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/slab.h>
  43 #include <linux/sunrpc/clnt.h>
  44 #include <linux/sunrpc/gss_api.h>
  45 #include <linux/nfs.h>
  46 #include <linux/nfs4.h>
  47 #include <linux/nfs_fs.h>
  48 #include <linux/nfs_page.h>
  49 #include <linux/nfs_mount.h>
  50 #include <linux/namei.h>
  51 #include <linux/mount.h>
  52 #include <linux/module.h>
  53 #include <linux/sunrpc/bc_xprt.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56
  57 #include "nfs4_fs.h"
  58 #include "delegation.h"
  59 #include "internal.h"
  60 #include "iostat.h"
  61 #include "callback.h"
  62 #include "pnfs.h"
  63
  64 #define NFSDBG_FACILITY         NFSDBG_PROC
  65
  66 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  67 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  68
  69 #define NFS4_MAX_LOOP_ON_RECOVER (10)
  70
  71 struct nfs4_opendata;
  72 static int _nfs4_proc_open(struct nfs4_opendata *data);
  73 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  74 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  75 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *);
  76 static int _nfs4_proc_lookup(struct rpc_clnt *client, struct inode *dir,
  77                              const struct qstr *name, struct nfs_fh *fhandle,
  78                              struct nfs_fattr *fattr);
  79 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr);
  80 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  81                             struct nfs_fattr *fattr, struct iattr *sattr,
  82                             struct nfs4_state *state);
  83
  84 /* Prevent leaks of NFSv4 errors into userland */
  85 static int nfs4_map_errors(int err)
  86 {
  87         if (err >= -1000)
  88                 return err;
  89         switch (err) {
  90         case -NFS4ERR_RESOURCE:
  91                 return -EREMOTEIO;
  92         case -NFS4ERR_WRONGSEC:
  93                 return -EPERM;
  94         case -NFS4ERR_BADOWNER:
  95         case -NFS4ERR_BADNAME:
  96                 return -EINVAL;
  97         default:
  98                 dprintk("%s could not handle NFSv4 error %d\n",
  99                                 __func__, -err);
 100                 break;
 101         }
 102         return -EIO;
 103 }
 104
 105 /*
 106  * This is our standard bitmap for GETATTR requests.
 107  */
 108 const u32 nfs4_fattr_bitmap[2] = {
 109         FATTR4_WORD0_TYPE
 110         | FATTR4_WORD0_CHANGE
 111         | FATTR4_WORD0_SIZE
 112         | FATTR4_WORD0_FSID
 113         | FATTR4_WORD0_FILEID,
 114         FATTR4_WORD1_MODE
 115         | FATTR4_WORD1_NUMLINKS
 116         | FATTR4_WORD1_OWNER
 117         | FATTR4_WORD1_OWNER_GROUP
 118         | FATTR4_WORD1_RAWDEV
 119         | FATTR4_WORD1_SPACE_USED
 120         | FATTR4_WORD1_TIME_ACCESS
 121         | FATTR4_WORD1_TIME_METADATA
 122         | FATTR4_WORD1_TIME_MODIFY
 123 };
 124
 125 const u32 nfs4_statfs_bitmap[2] = {
 126         FATTR4_WORD0_FILES_AVAIL
 127         | FATTR4_WORD0_FILES_FREE
 128         | FATTR4_WORD0_FILES_TOTAL,
 129         FATTR4_WORD1_SPACE_AVAIL
 130         | FATTR4_WORD1_SPACE_FREE
 131         | FATTR4_WORD1_SPACE_TOTAL
 132 };
 133
 134 const u32 nfs4_pathconf_bitmap[2] = {
 135         FATTR4_WORD0_MAXLINK
 136         | FATTR4_WORD0_MAXNAME,
 137         0
 138 };
 139
 140 const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
 141                         | FATTR4_WORD0_MAXREAD
 142                         | FATTR4_WORD0_MAXWRITE
 143                         | FATTR4_WORD0_LEASE_TIME,
 144                         FATTR4_WORD1_TIME_DELTA
 145                         | FATTR4_WORD1_FS_LAYOUT_TYPES
 146 };
 147
 148 const u32 nfs4_fs_locations_bitmap[2] = {
 149         FATTR4_WORD0_TYPE
 150         | FATTR4_WORD0_CHANGE
 151         | FATTR4_WORD0_SIZE
 152         | FATTR4_WORD0_FSID
 153         | FATTR4_WORD0_FILEID
 154         | FATTR4_WORD0_FS_LOCATIONS,
 155         FATTR4_WORD1_MODE
 156         | FATTR4_WORD1_NUMLINKS
 157         | FATTR4_WORD1_OWNER
 158         | FATTR4_WORD1_OWNER_GROUP
 159         | FATTR4_WORD1_RAWDEV
 160         | FATTR4_WORD1_SPACE_USED
 161         | FATTR4_WORD1_TIME_ACCESS
 162         | FATTR4_WORD1_TIME_METADATA
 163         | FATTR4_WORD1_TIME_MODIFY
 164         | FATTR4_WORD1_MOUNTED_ON_FILEID
 165 };
 166
 167 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 168                 struct nfs4_readdir_arg *readdir)
 169 {
 170         __be32 *start, *p;
 171
 172         BUG_ON(readdir->count < 80);
 173         if (cookie > 2) {
 174                 readdir->cookie = cookie;
 175                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 176                 return;
 177         }
 178
 179         readdir->cookie = 0;
 180         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 181         if (cookie == 2)
 182                 return;
 183
 184         /*
 185          * NFSv4 servers do not return entries for '.' and '..'
 186          * Therefore, we fake these entries here.  We let '.'
 187          * have cookie 0 and '..' have cookie 1.  Note that
 188          * when talking to the server, we always send cookie 0
 189          * instead of 1 or 2.
 190          */
 191         start = p = kmap_atomic(*readdir->pages, KM_USER0);
 192
 193         if (cookie == 0) {
 194                 *p++ = xdr_one;                                  /* next */
 195                 *p++ = xdr_zero;                   /* cookie, first word */
 196                 *p++ = xdr_one;                   /* cookie, second word */
 197                 *p++ = xdr_one;                             /* entry len */
 198                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 199                 p++;
 200                 *p++ = xdr_one;                         /* bitmap length */
 201                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 202                 *p++ = htonl(8);              /* attribute buffer length */
 203                 p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_inode));
 204         }
 205
 206         *p++ = xdr_one;                                  /* next */
 207         *p++ = xdr_zero;                   /* cookie, first word */
 208         *p++ = xdr_two;                   /* cookie, second word */
 209         *p++ = xdr_two;                             /* entry len */
 210         memcpy(p, "..\0\0", 4);                         /* entry */
 211         p++;
 212         *p++ = xdr_one;                         /* bitmap length */
 213         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 214         *p++ = htonl(8);              /* attribute buffer length */
 215         p = xdr_encode_hyper(p, NFS_FILEID(dentry->d_parent->d_inode));
 216
 217         readdir->pgbase = (char *)p - (char *)start;
 218         readdir->count -= readdir->pgbase;
 219         kunmap_atomic(start, KM_USER0);
 220 }
 221
 222 static int nfs4_wait_clnt_recover(struct nfs_client *clp)
 223 {
 224         int res;
 225
 226         might_sleep();
 227
 228         res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
 229                         nfs_wait_bit_killable, TASK_KILLABLE);
 230         return res;
 231 }
 232
 233 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 234 {
 235         int res = 0;
 236
 237         might_sleep();
 238
 239         if (*timeout <= 0)
 240                 *timeout = NFS4_POLL_RETRY_MIN;
 241         if (*timeout > NFS4_POLL_RETRY_MAX)
 242                 *timeout = NFS4_POLL_RETRY_MAX;
 243         schedule_timeout_killable(*timeout);
 244         if (fatal_signal_pending(current))
 245                 res = -ERESTARTSYS;
 246         *timeout <<= 1;
 247         return res;
 248 }
 249
 250 /* This is the error handling routine for processes that are allowed
 251  * to sleep.
 252  */
 253 static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 254 {
 255         struct nfs_client *clp = server->nfs_client;
 256         struct nfs4_state *state = exception->state;
 257         int ret = errorcode;
 258
 259         exception->retry = 0;
 260         switch(errorcode) {
 261                 case 0:
 262                         return 0;
 263                 case -NFS4ERR_ADMIN_REVOKED:
 264                 case -NFS4ERR_BAD_STATEID:
 265                 case -NFS4ERR_OPENMODE:
 266                         if (state == NULL)
 267                                 break;
 268                         nfs4_schedule_stateid_recovery(server, state);
 269                         goto wait_on_recovery;
 270                 case -NFS4ERR_STALE_STATEID:
 271                 case -NFS4ERR_STALE_CLIENTID:
 272                 case -NFS4ERR_EXPIRED:
 273                         nfs4_schedule_lease_recovery(clp);
 274                         goto wait_on_recovery;
 275 #if defined(CONFIG_NFS_V4_1)
 276                 case -NFS4ERR_BADSESSION:
 277                 case -NFS4ERR_BADSLOT:
 278                 case -NFS4ERR_BAD_HIGH_SLOT:
 279                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 280                 case -NFS4ERR_DEADSESSION:
 281                 case -NFS4ERR_SEQ_FALSE_RETRY:
 282                 case -NFS4ERR_SEQ_MISORDERED:
 283                         dprintk("%s ERROR: %d Reset session\n", __func__,
 284                                 errorcode);
 285                         nfs4_schedule_session_recovery(clp->cl_session);
 286                         exception->retry = 1;
 287                         break;
 288 #endif /* defined(CONFIG_NFS_V4_1) */
 289                 case -NFS4ERR_FILE_OPEN:
 290                         if (exception->timeout > HZ) {
 291                                 /* We have retried a decent amount, time to
 292                                  * fail
 293                                  */
 294                                 ret = -EBUSY;
 295                                 break;
 296                         }
 297                 case -NFS4ERR_GRACE:
 298                 case -NFS4ERR_DELAY:
 299                 case -EKEYEXPIRED:
 300                         ret = nfs4_delay(server->client, &exception->timeout);
 301                         if (ret != 0)
 302                                 break;
 303                 case -NFS4ERR_RETRY_UNCACHED_REP:
 304                 case -NFS4ERR_OLD_STATEID:
 305                         exception->retry = 1;
 306                         break;
 307                 case -NFS4ERR_BADOWNER:
 308                         /* The following works around a Linux server bug! */
 309                 case -NFS4ERR_BADNAME:
 310                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 311                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 312                                 exception->retry = 1;
 313                                 printk(KERN_WARNING "NFS: v4 server %s "
 314                                                 "does not accept raw "
 315                                                 "uid/gids. "
 316                                                 "Reenabling the idmapper.\n",
 317                                                 server->nfs_client->cl_hostname);
 318                         }
 319         }
 320         /* We failed to handle the error */
 321         return nfs4_map_errors(ret);
 322 wait_on_recovery:
 323         ret = nfs4_wait_clnt_recover(clp);
 324         if (ret == 0)
 325                 exception->retry = 1;
 326         return ret;
 327 }
 328
 329
 330 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 331 {
 332         spin_lock(&clp->cl_lock);
 333         if (time_before(clp->cl_last_renewal,timestamp))
 334                 clp->cl_last_renewal = timestamp;
 335         spin_unlock(&clp->cl_lock);
 336 }
 337
 338 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 339 {
 340         do_renew_lease(server->nfs_client, timestamp);
 341 }
 342
 343 #if defined(CONFIG_NFS_V4_1)
 344
 345 /*
 346  * nfs4_free_slot - free a slot and efficiently update slot table.
 347  *
 348  * freeing a slot is trivially done by clearing its respective bit
 349  * in the bitmap.
 350  * If the freed slotid equals highest_used_slotid we want to update it
 351  * so that the server would be able to size down the slot table if needed,
 352  * otherwise we know that the highest_used_slotid is still in use.
 353  * When updating highest_used_slotid there may be "holes" in the bitmap
 354  * so we need to scan down from highest_used_slotid to 0 looking for the now
 355  * highest slotid in use.
 356  * If none found, highest_used_slotid is set to -1.
 357  *
 358  * Must be called while holding tbl->slot_tbl_lock
 359  */
 360 static void
 361 nfs4_free_slot(struct nfs4_slot_table *tbl, struct nfs4_slot *free_slot)
 362 {
 363         int free_slotid = free_slot - tbl->slots;
 364         int slotid = free_slotid;
 365
 366         BUG_ON(slotid < 0 || slotid >= NFS4_MAX_SLOT_TABLE);
 367         /* clear used bit in bitmap */
 368         __clear_bit(slotid, tbl->used_slots);
 369
 370         /* update highest_used_slotid when it is freed */
 371         if (slotid == tbl->highest_used_slotid) {
 372                 slotid = find_last_bit(tbl->used_slots, tbl->max_slots);
 373                 if (slotid < tbl->max_slots)
 374                         tbl->highest_used_slotid = slotid;
 375                 else
 376                         tbl->highest_used_slotid = -1;
 377         }
 378         dprintk("%s: free_slotid %u highest_used_slotid %d\n", __func__,
 379                 free_slotid, tbl->highest_used_slotid);
 380 }
 381
 382 /*
 383  * Signal state manager thread if session fore channel is drained
 384  */
 385 static void nfs4_check_drain_fc_complete(struct nfs4_session *ses)
 386 {
 387         struct rpc_task *task;
 388
 389         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) {
 390                 task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq);
 391                 if (task)
 392                         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 393                 return;
 394         }
 395
 396         if (ses->fc_slot_table.highest_used_slotid != -1)
 397                 return;
 398
 399         dprintk("%s COMPLETE: Session Fore Channel Drained\n", __func__);
 400         complete(&ses->fc_slot_table.complete);
 401 }
 402
 403 /*
 404  * Signal state manager thread if session back channel is drained
 405  */
 406 void nfs4_check_drain_bc_complete(struct nfs4_session *ses)
 407 {
 408         if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state) ||
 409             ses->bc_slot_table.highest_used_slotid != -1)
 410                 return;
 411         dprintk("%s COMPLETE: Session Back Channel Drained\n", __func__);
 412         complete(&ses->bc_slot_table.complete);
 413 }
 414
 415 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 416 {
 417         struct nfs4_slot_table *tbl;
 418
 419         tbl = &res->sr_session->fc_slot_table;
 420         if (!res->sr_slot) {
 421                 /* just wake up the next guy waiting since
 422                  * we may have not consumed a slot after all */
 423                 dprintk("%s: No slot\n", __func__);
 424                 return;
 425         }
 426
 427         spin_lock(&tbl->slot_tbl_lock);
 428         nfs4_free_slot(tbl, res->sr_slot);
 429         nfs4_check_drain_fc_complete(res->sr_session);
 430         spin_unlock(&tbl->slot_tbl_lock);
 431         res->sr_slot = NULL;
 432 }
 433
 434 static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 435 {
 436         unsigned long timestamp;
 437         struct nfs_client *clp;
 438
 439         /*
 440          * sr_status remains 1 if an RPC level error occurred. The server
 441          * may or may not have processed the sequence operation..
 442          * Proceed as if the server received and processed the sequence
 443          * operation.
 444          */
 445         if (res->sr_status == 1)
 446                 res->sr_status = NFS_OK;
 447
 448         /* don't increment the sequence number if the task wasn't sent */
 449         if (!RPC_WAS_SENT(task))
 450                 goto out;
 451
 452         /* Check the SEQUENCE operation status */
 453         switch (res->sr_status) {
 454         case 0:
 455                 /* Update the slot's sequence and clientid lease timer */
 456                 ++res->sr_slot->seq_nr;
 457                 timestamp = res->sr_renewal_time;
 458                 clp = res->sr_session->clp;
 459                 do_renew_lease(clp, timestamp);
 460                 /* Check sequence flags */
 461                 if (res->sr_status_flags != 0)
 462                         nfs4_schedule_lease_recovery(clp);
 463                 break;
 464         case -NFS4ERR_DELAY:
 465                 /* The server detected a resend of the RPC call and
 466                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 467                  * of RFC5661.
 468                  */
 469                 dprintk("%s: slot=%td seq=%d: Operation in progress\n",
 470                         __func__,
 471                         res->sr_slot - res->sr_session->fc_slot_table.slots,
 472                         res->sr_slot->seq_nr);
 473                 goto out_retry;
 474         default:
 475                 /* Just update the slot sequence no. */
 476                 ++res->sr_slot->seq_nr;
 477         }
 478 out:
 479         /* The session may be reset by one of the error handlers. */
 480         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 481         nfs41_sequence_free_slot(res);
 482         return 1;
 483 out_retry:
 484         if (!rpc_restart_call(task))
 485                 goto out;
 486         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 487         return 0;
 488 }
 489
 490 static int nfs4_sequence_done(struct rpc_task *task,
 491                                struct nfs4_sequence_res *res)
 492 {
 493         if (res->sr_session == NULL)
 494                 return 1;
 495         return nfs41_sequence_done(task, res);
 496 }
 497
 498 /*
 499  * nfs4_find_slot - efficiently look for a free slot
 500  *
 501  * nfs4_find_slot looks for an unset bit in the used_slots bitmap.
 502  * If found, we mark the slot as used, update the highest_used_slotid,
 503  * and respectively set up the sequence operation args.
 504  * The slot number is returned if found, or NFS4_MAX_SLOT_TABLE otherwise.
 505  *
 506  * Note: must be called with under the slot_tbl_lock.
 507  */
 508 static u8
 509 nfs4_find_slot(struct nfs4_slot_table *tbl)
 510 {
 511         int slotid;
 512         u8 ret_id = NFS4_MAX_SLOT_TABLE;
 513         BUILD_BUG_ON((u8)NFS4_MAX_SLOT_TABLE != (int)NFS4_MAX_SLOT_TABLE);
 514
 515         dprintk("--> %s used_slots=%04lx highest_used=%d max_slots=%d\n",
 516                 __func__, tbl->used_slots[0], tbl->highest_used_slotid,
 517                 tbl->max_slots);
 518         slotid = find_first_zero_bit(tbl->used_slots, tbl->max_slots);
 519         if (slotid >= tbl->max_slots)
 520                 goto out;
 521         __set_bit(slotid, tbl->used_slots);
 522         if (slotid > tbl->highest_used_slotid)
 523                 tbl->highest_used_slotid = slotid;
 524         ret_id = slotid;
 525 out:
 526         dprintk("<-- %s used_slots=%04lx highest_used=%d slotid=%d \n",
 527                 __func__, tbl->used_slots[0], tbl->highest_used_slotid, ret_id);
 528         return ret_id;
 529 }
 530
 531 int nfs41_setup_sequence(struct nfs4_session *session,
 532                                 struct nfs4_sequence_args *args,
 533                                 struct nfs4_sequence_res *res,
 534                                 int cache_reply,
 535                                 struct rpc_task *task)
 536 {
 537         struct nfs4_slot *slot;
 538         struct nfs4_slot_table *tbl;
 539         u8 slotid;
 540
 541         dprintk("--> %s\n", __func__);
 542         /* slot already allocated? */
 543         if (res->sr_slot != NULL)
 544                 return 0;
 545
 546         tbl = &session->fc_slot_table;
 547
 548         spin_lock(&tbl->slot_tbl_lock);
 549         if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) &&
 550             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 551                 /*
 552                  * The state manager will wait until the slot table is empty.
 553                  * Schedule the reset thread
 554                  */
 555                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 556                 spin_unlock(&tbl->slot_tbl_lock);
 557                 dprintk("%s Schedule Session Reset\n", __func__);
 558                 return -EAGAIN;
 559         }
 560
 561         if (!rpc_queue_empty(&tbl->slot_tbl_waitq) &&
 562             !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) {
 563                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 564                 spin_unlock(&tbl->slot_tbl_lock);
 565                 dprintk("%s enforce FIFO order\n", __func__);
 566                 return -EAGAIN;
 567         }
 568
 569         slotid = nfs4_find_slot(tbl);
 570         if (slotid == NFS4_MAX_SLOT_TABLE) {
 571                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 572                 spin_unlock(&tbl->slot_tbl_lock);
 573                 dprintk("<-- %s: no free slots\n", __func__);
 574                 return -EAGAIN;
 575         }
 576         spin_unlock(&tbl->slot_tbl_lock);
 577
 578         rpc_task_set_priority(task, RPC_PRIORITY_NORMAL);
 579         slot = tbl->slots + slotid;
 580         args->sa_session = session;
 581         args->sa_slotid = slotid;
 582         args->sa_cache_this = cache_reply;
 583
 584         dprintk("<-- %s slotid=%d seqid=%d\n", __func__, slotid, slot->seq_nr);
 585
 586         res->sr_session = session;
 587         res->sr_slot = slot;
 588         res->sr_renewal_time = jiffies;
 589         res->sr_status_flags = 0;
 590         /*
 591          * sr_status is only set in decode_sequence, and so will remain
 592          * set to 1 if an rpc level failure occurs.
 593          */
 594         res->sr_status = 1;
 595         return 0;
 596 }
 597 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 598
 599 int nfs4_setup_sequence(const struct nfs_server *server,
 600                         struct nfs4_sequence_args *args,
 601                         struct nfs4_sequence_res *res,
 602                         int cache_reply,
 603                         struct rpc_task *task)
 604 {
 605         struct nfs4_session *session = nfs4_get_session(server);
 606         int ret = 0;
 607
 608         if (session == NULL) {
 609                 args->sa_session = NULL;
 610                 res->sr_session = NULL;
 611                 goto out;
 612         }
 613
 614         dprintk("--> %s clp %p session %p sr_slot %td\n",
 615                 __func__, session->clp, session, res->sr_slot ?
