recover_lock   = nfs4_lock_expired,
5243         .establish_clid = nfs4_init_clientid,
5244         .get_clid_cred  = nfs4_get_setclientid_cred,
5245 };
5246
5247 #if defined(CONFIG_NFS_V4_1)
5248 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5249         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5250         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5251         .recover_open   = nfs4_open_expired,
5252         .recover_lock   = nfs4_lock_expired,
5253         .establish_clid = nfs41_init_clientid,
5254         .get_clid_cred  = nfs4_get_exchange_id_cred,
5255 };
5256 #endif /* CONFIG_NFS_V4_1 */
5257
5258 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5259         .sched_state_renewal = nfs4_proc_async_renew,
5260         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5261         .renew_lease = nfs4_proc_renew,
5262 };
5263
5264 #if defined(CONFIG_NFS_V4_1)
5265 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5266         .sched_state_renewal = nfs41_proc_async_sequence,
5267         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5268         .renew_lease = nfs4_proc_sequence,
5269 };
5270 #endif
5271
5272 /*
5273  * Per minor version reboot and network partition recovery ops
5274  */
5275
5276 struct nfs4_state_recovery_ops *nfs4_reboot_recovery_ops[] = {
5277         &nfs40_reboot_recovery_ops,
5278 #if defined(CONFIG_NFS_V4_1)
5279         &nfs41_reboot_recovery_ops,
5280 #endif
5281 };
5282
5283 struct nfs4_state_recovery_ops *nfs4_nograce_recovery_ops[] = {
5284         &nfs40_nograce_recovery_ops,
5285 #if defined(CONFIG_NFS_V4_1)
5286         &nfs41_nograce_recovery_ops,
5287 #endif
5288 };
5289
5290 struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[] = {
5291         &nfs40_state_renewal_ops,
5292 #if defined(CONFIG_NFS_V4_1)
5293         &nfs41_state_renewal_ops,
5294 #endif
5295 };
5296
5297 static const struct inode_operations nfs4_file_inode_operations = {
5298         .permission     = nfs_permission,
5299         .getattr        = nfs_getattr,
5300         .setattr        = nfs_setattr,
5301         .getxattr       = nfs4_getxattr,
5302         .setxattr       = nfs4_setxattr,
5303         .listxattr      = nfs4_listxattr,
5304 };
5305
5306 const struct nfs_rpc_ops nfs_v4_clientops = {
5307         .version        = 4,                    /* protocol version */
5308         .dentry_ops     = &nfs4_dentry_operations,
5309         .dir_inode_ops  = &nfs4_dir_inode_operations,
5310         .file_inode_ops = &nfs4_file_inode_operations,
5311         .getroot        = nfs4_proc_get_root,
5312         .getattr        = nfs4_proc_getattr,
5313         .setattr        = nfs4_proc_setattr,
5314         .lookupfh       = nfs4_proc_lookupfh,
5315         .lookup         = nfs4_proc_lookup,
5316         .access         = nfs4_proc_access,
5317         .readlink       = nfs4_proc_readlink,
5318         .create         = nfs4_proc_create,
5319         .remove         = nfs4_proc_remove,
5320         .unlink_setup   = nfs4_proc_unlink_setup,
5321         .unlink_done    = nfs4_proc_unlink_done,
5322         .rename         = nfs4_proc_rename,
5323         .link           = nfs4_proc_link,
5324         .symlink        = nfs4_proc_symlink,
5325         .mkdir          = nfs4_proc_mkdir,
5326         .rmdir          = nfs4_proc_remove,
5327         .readdir        = nfs4_proc_readdir,
5328         .mknod          = nfs4_proc_mknod,
5329         .statfs         = nfs4_proc_statfs,
5330         .fsinfo         = nfs4_proc_fsinfo,
5331         .pathconf       = nfs4_proc_pathconf,
5332         .set_capabilities = nfs4_server_capabilities,
5333         .decode_dirent  = nfs4_decode_dirent,
5334         .read_setup     = nfs4_proc_read_setup,
5335         .read_done      = nfs4_read_done,
5336         .write_setup    = nfs4_proc_write_setup,
5337         .write_done     = nfs4_write_done,
5338         .commit_setup   = nfs4_proc_commit_setup,
5339         .commit_done    = nfs4_commit_done,
5340         .lock           = nfs4_proc_lock,
5341         .clear_acl_cache = nfs4_zap_acl_attr,
5342         .close_context  = nfs4_close_context,
5343 };
5344
5345 /*
5346  * Local variables:
5347  *  c-basic-offset: 8
5348  * End:
5349  */