 616                         res->sr_slot - session->fc_slot_table.slots : -1);
 617
 618         ret = nfs41_setup_sequence(session, args, res, cache_reply,
 619                                    task);
 620 out:
 621         dprintk("<-- %s status=%d\n", __func__, ret);
 622         return ret;
 623 }
 624
 625 struct nfs41_call_sync_data {
 626         const struct nfs_server *seq_server;
 627         struct nfs4_sequence_args *seq_args;
 628         struct nfs4_sequence_res *seq_res;
 629         int cache_reply;
 630 };
 631
 632 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 633 {
 634         struct nfs41_call_sync_data *data = calldata;
 635
 636         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 637
 638         if (nfs4_setup_sequence(data->seq_server, data->seq_args,
 639                                 data->seq_res, data->cache_reply, task))
 640                 return;
 641         rpc_call_start(task);
 642 }
 643
 644 static void nfs41_call_priv_sync_prepare(struct rpc_task *task, void *calldata)
 645 {
 646         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
 647         nfs41_call_sync_prepare(task, calldata);
 648 }
 649
 650 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 651 {
 652         struct nfs41_call_sync_data *data = calldata;
 653
 654         nfs41_sequence_done(task, data->seq_res);
 655 }
 656
 657 struct rpc_call_ops nfs41_call_sync_ops = {
 658         .rpc_call_prepare = nfs41_call_sync_prepare,
 659         .rpc_call_done = nfs41_call_sync_done,
 660 };
 661
 662 struct rpc_call_ops nfs41_call_priv_sync_ops = {
 663         .rpc_call_prepare = nfs41_call_priv_sync_prepare,
 664         .rpc_call_done = nfs41_call_sync_done,
 665 };
 666
 667 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 668                                    struct nfs_server *server,
 669                                    struct rpc_message *msg,
 670                                    struct nfs4_sequence_args *args,
 671                                    struct nfs4_sequence_res *res,
 672                                    int cache_reply,
 673                                    int privileged)
 674 {
 675         int ret;
 676         struct rpc_task *task;
 677         struct nfs41_call_sync_data data = {
 678                 .seq_server = server,
 679                 .seq_args = args,
 680                 .seq_res = res,
 681                 .cache_reply = cache_reply,
 682         };
 683         struct rpc_task_setup task_setup = {
 684                 .rpc_client = clnt,
 685                 .rpc_message = msg,
 686                 .callback_ops = &nfs41_call_sync_ops,
 687                 .callback_data = &data
 688         };
 689
 690         res->sr_slot = NULL;
 691         if (privileged)
 692                 task_setup.callback_ops = &nfs41_call_priv_sync_ops;
 693         task = rpc_run_task(&task_setup);
 694         if (IS_ERR(task))
 695                 ret = PTR_ERR(task);
 696         else {
 697                 ret = task->tk_status;
 698                 rpc_put_task(task);
 699         }
 700         return ret;
 701 }
 702
 703 int _nfs4_call_sync_session(struct rpc_clnt *clnt,
 704                             struct nfs_server *server,
 705                             struct rpc_message *msg,
 706                             struct nfs4_sequence_args *args,
 707                             struct nfs4_sequence_res *res,
 708                             int cache_reply)
 709 {
 710         return nfs4_call_sync_sequence(clnt, server, msg, args, res, cache_reply, 0);
 711 }
 712
 713 #else
 714 static int nfs4_sequence_done(struct rpc_task *task,
 715                                struct nfs4_sequence_res *res)
 716 {
 717         return 1;
 718 }
 719 #endif /* CONFIG_NFS_V4_1 */
 720
 721 int _nfs4_call_sync(struct rpc_clnt *clnt,
 722                     struct nfs_server *server,
 723                     struct rpc_message *msg,
 724                     struct nfs4_sequence_args *args,
 725                     struct nfs4_sequence_res *res,
 726                     int cache_reply)
 727 {
 728         args->sa_session = res->sr_session = NULL;
 729         return rpc_call_sync(clnt, msg, 0);
 730 }
 731
 732 static inline
 733 int nfs4_call_sync(struct rpc_clnt *clnt,
 734                    struct nfs_server *server,
 735                    struct rpc_message *msg,
 736                    struct nfs4_sequence_args *args,
 737                    struct nfs4_sequence_res *res,
 738                    int cache_reply)
 739 {
 740         return server->nfs_client->cl_mvops->call_sync(clnt, server, msg,
 741                                                 args, res, cache_reply);
 742 }
 743
 744 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 745 {
 746         struct nfs_inode *nfsi = NFS_I(dir);
 747
 748         spin_lock(&dir->i_lock);
 749         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_DATA;
 750         if (!cinfo->atomic || cinfo->before != nfsi->change_attr)
 751                 nfs_force_lookup_revalidate(dir);
 752         nfsi->change_attr = cinfo->after;
 753         spin_unlock(&dir->i_lock);
 754 }
 755
 756 struct nfs4_opendata {
 757         struct kref kref;
 758         struct nfs_openargs o_arg;
 759         struct nfs_openres o_res;
 760         struct nfs_open_confirmargs c_arg;
 761         struct nfs_open_confirmres c_res;
 762         struct nfs_fattr f_attr;
 763         struct nfs_fattr dir_attr;
 764         struct path path;
 765         struct dentry *dir;
 766         struct nfs4_state_owner *owner;
 767         struct nfs4_state *state;
 768         struct iattr attrs;
 769         unsigned long timestamp;
 770         unsigned int rpc_done : 1;
 771         int rpc_status;
 772         int cancelled;
 773 };
 774
 775
 776 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 777 {
 778         p->o_res.f_attr = &p->f_attr;
 779         p->o_res.dir_attr = &p->dir_attr;
 780         p->o_res.seqid = p->o_arg.seqid;
 781         p->c_res.seqid = p->c_arg.seqid;
 782         p->o_res.server = p->o_arg.server;
 783         nfs_fattr_init(&p->f_attr);
 784         nfs_fattr_init(&p->dir_attr);
 785 }
 786
 787 static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
 788                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 789                 const struct iattr *attrs,
 790                 gfp_t gfp_mask)
 791 {
 792         struct dentry *parent = dget_parent(path->dentry);
 793         struct inode *dir = parent->d_inode;
 794         struct nfs_server *server = NFS_SERVER(dir);
 795         struct nfs4_opendata *p;
 796
 797         p = kzalloc(sizeof(*p), gfp_mask);
 798         if (p == NULL)
 799                 goto err;
 800         p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid, gfp_mask);
 801         if (p->o_arg.seqid == NULL)
 802                 goto err_free;
 803         path_get(path);
 804         p->path = *path;
 805         p->dir = parent;
 806         p->owner = sp;
 807         atomic_inc(&sp->so_count);
 808         p->o_arg.fh = NFS_FH(dir);
 809         p->o_arg.open_flags = flags;
 810         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
 811         p->o_arg.clientid = server->nfs_client->cl_clientid;
 812         p->o_arg.id = sp->so_owner_id.id;
 813         p->o_arg.name = &p->path.dentry->d_name;
 814         p->o_arg.server = server;
 815         p->o_arg.bitmask = server->attr_bitmask;
 816         p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
 817         if (flags & O_CREAT) {
 818                 u32 *s;
 819
 820                 p->o_arg.u.attrs = &p->attrs;
 821                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
 822                 s = (u32 *) p->o_arg.u.verifier.data;
 823                 s[0] = jiffies;
 824                 s[1] = current->pid;
 825         }
 826         p->c_arg.fh = &p->o_res.fh;
 827         p->c_arg.stateid = &p->o_res.stateid;
 828         p->c_arg.seqid = p->o_arg.seqid;
 829         nfs4_init_opendata_res(p);
 830         kref_init(&p->kref);
 831         return p;
 832 err_free:
 833         kfree(p);
 834 err:
 835         dput(parent);
 836         return NULL;
 837 }
 838
 839 static void nfs4_opendata_free(struct kref *kref)
 840 {
 841         struct nfs4_opendata *p = container_of(kref,
 842                         struct nfs4_opendata, kref);
 843
 844         nfs_free_seqid(p->o_arg.seqid);
 845         if (p->state != NULL)
 846                 nfs4_put_open_state(p->state);
 847         nfs4_put_state_owner(p->owner);
 848         dput(p->dir);
 849         path_put(&p->path);
 850         kfree(p);
 851 }
 852
 853 static void nfs4_opendata_put(struct nfs4_opendata *p)
 854 {
 855         if (p != NULL)
 856                 kref_put(&p->kref, nfs4_opendata_free);
 857 }
 858
 859 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
 860 {
 861         int ret;
 862
 863         ret = rpc_wait_for_completion_task(task);
 864         return ret;
 865 }
 866
 867 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
 868 {
 869         int ret = 0;
 870
 871         if (open_mode & O_EXCL)
 872                 goto out;
 873         switch (mode & (FMODE_READ|FMODE_WRITE)) {
 874                 case FMODE_READ:
 875                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
 876                                 && state->n_rdonly != 0;
 877                         break;
 878                 case FMODE_WRITE:
 879                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
 880                                 && state->n_wronly != 0;
 881                         break;
 882                 case FMODE_READ|FMODE_WRITE:
 883                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
 884                                 && state->n_rdwr != 0;
 885         }
 886 out:
 887         return ret;
 888 }
 889
 890 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
 891 {
 892         if ((delegation->type & fmode) != fmode)
 893                 return 0;
 894         if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
 895                 return 0;
 896         nfs_mark_delegation_referenced(delegation);
 897         return 1;
 898 }
 899
 900 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
 901 {
 902         switch (fmode) {
 903                 case FMODE_WRITE:
 904                         state->n_wronly++;
 905                         break;
 906                 case FMODE_READ:
 907                         state->n_rdonly++;
 908                         break;
 909                 case FMODE_READ|FMODE_WRITE:
 910                         state->n_rdwr++;
 911         }
 912         nfs4_state_set_mode_locked(state, state->state | fmode);
 913 }
 914
 915 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 916 {
 917         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
 918                 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
 919         memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
 920         switch (fmode) {
 921                 case FMODE_READ:
 922                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
 923                         break;
 924                 case FMODE_WRITE:
 925                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
 926                         break;
 927                 case FMODE_READ|FMODE_WRITE:
 928                         set_bit(NFS_O_RDWR_STATE, &state->flags);
 929         }
 930 }
 931
 932 static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
 933 {
 934         write_seqlock(&state->seqlock);
 935         nfs_set_open_stateid_locked(state, stateid, fmode);
 936         write_sequnlock(&state->seqlock);
 937 }
 938
 939 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
 940 {
 941         /*
 942          * Protect the call to nfs4_state_set_mode_locked and
 943          * serialise the stateid update
 944          */
 945         write_seqlock(&state->seqlock);
 946         if (deleg_stateid != NULL) {
 947                 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
 948                 set_bit(NFS_DELEGATED_STATE, &state->flags);
 949         }
 950         if (open_stateid != NULL)
 951                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
 952         write_sequnlock(&state->seqlock);
 953         spin_lock(&state->owner->so_lock);
 954         update_open_stateflags(state, fmode);
 955         spin_unlock(&state->owner->so_lock);
 956 }
 957
 958 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
 959 {
 960         struct nfs_inode *nfsi = NFS_I(state->inode);
 961         struct nfs_delegation *deleg_cur;
 962         int ret = 0;
 963
 964         fmode &= (FMODE_READ|FMODE_WRITE);
 965
 966         rcu_read_lock();
 967         deleg_cur = rcu_dereference(nfsi->delegation);
 968         if (deleg_cur == NULL)
 969                 goto no_delegation;
 970
 971         spin_lock(&deleg_cur->lock);
 972         if (nfsi->delegation != deleg_cur ||
 973             (deleg_cur->type & fmode) != fmode)
 974                 goto no_delegation_unlock;
 975
 976         if (delegation == NULL)
 977                 delegation = &deleg_cur->stateid;
 978         else if (memcmp(deleg_cur->stateid.data, delegation->data, NFS4_STATEID_SIZE) != 0)
 979                 goto no_delegation_unlock;
 980
 981         nfs_mark_delegation_referenced(deleg_cur);
 982         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
 983         ret = 1;
 984 no_delegation_unlock:
 985         spin_unlock(&deleg_cur->lock);
 986 no_delegation:
 987         rcu_read_unlock();
 988
 989         if (!ret && open_stateid != NULL) {
 990                 __update_open_stateid(state, open_stateid, NULL, fmode);
 991                 ret = 1;
 992         }
 993
 994         return ret;
 995 }
 996
 997
 998 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
 999 {
1000         struct nfs_delegation *delegation;
1001
1002         rcu_read_lock();
1003         delegation = rcu_dereference(NFS_I(inode)->delegation);
1004         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1005                 rcu_read_unlock();
1006                 return;
1007         }
1008         rcu_read_unlock();
1009         nfs_inode_return_delegation(inode);
1010 }
1011
1012 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1013 {
1014         struct nfs4_state *state = opendata->state;
1015         struct nfs_inode *nfsi = NFS_I(state->inode);
1016         struct nfs_delegation *delegation;
1017         int open_mode = opendata->o_arg.open_flags & O_EXCL;
1018         fmode_t fmode = opendata->o_arg.fmode;
1019         nfs4_stateid stateid;
1020         int ret = -EAGAIN;
1021
1022         for (;;) {
1023                 if (can_open_cached(state, fmode, open_mode)) {
1024                         spin_lock(&state->owner->so_lock);
1025                         if (can_open_cached(state, fmode, open_mode)) {
1026                                 update_open_stateflags(state, fmode);
1027                                 spin_unlock(&state->owner->so_lock);
1028                                 goto out_return_state;
1029                         }
1030                         spin_unlock(&state->owner->so_lock);
1031                 }
1032                 rcu_read_lock();
1033                 delegation = rcu_dereference(nfsi->delegation);
1034                 if (delegation == NULL ||
1035                     !can_open_delegated(delegation, fmode)) {
1036                         rcu_read_unlock();
1037                         break;
1038                 }
1039                 /* Save the delegation */
1040                 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
1041                 rcu_read_unlock();
1042                 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1043                 if (ret != 0)
1044                         goto out;
1045                 ret = -EAGAIN;
1046
1047                 /* Try to update the stateid using the delegation */
1048                 if (update_open_stateid(state, NULL, &stateid, fmode))
1049                         goto out_return_state;
1050         }
1051 out:
1052         return ERR_PTR(ret);
1053 out_return_state:
1054         atomic_inc(&state->count);
1055         return state;
1056 }
1057
1058 static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1059 {
1060         struct inode *inode;
1061         struct nfs4_state *state = NULL;
1062         struct nfs_delegation *delegation;
1063         int ret;
1064
1065         if (!data->rpc_done) {
1066                 state = nfs4_try_open_cached(data);
1067                 goto out;
1068         }
1069
1070         ret = -EAGAIN;
1071         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1072                 goto err;
1073         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
1074         ret = PTR_ERR(inode);
1075         if (IS_ERR(inode))
1076                 goto err;
1077         ret = -ENOMEM;
1078         state = nfs4_get_open_state(inode, data->owner);
1079         if (state == NULL)
1080                 goto err_put_inode;
1081         if (data->o_res.delegation_type != 0) {
1082                 int delegation_flags = 0;
1083
1084                 rcu_read_lock();
1085                 delegation = rcu_dereference(NFS_I(inode)->delegation);
1086                 if (delegation)
1087                         delegation_flags = delegation->flags;
1088                 rcu_read_unlock();
1089                 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1090                         nfs_inode_set_delegation(state->inode,
1091                                         data->owner->so_cred,
1092                                         &data->o_res);
1093                 else
1094                         nfs_inode_reclaim_delegation(state->inode,
1095                                         data->owner->so_cred,
1096                                         &data->o_res);
1097         }
1098
1099         update_open_stateid(state, &data->o_res.stateid, NULL,
1100                         data->o_arg.fmode);
1101         iput(inode);
1102 out:
1103         return state;
1104 err_put_inode:
1105         iput(inode);
1106 err:
1107         return ERR_PTR(ret);
1108 }
1109
1110 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1111 {
1112         struct nfs_inode *nfsi = NFS_I(state->inode);
1113         struct nfs_open_context *ctx;
1114
1115         spin_lock(&state->inode->i_lock);
1116         list_for_each_entry(ctx, &nfsi->open_files, list) {
1117                 if (ctx->state != state)
1118                         continue;
1119                 get_nfs_open_context(ctx);
1120                 spin_unlock(&state->inode->i_lock);
1121                 return ctx;
1122         }
1123         spin_unlock(&state->inode->i_lock);
1124         return ERR_PTR(-ENOENT);
1125 }
1126
1127 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx, struct nfs4_state *state)
1128 {
1129         struct nfs4_opendata *opendata;
1130
1131         opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, 0, NULL, GFP_NOFS);
1132         if (opendata == NULL)
1133                 return ERR_PTR(-ENOMEM);
1134         opendata->state = state;
1135         atomic_inc(&state->count);
1136         return opendata;
1137 }
1138
1139 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1140 {
1141         struct nfs4_state *newstate;
1142         int ret;
1143
1144         opendata->o_arg.open_flags = 0;
1145         opendata->o_arg.fmode = fmode;
1146         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1147         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1148         nfs4_init_opendata_res(opendata);
1149         ret = _nfs4_recover_proc_open(opendata);
1150         if (ret != 0)
1151                 return ret;
1152         newstate = nfs4_opendata_to_nfs4_state(opendata);
1153         if (IS_ERR(newstate))
1154                 return PTR_ERR(newstate);
1155         nfs4_close_state(&opendata->path, newstate, fmode);
1156         *res = newstate;
1157         return 0;
1158 }
1159
1160 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1161 {
1162         struct nfs4_state *newstate;
1163         int ret;
1164
1165         /* memory barrier prior to reading state->n_* */
1166         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1167         smp_rmb();
1168         if (state->n_rdwr != 0) {
1169                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1170                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1171                 if (ret != 0)
1172                         return ret;
1173                 if (newstate != state)
1174                         return -ESTALE;
1175         }
1176         if (state->n_wronly != 0) {
1177                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1178                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1179                 if (ret != 0)
1180                         return ret;
1181                 if (newstate != state)
1182                         return -ESTALE;
1183         }
1184         if (state->n_rdonly != 0) {
1185                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1186                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1187                 if (ret != 0)
1188                         return ret;
1189                 if (newstate != state)
1190                         return -ESTALE;
1191         }
1192         /*
1193          * We may have performed cached opens for all three recoveries.
1194          * Check if we need to update the current stateid.
1195          */
1196         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1197             memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
1198                 write_seqlock(&state->seqlock);
1199                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1200                         memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
1201                 write_sequnlock(&state->seqlock);
1202         }
1203         return 0;
1204 }
1205
1206 /*
1207  * OPEN_RECLAIM:
1208  *      reclaim state on the server after a reboot.
1209  */
1210 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1211 {
1212         struct nfs_delegation *delegation;
1213         struct nfs4_opendata *opendata;
1214         fmode_t delegation_type = 0;
1215         int status;
1216
1217         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1218         if (IS_ERR(opendata))
1219                 return PTR_ERR(opendata);
1220         opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
1221         opendata->o_arg.fh = NFS_FH(state->inode);
1222         rcu_read_lock();
1223         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1224         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1225                 delegation_type = delegation->type;
1226         rcu_read_unlock();
1227         opendata->o_arg.u.delegation_type = delegation_type;
1228         status = nfs4_open_recover(opendata, state);
1229         nfs4_opendata_put(opendata);
1230         return status;
1231 }
1232
1233 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1234 {
1235         struct nfs_server *server = NFS_SERVER(state->inode);
1236         struct nfs4_exception exception = { };
1237         int err;
1238         do {
1239                 err = _nfs4_do_open_reclaim(ctx, state);
1240                 if (err != -NFS4ERR_DELAY)
1241                         break;
1242                 nfs4_handle_exception(server, err, &exception);
1243         } while (exception.retry);
1244         return err;
1245 }
1246
1247 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1248 {
1249         struct nfs_open_context *ctx;
1250         int ret;
1251
1252         ctx = nfs4_state_find_open_context(state);
1253         if (IS_ERR(ctx))
1254                 return PTR_ERR(ctx);
1255         ret = nfs4_do_open_reclaim(ctx, state);
1256         put_nfs_open_context(ctx);
1257         return ret;
1258 }
1259
1260 static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1261 {
1262         struct nfs4_opendata *opendata;
1263         int ret;
1264
1265         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1266         if (IS_ERR(opendata))
1267                 return PTR_ERR(opendata);
1268         opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
1269         memcpy(opendata->o_arg.u.delegation.data, stateid->data,
1270                         sizeof(opendata->o_arg.u.delegation.data));
1271         ret = nfs4_open_recover(opendata, state);
1272         nfs4_opendata_put(opendata);
1273         return ret;
1274 }
1275
1276 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1277 {
1278         struct nfs4_exception exception = { };
1279         struct nfs_server *server = NFS_SERVER(state->inode);
1280         int err;
1281         do {
1282                 err = _nfs4_open_delegation_recall(ctx, state, stateid);
1283                 switch (err) {
1284                         case 0:
1285                         case -ENOENT:
1286                         case -ESTALE:
1287                                 goto out;
1288                         case -NFS4ERR_BADSESSION:
1289                         case -NFS4ERR_BADSLOT:
1290                         case -NFS4ERR_BAD_HIGH_SLOT:
1291                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1292                         case -NFS4ERR_DEADSESSION:
1293                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
1294                                 goto out;
1295                         case -NFS4ERR_STALE_CLIENTID:
1296                         case -NFS4ERR_STALE_STATEID:
1297                         case -NFS4ERR_EXPIRED:
1298                                 /* Don't recall a delegation if it was lost */
1299                                 nfs4_schedule_lease_recovery(server->nfs_client);
1300                                 goto out;
1301                         case -ERESTARTSYS:
1302                                 /*
1303                                  * The show must go on: exit, but mark the
1304                                  * stateid as needing recovery.
1305                                  */
1306                         case -NFS4ERR_ADMIN_REVOKED:
1307                         case -NFS4ERR_BAD_STATEID:
1308                                 nfs4_schedule_stateid_recovery(server, state);
1309                         case -EKEYEXPIRED:
1310                                 /*
1311                                  * User RPCSEC_GSS context has expired.
1312                                  * We cannot recover this stateid now, so
1313                                  * skip it and allow recovery thread to
1314                                  * proceed.
1315                                  */
1316                         case -ENOMEM:
1317                                 err = 0;
1318                                 goto out;
1319                 }
1320                 err = nfs4_handle_exception(server, err, &exception);
1321         } while (exception.retry);
1322 out:
1323         return err;
1324 }
1325
1326 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1327 {
1328         struct nfs4_opendata *data = calldata;
1329
1330         data->rpc_status = task->tk_status;
1331         if (data->rpc_status == 0) {
1332                 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
1333                                 sizeof(data->o_res.stateid.data));
1334                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1335                 renew_lease(data->o_res.server, data->timestamp);
1336                 data->rpc_done = 1;
1337         }
1338 }
1339
1340 static void nfs4_open_confirm_release(void *calldata)
1341 {
1342         struct nfs4_opendata *data = calldata;
1343         struct nfs4_state *state = NULL;
1344
1345         /* If this request hasn't been cancelled, do nothing */
1346         if (data->cancelled == 0)
1347                 goto out_free;
1348         /* In case of error, no cleanup! */
1349         if (!data->rpc_done)
1350                 goto out_free;
1351         state = nfs4_opendata_to_nfs4_state(data);
1352         if (!IS_ERR(state))
1353                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1354 out_free:
1355         nfs4_opendata_put(data);
1356 }
1357
1358 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1359         .rpc_call_done = nfs4_open_confirm_done,
1360         .rpc_release = nfs4_open_confirm_release,
1361 };
1362
1363 /*
1364  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1365  */
1366 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1367 {
1368         struct nfs_server *server = NFS_SERVER(data->dir->d_inode);
1369         struct rpc_task *task;
1370         struct  rpc_message msg = {
1371                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1372                 .rpc_argp = &data->c_arg,
1373                 .rpc_resp = &data->c_res,
1374                 .rpc_cred = data->owner->so_cred,
1375         };
1376         struct rpc_task_setup task_setup_data = {
1377                 .rpc_client = server->client,
1378                 .rpc_message = &msg,
1379                 .callback_ops = &nfs4_open_confirm_ops,
1380                 .callback_data = data,
1381                 .workqueue = nfsiod_workqueue,
1382                 .flags = RPC_TASK_ASYNC,
1383         };
1384         int status;
1385
1386         kref_get(&data->kref);
1387         data->rpc_done = 0;
1388         data->rpc_status = 0;
1389         data->timestamp = jiffies;
1390         task = rpc_run_task(&task_setup_data);
1391         if (IS_ERR(task))
1392                 return PTR_ERR(task);
1393         status = nfs4_wait_for_completion_rpc_task(task);
1394         if (status != 0) {
1395                 data->cancelled = 1;
1396                 smp_wmb();
1397         } else
1398                 status = data->rpc_status;
1399         rpc_put_task(task);
1400         return status;
1401 }
1402
1403 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1404 {
1405         struct nfs4_opendata *data = calldata;
1406         struct nfs4_state_owner *sp = data->owner;
1407
1408         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1409                 return;
1410         /*
1411          * Check if we still need to send an OPEN call, or if we can use
1412          * a delegation instead.
1413          */
1414         if (data->state != NULL) {
1415                 struct nfs_delegation *delegation;
1416
1417                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1418                         goto out_no_action;
1419                 rcu_read_lock();
1420                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1421                 if (delegation != NULL &&
1422                     test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) == 0) {
1423                         rcu_read_unlock();
1424                         goto out_no_action;
1425                 }
1426                 rcu_read_unlock();
1427         }
1428         /* Update sequence id. */
1429         data->o_arg.id = sp->so_owner_id.id;
1430         data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid;
1431         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) {
1432                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1433                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1434         }
1435         data->timestamp = jiffies;
1436         if (nfs4_setup_sequence(data->o_arg.server,
1437                                 &data->o_arg.seq_args,
1438                                 &data->o_res.seq_res, 1, task))
1439                 return;
1440         rpc_call_start(task);
1441         return;
1442 out_no_action:
1443         task->tk_action = NULL;
1444
1445 }
1446
1447 static void nfs4_recover_open_prepare(struct rpc_task *task, void *calldata)
1448 {
1449         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
1450         nfs4_open_prepare(task, calldata);
1451 }
1452
1453 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1454 {
1455         struct nfs4_opendata *data = calldata;
1456
1457         data->rpc_status = task->tk_status;
1458
1459         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1460                 return;
1461
1462         if (task->tk_status == 0) {
1463                 switch (data->o_res.f_attr->mode & S_IFMT) {
1464                         case S_IFREG:
1465                                 break;
1466                         case S_IFLNK:
1467                                 data->rpc_status = -ELOOP;
1468                                 break;
1469                         case S_IFDIR:
1470                                 data->rpc_status = -EISDIR;
1471                                 break;
1472                         default:
1473                                 data->rpc_status = -ENOTDIR;
1474                 }
1475                 renew_lease(data->o_res.server, data->timestamp);
1476                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1477                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1478         }
1479         data->rpc_done = 1;
1480 }
1481
1482 static void nfs4_open_release(void *calldata)
1483 {
1484         struct nfs4_opendata *data = calldata;
1485         struct nfs4_state *state = NULL;
1486
1487         /* If this request hasn't been cancelled, do nothing */
1488         if (data->cancelled == 0)
1489                 goto out_free;
1490         /* In case of error, no cleanup! */
1491         if (data->rpc_status != 0 || !data->rpc_done)
1492                 goto out_free;
1493         /* In case we need an open_confirm, no cleanup! */
1494         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1495                 goto out_free;
1496         state = nfs4_opendata_to_nfs4_state(data);
1497         if (!IS_ERR(state))
1498                 nfs4_close_state(&data->path, state, data->o_arg.fmode);
1499 out_free:
1500         nfs4_opendata_put(data);
1501 }
1502
1503 static const struct rpc_call_ops nfs4_open_ops = {
1504         .rpc_call_prepare = nfs4_open_prepare,
1505         .rpc_call_done = nfs4_open_done,
1506         .rpc_release = nfs4_open_release,
1507 };
1508
1509 static const struct rpc_call_ops nfs4_recover_open_ops = {
1510         .rpc_call_prepare = nfs4_recover_open_prepare,
1511         .rpc_call_done = nfs4_open_done,
1512         .rpc_release = nfs4_open_release,
1513 };
1514
1515 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1516 {
1517         struct inode *dir = data->dir->d_inode;
1518         struct nfs_server *server = NFS_SERVER(dir);
1519         struct nfs_openargs *o_arg = &data->o_arg;
1520         struct nfs_openres *o_res = &data->o_res;
1521         struct rpc_task *task;
1522         struct rpc_message msg = {
1523                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1524                 .rpc_argp = o_arg,
1525                 .rpc_resp = o_res,
1526                 .rpc_cred = data->owner->so_cred,
1527         };
1528         struct rpc_task_setup task_setup_data = {
1529                 .rpc_client = server->client,
1530                 .rpc_message = &msg,
1531                 .callback_ops = &nfs4_open_ops,
1532                 .callback_data = data,
1533                 .workqueue = nfsiod_workqueue,
1534                 .flags = RPC_TASK_ASYNC,
1535         };
1536         int status;
1537
1538         kref_get(&data->kref);
1539         data->rpc_done = 0;
1540         data->rpc_status = 0;
1541         data->cancelled = 0;
1542         if (isrecover)
1543                 task_setup_data.callback_ops = &nfs4_recover_open_ops;
1544         task = rpc_run_task(&task_setup_data);
1545         if (IS_ERR(task))
1546                 return PTR_ERR(task);
1547         status = nfs4_wait_for_completion_rpc_task(task);
1548         if (status != 0) {
1549                 data->cancelled = 1;
1550                 smp_wmb();
1551         } else
1552                 status = data->rpc_status;
1553         rpc_put_task(task);
1554
1555         return status;
1556 }
1557
1558 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
1559 {
1560         struct inode *dir = data->dir->d_inode;
1561         struct nfs_openres *o_res = &data->o_res;
1562         int status;
1563
1564         status = nfs4_run_open_task(data, 1);
1565         if (status != 0 || !data->rpc_done)
1566                 return status;
1567
1568         nfs_refresh_inode(dir, o_res->dir_attr);
1569
1570         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1571                 status = _nfs4_proc_open_confirm(data);
1572                 if (status != 0)
1573                         return status;
1574         }
1575
1576         return status;
1577 }
1578
1579 /*
1580  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
1581  */
1582 static int _nfs4_proc_open(struct nfs4_opendata *data)
1583 {
1584         struct inode *dir = data->dir->d_inode;
1585         struct nfs_server *server = NFS_SERVER(dir);
1586         struct nfs_openargs *o_arg = &data->o_arg;
1587         struct nfs_openres *o_res = &data->o_res;
1588         int status;
1589
1590         status = nfs4_run_open_task(data, 0);
1591         if (status != 0 || !data->rpc_done)
1592                 return status;
1593
1594         if (o_arg->open_flags & O_CREAT) {
1595                 update_changeattr(dir, &o_res->cinfo);
1596                 nfs_post_op_update_inode(dir, o_res->dir_attr);
1597         } else
1598                 nfs_refresh_inode(dir, o_res->dir_attr);
1599         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
1600                 server->caps &= ~NFS_CAP_POSIX_LOCK;
1601         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
1602                 status = _nfs4_proc_open_confirm(data);
1603                 if (status != 0)
1604                         return status;
1605         }
1606         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
1607                 _nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr);
1608         return 0;
1609 }
1610
1611 static int nfs4_client_recover_expired_lease(struct nfs_client *clp)
1612 {
1613         unsigned int loop;
1614         int ret;
1615
1616         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
1617                 ret = nfs4_wait_clnt_recover(clp);
1618                 if (ret != 0)
1619                         break;
1620                 if (!test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) &&
1621                     !test_bit(NFS4CLNT_CHECK_LEASE,&clp->cl_state))
1622                         break;
1623                 nfs4_schedule_state_manager(clp);
1624                 ret = -EIO;
1625         }
1626         return ret;
1627 }
1628
1629 static int nfs4_recover_expired_lease(struct nfs_server *server)
1630 {
1631         return nfs4_client_recover_expired_lease(server->nfs_client);
1632 }
1633
1634 /*
1635  * OPEN_EXPIRED:
1636  *      reclaim state on the server after a network partition.
1637  *      Assumes caller holds the appropriate lock
1638  */
1639 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1640 {
1641         struct nfs4_opendata *opendata;
1642         int ret;
1643
1644         opendata = nfs4_open_recoverdata_alloc(ctx, state);
1645         if (IS_ERR(opendata))
1646                 return PTR_ERR(opendata);
1647         ret = nfs4_open_recover(opendata, state);
1648         if (ret == -ESTALE)
1649                 d_drop(ctx->path.dentry);
1650         nfs4_opendata_put(opendata);
1651         return ret;
1652 }
1653
1654 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
1655 {
1656         struct nfs_server *server = NFS_SERVER(state->inode);
1657         struct nfs4_exception exception = { };
1658         int err;
1659
1660         do {
1661                 err = _nfs4_open_expired(ctx, state);
1662                 switch (err) {
1663                 default:
1664                         goto out;
1665                 case -NFS4ERR_GRACE:
1666                 case -NFS4ERR_DELAY:
1667                         nfs4_handle_exception(server, err, &exception);
1668                         err = 0;
1669                 }
1670         } while (exception.retry);
1671 out:
1672         return err;
1673 }
1674
1675 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
1676 {
1677         struct nfs_open_context *ctx;
1678         int ret;
1679
1680         ctx = nfs4_state_find_open_context(state);
1681         if (IS_ERR(ctx))
1682                 return PTR_ERR(ctx);
1683         ret = nfs4_do_open_expired(ctx, state);
1684         put_nfs_open_context(ctx);
1685         return ret;
1686 }
1687
1688 /*
1689  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1690  * fields corresponding to attributes that were used to store the verifier.
1691  * Make sure we clobber those fields in the later setattr call
1692  */
1693 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
1694 {
1695         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
1696             !(sattr->ia_valid & ATTR_ATIME_SET))
1697                 sattr->ia_valid |= ATTR_ATIME;
1698
1699         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
1700             !(sattr->ia_valid & ATTR_MTIME_SET))
1701                 sattr->ia_valid |= ATTR_MTIME;
1702 }
1703
1704 /*
1705  * Returns a referenced nfs4_state
1706  */
1707 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)
1708 {
1709         struct nfs4_state_owner  *sp;
1710         struct nfs4_state     *state = NULL;
1711         struct nfs_server       *server = NFS_SERVER(dir);
1712         struct nfs4_opendata *opendata;
1713         int status;
1714
1715         /* Protect against reboot recovery conflicts */
1716         status = -ENOMEM;
1717         if (!(sp = nfs4_get_state_owner(server, cred))) {
1718                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
1719                 goto out_err;
1720         }
1721         status = nfs4_recover_expired_lease(server);
1722         if (status != 0)
1723                 goto err_put_state_owner;
1724         if (path->dentry->d_inode != NULL)
1725                 nfs4_return_incompatible_delegation(path->dentry->d_inode, fmode);
1726         status = -ENOMEM;
1727         opendata = nfs4_opendata_alloc(path, sp, fmode, flags, sattr, GFP_KERNEL);
1728         if (opendata == NULL)
1729                 goto err_put_state_owner;
1730
1731         if (path->dentry->d_inode != NULL)
1732                 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1733
1734         status = _nfs4_proc_open(opendata);
1735         if (status != 0)
1736                 goto err_opendata_put;
1737
1738         state = nfs4_opendata_to_nfs4_state(opendata);
1739         status = PTR_ERR(state);
1740         if (IS_ERR(state))
1741                 goto err_opendata_put;
1742         if (server->caps & NFS_CAP_POSIX_LOCK)
1743                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
1744
1745         if (opendata->o_arg.open_flags & O_EXCL) {
1746                 nfs4_exclusive_attrset(opendata, sattr);
1747
1748                 nfs_fattr_init(opendata->o_res.f_attr);
1749                 status = nfs4_do_setattr(state->inode, cred,
1750                                 opendata->o_res.f_attr, sattr,
1751                                 state);
1752                 if (status == 0)
1753                         nfs_setattr_update_inode(state->inode, sattr);
1754                 nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
1755         }
1756         nfs4_opendata_put(opendata);
1757         nfs4_put_state_owner(sp);
1758         *res = state;
1759         return 0;
1760 err_opendata_put:
1761         nfs4_opendata_put(opendata);
1762 err_put_state_owner:
1763         nfs4_put_state_owner(sp);
1764 out_err:
1765         *res = NULL;
1766         return status;
1767 }
1768
1769
1770 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)
1771 {
1772         struct nfs4_exception exception = { };
1773         struct nfs4_state *res;
1774         int status;
1775
1776         do {
1777                 status = _nfs4_do_open(dir, path, fmode, flags, sattr, cred, &res);
1778                 if (status == 0)
1779                         break;
1780                 /* NOTE: BAD_SEQID means the server and client disagree about the
1781                  * book-keeping w.r.t. state-changing operations
1782                  * (OPEN/CLOSE/LOCK/LOCKU...)
1783                  * It is actually a sign of a bug on the client or on the server.
1784                  *
1785                  * If we receive a BAD_SEQID error in the particular case of
1786                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
1787                  * have unhashed the old state_owner for us, and that we can
1788                  * therefore safely retry using a new one. We should still warn
1789                  * the user though...
1790                  */
1791                 if (status == -NFS4ERR_BAD_SEQID) {
1792                         printk(KERN_WARNING "NFS: v4 server %s "
1793                                         " returned a bad sequence-id error!\n",
1794                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
1795                         exception.retry = 1;
1796                         continue;
1797                 }
1798                 /*
1799                  * BAD_STATEID on OPEN means that the server cancelled our
1800                  * state before it received the OPEN_CONFIRM.
1801                  * Recover by retrying the request as per the discussion
1802                  * on Page 181 of RFC3530.
1803                  */
1804                 if (status == -NFS4ERR_BAD_STATEID) {
1805                         exception.retry = 1;
1806                         continue;
1807                 }
1808                 if (status == -EAGAIN) {
1809                         /* We must have found a delegation */
1810                         exception.retry = 1;
1811                         continue;
1812                 }
1813                 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1814                                         status, &exception));
1815         } while (exception.retry);
1816         return res;
1817 }
1818
1819 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1820                             struct nfs_fattr *fattr, struct iattr *sattr,
1821                             struct nfs4_state *state)
1822 {
1823         struct nfs_server *server = NFS_SERVER(inode);
1824         struct nfs_setattrargs  arg = {
1825                 .fh             = NFS_FH(inode),
1826                 .iap            = sattr,
1827                 .server         = server,
1828                 .bitmask = server->attr_bitmask,
1829         };
1830         struct nfs_setattrres  res = {
1831                 .fattr          = fattr,
1832                 .server         = server,
1833         };
1834         struct rpc_message msg = {
1835                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
1836                 .rpc_argp       = &arg,
1837                 .rpc_resp       = &res,
1838                 .rpc_cred       = cred,
1839         };
1840         unsigned long timestamp = jiffies;
1841         int status;
1842
1843         nfs_fattr_init(fattr);
1844
1845         if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) {
1846                 /* Use that stateid */
1847         } else if (state != NULL) {
1848                 nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid);
1849         } else
1850                 memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));
1851
1852         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
1853         if (status == 0 && state != NULL)
1854                 renew_lease(server, timestamp);
1855         return status;
1856 }
1857
1858 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
1859                            struct nfs_fattr *fattr, struct iattr *sattr,
1860                            struct nfs4_state *state)
1861 {
1862         struct nfs_server *server = NFS_SERVER(inode);
1863         struct nfs4_exception exception = { };
1864         int err;
1865         do {
1866                 err = nfs4_handle_exception(server,
1867                                 _nfs4_do_setattr(inode, cred, fattr, sattr, state),
1868                                 &exception);
1869         } while (exception.retry);
1870         return err;
1871 }
1872
1873 struct nfs4_closedata {
1874         struct path path;
1875         struct inode *inode;
1876         struct nfs4_state *state;
1877         struct nfs_closeargs arg;
1878         struct nfs_closeres res;
1879         struct nfs_fattr fattr;
1880         unsigned long timestamp;
1881         bool roc;
1882         u32 roc_barrier;
1883 };
1884
1885 static void nfs4_free_closedata(void *data)
1886 {
1887         struct nfs4_closedata *calldata = data;
1888         struct nfs4_state_owner *sp = calldata->state->owner;
1889
1890         if (calldata->roc)
1891                 pnfs_roc_release(calldata->state->inode);
1892         nfs4_put_open_state(calldata->state);
1893         nfs_free_seqid(calldata->arg.seqid);
1894         nfs4_put_state_owner(sp);
1895         path_put(&calldata->path);
1896         kfree(calldata);
1897 }
1898
1899 static void nfs4_close_clear_stateid_flags(struct nfs4_state *state,
1900                 fmode_t fmode)
1901 {
1902         spin_lock(&state->owner->so_lock);
1903         if (!(fmode & FMODE_READ))
1904                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1905         if (!(fmode & FMODE_WRITE))
1906                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1907         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1908         spin_unlock(&state->owner->so_lock);
1909 }
1910
1911 static void nfs4_close_done(struct rpc_task *task, void *data)
1912 {
1913         struct nfs4_closedata *calldata = data;
1914         struct nfs4_state *state = calldata->state;
1915         struct nfs_server *server = NFS_SERVER(calldata->inode);
1916
1917         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
1918                 return;
1919         /* hmm. we are done with the inode, and in the process of freeing
1920          * the state_owner. we keep this around to process errors
1921          */
1922         switch (task->tk_status) {
1923                 case 0:
1924                         if (calldata->roc)
1925                                 pnfs_roc_set_barrier(state->inode,
1926                                                      calldata->roc_barrier);
1927                         nfs_set_open_stateid(state, &calldata->res.stateid, 0);
1928                         renew_lease(server, calldata->timestamp);
1929                         nfs4_close_clear_stateid_flags(state,
1930                                         calldata->arg.fmode);
1931                         break;
1932                 case -NFS4ERR_STALE_STATEID:
1933                 case -NFS4ERR_OLD_STATEID:
1934                 case -NFS4ERR_BAD_STATEID:
1935                 case -NFS4ERR_EXPIRED:
1936                         if (calldata->arg.fmode == 0)
1937                                 break;
1938                 default:
1939                         if (nfs4_async_handle_error(task, server, state) == -EAGAIN)
1940                                 rpc_restart_call_prepare(task);
1941         }
1942         nfs_release_seqid(calldata->arg.seqid);
1943         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
1944 }
1945
1946 static void nfs4_close_prepare(struct rpc_task *task, void *data)
1947 {
1948         struct nfs4_closedata *calldata = data;
1949         struct nfs4_state *state = calldata->state;
1950         int call_close = 0;
1951
1952         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1953                 return;
1954
1955         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
1956         calldata->arg.fmode = FMODE_READ|FMODE_WRITE;
1957         spin_lock(&state->owner->so_lock);
1958         /* Calculate the change in open mode */
1959         if (state->n_rdwr == 0) {
1960                 if (state->n_rdonly == 0) {
1961                         call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
1962                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1963                         calldata->arg.fmode &= ~FMODE_READ;
1964                 }
1965                 if (state->n_wronly == 0) {
1966                         call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
1967                         call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
1968                         calldata->arg.fmode &= ~FMODE_WRITE;
1969                 }
1970         }
1971         spin_unlock(&state->owner->so_lock);
1972
1973         if (!call_close) {
1974                 /* Note: exit _without_ calling nfs4_close_done */
1975                 task->tk_action = NULL;
1976                 return;
1977         }
1978
1979         if (calldata->arg.fmode == 0) {
1980                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
1981                 if (calldata->roc &&
1982                     pnfs_roc_drain(calldata->inode, &calldata->roc_barrier)) {
1983                         rpc_sleep_on(&NFS_SERVER(calldata->inode)->roc_rpcwaitq,
1984                                      task, NULL);
1985                         return;
1986                 }
1987         }
1988
1989         nfs_fattr_init(calldata->res.fattr);
1990         calldata->timestamp = jiffies;
1991         if (nfs4_setup_sequence(NFS_SERVER(calldata->inode),
1992                                 &calldata->arg.seq_args, &calldata->res.seq_res,
1993                                 1, task))
1994                 return;
1995         rpc_call_start(task);
1996 }
1997
1998 static const struct rpc_call_ops nfs4_close_ops = {
1999         .rpc_call_prepare = nfs4_close_prepare,
2000         .rpc_call_done = nfs4_close_done,
2001         .rpc_release = nfs4_free_closedata,
2002 };
2003
2004 /*
2005  * It is possible for data to be read/written from a mem-mapped file
2006  * after the sys_close call (which hits the vfs layer as a flush).
2007  * This means that we can't safely call nfsv4 close on a file until
2008  * the inode is cleared. This in turn means that we are not good
2009  * NFSv4 citizens - we do not indicate to the server to update the file's
2010  * share state even when we are done with one of the three share
2011  * stateid's in the inode.
2012  *
2013  * NOTE: Caller must be holding the sp->so_owner semaphore!
2014  */
2015 int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_mask, int wait, bool roc)
2016 {
2017         struct nfs_server *server = NFS_SERVER(state->inode);
2018         struct nfs4_closedata *calldata;
2019         struct nfs4_state_owner *sp = state->owner;
2020         struct rpc_task *task;
2021         struct rpc_message msg = {
2022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2023                 .rpc_cred = state->owner->so_cred,
2024         };
2025         struct rpc_task_setup task_setup_data = {
2026                 .rpc_client = server->client,
2027                 .rpc_message = &msg,
2028                 .callback_ops = &nfs4_close_ops,
2029                 .workqueue = nfsiod_workqueue,
2030                 .flags = RPC_TASK_ASYNC,
2031         };
2032         int status = -ENOMEM;
2033
2034         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2035         if (calldata == NULL)
2036                 goto out;
2037         calldata->inode = state->inode;
2038         calldata->state = state;
2039         calldata->arg.fh = NFS_FH(state->inode);
2040         calldata->arg.stateid = &state->open_stateid;
2041         /* Serialization for the sequence id */
2042         calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid, gfp_mask);
2043         if (calldata->arg.seqid == NULL)
2044                 goto out_free_calldata;
2045         calldata->arg.fmode = 0;
2046         calldata->arg.bitmask = server->cache_consistency_bitmask;
2047         calldata->res.fattr = &calldata->fattr;
2048         calldata->res.seqid = calldata->arg.seqid;
2049         calldata->res.server = server;
2050         calldata->roc = roc;
2051         path_get(path);
2052         calldata->path = *path;
2053
2054         msg.rpc_argp = &calldata->arg;
2055         msg.rpc_resp = &calldata->res;
2056         task_setup_data.callback_data = calldata;
2057         task = rpc_run_task(&task_setup_data);
2058         if (IS_ERR(task))
2059                 return PTR_ERR(task);
2060         status = 0;
2061         if (wait)
2062                 status = rpc_wait_for_completion_task(task);
2063         rpc_put_task(task);
2064         return status;
2065 out_free_calldata:
2066         kfree(calldata);
2067 out:
2068         if (roc)
2069                 pnfs_roc_release(state->inode);
2070         nfs4_put_open_state(state);
2071         nfs4_put_state_owner(sp);
2072         return status;
2073 }
2074
2075 static struct inode *
2076 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx, int open_flags, struct iattr *attr)
2077 {
2078         struct nfs4_state *state;
2079
2080         /* Protect against concurrent sillydeletes */
2081         state = nfs4_do_open(dir, &ctx->path, ctx->mode, open_flags, attr, ctx->cred);
2082         if (IS_ERR(state))
2083                 return ERR_CAST(state);
2084         ctx->state = state;
2085         return igrab(state->inode);
2086 }
2087
2088 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2089 {
2090         if (ctx->state == NULL)
2091                 return;
2092         if (is_sync)
2093                 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
2094         else
2095                 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
2096 }
2097
2098 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2099 {
2100         struct nfs4_server_caps_arg args = {
2101                 .fhandle = fhandle,
2102         };
2103         struct nfs4_server_caps_res res = {};
2104         struct rpc_message msg = {
2105                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2106                 .rpc_argp = &args,
2107                 .rpc_resp = &res,
2108         };
2109         int status;
2110
2111         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2112         if (status == 0) {
2113                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2114                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2115                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2116                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2117                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2118                                 NFS_CAP_CTIME|NFS_CAP_MTIME);
2119                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL)
2120                         server->caps |= NFS_CAP_ACLS;
2121                 if (res.has_links != 0)
2122                         server->caps |= NFS_CAP_HARDLINKS;
2123                 if (res.has_symlinks != 0)
2124                         server->caps |= NFS_CAP_SYMLINKS;
2125                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2126                         server->caps |= NFS_CAP_FILEID;
2127                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2128                         server->caps |= NFS_CAP_MODE;
2129                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2130                         server->caps |= NFS_CAP_NLINK;
2131                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2132                         server->caps |= NFS_CAP_OWNER;
2133                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2134                         server->caps |= NFS_CAP_OWNER_GROUP;
2135                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2136                         server->caps |= NFS_CAP_ATIME;
2137                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2138                         server->caps |= NFS_CAP_CTIME;
2139                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2140                         server->caps |= NFS_CAP_MTIME;
2141
2142                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2143                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2144                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2145                 server->acl_bitmask = res.acl_bitmask;
2146         }
2147
2148         return status;
2149 }
2150
2151 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2152 {
2153         struct nfs4_exception exception = { };
2154         int err;
2155         do {
2156                 err = nfs4_handle_exception(server,
2157                                 _nfs4_server_capabilities(server, fhandle),
2158                                 &exception);
2159         } while (exception.retry);
2160         return err;
2161 }
2162
2163 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2164                 struct nfs_fsinfo *info)
2165 {
2166         struct nfs4_lookup_root_arg args = {
2167                 .bitmask = nfs4_fattr_bitmap,
2168         };
2169         struct nfs4_lookup_res res = {
2170                 .server = server,
2171                 .fattr = info->fattr,
2172                 .fh = fhandle,
2173         };
2174         struct rpc_message msg = {
2175                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2176                 .rpc_argp = &args,
2177                 .rpc_resp = &res,
2178         };
2179
2180         nfs_fattr_init(info->fattr);
2181         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2182 }
2183
2184 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2185                 struct nfs_fsinfo *info)
2186 {
2187         struct nfs4_exception exception = { };
2188         int err;
2189         do {
2190                 err = _nfs4_lookup_root(server, fhandle, info);
2191                 switch (err) {
2192                 case 0:
2193                 case -NFS4ERR_WRONGSEC:
2194                         break;
2195                 default:
2196                         err = nfs4_handle_exception(server, err, &exception);
2197                 }
2198         } while (exception.retry);
2199         return err;
2200 }
2201
2202 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2203                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
2204 {
2205         struct rpc_auth *auth;
2206         int ret;
2207
2208         auth = rpcauth_create(flavor, server->client);
2209         if (!auth) {
2210                 ret = -EIO;
2211                 goto out;
2212         }
2213         ret = nfs4_lookup_root(server, fhandle, info);
2214 out:
2215         return ret;
2216 }
2217
2218 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
2219                               struct nfs_fsinfo *info)
2220 {
2221         int i, len, status = 0;
2222         rpc_authflavor_t flav_array[NFS_MAX_SECFLAVORS];
2223
2224         len = gss_mech_list_pseudoflavors(&flav_array[0]);
2225         flav_array[len] = RPC_AUTH_NULL;
2226         len += 1;
2227
2228         for (i = 0; i < len; i++) {
2229                 status = nfs4_lookup_root_sec(server, fhandle, info, flav_array[i]);
2230                 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
2231                         continue;
2232                 break;
2233         }
2234         /*
2235          * -EACCESS could mean that the user doesn't have correct permissions
2236          * to access the mount.  It could also mean that we tried to mount
2237          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
2238          * existing mount programs don't handle -EACCES very well so it should
2239          * be mapped to -EPERM instead.
2240          */
2241         if (status == -EACCES)
2242                 status = -EPERM;
2243         return status;
2244 }
2245
2246 /*
2247  * get the file handle for the "/" directory on the server
2248  */
2249 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *fhandle,
2250                               struct nfs_fsinfo *info)
2251 {
2252         int status = nfs4_lookup_root(server, fhandle, info);
2253         if ((status == -NFS4ERR_WRONGSEC) && !(server->flags & NFS_MOUNT_SECFLAVOUR))
2254                 /*
2255                  * A status of -NFS4ERR_WRONGSEC will be mapped to -EPERM
2256                  * by nfs4_map_errors() as this function exits.
2257                  */
2258                 status = nfs4_find_root_sec(server, fhandle, info);
2259         if (status == 0)
2260                 status = nfs4_server_capabilities(server, fhandle);
2261         if (status == 0)
2262                 status = nfs4_do_fsinfo(server, fhandle, info);
2263         return nfs4_map_errors(status);
2264 }
2265
2266 /*
2267  * Get locations and (maybe) other attributes of a referral.
2268  * Note that we'll actually follow the referral later when
2269  * we detect fsid mismatch in inode revalidation
2270  */
2271 static int nfs4_get_referral(struct inode *dir, const struct qstr *name, struct nfs_fattr *fattr, struct nfs_fh *fhandle)
2272 {
2273         int status = -ENOMEM;
2274         struct page *page = NULL;
2275         struct nfs4_fs_locations *locations = NULL;
2276
2277         page = alloc_page(GFP_KERNEL);
2278         if (page == NULL)
2279                 goto out;
2280         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2281         if (locations == NULL)
2282                 goto out;
2283
2284         status = nfs4_proc_fs_locations(dir, name, locations, page);
2285         if (status != 0)
2286                 goto out;
2287         /* Make sure server returned a different fsid for the referral */
2288         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
2289                 dprintk("%s: server did not return a different fsid for a referral at %s\n", __func__, name->name);
2290                 status = -EIO;
2291                 goto out;
2292         }
2293
2294         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
2295         fattr->valid |= NFS_ATTR_FATTR_V4_REFERRAL;
2296         if (!fattr->mode)
2297                 fattr->mode = S_IFDIR;
2298         memset(fhandle, 0, sizeof(struct nfs_fh));
2299 out:
2300         if (page)
2301                 __free_page(page);
2302         kfree(locations);
2303         return status;
2304 }
2305
2306 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2307 {
2308         struct nfs4_getattr_arg args = {
2309                 .fh = fhandle,
2310                 .bitmask = server->attr_bitmask,
2311         };
2312         struct nfs4_getattr_res res = {
2313                 .fattr = fattr,
2314                 .server = server,
2315         };
2316         struct rpc_message msg = {
2317                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
2318                 .rpc_argp = &args,
2319                 .rpc_resp = &res,
2320         };
2321
2322         nfs_fattr_init(fattr);
2323         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2324 }
2325
2326 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2327 {
2328         struct nfs4_exception exception = { };
2329         int err;
2330         do {
2331                 err = nfs4_handle_exception(server,
2332                                 _nfs4_proc_getattr(server, fhandle, fattr),
2333                                 &exception);
2334         } while (exception.retry);
2335         return err;
2336 }
2337
2338 /*
2339  * The file is not closed if it is opened due to the a request to change
2340  * the size of the file. The open call will not be needed once the
2341  * VFS layer lookup-intents are implemented.
2342  *
2343  * Close is called when the inode is destroyed.
2344  * If we haven't opened the file for O_WRONLY, we
2345  * need to in the size_change case to obtain a stateid.
2346  *
2347  * Got race?
2348  * Because OPEN is always done by name in nfsv4, it is
2349  * possible that we opened a different file by the same
2350  * name.  We can recognize this race condition, but we
2351  * can't do anything about it besides returning an error.
2352  *
2353  * This will be fixed with VFS changes (lookup-intent).
2354  */
2355 static int
2356 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
2357                   struct iattr *sattr)
2358 {
2359         struct inode *inode = dentry->d_inode;
2360         struct rpc_cred *cred = NULL;
2361         struct nfs4_state *state = NULL;
2362         int status;
2363
2364         nfs_fattr_init(fattr);
2365
2366         /* Search for an existing open(O_WRITE) file */
2367         if (sattr->ia_valid & ATTR_FILE) {
2368                 struct nfs_open_context *ctx;
2369
2370                 ctx = nfs_file_open_context(sattr->ia_file);
2371                 if (ctx) {
2372                         cred = ctx->cred;
2373                         state = ctx->state;
2374                 }
2375         }
2376
2377         status = nfs4_do_setattr(inode, cred, fattr, sattr, state);
2378         if (status == 0)
2379                 nfs_setattr_update_inode(inode, sattr);
2380         return status;
2381 }
2382
2383 static int _nfs4_proc_lookupfh(struct rpc_clnt *clnt, struct nfs_server *server,
2384                 const struct nfs_fh *dirfh, const struct qstr *name,
2385                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2386 {
2387         int                    status;
2388         struct nfs4_lookup_arg args = {
2389                 .bitmask = server->attr_bitmask,
2390                 .dir_fh = dirfh,
2391                 .name = name,
2392         };
2393         struct nfs4_lookup_res res = {
2394                 .server = server,
2395                 .fattr = fattr,
2396                 .fh = fhandle,
2397         };
2398         struct rpc_message msg = {
2399                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
2400                 .rpc_argp = &args,
2401                 .rpc_resp = &res,
2402         };
2403
2404         nfs_fattr_init(fattr);
2405
2406         dprintk("NFS call  lookupfh %s\n", name->name);
2407         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
2408         dprintk("NFS reply lookupfh: %d\n", status);
2409         return status;
2410 }
2411
2412 static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
2413                               struct qstr *name, struct nfs_fh *fhandle,
2414                               struct nfs_fattr *fattr)
2415 {
2416         struct nfs4_exception exception = { };
2417         int err;
2418         do {
2419                 err = _nfs4_proc_lookupfh(server->client, server, dirfh, name, fhandle, fattr);
2420                 /* FIXME: !!!! */
2421                 if (err == -NFS4ERR_MOVED) {
2422                         err = -EREMOTE;
2423                         break;
2424                 }
2425                 err = nfs4_handle_exception(server, err, &exception);
2426         } while (exception.retry);
2427         return err;
2428 }
2429
2430 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
2431                 const struct qstr *name, struct nfs_fh *fhandle,
2432                 struct nfs_fattr *fattr)
2433 {
2434         int status;
2435
2436         dprintk("NFS call  lookup %s\n", name->name);
2437         status = _nfs4_proc_lookupfh(clnt, NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
2438         if (status == -NFS4ERR_MOVED)
2439                 status = nfs4_get_referral(dir, name, fattr, fhandle);
2440         dprintk("NFS reply lookup: %d\n", status);
2441         return status;
2442 }
2443
2444 void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr, struct nfs_fh *fh)
2445 {
2446         memset(fh, 0, sizeof(struct nfs_fh));
2447         fattr->fsid.major = 1;
2448         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
2449                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_FSID | NFS_ATTR_FATTR_MOUNTPOINT;
2450         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
2451         fattr->nlink = 2;
2452 }
2453
2454 static int nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir, struct qstr *name,
2455                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
2456 {
2457         struct nfs4_exception exception = { };
2458         int err;
2459         do {
2460                 err = nfs4_handle_exception(NFS_SERVER(dir),
2461                                 _nfs4_proc_lookup(clnt, dir, name, fhandle, fattr),
2462                                 &exception);
2463                 if (err == -EPERM)
2464                         nfs_fixup_secinfo_attributes(fattr, fhandle);
2465         } while (exception.retry);
2466         return err;
2467 }
2468
2469 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2470 {
2471         struct nfs_server *server = NFS_SERVER(inode);
2472         struct nfs4_accessargs args = {
2473                 .fh = NFS_FH(inode),
2474                 .bitmask = server->attr_bitmask,
2475         };
2476         struct nfs4_accessres res = {
2477                 .server = server,
2478         };
2479         struct rpc_message msg = {
2480                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
2481                 .rpc_argp = &args,
2482                 .rpc_resp = &res,
2483                 .rpc_cred = entry->cred,
2484         };
2485         int mode = entry->mask;
2486         int status;
2487
2488         /*
2489          * Determine which access bits we want to ask for...
2490          */
2491         if (mode & MAY_READ)
2492                 args.access |= NFS4_ACCESS_READ;
2493         if (S_ISDIR(inode->i_mode)) {
2494                 if (mode & MAY_WRITE)
2495                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
2496                 if (mode & MAY_EXEC)
2497                         args.access |= NFS4_ACCESS_LOOKUP;
2498         } else {
2499                 if (mode & MAY_WRITE)
2500                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
2501                 if (mode & MAY_EXEC)
2502                         args.access |= NFS4_ACCESS_EXECUTE;
2503         }
2504
2505         res.fattr = nfs_alloc_fattr();
2506         if (res.fattr == NULL)
2507                 return -ENOMEM;
2508
2509         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2510         if (!status) {
2511                 entry->mask = 0;
2512                 if (res.access & NFS4_ACCESS_READ)
2513                         entry->mask |= MAY_READ;
2514                 if (res.access & (NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE))
2515                         entry->mask |= MAY_WRITE;
2516                 if (res.access & (NFS4_ACCESS_LOOKUP|NFS4_ACCESS_EXECUTE))
2517                         entry->mask |= MAY_EXEC;
2518                 nfs_refresh_inode(inode, res.fattr);
2519         }
2520         nfs_free_fattr(res.fattr);
2521         return status;
2522 }
2523
2524 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
2525 {
2526         struct nfs4_exception exception = { };
2527         int err;
2528         do {
2529                 err = nfs4_handle_exception(NFS_SERVER(inode),
2530                                 _nfs4_proc_access(inode, entry),
2531                                 &exception);
2532         } while (exception.retry);
2533         return err;
2534 }
2535
2536 /*
2537  * TODO: For the time being, we don't try to get any attributes
2538  * along with any of the zero-copy operations READ, READDIR,
2539  * READLINK, WRITE.
2540  *
2541  * In the case of the first three, we want to put the GETATTR
2542  * after the read-type operation -- this is because it is hard
2543  * to predict the length of a GETATTR response in v4, and thus
2544  * align the READ data correctly.  This means that the GETATTR
2545  * may end up partially falling into the page cache, and we should
2546  * shift it into the 'tail' of the xdr_buf before processing.
2547  * To do this efficiently, we need to know the total length
2548  * of data received, which doesn't seem to be available outside
2549  * of the RPC layer.
2550  *
2551  * In the case of WRITE, we also want to put the GETATTR after
2552  * the operation -- in this case because we want to make sure
2553  * we get the post-operation mtime and size.  This means that
2554  * we can't use xdr_encode_pages() as written: we need a variant
2555  * of it which would leave room in the 'tail' iovec.
2556  *
2557  * Both of these changes to the XDR layer would in fact be quite
2558  * minor, but I decided to leave them for a subsequent patch.
2559  */
2560 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
2561                 unsigned int pgbase, unsigned int pglen)
2562 {
2563         struct nfs4_readlink args = {
2564                 .fh       = NFS_FH(inode),
2565                 .pgbase   = pgbase,
2566                 .pglen    = pglen,
2567                 .pages    = &page,
2568         };
2569         struct nfs4_readlink_res res;
2570         struct rpc_message msg = {
2571                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
2572                 .rpc_argp = &args,
2573                 .rpc_resp = &res,
2574         };
2575
2576         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
2577 }
2578
2579 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
2580                 unsigned int pgbase, unsigned int pglen)
2581 {
2582         struct nfs4_exception exception = { };
2583         int err;
2584         do {
2585                 err = nfs4_handle_exception(NFS_SERVER(inode),
2586                                 _nfs4_proc_readlink(inode, page, pgbase, pglen),
2587                                 &exception);
2588         } while (exception.retry);
2589         return err;
2590 }
2591
2592 /*
2593  * Got race?
2594  * We will need to arrange for the VFS layer to provide an atomic open.
2595  * Until then, this create/open method is prone to inefficiency and race
2596  * conditions due to the lookup, create, and open VFS calls from sys_open()
2597  * placed on the wire.
2598  *
2599  * Given the above sorry state of affairs, I'm simply sending an OPEN.
2600  * The file will be opened again in the subsequent VFS open call
2601  * (nfs4_proc_file_open).
2602  *
2603  * The open for read will just hang around to be used by any process that
2604  * opens the file O_RDONLY. This will all be resolved with the VFS changes.
2605  */
2606
2607 static int
2608 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
2609                  int flags, struct nfs_open_context *ctx)
2610 {
2611         struct path my_path = {
2612                 .dentry = dentry,
2613         };
2614         struct path *path = &my_path;
2615         struct nfs4_state *state;
2616         struct rpc_cred *cred = NULL;
2617         fmode_t fmode = 0;
2618         int status = 0;
2619
2620         if (ctx != NULL) {
2621                 cred = ctx->cred;
2622                 path = &ctx->path;
2623                 fmode = ctx->mode;
2624         }
2625         sattr->ia_mode &= ~current_umask();
2626         state = nfs4_do_open(dir, path, fmode, flags, sattr, cred);
2627         d_drop(dentry);
2628         if (IS_ERR(state)) {
2629                 status = PTR_ERR(state);
2630                 goto out;
2631         }
2632         d_add(dentry, igrab(state->inode));
2633         nfs_set_verifier(dentry, nfs_save_change_attribute(dir));
2634         if (ctx != NULL)
2635                 ctx->state = state;
2636         else
2637                 nfs4_close_sync(path, state, fmode);
2638 out:
2639         return status;
2640 }
2641
2642 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
2643 {
2644         struct nfs_server *server = NFS_SERVER(dir);
2645         struct nfs_removeargs args = {
2646                 .fh = NFS_FH(dir),
2647                 .name.len = name->len,
2648                 .name.name = name->name,
2649                 .bitmask = server->attr_bitmask,
2650         };
2651         struct nfs_removeres res = {
2652                 .server = server,
2653         };
2654         struct rpc_message msg = {
2655                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
2656                 .rpc_argp = &args,
2657                 .rpc_resp = &res,
2658         };
2659         int status = -ENOMEM;
2660
2661         res.dir_attr = nfs_alloc_fattr();
2662         if (res.dir_attr == NULL)
2663                 goto out;
2664
2665         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
2666         if (status == 0) {
2667                 update_changeattr(dir, &res.cinfo);
2668                 nfs_post_op_update_inode(dir, res.dir_attr);
2669         }
2670         nfs_free_fattr(res.dir_attr);
2671 out:
2672         return status;
2673 }
2674
2675 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
2676 {
2677         struct nfs4_exception exception = { };
2678         int err;
2679         do {
2680                 err = nfs4_handle_exception(NFS_SERVER(dir),
2681                                 _nfs4_proc_remove(dir, name),
2682                                 &exception);
2683         } while (exception.retry);
2684         return err;
2685 }
2686
2687 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2688 {
2689         struct nfs_server *server = NFS_SERVER(dir);
2690         struct nfs_removeargs *args = msg->rpc_argp;
2691         struct nfs_removeres *res = msg->rpc_resp;
2692
2693         args->bitmask = server->cache_consistency_bitmask;
2694         res->server = server;
2695         res->seq_res.sr_slot = NULL;
2696         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2697 }
2698
2699 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
2700 {
2701         struct nfs_removeres *res = task->tk_msg.rpc_resp;
2702
2703         if (!nfs4_sequence_done(task, &res->seq_res))
2704                 return 0;
2705         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2706                 return 0;
2707         update_changeattr(dir, &res->cinfo);
2708         nfs_post_op_update_inode(dir, res->dir_attr);
2709         return 1;
2710 }
2711
2712 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
2713 {
2714         struct nfs_server *server = NFS_SERVER(dir);
2715         struct nfs_renameargs *arg = msg->rpc_argp;
2716         struct nfs_renameres *res = msg->rpc_resp;
2717
2718         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
2719         arg->bitmask = server->attr_bitmask;
2720         res->server = server;
2721 }
2722
2723 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
2724                                  struct inode *new_dir)
2725 {
2726         struct nfs_renameres *res = task->tk_msg.rpc_resp;
2727
2728         if (!nfs4_sequence_done(task, &res->seq_res))
2729                 return 0;
2730         if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN)
2731                 return 0;
2732
2733         update_changeattr(old_dir, &res->old_cinfo);
2734         nfs_post_op_update_inode(old_dir, res->old_fattr);
2735         update_changeattr(new_dir, &res->new_cinfo);
2736         nfs_post_op_update_inode(new_dir, res->new_fattr);
2737         return 1;
2738 }
2739
2740 static int _nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2741                 struct inode *new_dir, struct qstr *new_name)
2742 {
2743         struct nfs_server *server = NFS_SERVER(old_dir);
2744         struct nfs_renameargs arg = {
2745                 .old_dir = NFS_FH(old_dir),
2746                 .new_dir = NFS_FH(new_dir),
2747                 .old_name = old_name,
2748                 .new_name = new_name,
2749                 .bitmask = server->attr_bitmask,
2750         };
2751         struct nfs_renameres res = {
2752                 .server = server,
2753         };
2754         struct rpc_message msg = {
2755                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME],
2756                 .rpc_argp = &arg,
2757                 .rpc_resp = &res,
2758         };
2759         int status = -ENOMEM;
2760
2761         res.old_fattr = nfs_alloc_fattr();
2762         res.new_fattr = nfs_alloc_fattr();
2763         if (res.old_fattr == NULL || res.new_fattr == NULL)
2764                 goto out;
2765
2766         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2767         if (!status) {
2768                 update_changeattr(old_dir, &res.old_cinfo);
2769                 nfs_post_op_update_inode(old_dir, res.old_fattr);
2770                 update_changeattr(new_dir, &res.new_cinfo);
2771                 nfs_post_op_update_inode(new_dir, res.new_fattr);
2772         }
2773 out:
2774         nfs_free_fattr(res.new_fattr);
2775         nfs_free_fattr(res.old_fattr);
2776         return status;
2777 }
2778
2779 static int nfs4_proc_rename(struct inode *old_dir, struct qstr *old_name,
2780                 struct inode *new_dir, struct qstr *new_name)
2781 {
2782         struct nfs4_exception exception = { };
2783         int err;
2784         do {
2785                 err = nfs4_handle_exception(NFS_SERVER(old_dir),
2786                                 _nfs4_proc_rename(old_dir, old_name,
2787                                         new_dir, new_name),
2788                                 &exception);
2789         } while (exception.retry);
2790         return err;
2791 }
2792
2793 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2794 {
2795         struct nfs_server *server = NFS_SERVER(inode);
2796         struct nfs4_link_arg arg = {
2797                 .fh     = NFS_FH(inode),
2798                 .dir_fh = NFS_FH(dir),
2799                 .name   = name,
2800                 .bitmask = server->attr_bitmask,
2801         };
2802         struct nfs4_link_res res = {
2803                 .server = server,
2804         };
2805         struct rpc_message msg = {
2806                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
2807                 .rpc_argp = &arg,
2808                 .rpc_resp = &res,
2809         };
2810         int status = -ENOMEM;
2811
2812         res.fattr = nfs_alloc_fattr();
2813         res.dir_attr = nfs_alloc_fattr();
2814         if (res.fattr == NULL || res.dir_attr == NULL)
2815                 goto out;
2816
2817         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2818         if (!status) {
2819                 update_changeattr(dir, &res.cinfo);
2820                 nfs_post_op_update_inode(dir, res.dir_attr);
2821                 nfs_post_op_update_inode(inode, res.fattr);
2822         }
2823 out:
2824         nfs_free_fattr(res.dir_attr);
2825         nfs_free_fattr(res.fattr);
2826         return status;
2827 }
2828
2829 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
2830 {
2831         struct nfs4_exception exception = { };
2832         int err;
2833         do {
2834                 err = nfs4_handle_exception(NFS_SERVER(inode),
2835                                 _nfs4_proc_link(inode, dir, name),
2836                                 &exception);
2837         } while (exception.retry);
2838         return err;
2839 }
2840
2841 struct nfs4_createdata {
2842         struct rpc_message msg;
2843         struct nfs4_create_arg arg;
2844         struct nfs4_create_res res;
2845         struct nfs_fh fh;
2846         struct nfs_fattr fattr;
2847         struct nfs_fattr dir_fattr;
2848 };
2849
2850 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
2851                 struct qstr *name, struct iattr *sattr, u32 ftype)
2852 {
2853         struct nfs4_createdata *data;
2854
2855         data = kzalloc(sizeof(*data), GFP_KERNEL);
2856         if (data != NULL) {
2857                 struct nfs_server *server = NFS_SERVER(dir);
2858
2859                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
2860                 data->msg.rpc_argp = &data->arg;
2861                 data->msg.rpc_resp = &data->res;
2862                 data->arg.dir_fh = NFS_FH(dir);
2863                 data->arg.server = server;
2864                 data->arg.name = name;
2865                 data->arg.attrs = sattr;
2866                 data->arg.ftype = ftype;
2867                 data->arg.bitmask = server->attr_bitmask;
2868                 data->res.server = server;
2869                 data->res.fh = &data->fh;
2870                 data->res.fattr = &data->fattr;
2871                 data->res.dir_fattr = &data->dir_fattr;
2872                 nfs_fattr_init(data->res.fattr);
2873                 nfs_fattr_init(data->res.dir_fattr);
2874         }
2875         return data;
2876 }
2877
2878 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
2879 {
2880         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
2881                                     &data->arg.seq_args, &data->res.seq_res, 1);
2882         if (status == 0) {
2883                 update_changeattr(dir, &data->res.dir_cinfo);
2884                 nfs_post_op_update_inode(dir, data->res.dir_fattr);
2885                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
2886         }
2887         return status;
2888 }
2889
2890 static void nfs4_free_createdata(struct nfs4_createdata *data)
2891 {
2892         kfree(data);
2893 }
2894
2895 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2896                 struct page *page, unsigned int len, struct iattr *sattr)
2897 {
2898         struct nfs4_createdata *data;
2899         int status = -ENAMETOOLONG;
2900
2901         if (len > NFS4_MAXPATHLEN)
2902                 goto out;
2903
2904         status = -ENOMEM;
2905         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
2906         if (data == NULL)
2907                 goto out;
2908
2909         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
2910         data->arg.u.symlink.pages = &page;
2911         data->arg.u.symlink.len = len;
2912
2913         status = nfs4_do_create(dir, dentry, data);
2914
2915         nfs4_free_createdata(data);
2916 out:
2917         return status;
2918 }
2919
2920 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
2921                 struct page *page, unsigned int len, struct iattr *sattr)
2922 {
2923         struct nfs4_exception exception = { };
2924         int err;
2925         do {
2926                 err = nfs4_handle_exception(NFS_SERVER(dir),
2927                                 _nfs4_proc_symlink(dir, dentry, page,
2928                                                         len, sattr),
2929                                 &exception);
2930         } while (exception.retry);
2931         return err;
2932 }
2933
2934 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2935                 struct iattr *sattr)
2936 {
2937         struct nfs4_createdata *data;
2938         int status = -ENOMEM;
2939
2940         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
2941         if (data == NULL)
2942                 goto out;
2943
2944         status = nfs4_do_create(dir, dentry, data);
2945
2946         nfs4_free_createdata(data);
2947 out:
2948         return status;
2949 }
2950
2951 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
2952                 struct iattr *sattr)
2953 {
2954         struct nfs4_exception exception = { };
2955         int err;
2956
2957         sattr->ia_mode &= ~current_umask();
2958         do {
2959                 err = nfs4_handle_exception(NFS_SERVER(dir),
2960                                 _nfs4_proc_mkdir(dir, dentry, sattr),
2961                                 &exception);
2962         } while (exception.retry);
2963         return err;
2964 }
2965
2966 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2967                 u64 cookie, struct page **pages, unsigned int count, int plus)
2968 {
2969         struct inode            *dir = dentry->d_inode;
2970         struct nfs4_readdir_arg args = {
2971                 .fh = NFS_FH(dir),
2972                 .pages = pages,
2973                 .pgbase = 0,
2974                 .count = count,
2975                 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask,
2976                 .plus = plus,
2977         };
2978         struct nfs4_readdir_res res;
2979         struct rpc_message msg = {
2980                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
2981                 .rpc_argp = &args,
2982                 .rpc_resp = &res,
2983                 .rpc_cred = cred,
2984         };
2985         int                     status;
2986
2987         dprintk("%s: dentry = %s/%s, cookie = %Lu\n", __func__,
2988                         dentry->d_parent->d_name.name,
2989                         dentry->d_name.name,
2990                         (unsigned long long)cookie);
2991         nfs4_setup_readdir(cookie, NFS_COOKIEVERF(dir), dentry, &args);
2992         res.pgbase = args.pgbase;
2993         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
2994         if (status >= 0) {
2995                 memcpy(NFS_COOKIEVERF(dir), res.verifier.data, NFS4_VERIFIER_SIZE);
2996                 status += args.pgbase;
2997         }
2998
2999         nfs_invalidate_atime(dir);
3000
3001         dprintk("%s: returns %d\n", __func__, status);
3002         return status;
3003 }
3004
3005 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3006                 u64 cookie, struct page **pages, unsigned int count, int plus)
3007 {
3008         struct nfs4_exception exception = { };
3009         int err;
3010         do {
3011                 err = nfs4_handle_exception(NFS_SERVER(dentry->d_inode),
3012                                 _nfs4_proc_readdir(dentry, cred, cookie,
3013                                         pages, count, plus),
3014                                 &exception);
3015         } while (exception.retry);
3016         return err;
3017 }
3018
3019 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3020                 struct iattr *sattr, dev_t rdev)
3021 {
3022         struct nfs4_createdata *data;
3023         int mode = sattr->ia_mode;
3024         int status = -ENOMEM;
3025
3026         BUG_ON(!(sattr->ia_valid & ATTR_MODE));
3027         BUG_ON(!S_ISFIFO(mode) && !S_ISBLK(mode) && !S_ISCHR(mode) && !S_ISSOCK(mode));
3028
3029         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3030         if (data == NULL)
3031                 goto out;
3032
3033         if (S_ISFIFO(mode))
3034                 data->arg.ftype = NF4FIFO;
3035         else if (S_ISBLK(mode)) {
3036                 data->arg.ftype = NF4BLK;
3037                 data->arg.u.device.specdata1 = MAJOR(rdev);
3038                 data->arg.u.device.specdata2 = MINOR(rdev);
3039         }
3040         else if (S_ISCHR(mode)) {
3041                 data->arg.ftype = NF4CHR;
3042                 data->arg.u.device.specdata1 = MAJOR(rdev);
3043                 data->arg.u.device.specdata2 = MINOR(rdev);
3044         }
3045
3046         status = nfs4_do_create(dir, dentry, data);
3047
3048         nfs4_free_createdata(data);
3049 out:
3050         return status;
3051 }
3052
3053 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3054                 struct iattr *sattr, dev_t rdev)
3055 {
3056         struct nfs4_exception exception = { };
3057         int err;
3058
3059         sattr->ia_mode &= ~current_umask();
3060         do {
3061                 err = nfs4_handle_exception(NFS_SERVER(dir),
3062                                 _nfs4_proc_mknod(dir, dentry, sattr, rdev),
3063                                 &exception);
3064         } while (exception.retry);
3065         return err;
3066 }
3067
3068 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3069                  struct nfs_fsstat *fsstat)
3070 {
3071         struct nfs4_statfs_arg args = {
3072                 .fh = fhandle,
3073                 .bitmask = server->attr_bitmask,
3074         };
3075         struct nfs4_statfs_res res = {
3076                 .fsstat = fsstat,
3077         };
3078         struct rpc_message msg = {
3079                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
3080                 .rpc_argp = &args,
3081                 .rpc_resp = &res,
3082         };
3083
3084         nfs_fattr_init(fsstat->fattr);
3085         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3086 }
3087
3088 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
3089 {
3090         struct nfs4_exception exception = { };
3091         int err;
3092         do {
3093                 err = nfs4_handle_exception(server,
3094                                 _nfs4_proc_statfs(server, fhandle, fsstat),
3095                                 &exception);
3096         } while (exception.retry);
3097         return err;
3098 }
3099
3100 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
3101                 struct nfs_fsinfo *fsinfo)
3102 {
3103         struct nfs4_fsinfo_arg args = {
3104                 .fh = fhandle,
3105                 .bitmask = server->attr_bitmask,
3106         };
3107         struct nfs4_fsinfo_res res = {
3108                 .fsinfo = fsinfo,
3109         };
3110         struct rpc_message msg = {
3111                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
3112                 .rpc_argp = &args,
3113                 .rpc_resp = &res,
3114         };
3115
3116         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3117 }
3118
3119 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3120 {
3121         struct nfs4_exception exception = { };
3122         int err;
3123
3124         do {
3125                 err = nfs4_handle_exception(server,
3126                                 _nfs4_do_fsinfo(server, fhandle, fsinfo),
3127                                 &exception);
3128         } while (exception.retry);
3129         return err;
3130 }
3131
3132 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
3133 {
3134         nfs_fattr_init(fsinfo->fattr);
3135         return nfs4_do_fsinfo(server, fhandle, fsinfo);
3136 }
3137
3138 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3139                 struct nfs_pathconf *pathconf)
3140 {
3141         struct nfs4_pathconf_arg args = {
3142                 .fh = fhandle,
3143                 .bitmask = server->attr_bitmask,
3144         };
3145         struct nfs4_pathconf_res res = {
3146                 .pathconf = pathconf,
3147         };
3148         struct rpc_message msg = {
3149                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
3150                 .rpc_argp = &args,
3151                 .rpc_resp = &res,
3152         };
3153
3154         /* None of the pathconf attributes are mandatory to implement */
3155         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
3156                 memset(pathconf, 0, sizeof(*pathconf));
3157                 return 0;
3158         }
3159
3160         nfs_fattr_init(pathconf->fattr);
3161         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3162 }
3163
3164 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
3165                 struct nfs_pathconf *pathconf)
3166 {
3167         struct nfs4_exception exception = { };
3168         int err;
3169
3170         do {
3171                 err = nfs4_handle_exception(server,
3172                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
3173                                 &exception);
3174         } while (exception.retry);
3175         return err;
3176 }
3177
3178 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
3179 {
3180         struct nfs_server *server = NFS_SERVER(data->inode);
3181
3182         if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) {
3183                 nfs_restart_rpc(task, server->nfs_client);
3184                 return -EAGAIN;
3185         }
3186
3187         nfs_invalidate_atime(data->inode);
3188         if (task->tk_status > 0)
3189                 renew_lease(server, data->timestamp);
3190         return 0;
3191 }
3192
3193 static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data)
3194 {
3195
3196         dprintk("--> %s\n", __func__);
3197
3198         if (!nfs4_sequence_done(task, &data->res.seq_res))
3199                 return -EAGAIN;
3200
3201         return data->read_done_cb(task, data);
3202 }
3203
3204 static void nfs4_proc_read_setup(struct nfs_read_data *data, struct rpc_message *msg)
3205 {
3206         data->timestamp   = jiffies;
3207         data->read_done_cb = nfs4_read_done_cb;
3208         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
3209 }
3210
3211 /* Reset the the nfs_read_data to send the read to the MDS. */
3212 void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
3213 {
3214         dprintk("%s Reset task for i/o through\n", __func__);
3215         put_lseg(data->lseg);
3216         data->lseg = NULL;
3217         /* offsets will differ in the dense stripe case */
3218         data->args.offset = data->mds_offset;
3219         data->ds_clp = NULL;
3220         data->args.fh     = NFS_FH(data->inode);
3221         data->read_done_cb = nfs4_read_done_cb;
3222         task->tk_ops = data->mds_ops;
3223         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3224 }
3225 EXPORT_SYMBOL_GPL(nfs4_reset_read);
3226
3227 static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3228 {
3229         struct inode *inode = data->inode;
3230
3231         if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) {
3232                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3233                 return -EAGAIN;
3234         }
3235         if (task->tk_status >= 0) {
3236                 renew_lease(NFS_SERVER(inode), data->timestamp);
3237                 nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
3238         }
3239         return 0;
3240 }
3241
3242 static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data)
3243 {
3244         if (!nfs4_sequence_done(task, &data->res.seq_res))
3245                 return -EAGAIN;
3246         return data->write_done_cb(task, data);
3247 }
3248
3249 /* Reset the the nfs_write_data to send the write to the MDS. */
3250 void nfs4_reset_write(struct rpc_task *task, struct nfs_write_data *data)
3251 {
3252         dprintk("%s Reset task for i/o through\n", __func__);
3253         put_lseg(data->lseg);
3254         data->lseg          = NULL;
3255         data->ds_clp        = NULL;
3256         data->write_done_cb = nfs4_write_done_cb;
3257         data->args.fh       = NFS_FH(data->inode);
3258         data->args.bitmask  = data->res.server->cache_consistency_bitmask;
3259         data->args.offset   = data->mds_offset;
3260         data->res.fattr     = &data->fattr;
3261         task->tk_ops        = data->mds_ops;
3262         rpc_task_reset_client(task, NFS_CLIENT(data->inode));
3263 }
3264 EXPORT_SYMBOL_GPL(nfs4_reset_write);
3265
3266 static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_message *msg)
3267 {
3268         struct nfs_server *server = NFS_SERVER(data->inode);
3269
3270         if (data->lseg) {
3271                 data->args.bitmask = NULL;
3272                 data->res.fattr = NULL;
3273         } else
3274                 data->args.bitmask = server->cache_consistency_bitmask;
3275         if (!data->write_done_cb)
3276                 data->write_done_cb = nfs4_write_done_cb;
3277         data->res.server = server;
3278         data->timestamp   = jiffies;
3279
3280         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
3281 }
3282
3283 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *data)
3284 {
3285         struct inode *inode = data->inode;
3286
3287         if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) {
3288                 nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client);
3289                 return -EAGAIN;
3290         }
3291         nfs_refresh_inode(inode, data->res.fattr);
3292         return 0;
3293 }
3294
3295 static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data)
3296 {
3297         if (!nfs4_sequence_done(task, &data->res.seq_res))
3298                 return -EAGAIN;
3299         return data->write_done_cb(task, data);
3300 }
3301
3302 static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_message *msg)
3303 {
3304         struct nfs_server *server = NFS_SERVER(data->inode);
3305
3306         if (data->lseg) {
3307                 data->args.bitmask = NULL;
3308                 data->res.fattr = NULL;
3309         } else
3310                 data->args.bitmask = server->cache_consistency_bitmask;
3311         if (!data->write_done_cb)
3312                 data->write_done_cb = nfs4_commit_done_cb;
3313         data->res.server = server;
3314         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
3315 }
3316
3317 struct nfs4_renewdata {
3318         struct nfs_client       *client;
3319         unsigned long           timestamp;
3320 };
3321
3322 /*
3323  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
3324  * standalone procedure for queueing an asynchronous RENEW.
3325  */
3326 static void nfs4_renew_release(void *calldata)
3327 {
3328         struct nfs4_renewdata *data = calldata;
3329         struct nfs_client *clp = data->client;
3330
3331         if (atomic_read(&clp->cl_count) > 1)
3332                 nfs4_schedule_state_renewal(clp);
3333         nfs_put_client(clp);
3334         kfree(data);
3335 }
3336
3337 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
3338 {
3339         struct nfs4_renewdata *data = calldata;
3340         struct nfs_client *clp = data->client;
3341         unsigned long timestamp = data->timestamp;
3342
3343         if (task->tk_status < 0) {
3344                 /* Unless we're shutting down, schedule state recovery! */
3345                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) != 0)
3346                         nfs4_schedule_lease_recovery(clp);
3347                 return;
3348         }
3349         do_renew_lease(clp, timestamp);
3350 }
3351
3352 static const struct rpc_call_ops nfs4_renew_ops = {
3353         .rpc_call_done = nfs4_renew_done,
3354         .rpc_release = nfs4_renew_release,
3355 };
3356
3357 int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred)
3358 {
3359         struct rpc_message msg = {
3360                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3361                 .rpc_argp       = clp,
3362                 .rpc_cred       = cred,
3363         };
3364         struct nfs4_renewdata *data;
3365
3366         if (!atomic_inc_not_zero(&clp->cl_count))
3367                 return -EIO;
3368         data = kmalloc(sizeof(*data), GFP_KERNEL);
3369         if (data == NULL)
3370                 return -ENOMEM;
3371         data->client = clp;
3372         data->timestamp = jiffies;
3373         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT,
3374                         &nfs4_renew_ops, data);
3375 }
3376
3377 int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
3378 {
3379         struct rpc_message msg = {
3380                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
3381                 .rpc_argp       = clp,
3382                 .rpc_cred       = cred,
3383         };
3384         unsigned long now = jiffies;
3385         int status;
3386
3387         status = rpc_call_sync(clp->cl_rpcclient, &msg, 0);
3388         if (status < 0)
3389                 return status;
3390         do_renew_lease(clp, now);
3391         return 0;
3392 }
3393
3394 static inline int nfs4_server_supports_acls(struct nfs_server *server)
3395 {
3396         return (server->caps & NFS_CAP_ACLS)
3397                 && (server->acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3398                 && (server->acl_bitmask & ACL4_SUPPORT_DENY_ACL);
3399 }
3400
3401 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_CACHE_SIZE, and that
3402  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_CACHE_SIZE) bytes on
3403  * the stack.
3404  */
3405 #define NFS4ACL_MAXPAGES (XATTR_SIZE_MAX >> PAGE_CACHE_SHIFT)
3406
3407 static void buf_to_pages(const void *buf, size_t buflen,
3408                 struct page **pages, unsigned int *pgbase)
3409 {
3410         const void *p = buf;
3411
3412         *pgbase = offset_in_page(buf);
3413         p -= *pgbase;
3414         while (p < buf + buflen) {
3415                 *(pages++) = virt_to_page(p);
3416                 p += PAGE_CACHE_SIZE;
3417         }
3418 }
3419
3420 static int buf_to_pages_noslab(const void *buf, size_t buflen,
3421                 struct page **pages, unsigned int *pgbase)
3422 {
3423         struct page *newpage, **spages;
3424         int rc = 0;
3425         size_t len;
3426         spages = pages;
3427
3428         do {
3429                 len = min_t(size_t, PAGE_CACHE_SIZE, buflen);
3430                 newpage = alloc_page(GFP_KERNEL);
3431
3432                 if (newpage == NULL)
3433                         goto unwind;
3434                 memcpy(page_address(newpage), buf, len);
3435                 buf += len;
3436                 buflen -= len;
3437                 *pages++ = newpage;
3438                 rc++;
3439         } while (buflen != 0);
3440
3441         return rc;
3442
3443 unwind:
3444         for(; rc > 0; rc--)
3445                 __free_page(spages[rc-1]);
3446         return -ENOMEM;
3447 }
3448
3449 struct nfs4_cached_acl {
3450         int cached;
3451         size_t len;
3452         char data[0];
3453 };
3454
3455 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
3456 {
3457         struct nfs_inode *nfsi = NFS_I(inode);
3458
3459         spin_lock(&inode->i_lock);
3460         kfree(nfsi->nfs4_acl);
3461         nfsi->nfs4_acl = acl;
3462         spin_unlock(&inode->i_lock);
3463 }
3464
3465 static void nfs4_zap_acl_attr(struct inode *inode)
3466 {
3467         nfs4_set_cached_acl(inode, NULL);
3468 }
3469
3470 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
3471 {
3472         struct nfs_inode *nfsi = NFS_I(inode);
3473         struct nfs4_cached_acl *acl;
3474         int ret = -ENOENT;
3475
3476         spin_lock(&inode->i_lock);
3477         acl = nfsi->nfs4_acl;
3478         if (acl == NULL)
3479                 goto out;
3480         if (buf == NULL) /* user is just asking for length */
3481                 goto out_len;
3482         if (acl->cached == 0)
3483                 goto out;
3484         ret = -ERANGE; /* see getxattr(2) man page */
3485         if (acl->len > buflen)
3486                 goto out;
3487         memcpy(buf, acl->data, acl->len);
3488 out_len:
3489         ret = acl->len;
3490 out:
3491         spin_unlock(&inode->i_lock);
3492         return ret;
3493 }
3494
3495 static void nfs4_write_cached_acl(struct inode *inode, const char *buf, size_t acl_len)
3496 {
3497         struct nfs4_cached_acl *acl;
3498
3499         if (buf && acl_len <= PAGE_SIZE) {
3500                 acl = kmalloc(sizeof(*acl) + acl_len, GFP_KERNEL);
3501                 if (acl == NULL)
3502                         goto out;
3503                 acl->cached = 1;
3504                 memcpy(acl->data, buf, acl_len);
3505         } else {
3506                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
3507                 if (acl == NULL)
3508                         goto out;
3509                 acl->cached = 0;
3510         }
3511         acl->len = acl_len;
3512 out:
3513         nfs4_set_cached_acl(inode, acl);
3514 }
3515
3516 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3517 {
3518         struct page *pages[NFS4ACL_MAXPAGES];
3519         struct nfs_getaclargs args = {
3520                 .fh = NFS_FH(inode),
3521                 .acl_pages = pages,
3522                 .acl_len = buflen,
3523         };
3524         struct nfs_getaclres res = {
3525                 .acl_len = buflen,
3526         };
3527         void *resp_buf;
3528         struct rpc_message msg = {
3529                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
3530                 .rpc_argp = &args,
3531                 .rpc_resp = &res,
3532         };
3533         struct page *localpage = NULL;
3534         int ret;
3535
3536         if (buflen < PAGE_SIZE) {
3537                 /* As long as we're doing a round trip to the server anyway,
3538                  * let's be prepared for a page of acl data. */
3539                 localpage = alloc_page(GFP_KERNEL);
3540                 resp_buf = page_address(localpage);
3541                 if (localpage == NULL)
3542                         return -ENOMEM;
3543                 args.acl_pages[0] = localpage;
3544                 args.acl_pgbase = 0;
3545                 args.acl_len = PAGE_SIZE;
3546         } else {
3547                 resp_buf = buf;
3548                 buf_to_pages(buf, buflen, args.acl_pages, &args.acl_pgbase);
3549         }
3550         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3551         if (ret)
3552                 goto out_free;
3553         if (res.acl_len > args.acl_len)
3554                 nfs4_write_cached_acl(inode, NULL, res.acl_len);
3555         else
3556                 nfs4_write_cached_acl(inode, resp_buf, res.acl_len);
3557         if (buf) {
3558                 ret = -ERANGE;
3559                 if (res.acl_len > buflen)
3560                         goto out_free;
3561                 if (localpage)
3562                         memcpy(buf, resp_buf, res.acl_len);
3563         }
3564         ret = res.acl_len;
3565 out_free:
3566         if (localpage)
3567                 __free_page(localpage);
3568         return ret;
3569 }
3570
3571 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
3572 {
3573         struct nfs4_exception exception = { };
3574         ssize_t ret;
3575         do {
3576                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
3577                 if (ret >= 0)
3578                         break;
3579                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
3580         } while (exception.retry);
3581         return ret;
3582 }
3583
3584 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
3585 {
3586         struct nfs_server *server = NFS_SERVER(inode);
3587         int ret;
3588
3589         if (!nfs4_server_supports_acls(server))
3590                 return -EOPNOTSUPP;
3591         ret = nfs_revalidate_inode(server, inode);
3592         if (ret < 0)
3593                 return ret;
3594         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
3595                 nfs_zap_acl_cache(inode);
3596         ret = nfs4_read_cached_acl(inode, buf, buflen);
3597         if (ret != -ENOENT)
3598                 return ret;
3599         return nfs4_get_acl_uncached(inode, buf, buflen);
3600 }
3601
3602 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3603 {
3604         struct nfs_server *server = NFS_SERVER(inode);
3605         struct page *pages[NFS4ACL_MAXPAGES];
3606         struct nfs_setaclargs arg = {
3607                 .fh             = NFS_FH(inode),
3608                 .acl_pages      = pages,
3609                 .acl_len        = buflen,
3610         };
3611         struct nfs_setaclres res;
3612         struct rpc_message msg = {
3613                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
3614                 .rpc_argp       = &arg,
3615                 .rpc_resp       = &res,
3616         };
3617         int ret, i;
3618
3619         if (!nfs4_server_supports_acls(server))
3620                 return -EOPNOTSUPP;
3621         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
3622         if (i < 0)
3623                 return i;
3624         nfs_inode_return_delegation(inode);
3625         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3626
3627         /*
3628          * Free each page after tx, so the only ref left is
3629          * held by the network stack
3630          */
3631         for (; i > 0; i--)
3632                 put_page(pages[i-1]);
3633
3634         /*
3635          * Acl update can result in inode attribute update.
3636          * so mark the attribute cache invalid.
3637          */
3638         spin_lock(&inode->i_lock);
3639         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
3640         spin_unlock(&inode->i_lock);
3641         nfs_access_zap_cache(inode);
3642         nfs_zap_acl_cache(inode);
3643         return ret;
3644 }
3645
3646 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
3647 {
3648         struct nfs4_exception exception = { };
3649         int err;
3650         do {
3651                 err = nfs4_handle_exception(NFS_SERVER(inode),
3652                                 __nfs4_proc_set_acl(inode, buf, buflen),
3653                                 &exception);
3654         } while (exception.retry);
3655         return err;
3656 }
3657
3658 static int
3659 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state)
3660 {
3661         struct nfs_client *clp = server->nfs_client;
3662
3663         if (task->tk_status >= 0)
3664                 return 0;
3665         switch(task->tk_status) {
3666                 case -NFS4ERR_ADMIN_REVOKED:
3667                 case -NFS4ERR_BAD_STATEID:
3668                 case -NFS4ERR_OPENMODE:
3669                         if (state == NULL)
3670                                 break;
3671                         nfs4_schedule_stateid_recovery(server, state);
3672                         goto wait_on_recovery;
3673                 case -NFS4ERR_STALE_STATEID:
3674                 case -NFS4ERR_STALE_CLIENTID:
3675                 case -NFS4ERR_EXPIRED:
3676                         nfs4_schedule_lease_recovery(clp);
3677                         goto wait_on_recovery;
3678 #if defined(CONFIG_NFS_V4_1)
3679                 case -NFS4ERR_BADSESSION:
3680                 case -NFS4ERR_BADSLOT:
3681                 case -NFS4ERR_BAD_HIGH_SLOT:
3682                 case -NFS4ERR_DEADSESSION:
3683                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
3684                 case -NFS4ERR_SEQ_FALSE_RETRY:
3685                 case -NFS4ERR_SEQ_MISORDERED:
3686                         dprintk("%s ERROR %d, Reset session\n", __func__,
3687                                 task->tk_status);
3688                         nfs4_schedule_session_recovery(clp->cl_session);
3689                         task->tk_status = 0;
3690                         return -EAGAIN;
3691 #endif /* CONFIG_NFS_V4_1 */
3692                 case -NFS4ERR_DELAY:
3693                         nfs_inc_server_stats(server, NFSIOS_DELAY);
3694                 case -NFS4ERR_GRACE:
3695                 case -EKEYEXPIRED:
3696                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
3697                         task->tk_status = 0;
3698                         return -EAGAIN;
3699                 case -NFS4ERR_RETRY_UNCACHED_REP:
3700                 case -NFS4ERR_OLD_STATEID:
3701                         task->tk_status = 0;
3702                         return -EAGAIN;
3703         }
3704         task->tk_status = nfs4_map_errors(task->tk_status);
3705         return 0;
3706 wait_on_recovery:
3707         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
3708         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
3709                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
3710         task->tk_status = 0;
3711         return -EAGAIN;
3712 }
3713
3714 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
3715                 unsigned short port, struct rpc_cred *cred,
3716                 struct nfs4_setclientid_res *res)
3717 {
3718         nfs4_verifier sc_verifier;
3719         struct nfs4_setclientid setclientid = {
3720                 .sc_verifier = &sc_verifier,
3721                 .sc_prog = program,
3722                 .sc_cb_ident = clp->cl_cb_ident,
3723         };
3724         struct rpc_message msg = {
3725                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
3726                 .rpc_argp = &setclientid,
3727                 .rpc_resp = res,
3728                 .rpc_cred = cred,
3729         };
3730         __be32 *p;
3731         int loop = 0;
3732         int status;
3733
3734         p = (__be32*)sc_verifier.data;
3735         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
3736         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
3737
3738         for(;;) {
3739                 setclientid.sc_name_len = scnprintf(setclientid.sc_name,
3740                                 sizeof(setclientid.sc_name), "%s/%s %s %s %u",
3741                                 clp->cl_ipaddr,
3742                                 rpc_peeraddr2str(clp->cl_rpcclient,
3743                                                         RPC_DISPLAY_ADDR),
3744                                 rpc_peeraddr2str(clp->cl_rpcclient,
3745                                                         RPC_DISPLAY_PROTO),
3746                                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
3747                                 clp->cl_id_uniquifier);
3748                 setclientid.sc_netid_len = scnprintf(setclientid.sc_netid,
3749                                 sizeof(setclientid.sc_netid),
3750                                 rpc_peeraddr2str(clp->cl_rpcclient,
3751                                                         RPC_DISPLAY_NETID));
3752                 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
3753                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
3754                                 clp->cl_ipaddr, port >> 8, port & 255);
3755
3756                 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3757                 if (status != -NFS4ERR_CLID_INUSE)
3758                         break;
3759                 if (loop != 0) {
3760                         ++clp->cl_id_uniquifier;
3761                         break;
3762                 }
3763                 ++loop;
3764                 ssleep(clp->cl_lease_time / HZ + 1);
3765         }
3766         return status;
3767 }
3768
3769 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
3770                 struct nfs4_setclientid_res *arg,
3771                 struct rpc_cred *cred)
3772 {
3773         struct nfs_fsinfo fsinfo;
3774         struct rpc_message msg = {
3775                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
3776                 .rpc_argp = arg,
3777                 .rpc_resp = &fsinfo,
3778                 .rpc_cred = cred,
3779         };
3780         unsigned long now;
3781         int status;
3782
3783         now = jiffies;
3784         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
3785         if (status == 0) {
3786                 spin_lock(&clp->cl_lock);
3787                 clp->cl_lease_time = fsinfo.lease_time * HZ;
3788                 clp->cl_last_renewal = now;
3789                 spin_unlock(&clp->cl_lock);
3790         }
3791         return status;
3792 }
3793
3794 struct nfs4_delegreturndata {
3795         struct nfs4_delegreturnargs args;
3796         struct nfs4_delegreturnres res;
3797         struct nfs_fh fh;
3798         nfs4_stateid stateid;
3799         unsigned long timestamp;
3800         struct nfs_fattr fattr;
3801         int rpc_status;
3802 };
3803
3804 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
3805 {
3806         struct nfs4_delegreturndata *data = calldata;
3807
3808         if (!nfs4_sequence_done(task, &data->res.seq_res))
3809                 return;
3810
3811         switch (task->tk_status) {
3812         case -NFS4ERR_STALE_STATEID:
3813         case -NFS4ERR_EXPIRED:
3814         case 0:
3815                 renew_lease(data->res.server, data->timestamp);
3816                 break;
3817         default:
3818                 if (nfs4_async_handle_error(task, data->res.server, NULL) ==
3819                                 -EAGAIN) {
3820                         nfs_restart_rpc(task, data->res.server->nfs_client);
3821                         return;
3822                 }
3823         }
3824         data->rpc_status = task->tk_status;
3825 }
3826
3827 static void nfs4_delegreturn_release(void *calldata)
3828 {
3829         kfree(calldata);
3830 }
3831
3832 #if defined(CONFIG_NFS_V4_1)
3833 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
3834 {
3835         struct nfs4_delegreturndata *d_data;
3836
3837         d_data = (struct nfs4_delegreturndata *)data;
3838
3839         if (nfs4_setup_sequence(d_data->res.server,
3840                                 &d_data->args.seq_args,
3841                                 &d_data->res.seq_res, 1, task))
3842                 return;
3843         rpc_call_start(task);
3844 }
3845 #endif /* CONFIG_NFS_V4_1 */
3846
3847 static const struct rpc_call_ops nfs4_delegreturn_ops = {
3848 #if defined(CONFIG_NFS_V4_1)
3849         .rpc_call_prepare = nfs4_delegreturn_prepare,
3850 #endif /* CONFIG_NFS_V4_1 */
3851         .rpc_call_done = nfs4_delegreturn_done,
3852         .rpc_release = nfs4_delegreturn_release,
3853 };
3854
3855 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3856 {
3857         struct nfs4_delegreturndata *data;
3858         struct nfs_server *server = NFS_SERVER(inode);
3859         struct rpc_task *task;
3860         struct rpc_message msg = {
3861                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
3862                 .rpc_cred = cred,
3863         };
3864         struct rpc_task_setup task_setup_data = {
3865                 .rpc_client = server->client,
3866                 .rpc_message = &msg,
3867                 .callback_ops = &nfs4_delegreturn_ops,
3868                 .flags = RPC_TASK_ASYNC,
3869         };
3870         int status = 0;
3871
3872         data = kzalloc(sizeof(*data), GFP_NOFS);
3873         if (data == NULL)
3874                 return -ENOMEM;
3875         data->args.fhandle = &data->fh;
3876         data->args.stateid = &data->stateid;
3877         data->args.bitmask = server->attr_bitmask;
3878         nfs_copy_fh(&data->fh, NFS_FH(inode));
3879         memcpy(&data->stateid, stateid, sizeof(data->stateid));
3880         data->res.fattr = &data->fattr;
3881         data->res.server = server;
3882         nfs_fattr_init(data->res.fattr);
3883         data->timestamp = jiffies;
3884         data->rpc_status = 0;
3885
3886         task_setup_data.callback_data = data;
3887         msg.rpc_argp = &data->args;
3888         msg.rpc_resp = &data->res;
3889         task = rpc_run_task(&task_setup_data);
3890         if (IS_ERR(task))
3891                 return PTR_ERR(task);
3892         if (!issync)
3893                 goto out;
3894         status = nfs4_wait_for_completion_rpc_task(task);
3895         if (status != 0)
3896                 goto out;
3897         status = data->rpc_status;
3898         if (status != 0)
3899                 goto out;
3900         nfs_refresh_inode(inode, &data->fattr);
3901 out:
3902         rpc_put_task(task);
3903         return status;
3904 }
3905
3906 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
3907 {
3908         struct nfs_server *server = NFS_SERVER(inode);
3909         struct nfs4_exception exception = { };
3910         int err;
3911         do {
3912                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
3913                 switch (err) {
3914                         case -NFS4ERR_STALE_STATEID:
3915                         case -NFS4ERR_EXPIRED:
3916                         case 0:
3917                                 return 0;
3918                 }
3919                 err = nfs4_handle_exception(server, err, &exception);
3920         } while (exception.retry);
3921         return err;
3922 }
3923
3924 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
3925 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
3926
3927 /*
3928  * sleep, with exponential backoff, and retry the LOCK operation.
3929  */
3930 static unsigned long
3931 nfs4_set_lock_task_retry(unsigned long timeout)
3932 {
3933         schedule_timeout_killable(timeout);
3934         timeout <<= 1;
3935         if (timeout > NFS4_LOCK_MAXTIMEOUT)
3936                 return NFS4_LOCK_MAXTIMEOUT;
3937         return timeout;
3938 }
3939
3940 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3941 {
3942         struct inode *inode = state->inode;
3943         struct nfs_server *server = NFS_SERVER(inode);
3944         struct nfs_client *clp = server->nfs_client;
3945         struct nfs_lockt_args arg = {
3946                 .fh = NFS_FH(inode),
3947                 .fl = request,
3948         };
3949         struct nfs_lockt_res res = {
3950                 .denied = request,
3951         };
3952         struct rpc_message msg = {
3953                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
3954                 .rpc_argp       = &arg,
3955                 .rpc_resp       = &res,
3956                 .rpc_cred       = state->owner->so_cred,
3957         };
3958         struct nfs4_lock_state *lsp;
3959         int status;
3960
3961         arg.lock_owner.clientid = clp->cl_clientid;
3962         status = nfs4_set_lock_state(state, request);
3963         if (status != 0)
3964                 goto out;
3965         lsp = request->fl_u.nfs4_fl.owner;
3966         arg.lock_owner.id = lsp->ls_id.id;
3967         arg.lock_owner.s_dev = server->s_dev;
3968         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3969         switch (status) {
3970                 case 0:
3971                         request->fl_type = F_UNLCK;
3972                         break;
3973                 case -NFS4ERR_DENIED:
3974                         status = 0;
3975         }
3976         request->fl_ops->fl_release_private(request);
3977 out:
3978         return status;
3979 }
3980
3981 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
3982 {
3983         struct nfs4_exception exception = { };
3984         int err;
3985
3986         do {
3987                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
3988                                 _nfs4_proc_getlk(state, cmd, request),
3989                                 &exception);
3990         } while (exception.retry);
3991         return err;
3992 }
3993
3994 static int do_vfs_lock(struct file *file, struct file_lock *fl)
3995 {
3996         int res = 0;
3997         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
3998                 case FL_POSIX:
3999                         res = posix_lock_file_wait(file, fl);
4000                         break;
4001                 case FL_FLOCK:
4002                         res = flock_lock_file_wait(file, fl);
4003                         break;
4004                 default:
4005                         BUG();
4006         }
4007         return res;
4008 }
4009
4010 struct nfs4_unlockdata {
4011         struct nfs_locku_args arg;
4012         struct nfs_locku_res res;
4013         struct nfs4_lock_state *lsp;
4014         struct nfs_open_context *ctx;
4015         struct file_lock fl;
4016         const struct nfs_server *server;
4017         unsigned long timestamp;
4018 };
4019
4020 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
4021                 struct nfs_open_context *ctx,
4022                 struct nfs4_lock_state *lsp,
4023                 struct nfs_seqid *seqid)
4024 {
4025         struct nfs4_unlockdata *p;
4026         struct inode *inode = lsp->ls_state->inode;
4027
4028         p = kzalloc(sizeof(*p), GFP_NOFS);
4029         if (p == NULL)
4030                 return NULL;
4031         p->arg.fh = NFS_FH(inode);
4032         p->arg.fl = &p->fl;
4033         p->arg.seqid = seqid;
4034         p->res.seqid = seqid;
4035         p->arg.stateid = &lsp->ls_stateid;
4036         p->lsp = lsp;
4037         atomic_inc(&lsp->ls_count);
4038         /* Ensure we don't close file until we're done freeing locks! */
4039         p->ctx = get_nfs_open_context(ctx);
4040         memcpy(&p->fl, fl, sizeof(p->fl));
4041         p->server = NFS_SERVER(inode);
4042         return p;
4043 }
4044
4045 static void nfs4_locku_release_calldata(void *data)
4046 {
4047         struct nfs4_unlockdata *calldata = data;
4048         nfs_free_seqid(calldata->arg.seqid);
4049         nfs4_put_lock_state(calldata->lsp);
4050         put_nfs_open_context(calldata->ctx);
4051         kfree(calldata);
4052 }
4053
4054 static void nfs4_locku_done(struct rpc_task *task, void *data)
4055 {
4056         struct nfs4_unlockdata *calldata = data;
4057
4058         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
4059                 return;
4060         switch (task->tk_status) {
4061                 case 0:
4062                         memcpy(calldata->lsp->ls_stateid.data,
4063                                         calldata->res.stateid.data,
4064                                         sizeof(calldata->lsp->ls_stateid.data));
4065                         renew_lease(calldata->server, calldata->timestamp);
4066                         break;
4067                 case -NFS4ERR_BAD_STATEID:
4068                 case -NFS4ERR_OLD_STATEID:
4069                 case -NFS4ERR_STALE_STATEID:
4070                 case -NFS4ERR_EXPIRED:
4071                         break;
4072                 default:
4073                         if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN)
4074                                 nfs_restart_rpc(task,
4075                                                  calldata->server->nfs_client);
4076         }
4077 }
4078
4079 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
4080 {
4081         struct nfs4_unlockdata *calldata = data;
4082
4083         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
4084                 return;
4085         if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) {
4086                 /* Note: exit _without_ running nfs4_locku_done */
4087                 task->tk_action = NULL;
4088                 return;
4089         }
4090         calldata->timestamp = jiffies;
4091         if (nfs4_setup_sequence(calldata->server,
4092                                 &calldata->arg.seq_args,
4093                                 &calldata->res.seq_res, 1, task))
4094                 return;
4095         rpc_call_start(task);
4096 }
4097
4098 static const struct rpc_call_ops nfs4_locku_ops = {
4099         .rpc_call_prepare = nfs4_locku_prepare,
4100         .rpc_call_done = nfs4_locku_done,
4101         .rpc_release = nfs4_locku_release_calldata,
4102 };
4103
4104 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
4105                 struct nfs_open_context *ctx,
4106                 struct nfs4_lock_state *lsp,
4107                 struct nfs_seqid *seqid)
4108 {
4109         struct nfs4_unlockdata *data;
4110         struct rpc_message msg = {
4111                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
4112                 .rpc_cred = ctx->cred,
4113         };
4114         struct rpc_task_setup task_setup_data = {
4115                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
4116                 .rpc_message = &msg,
4117                 .callback_ops = &nfs4_locku_ops,
4118                 .workqueue = nfsiod_workqueue,
4119                 .flags = RPC_TASK_ASYNC,
4120         };
4121
4122         /* Ensure this is an unlock - when canceling a lock, the
4123          * canceled lock is passed in, and it won't be an unlock.
4124          */
4125         fl->fl_type = F_UNLCK;
4126
4127         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
4128         if (data == NULL) {
4129                 nfs_free_seqid(seqid);
4130                 return ERR_PTR(-ENOMEM);
4131         }
4132
4133         msg.rpc_argp = &data->arg;
4134         msg.rpc_resp = &data->res;
4135         task_setup_data.callback_data = data;
4136         return rpc_run_task(&task_setup_data);
4137 }
4138
4139 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
4140 {
4141         struct nfs_inode *nfsi = NFS_I(state->inode);
4142         struct nfs_seqid *seqid;
4143         struct nfs4_lock_state *lsp;
4144         struct rpc_task *task;
4145         int status = 0;
4146         unsigned char fl_flags = request->fl_flags;
4147
4148         status = nfs4_set_lock_state(state, request);
4149         /* Unlock _before_ we do the RPC call */
4150         request->fl_flags |= FL_EXISTS;
4151         down_read(&nfsi->rwsem);
4152         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
4153                 up_read(&nfsi->rwsem);
4154                 goto out;
4155         }
4156         up_read(&nfsi->rwsem);
4157         if (status != 0)
4158                 goto out;
4159         /* Is this a delegated lock? */
4160         if (test_bit(NFS_DELEGATED_STATE, &state->flags))
4161                 goto out;
4162         lsp = request->fl_u.nfs4_fl.owner;
4163         seqid = nfs_alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
4164         status = -ENOMEM;
4165         if (seqid == NULL)
4166                 goto out;
4167         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
4168         status = PTR_ERR(task);
4169         if (IS_ERR(task))
4170                 goto out;
4171         status = nfs4_wait_for_completion_rpc_task(task);
4172         rpc_put_task(task);
4173 out:
4174         request->fl_flags = fl_flags;
4175         return status;
4176 }
4177
4178 struct nfs4_lockdata {
4179         struct nfs_lock_args arg;
4180         struct nfs_lock_res res;
4181         struct nfs4_lock_state *lsp;
4182         struct nfs_open_context *ctx;
4183         struct file_lock fl;
4184         unsigned long timestamp;
4185         int rpc_status;
4186         int cancelled;
4187         struct nfs_server *server;
4188 };
4189
4190 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
4191                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
4192                 gfp_t gfp_mask)
4193 {
4194         struct nfs4_lockdata *p;
4195         struct inode *inode = lsp->ls_state->inode;
4196         struct nfs_server *server = NFS_SERVER(inode);
4197
4198         p = kzalloc(sizeof(*p), gfp_mask);
4199         if (p == NULL)
4200                 return NULL;
4201
4202         p->arg.fh = NFS_FH(inode);
4203         p->arg.fl = &p->fl;
4204         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
4205         if (p->arg.open_seqid == NULL)
4206                 goto out_free;
4207         p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid, gfp_mask);
4208         if (p->arg.lock_seqid == NULL)
4209                 goto out_free_seqid;
4210         p->arg.lock_stateid = &lsp->ls_stateid;
4211         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
4212         p->arg.lock_owner.id = lsp->ls_id.id;
4213         p->arg.lock_owner.s_dev = server->s_dev;
4214         p->res.lock_seqid = p->arg.lock_seqid;
4215         p->lsp = lsp;
4216         p->server = server;
4217         atomic_inc(&lsp->ls_count);
4218         p->ctx = get_nfs_open_context(ctx);
4219         memcpy(&p->fl, fl, sizeof(p->fl));
4220         return p;
4221 out_free_seqid:
4222         nfs_free_seqid(p->arg.open_seqid);
4223 out_free:
4224         kfree(p);
4225         return NULL;
4226 }
4227
4228 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
4229 {
4230         struct nfs4_lockdata *data = calldata;
4231         struct nfs4_state *state = data->lsp->ls_state;
4232
4233         dprintk("%s: begin!\n", __func__);
4234         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
4235                 return;
4236         /* Do we need to do an open_to_lock_owner? */
4237         if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
4238                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0)
4239                         return;
4240                 data->arg.open_stateid = &state->stateid;
4241                 data->arg.new_lock_owner = 1;
4242                 data->res.open_seqid = data->arg.open_seqid;
4243         } else
4244                 data->arg.new_lock_owner = 0;
4245         data->timestamp = jiffies;
4246         if (nfs4_setup_sequence(data->server,
4247                                 &data->arg.seq_args,
4248                                 &data->res.seq_res, 1, task))
4249                 return;
4250         rpc_call_start(task);
4251         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
4252 }
4253
4254 static void nfs4_recover_lock_prepare(struct rpc_task *task, void *calldata)
4255 {
4256         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4257         nfs4_lock_prepare(task, calldata);
4258 }
4259
4260 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
4261 {
4262         struct nfs4_lockdata *data = calldata;
4263
4264         dprintk("%s: begin!\n", __func__);
4265
4266         if (!nfs4_sequence_done(task, &data->res.seq_res))
4267                 return;
4268
4269         data->rpc_status = task->tk_status;
4270         if (data->arg.new_lock_owner != 0) {
4271                 if (data->rpc_status == 0)
4272                         nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
4273                 else
4274                         goto out;
4275         }
4276         if (data->rpc_status == 0) {
4277                 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
4278                                         sizeof(data->lsp->ls_stateid.data));
4279                 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
4280                 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
4281         }
4282 out:
4283         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
4284 }
4285
4286 static void nfs4_lock_release(void *calldata)
4287 {
4288         struct nfs4_lockdata *data = calldata;
4289
4290         dprintk("%s: begin!\n", __func__);
4291         nfs_free_seqid(data->arg.open_seqid);
4292         if (data->cancelled != 0) {
4293                 struct rpc_task *task;
4294                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
4295                                 data->arg.lock_seqid);
4296                 if (!IS_ERR(task))
4297                         rpc_put_task_async(task);
4298                 dprintk("%s: cancelling lock!\n", __func__);
4299         } else
4300                 nfs_free_seqid(data->arg.lock_seqid);
4301         nfs4_put_lock_state(data->lsp);
4302         put_nfs_open_context(data->ctx);
4303         kfree(data);
4304         dprintk("%s: done!\n", __func__);
4305 }
4306
4307 static const struct rpc_call_ops nfs4_lock_ops = {
4308         .rpc_call_prepare = nfs4_lock_prepare,
4309         .rpc_call_done = nfs4_lock_done,
4310         .rpc_release = nfs4_lock_release,
4311 };
4312
4313 static const struct rpc_call_ops nfs4_recover_lock_ops = {
4314         .rpc_call_prepare = nfs4_recover_lock_prepare,
4315         .rpc_call_done = nfs4_lock_done,
4316         .rpc_release = nfs4_lock_release,
4317 };
4318
4319 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
4320 {
4321         switch (error) {
4322         case -NFS4ERR_ADMIN_REVOKED:
4323         case -NFS4ERR_BAD_STATEID:
4324                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4325                 if (new_lock_owner != 0 ||
4326                    (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
4327                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
4328                 break;
4329         case -NFS4ERR_STALE_STATEID:
4330                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
4331         case -NFS4ERR_EXPIRED:
4332                 nfs4_schedule_lease_recovery(server->nfs_client);
4333         };
4334 }
4335
4336 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
4337 {
4338         struct nfs4_lockdata *data;
4339         struct rpc_task *task;
4340         struct rpc_message msg = {
4341                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
4342                 .rpc_cred = state->owner->so_cred,
4343         };
4344         struct rpc_task_setup task_setup_data = {
4345                 .rpc_client = NFS_CLIENT(state->inode),
4346                 .rpc_message = &msg,
4347                 .callback_ops = &nfs4_lock_ops,
4348                 .workqueue = nfsiod_workqueue,
4349                 .flags = RPC_TASK_ASYNC,
4350         };
4351         int ret;
4352
4353         dprintk("%s: begin!\n", __func__);
4354         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
4355                         fl->fl_u.nfs4_fl.owner,
4356                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
4357         if (data == NULL)
4358                 return -ENOMEM;
4359         if (IS_SETLKW(cmd))
4360                 data->arg.block = 1;
4361         if (recovery_type > NFS_LOCK_NEW) {
4362                 if (recovery_type == NFS_LOCK_RECLAIM)
4363                         data->arg.reclaim = NFS_LOCK_RECLAIM;
4364                 task_setup_data.callback_ops = &nfs4_recover_lock_ops;
4365         }
4366         msg.rpc_argp = &data->arg;
4367         msg.rpc_resp = &data->res;
4368         task_setup_data.callback_data = data;
4369         task = rpc_run_task(&task_setup_data);
4370         if (IS_ERR(task))
4371                 return PTR_ERR(task);
4372         ret = nfs4_wait_for_completion_rpc_task(task);
4373         if (ret == 0) {
4374                 ret = data->rpc_status;
4375                 if (ret)
4376                         nfs4_handle_setlk_error(data->server, data->lsp,
4377                                         data->arg.new_lock_owner, ret);
4378         } else
4379                 data->cancelled = 1;
4380         rpc_put_task(task);
4381         dprintk("%s: done, ret = %d!\n", __func__, ret);
4382         return ret;
4383 }
4384
4385 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
4386 {
4387         struct nfs_server *server = NFS_SERVER(state->inode);
4388         struct nfs4_exception exception = { };
4389         int err;
4390
4391         do {
4392                 /* Cache the lock if possible... */
4393                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4394                         return 0;
4395                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
4396                 if (err != -NFS4ERR_DELAY)
4397                         break;
4398                 nfs4_handle_exception(server, err, &exception);
4399         } while (exception.retry);
4400         return err;
4401 }
4402
4403 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
4404 {
4405         struct nfs_server *server = NFS_SERVER(state->inode);
4406         struct nfs4_exception exception = { };
4407         int err;
4408
4409         err = nfs4_set_lock_state(state, request);
4410         if (err != 0)
4411                 return err;
4412         do {
4413                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
4414                         return 0;
4415                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
4416                 switch (err) {
4417                 default:
4418                         goto out;
4419                 case -NFS4ERR_GRACE:
4420                 case -NFS4ERR_DELAY:
4421                         nfs4_handle_exception(server, err, &exception);
4422                         err = 0;
4423                 }
4424         } while (exception.retry);
4425 out:
4426         return err;
4427 }
4428
4429 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4430 {
4431         struct nfs_inode *nfsi = NFS_I(state->inode);
4432         unsigned char fl_flags = request->fl_flags;
4433         int status = -ENOLCK;
4434
4435         if ((fl_flags & FL_POSIX) &&
4436                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
4437                 goto out;
4438         /* Is this a delegated open? */
4439         status = nfs4_set_lock_state(state, request);
4440         if (status != 0)
4441                 goto out;
4442         request->fl_flags |= FL_ACCESS;
4443         status = do_vfs_lock(request->fl_file, request);
4444         if (status < 0)
4445                 goto out;
4446         down_read(&nfsi->rwsem);
4447         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
4448                 /* Yes: cache locks! */
4449                 /* ...but avoid races with delegation recall... */
4450                 request->fl_flags = fl_flags & ~FL_SLEEP;
4451                 status = do_vfs_lock(request->fl_file, request);
4452                 goto out_unlock;
4453         }
4454         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
4455         if (status != 0)
4456                 goto out_unlock;
4457         /* Note: we always want to sleep here! */
4458         request->fl_flags = fl_flags | FL_SLEEP;
4459         if (do_vfs_lock(request->fl_file, request) < 0)
4460                 printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __func__);
4461 out_unlock:
4462         up_read(&nfsi->rwsem);
4463 out:
4464         request->fl_flags = fl_flags;
4465         return status;
4466 }
4467
4468 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
4469 {
4470         struct nfs4_exception exception = { };
4471         int err;
4472
4473         do {
4474                 err = _nfs4_proc_setlk(state, cmd, request);
4475                 if (err == -NFS4ERR_DENIED)
4476                         err = -EAGAIN;
4477                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
4478                                 err, &exception);
4479         } while (exception.retry);
4480         return err;
4481 }
4482
4483 static int
4484 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
4485 {
4486         struct nfs_open_context *ctx;
4487         struct nfs4_state *state;
4488         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
4489         int status;
4490
4491         /* verify open state */
4492         ctx = nfs_file_open_context(filp);
4493         state = ctx->state;
4494
4495         if (request->fl_start < 0 || request->fl_end < 0)
4496                 return -EINVAL;
4497
4498         if (IS_GETLK(cmd)) {
4499                 if (state != NULL)
4500                         return nfs4_proc_getlk(state, F_GETLK, request);
4501                 return 0;
4502         }
4503
4504         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
4505                 return -EINVAL;
4506
4507         if (request->fl_type == F_UNLCK) {
4508                 if (state != NULL)
4509                         return nfs4_proc_unlck(state, cmd, request);
4510                 return 0;
4511         }
4512
4513         if (state == NULL)
4514                 return -ENOLCK;
4515         do {
4516                 status = nfs4_proc_setlk(state, cmd, request);
4517                 if ((status != -EAGAIN) || IS_SETLK(cmd))
4518                         break;
4519                 timeout = nfs4_set_lock_task_retry(timeout);
4520                 status = -ERESTARTSYS;
4521                 if (signalled())
4522                         break;
4523         } while(status < 0);
4524         return status;
4525 }
4526
4527 int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl)
4528 {
4529         struct nfs_server *server = NFS_SERVER(state->inode);
4530         struct nfs4_exception exception = { };
4531         int err;
4532
4533         err = nfs4_set_lock_state(state, fl);
4534         if (err != 0)
4535                 goto out;
4536         do {
4537                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
4538                 switch (err) {
4539                         default:
4540                                 printk(KERN_ERR "%s: unhandled error %d.\n",
4541                                                 __func__, err);
4542                         case 0:
4543                         case -ESTALE:
4544                                 goto out;
4545                         case -NFS4ERR_EXPIRED:
4546                         case -NFS4ERR_STALE_CLIENTID:
4547                         case -NFS4ERR_STALE_STATEID:
4548                                 nfs4_schedule_lease_recovery(server->nfs_client);
4549                                 goto out;
4550                         case -NFS4ERR_BADSESSION:
4551                         case -NFS4ERR_BADSLOT:
4552                         case -NFS4ERR_BAD_HIGH_SLOT:
4553                         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4554                         case -NFS4ERR_DEADSESSION:
4555                                 nfs4_schedule_session_recovery(server->nfs_client->cl_session);
4556                                 goto out;
4557                         case -ERESTARTSYS:
4558                                 /*
4559                                  * The show must go on: exit, but mark the
4560                                  * stateid as needing recovery.
4561                                  */
4562                         case -NFS4ERR_ADMIN_REVOKED:
4563                         case -NFS4ERR_BAD_STATEID:
4564                         case -NFS4ERR_OPENMODE:
4565                                 nfs4_schedule_stateid_recovery(server, state);
4566                                 err = 0;
4567                                 goto out;
4568                         case -EKEYEXPIRED:
4569                                 /*
4570                                  * User RPCSEC_GSS context has expired.
4571                                  * We cannot recover this stateid now, so
4572                                  * skip it and allow recovery thread to
4573                                  * proceed.
4574                                  */
4575                                 err = 0;
4576                                 goto out;
4577                         case -ENOMEM:
4578                         case -NFS4ERR_DENIED:
4579                                 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
4580                                 err = 0;
4581                                 goto out;
4582                         case -NFS4ERR_DELAY:
4583                                 break;
4584                 }
4585                 err = nfs4_handle_exception(server, err, &exception);
4586         } while (exception.retry);
4587 out:
4588         return err;
4589 }
4590
4591 static void nfs4_release_lockowner_release(void *calldata)
4592 {
4593         kfree(calldata);
4594 }
4595
4596 const struct rpc_call_ops nfs4_release_lockowner_ops = {
4597         .rpc_release = nfs4_release_lockowner_release,
4598 };
4599
4600 void nfs4_release_lockowner(const struct nfs4_lock_state *lsp)
4601 {
4602         struct nfs_server *server = lsp->ls_state->owner->so_server;
4603         struct nfs_release_lockowner_args *args;
4604         struct rpc_message msg = {
4605                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
4606         };
4607
4608         if (server->nfs_client->cl_mvops->minor_version != 0)
4609                 return;
4610         args = kmalloc(sizeof(*args), GFP_NOFS);
4611         if (!args)
4612                 return;
4613         args->lock_owner.clientid = server->nfs_client->cl_clientid;
4614         args->lock_owner.id = lsp->ls_id.id;
4615         args->lock_owner.s_dev = server->s_dev;
4616         msg.rpc_argp = args;
4617         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args);
4618 }
4619
4620 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
4621
4622 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
4623                                    const void *buf, size_t buflen,
4624                                    int flags, int type)
4625 {
4626         if (strcmp(key, "") != 0)
4627                 return -EINVAL;
4628
4629         return nfs4_proc_set_acl(dentry->d_inode, buf, buflen);
4630 }
4631
4632 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
4633                                    void *buf, size_t buflen, int type)
4634 {
4635         if (strcmp(key, "") != 0)
4636                 return -EINVAL;
4637
4638         return nfs4_proc_get_acl(dentry->d_inode, buf, buflen);
4639 }
4640
4641 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
4642                                        size_t list_len, const char *name,
4643                                        size_t name_len, int type)
4644 {
4645         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
4646
4647         if (!nfs4_server_supports_acls(NFS_SERVER(dentry->d_inode)))
4648                 return 0;
4649
4650         if (list && len <= list_len)
4651                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
4652         return len;
4653 }
4654
4655 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
4656 {
4657         if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
4658                 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
4659                 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
4660                 return;
4661
4662         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4663                 NFS_ATTR_FATTR_NLINK;
4664         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4665         fattr->nlink = 2;
4666 }
4667
4668 int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
4669                 struct nfs4_fs_locations *fs_locations, struct page *page)
4670 {
4671         struct nfs_server *server = NFS_SERVER(dir);
4672         u32 bitmask[2] = {
4673                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
4674                 [1] = FATTR4_WORD1_MOUNTED_ON_FILEID,
4675         };
4676         struct nfs4_fs_locations_arg args = {
4677                 .dir_fh = NFS_FH(dir),
4678                 .name = name,
4679                 .page = page,
4680                 .bitmask = bitmask,
4681         };
4682         struct nfs4_fs_locations_res res = {
4683                 .fs_locations = fs_locations,
4684         };
4685         struct rpc_message msg = {
4686                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
4687                 .rpc_argp = &args,
4688                 .rpc_resp = &res,
4689         };
4690         int status;
4691
4692         dprintk("%s: start\n", __func__);
4693         nfs_fattr_init(&fs_locations->fattr);
4694         fs_locations->server = server;
4695         fs_locations->nlocations = 0;
4696         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4697         nfs_fixup_referral_attributes(&fs_locations->fattr);
4698         dprintk("%s: returned status = %d\n", __func__, status);
4699         return status;
4700 }
4701
4702 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4703 {
4704         int status;
4705         struct nfs4_secinfo_arg args = {
4706                 .dir_fh = NFS_FH(dir),
4707                 .name   = name,
4708         };
4709         struct nfs4_secinfo_res res = {
4710                 .flavors     = flavors,
4711         };
4712         struct rpc_message msg = {
4713                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
4714                 .rpc_argp = &args,
4715                 .rpc_resp = &res,
4716         };
4717
4718         dprintk("NFS call  secinfo %s\n", name->name);
4719         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4720         dprintk("NFS reply  secinfo: %d\n", status);
4721         return status;
4722 }
4723
4724 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors)
4725 {
4726         struct nfs4_exception exception = { };
4727         int err;
4728         do {
4729                 err = nfs4_handle_exception(NFS_SERVER(dir),
4730                                 _nfs4_proc_secinfo(dir, name, flavors),
4731                                 &exception);
4732         } while (exception.retry);
4733         return err;
4734 }
4735
4736 #ifdef CONFIG_NFS_V4_1
4737 /*
4738  * Check the exchange flags returned by the server for invalid flags, having
4739  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
4740  * DS flags set.
4741  */
4742 static int nfs4_check_cl_exchange_flags(u32 flags)
4743 {
4744         if (flags & ~EXCHGID4_FLAG_MASK_R)
4745                 goto out_inval;
4746         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
4747             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
4748                 goto out_inval;
4749         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
4750                 goto out_inval;
4751         return NFS_OK;
4752 out_inval:
4753         return -NFS4ERR_INVAL;
4754 }
4755
4756 /*
4757  * nfs4_proc_exchange_id()
4758  *
4759  * Since the clientid has expired, all compounds using sessions
4760  * associated with the stale clientid will be returning
4761  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
4762  * be in some phase of session reset.
4763  */
4764 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
4765 {
4766         nfs4_verifier verifier;
4767         struct nfs41_exchange_id_args args = {
4768                 .client = clp,
4769                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER,
4770         };
4771         struct nfs41_exchange_id_res res = {
4772                 .client = clp,
4773         };
4774         int status;
4775         struct rpc_message msg = {
4776                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
4777                 .rpc_argp = &args,
4778                 .rpc_resp = &res,
4779                 .rpc_cred = cred,
4780         };
4781         __be32 *p;
4782
4783         dprintk("--> %s\n", __func__);
4784         BUG_ON(clp == NULL);
4785
4786         p = (u32 *)verifier.data;
4787         *p++ = htonl((u32)clp->cl_boot_time.tv_sec);
4788         *p = htonl((u32)clp->cl_boot_time.tv_nsec);
4789         args.verifier = &verifier;
4790
4791         args.id_len = scnprintf(args.id, sizeof(args.id),
4792                                 "%s/%s.%s/%u",
4793                                 clp->cl_ipaddr,
4794                                 init_utsname()->nodename,
4795                                 init_utsname()->domainname,
4796                                 clp->cl_rpcclient->cl_auth->au_flavor);
4797
4798         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4799         if (!status)
4800                 status = nfs4_check_cl_exchange_flags(clp->cl_exchange_flags);
4801         dprintk("<-- %s status= %d\n", __func__, status);
4802         return status;
4803 }
4804
4805 struct nfs4_get_lease_time_data {
4806         struct nfs4_get_lease_time_args *args;
4807         struct nfs4_get_lease_time_res *res;
4808         struct nfs_client *clp;
4809 };
4810
4811 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
4812                                         void *calldata)
4813 {
4814         int ret;
4815         struct nfs4_get_lease_time_data *data =
4816                         (struct nfs4_get_lease_time_data *)calldata;
4817
4818         dprintk("--> %s\n", __func__);
4819         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
4820         /* just setup sequence, do not trigger session recovery
4821            since we're invoked within one */
4822         ret = nfs41_setup_sequence(data->clp->cl_session,
4823                                    &data->args->la_seq_args,
4824                                    &data->res->lr_seq_res, 0, task);
4825
4826         BUG_ON(ret == -EAGAIN);
4827         rpc_call_start(task);
4828         dprintk("<-- %s\n", __func__);
4829 }
4830
4831 /*
4832  * Called from nfs4_state_manager thread for session setup, so don't recover
4833  * from sequence operation or clientid errors.
4834  */
4835 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
4836 {
4837         struct nfs4_get_lease_time_data *data =
4838                         (struct nfs4_get_lease_time_data *)calldata;
4839
4840         dprintk("--> %s\n", __func__);
4841         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
4842                 return;
4843         switch (task->tk_status) {
4844         case -NFS4ERR_DELAY:
4845         case -NFS4ERR_GRACE:
4846                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
4847                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
4848                 task->tk_status = 0;
4849                 /* fall through */
4850         case -NFS4ERR_RETRY_UNCACHED_REP:
4851                 nfs_restart_rpc(task, data->clp);
4852                 return;
4853         }
4854         dprintk("<-- %s\n", __func__);
4855 }
4856
4857 struct rpc_call_ops nfs4_get_lease_time_ops = {
4858         .rpc_call_prepare = nfs4_get_lease_time_prepare,
4859         .rpc_call_done = nfs4_get_lease_time_done,
4860 };
4861
4862 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
4863 {
4864         struct rpc_task *task;
4865         struct nfs4_get_lease_time_args args;
4866         struct nfs4_get_lease_time_res res = {
4867                 .lr_fsinfo = fsinfo,
4868         };
4869         struct nfs4_get_lease_time_data data = {
4870                 .args = &args,
4871                 .res = &res,
4872                 .clp = clp,
4873         };
4874         struct rpc_message msg = {
4875                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
4876                 .rpc_argp = &args,
4877                 .rpc_resp = &res,
4878         };
4879         struct rpc_task_setup task_setup = {
4880                 .rpc_client = clp->cl_rpcclient,
4881                 .rpc_message = &msg,
4882                 .callback_ops = &nfs4_get_lease_time_ops,
4883                 .callback_data = &data,
4884                 .flags = RPC_TASK_TIMEOUT,
4885         };
4886         int status;
4887
4888         dprintk("--> %s\n", __func__);
4889         task = rpc_run_task(&task_setup);
4890
4891         if (IS_ERR(task))
4892                 status = PTR_ERR(task);
4893         else {
4894                 status = task->tk_status;
4895                 rpc_put_task(task);
4896         }
4897         dprintk("<-- %s return %d\n", __func__, status);
4898
4899         return status;
4900 }
4901
4902 /*
4903  * Reset a slot table
4904  */
4905 static int nfs4_reset_slot_table(struct nfs4_slot_table *tbl, u32 max_reqs,
4906                                  int ivalue)
4907 {
4908         struct nfs4_slot *new = NULL;
4909         int i;
4910         int ret = 0;
4911
4912         dprintk("--> %s: max_reqs=%u, tbl->max_slots %d\n", __func__,
4913                 max_reqs, tbl->max_slots);
4914
4915         /* Does the newly negotiated max_reqs match the existing slot table? */
4916         if (max_reqs != tbl->max_slots) {
4917                 ret = -ENOMEM;
4918                 new = kmalloc(max_reqs * sizeof(struct nfs4_slot),
4919                               GFP_NOFS);
4920                 if (!new)
4921                         goto out;
4922                 ret = 0;
4923                 kfree(tbl->slots);
4924         }
4925         spin_lock(&tbl->slot_tbl_lock);
4926         if (new) {
4927                 tbl->slots = new;
4928                 tbl->max_slots = max_reqs;
4929         }
4930         for (i = 0; i < tbl->max_slots; ++i)
4931                 tbl->slots[i].seq_nr = ivalue;
4932         spin_unlock(&tbl->slot_tbl_lock);
4933         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4934                 tbl, tbl->slots, tbl->max_slots);
4935 out:
4936         dprintk("<-- %s: return %d\n", __func__, ret);
4937         return ret;
4938 }
4939
4940 /*
4941  * Reset the forechannel and backchannel slot tables
4942  */
4943 static int nfs4_reset_slot_tables(struct nfs4_session *session)
4944 {
4945         int status;
4946
4947         status = nfs4_reset_slot_table(&session->fc_slot_table,
4948                         session->fc_attrs.max_reqs, 1);
4949         if (status)
4950                 return status;
4951
4952         status = nfs4_reset_slot_table(&session->bc_slot_table,
4953                         session->bc_attrs.max_reqs, 0);
4954         return status;
4955 }
4956
4957 /* Destroy the slot table */
4958 static void nfs4_destroy_slot_tables(struct nfs4_session *session)
4959 {
4960         if (session->fc_slot_table.slots != NULL) {
4961                 kfree(session->fc_slot_table.slots);
4962                 session->fc_slot_table.slots = NULL;
4963         }
4964         if (session->bc_slot_table.slots != NULL) {
4965                 kfree(session->bc_slot_table.slots);
4966                 session->bc_slot_table.slots = NULL;
4967         }
4968         return;
4969 }
4970
4971 /*
4972  * Initialize slot table
4973  */
4974 static int nfs4_init_slot_table(struct nfs4_slot_table *tbl,
4975                 int max_slots, int ivalue)
4976 {
4977         struct nfs4_slot *slot;
4978         int ret = -ENOMEM;
4979
4980         BUG_ON(max_slots > NFS4_MAX_SLOT_TABLE);
4981
4982         dprintk("--> %s: max_reqs=%u\n", __func__, max_slots);
4983
4984         slot = kcalloc(max_slots, sizeof(struct nfs4_slot), GFP_NOFS);
4985         if (!slot)
4986                 goto out;
4987         ret = 0;
4988
4989         spin_lock(&tbl->slot_tbl_lock);
4990         tbl->max_slots = max_slots;
4991         tbl->slots = slot;
4992         tbl->highest_used_slotid = -1;  /* no slot is currently used */
4993         spin_unlock(&tbl->slot_tbl_lock);
4994         dprintk("%s: tbl=%p slots=%p max_slots=%d\n", __func__,
4995                 tbl, tbl->slots, tbl->max_slots);
4996 out:
4997         dprintk("<-- %s: return %d\n", __func__, ret);
4998         return ret;
4999 }
5000
5001 /*
5002  * Initialize the forechannel and backchannel tables
5003  */
5004 static int nfs4_init_slot_tables(struct nfs4_session *session)
5005 {
5006         struct nfs4_slot_table *tbl;
5007         int status = 0;
5008
5009         tbl = &session->fc_slot_table;
5010         if (tbl->slots == NULL) {
5011                 status = nfs4_init_slot_table(tbl,
5012                                 session->fc_attrs.max_reqs, 1);
5013                 if (status)
5014                         return status;
5015         }
5016
5017         tbl = &session->bc_slot_table;
5018         if (tbl->slots == NULL) {
5019                 status = nfs4_init_slot_table(tbl,
5020                                 session->bc_attrs.max_reqs, 0);
5021                 if (status)
5022                         nfs4_destroy_slot_tables(session);
5023         }
5024
5025         return status;
5026 }
5027
5028 struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp)
5029 {
5030         struct nfs4_session *session;
5031         struct nfs4_slot_table *tbl;
5032
5033         session = kzalloc(sizeof(struct nfs4_session), GFP_NOFS);
5034         if (!session)
5035                 return NULL;
5036
5037         tbl = &session->fc_slot_table;
5038         tbl->highest_used_slotid = -1;
5039         spin_lock_init(&tbl->slot_tbl_lock);
5040         rpc_init_priority_wait_queue(&tbl->slot_tbl_waitq, "ForeChannel Slot table");
5041         init_completion(&tbl->complete);
5042
5043         tbl = &session->bc_slot_table;
5044         tbl->highest_used_slotid = -1;
5045         spin_lock_init(&tbl->slot_tbl_lock);
5046         rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table");
5047         init_completion(&tbl->complete);
5048
5049         session->session_state = 1<<NFS4_SESSION_INITING;
5050
5051         session->clp = clp;
5052         return session;
5053 }
5054
5055 void nfs4_destroy_session(struct nfs4_session *session)
5056 {
5057         nfs4_proc_destroy_session(session);
5058         dprintk("%s Destroy backchannel for xprt %p\n",
5059                 __func__, session->clp->cl_rpcclient->cl_xprt);
5060         xprt_destroy_backchannel(session->clp->cl_rpcclient->cl_xprt,
5061                                 NFS41_BC_MIN_CALLBACKS);
5062         nfs4_destroy_slot_tables(session);
5063         kfree(session);
5064 }
5065
5066 /*
5067  * Initialize the values to be used by the client in CREATE_SESSION
5068  * If nfs4_init_session set the fore channel request and response sizes,
5069  * use them.
5070  *
5071  * Set the back channel max_resp_sz_cached to zero to force the client to
5072  * always set csa_cachethis to FALSE because the current implementation
5073  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
5074  */
5075 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
5076 {
5077         struct nfs4_session *session = args->client->cl_session;
5078         unsigned int mxrqst_sz = session->fc_attrs.max_rqst_sz,
5079                      mxresp_sz = session->fc_attrs.max_resp_sz;
5080
5081         if (mxrqst_sz == 0)
5082                 mxrqst_sz = NFS_MAX_FILE_IO_SIZE;
5083         if (mxresp_sz == 0)
5084                 mxresp_sz = NFS_MAX_FILE_IO_SIZE;
5085         /* Fore channel attributes */
5086         args->fc_attrs.headerpadsz = 0;
5087         args->fc_attrs.max_rqst_sz = mxrqst_sz;
5088         args->fc_attrs.max_resp_sz = mxresp_sz;
5089         args->fc_attrs.max_ops = NFS4_MAX_OPS;
5090         args->fc_attrs.max_reqs = session->clp->cl_rpcclient->cl_xprt->max_reqs;
5091
5092         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
5093                 "max_ops=%u max_reqs=%u\n",
5094                 __func__,
5095                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
5096                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
5097
5098         /* Back channel attributes */
5099         args->bc_attrs.headerpadsz = 0;
5100         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
5101         args->bc_attrs.max_resp_sz = PAGE_SIZE;
5102         args->bc_attrs.max_resp_sz_cached = 0;
5103         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
5104         args->bc_attrs.max_reqs = 1;
5105
5106         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
5107                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
5108                 __func__,
5109                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
5110                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
5111                 args->bc_attrs.max_reqs);
5112 }
5113
5114 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5115 {
5116         struct nfs4_channel_attrs *sent = &args->fc_attrs;
5117         struct nfs4_channel_attrs *rcvd = &session->fc_attrs;
5118
5119         if (rcvd->headerpadsz > sent->headerpadsz)
5120                 return -EINVAL;
5121         if (rcvd->max_resp_sz > sent->max_resp_sz)
5122                 return -EINVAL;
5123         /*
5124          * Our requested max_ops is the minimum we need; we're not
5125          * prepared to break up compounds into smaller pieces than that.
5126          * So, no point even trying to continue if the server won't
5127          * cooperate:
5128          */
5129         if (rcvd->max_ops < sent->max_ops)
5130                 return -EINVAL;
5131         if (rcvd->max_reqs == 0)
5132                 return -EINVAL;
5133         return 0;
5134 }
5135
5136 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args, struct nfs4_session *session)
5137 {
5138         struct nfs4_channel_attrs *sent = &args->bc_attrs;
5139         struct nfs4_channel_attrs *rcvd = &session->bc_attrs;
5140
5141         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
5142                 return -EINVAL;
5143         if (rcvd->max_resp_sz < sent->max_resp_sz)
5144                 return -EINVAL;
5145         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
5146                 return -EINVAL;
5147         /* These would render the backchannel useless: */
5148         if (rcvd->max_ops  == 0)
5149                 return -EINVAL;
5150         if (rcvd->max_reqs == 0)
5151                 return -EINVAL;
5152         return 0;
5153 }
5154
5155 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
5156                                      struct nfs4_session *session)
5157 {
5158         int ret;
5159
5160         ret = nfs4_verify_fore_channel_attrs(args, session);
5161         if (ret)
5162                 return ret;
5163         return nfs4_verify_back_channel_attrs(args, session);
5164 }
5165
5166 static int _nfs4_proc_create_session(struct nfs_client *clp)
5167 {
5168         struct nfs4_session *session = clp->cl_session;
5169         struct nfs41_create_session_args args = {
5170                 .client = clp,
5171                 .cb_program = NFS4_CALLBACK,
5172         };
5173         struct nfs41_create_session_res res = {
5174                 .client = clp,
5175         };
5176         struct rpc_message msg = {
5177                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
5178                 .rpc_argp = &args,
5179                 .rpc_resp = &res,
5180         };
5181         int status;
5182
5183         nfs4_init_channel_attrs(&args);
5184         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
5185
5186         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5187
5188         if (!status)
5189                 /* Verify the session's negotiated channel_attrs values */
5190                 status = nfs4_verify_channel_attrs(&args, session);
5191         if (!status) {
5192                 /* Increment the clientid slot sequence id */
5193                 clp->cl_seqid++;
5194         }
5195
5196         return status;
5197 }
5198
5199 /*
5200  * Issues a CREATE_SESSION operation to the server.
5201  * It is the responsibility of the caller to verify the session is
5202  * expired before calling this routine.
5203  */
5204 int nfs4_proc_create_session(struct nfs_client *clp)
5205 {
5206         int status;
5207         unsigned *ptr;
5208         struct nfs4_session *session = clp->cl_session;
5209
5210         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
5211
5212         status = _nfs4_proc_create_session(clp);
5213         if (status)
5214                 goto out;
5215
5216         /* Init and reset the fore channel */
5217         status = nfs4_init_slot_tables(session);
5218         dprintk("slot table initialization returned %d\n", status);
5219         if (status)
5220                 goto out;
5221         status = nfs4_reset_slot_tables(session);
5222         dprintk("slot table reset returned %d\n", status);
5223         if (status)
5224                 goto out;
5225
5226         ptr = (unsigned *)&session->sess_id.data[0];
5227         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
5228                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
5229 out:
5230         dprintk("<-- %s\n", __func__);
5231         return status;
5232 }
5233
5234 /*
5235  * Issue the over-the-wire RPC DESTROY_SESSION.
5236  * The caller must serialize access to this routine.
5237  */
5238 int nfs4_proc_destroy_session(struct nfs4_session *session)
5239 {
5240         int status = 0;
5241         struct rpc_message msg;
5242
5243         dprintk("--> nfs4_proc_destroy_session\n");
5244
5245         /* session is still being setup */
5246         if (session->clp->cl_cons_state != NFS_CS_READY)
5247                 return status;
5248
5249         msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION];
5250         msg.rpc_argp = session;
5251         msg.rpc_resp = NULL;
5252         msg.rpc_cred = NULL;
5253         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5254
5255         if (status)
5256                 printk(KERN_WARNING
5257                         "Got error %d from the server on DESTROY_SESSION. "
5258                         "Session has been destroyed regardless...\n", status);
5259
5260         dprintk("<-- nfs4_proc_destroy_session\n");
5261         return status;
5262 }
5263
5264 int nfs4_init_session(struct nfs_server *server)
5265 {
5266         struct nfs_client *clp = server->nfs_client;
5267         struct nfs4_session *session;
5268         unsigned int rsize, wsize;
5269         int ret;
5270
5271         if (!nfs4_has_session(clp))
5272                 return 0;
5273
5274         session = clp->cl_session;
5275         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5276                 return 0;
5277
5278         rsize = server->rsize;
5279         if (rsize == 0)
5280                 rsize = NFS_MAX_FILE_IO_SIZE;
5281         wsize = server->wsize;
5282         if (wsize == 0)
5283                 wsize = NFS_MAX_FILE_IO_SIZE;
5284
5285         session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead;
5286         session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead;
5287
5288         ret = nfs4_recover_expired_lease(server);
5289         if (!ret)
5290                 ret = nfs4_check_client_ready(clp);
5291         return ret;
5292 }
5293
5294 int nfs4_init_ds_session(struct nfs_client *clp)
5295 {
5296         struct nfs4_session *session = clp->cl_session;
5297         int ret;
5298
5299         if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state))
5300                 return 0;
5301
5302         ret = nfs4_client_recover_expired_lease(clp);
5303         if (!ret)
5304                 /* Test for the DS role */
5305                 if (!is_ds_client(clp))
5306                         ret = -ENODEV;
5307         if (!ret)
5308                 ret = nfs4_check_client_ready(clp);
5309         return ret;
5310
5311 }
5312 EXPORT_SYMBOL_GPL(nfs4_init_ds_session);
5313
5314
5315 /*
5316  * Renew the cl_session lease.
5317  */
5318 struct nfs4_sequence_data {
5319         struct nfs_client *clp;
5320         struct nfs4_sequence_args args;
5321         struct nfs4_sequence_res res;
5322 };
5323
5324 static void nfs41_sequence_release(void *data)
5325 {
5326         struct nfs4_sequence_data *calldata = data;
5327         struct nfs_client *clp = calldata->clp;
5328
5329         if (atomic_read(&clp->cl_count) > 1)
5330                 nfs4_schedule_state_renewal(clp);
5331         nfs_put_client(clp);
5332         kfree(calldata);
5333 }
5334
5335 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5336 {
5337         switch(task->tk_status) {
5338         case -NFS4ERR_DELAY:
5339                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5340                 return -EAGAIN;
5341         default:
5342                 nfs4_schedule_lease_recovery(clp);
5343         }
5344         return 0;
5345 }
5346
5347 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
5348 {
5349         struct nfs4_sequence_data *calldata = data;
5350         struct nfs_client *clp = calldata->clp;
5351
5352         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
5353                 return;
5354
5355         if (task->tk_status < 0) {
5356                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
5357                 if (atomic_read(&clp->cl_count) == 1)
5358                         goto out;
5359
5360                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
5361                         rpc_restart_call_prepare(task);
5362                         return;
5363                 }
5364         }
5365         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
5366 out:
5367         dprintk("<-- %s\n", __func__);
5368 }
5369
5370 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
5371 {
5372         struct nfs4_sequence_data *calldata = data;
5373         struct nfs_client *clp = calldata->clp;
5374         struct nfs4_sequence_args *args;
5375         struct nfs4_sequence_res *res;
5376
5377         args = task->tk_msg.rpc_argp;
5378         res = task->tk_msg.rpc_resp;
5379
5380         if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task))
5381                 return;
5382         rpc_call_start(task);
5383 }
5384
5385 static const struct rpc_call_ops nfs41_sequence_ops = {
5386         .rpc_call_done = nfs41_sequence_call_done,
5387         .rpc_call_prepare = nfs41_sequence_prepare,
5388         .rpc_release = nfs41_sequence_release,
5389 };
5390
5391 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5392 {
5393         struct nfs4_sequence_data *calldata;
5394         struct rpc_message msg = {
5395                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
5396                 .rpc_cred = cred,
5397         };
5398         struct rpc_task_setup task_setup_data = {
5399                 .rpc_client = clp->cl_rpcclient,
5400                 .rpc_message = &msg,
5401                 .callback_ops = &nfs41_sequence_ops,
5402                 .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT,
5403         };
5404
5405         if (!atomic_inc_not_zero(&clp->cl_count))
5406                 return ERR_PTR(-EIO);
5407         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5408         if (calldata == NULL) {
5409                 nfs_put_client(clp);
5410                 return ERR_PTR(-ENOMEM);
5411         }
5412         msg.rpc_argp = &calldata->args;
5413         msg.rpc_resp = &calldata->res;
5414         calldata->clp = clp;
5415         task_setup_data.callback_data = calldata;
5416
5417         return rpc_run_task(&task_setup_data);
5418 }
5419
5420 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5421 {
5422         struct rpc_task *task;
5423         int ret = 0;
5424
5425         task = _nfs41_proc_sequence(clp, cred);
5426         if (IS_ERR(task))
5427                 ret = PTR_ERR(task);
5428         else
5429                 rpc_put_task_async(task);
5430         dprintk("<-- %s status=%d\n", __func__, ret);
5431         return ret;
5432 }
5433
5434 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
5435 {
5436         struct rpc_task *task;
5437         int ret;
5438
5439         task = _nfs41_proc_sequence(clp, cred);
5440         if (IS_ERR(task)) {
5441                 ret = PTR_ERR(task);
5442                 goto out;
5443         }
5444         ret = rpc_wait_for_completion_task(task);
5445         if (!ret) {
5446                 struct nfs4_sequence_res *res = task->tk_msg.rpc_resp;
5447
5448                 if (task->tk_status == 0)
5449                         nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
5450                 ret = task->tk_status;
5451         }
5452         rpc_put_task(task);
5453 out:
5454         dprintk("<-- %s status=%d\n", __func__, ret);
5455         return ret;
5456 }
5457
5458 struct nfs4_reclaim_complete_data {
5459         struct nfs_client *clp;
5460         struct nfs41_reclaim_complete_args arg;
5461         struct nfs41_reclaim_complete_res res;
5462 };
5463
5464 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
5465 {
5466         struct nfs4_reclaim_complete_data *calldata = data;
5467
5468         rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED);
5469         if (nfs41_setup_sequence(calldata->clp->cl_session,
5470                                 &calldata->arg.seq_args,
5471                                 &calldata->res.seq_res, 0, task))
5472                 return;
5473
5474         rpc_call_start(task);
5475 }
5476
5477 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
5478 {
5479         switch(task->tk_status) {
5480         case 0:
5481         case -NFS4ERR_COMPLETE_ALREADY:
5482         case -NFS4ERR_WRONG_CRED: /* What to do here? */
5483                 break;
5484         case -NFS4ERR_DELAY:
5485                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
5486                 /* fall through */
5487         case -NFS4ERR_RETRY_UNCACHED_REP:
5488                 return -EAGAIN;
5489         default:
5490                 nfs4_schedule_lease_recovery(clp);
5491         }
5492         return 0;
5493 }
5494
5495 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
5496 {
5497         struct nfs4_reclaim_complete_data *calldata = data;
5498         struct nfs_client *clp = calldata->clp;
5499         struct nfs4_sequence_res *res = &calldata->res.seq_res;
5500
5501         dprintk("--> %s\n", __func__);
5502         if (!nfs41_sequence_done(task, res))
5503                 return;
5504
5505         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
5506                 rpc_restart_call_prepare(task);
5507                 return;
5508         }
5509         dprintk("<-- %s\n", __func__);
5510 }
5511
5512 static void nfs4_free_reclaim_complete_data(void *data)
5513 {
5514         struct nfs4_reclaim_complete_data *calldata = data;
5515
5516         kfree(calldata);
5517 }
5518
5519 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
5520         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
5521         .rpc_call_done = nfs4_reclaim_complete_done,
5522         .rpc_release = nfs4_free_reclaim_complete_data,
5523 };
5524
5525 /*
5526  * Issue a global reclaim complete.
5527  */
5528 static int nfs41_proc_reclaim_complete(struct nfs_client *clp)
5529 {
5530         struct nfs4_reclaim_complete_data *calldata;
5531         struct rpc_task *task;
5532         struct rpc_message msg = {
5533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
5534         };
5535         struct rpc_task_setup task_setup_data = {
5536                 .rpc_client = clp->cl_rpcclient,
5537                 .rpc_message = &msg,
5538                 .callback_ops = &nfs4_reclaim_complete_call_ops,
5539                 .flags = RPC_TASK_ASYNC,
5540         };
5541         int status = -ENOMEM;
5542
5543         dprintk("--> %s\n", __func__);
5544         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
5545         if (calldata == NULL)
5546                 goto out;
5547         calldata->clp = clp;
5548         calldata->arg.one_fs = 0;
5549
5550         msg.rpc_argp = &calldata->arg;
5551         msg.rpc_resp = &calldata->res;
5552         task_setup_data.callback_data = calldata;
5553         task = rpc_run_task(&task_setup_data);
5554         if (IS_ERR(task)) {
5555                 status = PTR_ERR(task);
5556                 goto out;
5557         }
5558         status = nfs4_wait_for_completion_rpc_task(task);
5559         if (status == 0)
5560                 status = task->tk_status;
5561         rpc_put_task(task);
5562         return 0;
5563 out:
5564         dprintk("<-- %s status=%d\n", __func__, status);
5565         return status;
5566 }
5567
5568 static void
5569 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
5570 {
5571         struct nfs4_layoutget *lgp = calldata;
5572         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5573
5574         dprintk("--> %s\n", __func__);
5575         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
5576          * right now covering the LAYOUTGET we are about to send.
5577          * However, that is not so catastrophic, and there seems
5578          * to be no way to prevent it completely.
5579          */
5580         if (nfs4_setup_sequence(server, &lgp->args.seq_args,
5581                                 &lgp->res.seq_res, 0, task))
5582                 return;
5583         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
5584                                           NFS_I(lgp->args.inode)->layout,
5585                                           lgp->args.ctx->state)) {
5586                 rpc_exit(task, NFS4_OK);
5587                 return;
5588         }
5589         rpc_call_start(task);
5590 }
5591
5592 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
5593 {
5594         struct nfs4_layoutget *lgp = calldata;
5595         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5596
5597         dprintk("--> %s\n", __func__);
5598
5599         if (!nfs4_sequence_done(task, &lgp->res.seq_res))
5600                 return;
5601
5602         switch (task->tk_status) {
5603         case 0:
5604                 break;
5605         case -NFS4ERR_LAYOUTTRYLATER:
5606         case -NFS4ERR_RECALLCONFLICT:
5607                 task->tk_status = -NFS4ERR_DELAY;
5608                 /* Fall through */
5609         default:
5610                 if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5611                         rpc_restart_call_prepare(task);
5612                         return;
5613                 }
5614         }
5615         dprintk("<-- %s\n", __func__);
5616 }
5617
5618 static void nfs4_layoutget_release(void *calldata)
5619 {
5620         struct nfs4_layoutget *lgp = calldata;
5621
5622         dprintk("--> %s\n", __func__);
5623         put_nfs_open_context(lgp->args.ctx);
5624         kfree(calldata);
5625         dprintk("<-- %s\n", __func__);
5626 }
5627
5628 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
5629         .rpc_call_prepare = nfs4_layoutget_prepare,
5630         .rpc_call_done = nfs4_layoutget_done,
5631         .rpc_release = nfs4_layoutget_release,
5632 };
5633
5634 int nfs4_proc_layoutget(struct nfs4_layoutget *lgp)
5635 {
5636         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
5637         struct rpc_task *task;
5638         struct rpc_message msg = {
5639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
5640                 .rpc_argp = &lgp->args,
5641                 .rpc_resp = &lgp->res,
5642         };
5643         struct rpc_task_setup task_setup_data = {
5644                 .rpc_client = server->client,
5645                 .rpc_message = &msg,
5646                 .callback_ops = &nfs4_layoutget_call_ops,
5647                 .callback_data = lgp,
5648                 .flags = RPC_TASK_ASYNC,
5649         };
5650         int status = 0;
5651
5652         dprintk("--> %s\n", __func__);
5653
5654         lgp->res.layoutp = &lgp->args.layout;
5655         lgp->res.seq_res.sr_slot = NULL;
5656         task = rpc_run_task(&task_setup_data);
5657         if (IS_ERR(task))
5658                 return PTR_ERR(task);
5659         status = nfs4_wait_for_completion_rpc_task(task);
5660         if (status == 0)
5661                 status = task->tk_status;
5662         if (status == 0)
5663                 status = pnfs_layout_process(lgp);
5664         rpc_put_task(task);
5665         dprintk("<-- %s status=%d\n", __func__, status);
5666         return status;
5667 }
5668
5669 static int
5670 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5671 {
5672         struct nfs4_getdeviceinfo_args args = {
5673                 .pdev = pdev,
5674         };
5675         struct nfs4_getdeviceinfo_res res = {
5676                 .pdev = pdev,
5677         };
5678         struct rpc_message msg = {
5679                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
5680                 .rpc_argp = &args,
5681                 .rpc_resp = &res,
5682         };
5683         int status;
5684
5685         dprintk("--> %s\n", __func__);
5686         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5687         dprintk("<-- %s status=%d\n", __func__, status);
5688
5689         return status;
5690 }
5691
5692 int nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
5693 {
5694         struct nfs4_exception exception = { };
5695         int err;
5696
5697         do {
5698                 err = nfs4_handle_exception(server,
5699                                         _nfs4_proc_getdeviceinfo(server, pdev),
5700                                         &exception);
5701         } while (exception.retry);
5702         return err;
5703 }
5704 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
5705
5706 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
5707 {
5708         struct nfs4_layoutcommit_data *data = calldata;
5709         struct nfs_server *server = NFS_SERVER(data->args.inode);
5710
5711         if (nfs4_setup_sequence(server, &data->args.seq_args,
5712                                 &data->res.seq_res, 1, task))
5713                 return;
5714         rpc_call_start(task);
5715 }
5716
5717 static void
5718 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
5719 {
5720         struct nfs4_layoutcommit_data *data = calldata;
5721         struct nfs_server *server = NFS_SERVER(data->args.inode);
5722
5723         if (!nfs4_sequence_done(task, &data->res.seq_res))
5724                 return;
5725
5726         switch (task->tk_status) { /* Just ignore these failures */
5727         case NFS4ERR_DELEG_REVOKED: /* layout was recalled */
5728         case NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
5729         case NFS4ERR_BADLAYOUT:     /* no layout */
5730         case NFS4ERR_GRACE:         /* loca_recalim always false */
5731                 task->tk_status = 0;
5732         }
5733
5734         if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) {
5735                 nfs_restart_rpc(task, server->nfs_client);
5736                 return;
5737         }
5738
5739         if (task->tk_status == 0)
5740                 nfs_post_op_update_inode_force_wcc(data->args.inode,
5741                                                    data->res.fattr);
5742 }
5743
5744 static void nfs4_layoutcommit_release(void *calldata)
5745 {
5746         struct nfs4_layoutcommit_data *data = calldata;
5747
5748         /* Matched by references in pnfs_set_layoutcommit */
5749         put_lseg(data->lseg);
5750         put_rpccred(data->cred);
5751         kfree(data);
5752 }
5753
5754 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
5755         .rpc_call_prepare = nfs4_layoutcommit_prepare,
5756         .rpc_call_done = nfs4_layoutcommit_done,
5757         .rpc_release = nfs4_layoutcommit_release,
5758 };
5759
5760 int
5761 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
5762 {
5763         struct rpc_message msg = {
5764                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
5765                 .rpc_argp = &data->args,
5766                 .rpc_resp = &data->res,
5767                 .rpc_cred = data->cred,
5768         };
5769         struct rpc_task_setup task_setup_data = {
5770                 .task = &data->task,
5771                 .rpc_client = NFS_CLIENT(data->args.inode),
5772                 .rpc_message = &msg,
5773                 .callback_ops = &nfs4_layoutcommit_ops,
5774                 .callback_data = data,
5775                 .flags = RPC_TASK_ASYNC,
5776         };
5777         struct rpc_task *task;
5778         int status = 0;
5779
5780         dprintk("NFS: %4d initiating layoutcommit call. sync %d "
5781                 "lbw: %llu inode %lu\n",
5782                 data->task.tk_pid, sync,
5783                 data->args.lastbytewritten,
5784                 data->args.inode->i_ino);
5785
5786         task = rpc_run_task(&task_setup_data);
5787         if (IS_ERR(task))
5788                 return PTR_ERR(task);
5789         if (sync == false)
5790                 goto out;
5791         status = nfs4_wait_for_completion_rpc_task(task);
5792         if (status != 0)
5793                 goto out;
5794         status = task->tk_status;
5795 out:
5796         dprintk("%s: status %d\n", __func__, status);
5797         rpc_put_task(task);
5798         return status;
5799 }
5800 #endif /* CONFIG_NFS_V4_1 */
5801
5802 struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
5803         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5804         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5805         .recover_open   = nfs4_open_reclaim,
5806         .recover_lock   = nfs4_lock_reclaim,
5807         .establish_clid = nfs4_init_clientid,
5808         .get_clid_cred  = nfs4_get_setclientid_cred,
5809 };
5810
5811 #if defined(CONFIG_NFS_V4_1)
5812 struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
5813         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
5814         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
5815         .recover_open   = nfs4_open_reclaim,
5816         .recover_lock   = nfs4_lock_reclaim,
5817         .establish_clid = nfs41_init_clientid,
5818         .get_clid_cred  = nfs4_get_exchange_id_cred,
5819         .reclaim_complete = nfs41_proc_reclaim_complete,
5820 };
5821 #endif /* CONFIG_NFS_V4_1 */
5822
5823 struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
5824         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5825         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5826         .recover_open   = nfs4_open_expired,
5827         .recover_lock   = nfs4_lock_expired,
5828         .establish_clid = nfs4_init_clientid,
5829         .get_clid_cred  = nfs4_get_setclientid_cred,
5830 };
5831
5832 #if defined(CONFIG_NFS_V4_1)
5833 struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
5834         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
5835         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
5836         .recover_open   = nfs4_open_expired,
5837         .recover_lock   = nfs4_lock_expired,
5838         .establish_clid = nfs41_init_clientid,
5839         .get_clid_cred  = nfs4_get_exchange_id_cred,
5840 };
5841 #endif /* CONFIG_NFS_V4_1 */
5842
5843 struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
5844         .sched_state_renewal = nfs4_proc_async_renew,
5845         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
5846         .renew_lease = nfs4_proc_renew,
5847 };
5848
5849 #if defined(CONFIG_NFS_V4_1)
5850 struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
5851         .sched_state_renewal = nfs41_proc_async_sequence,
5852         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
5853         .renew_lease = nfs4_proc_sequence,
5854 };
5855 #endif
5856
5857 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
5858         .minor_version = 0,
5859         .call_sync = _nfs4_call_sync,
5860         .validate_stateid = nfs4_validate_delegation_stateid,
5861         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
5862         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
5863         .state_renewal_ops = &nfs40_state_renewal_ops,
5864 };
5865
5866 #if defined(CONFIG_NFS_V4_1)
5867 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
5868         .minor_version = 1,
5869         .call_sync = _nfs4_call_sync_session,
5870         .validate_stateid = nfs41_validate_delegation_stateid,
5871         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
5872         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
5873         .state_renewal_ops = &nfs41_state_renewal_ops,
5874 };
5875 #endif
5876
5877 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
5878         [0] = &nfs_v4_0_minor_ops,
5879 #if defined(CONFIG_NFS_V4_1)
5880         [1] = &nfs_v4_1_minor_ops,
5881 #endif
5882 };
5883
5884 static const struct inode_operations nfs4_file_inode_operations = {
5885         .permission     = nfs_permission,
5886         .getattr        = nfs_getattr,
5887         .setattr        = nfs_setattr,
5888         .getxattr       = generic_getxattr,
5889         .setxattr       = generic_setxattr,
5890         .listxattr      = generic_listxattr,
5891         .removexattr    = generic_removexattr,
5892 };
5893
5894 const struct nfs_rpc_ops nfs_v4_clientops = {
5895         .version        = 4,                    /* protocol version */
5896         .dentry_ops     = &nfs4_dentry_operations,
5897         .dir_inode_ops  = &nfs4_dir_inode_operations,
5898         .file_inode_ops = &nfs4_file_inode_operations,
5899         .getroot        = nfs4_proc_get_root,
5900         .getattr        = nfs4_proc_getattr,
5901         .setattr        = nfs4_proc_setattr,
5902         .lookupfh       = nfs4_proc_lookupfh,
5903         .lookup         = nfs4_proc_lookup,
5904         .access         = nfs4_proc_access,
5905         .readlink       = nfs4_proc_readlink,
5906         .create         = nfs4_proc_create,
5907         .remove         = nfs4_proc_remove,
5908         .unlink_setup   = nfs4_proc_unlink_setup,
5909         .unlink_done    = nfs4_proc_unlink_done,
5910         .rename         = nfs4_proc_rename,
5911         .rename_setup   = nfs4_proc_rename_setup,
5912         .rename_done    = nfs4_proc_rename_done,
5913         .link           = nfs4_proc_link,
5914         .symlink        = nfs4_proc_symlink,
5915         .mkdir          = nfs4_proc_mkdir,
5916         .rmdir          = nfs4_proc_remove,
5917         .readdir        = nfs4_proc_readdir,
5918         .mknod          = nfs4_proc_mknod,
5919         .statfs         = nfs4_proc_statfs,
5920         .fsinfo         = nfs4_proc_fsinfo,
5921         .pathconf       = nfs4_proc_pathconf,
5922         .set_capabilities = nfs4_server_capabilities,
5923         .decode_dirent  = nfs4_decode_dirent,
5924         .read_setup     = nfs4_proc_read_setup,
5925         .read_done      = nfs4_read_done,
5926         .write_setup    = nfs4_proc_write_setup,
5927         .write_done     = nfs4_write_done,
5928         .commit_setup   = nfs4_proc_commit_setup,
5929         .commit_done    = nfs4_commit_done,
5930         .lock           = nfs4_proc_lock,
5931         .clear_acl_cache = nfs4_zap_acl_attr,
5932         .close_context  = nfs4_close_context,
5933         .open_context   = nfs4_atomic_open,
5934         .init_client    = nfs4_init_client,
5935         .secinfo        = nfs4_proc_secinfo,
5936 };
5937
5938 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
5939         .prefix = XATTR_NAME_NFSV4_ACL,
5940         .list   = nfs4_xattr_list_nfs4_acl,
5941         .get    = nfs4_xattr_get_nfs4_acl,
5942         .set    = nfs4_xattr_set_nfs4_acl,
5943 };
5944
5945 const struct xattr_handler *nfs4_xattr_handlers[] = {
5946         &nfs4_xattr_nfs4_acl_handler,
5947         NULL
5948 };
5949
5950 /*
5951  * Local variables:
5952  *  c-basic-offset: 8
5953  * End:
5954  */