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/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 struct nfs4_opendata;
  78 static int _nfs4_proc_open(struct nfs4_opendata *data);
  79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  81 static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *, struct nfs4_state *, long *);
  82 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  83 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  84 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  85 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  86                             struct nfs_fattr *fattr, struct iattr *sattr,
  87                             struct nfs4_state *state, struct nfs4_label *ilabel,
  88                             struct nfs4_label *olabel);
  89 #ifdef CONFIG_NFS_V4_1
  90 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
  91                 struct rpc_cred *);
  92 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
  93                 struct rpc_cred *);
  94 #endif
  95
  96 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
  97 static inline struct nfs4_label *
  98 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
  99         struct iattr *sattr, struct nfs4_label *label)
 100 {
 101         int err;
 102
 103         if (label == NULL)
 104                 return NULL;
 105
 106         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 107                 return NULL;
 108
 109         err = security_dentry_init_security(dentry, sattr->ia_mode,
 110                                 &dentry->d_name, (void **)&label->label, &label->len);
 111         if (err == 0)
 112                 return label;
 113
 114         return NULL;
 115 }
 116 static inline void
 117 nfs4_label_release_security(struct nfs4_label *label)
 118 {
 119         if (label)
 120                 security_release_secctx(label->label, label->len);
 121 }
 122 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 123 {
 124         if (label)
 125                 return server->attr_bitmask;
 126
 127         return server->attr_bitmask_nl;
 128 }
 129 #else
 130 static inline struct nfs4_label *
 131 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 132         struct iattr *sattr, struct nfs4_label *l)
 133 { return NULL; }
 134 static inline void
 135 nfs4_label_release_security(struct nfs4_label *label)
 136 { return; }
 137 static inline u32 *
 138 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 139 { return server->attr_bitmask; }
 140 #endif
 141
 142 /* Prevent leaks of NFSv4 errors into userland */
 143 static int nfs4_map_errors(int err)
 144 {
 145         if (err >= -1000)
 146                 return err;
 147         switch (err) {
 148         case -NFS4ERR_RESOURCE:
 149         case -NFS4ERR_LAYOUTTRYLATER:
 150         case -NFS4ERR_RECALLCONFLICT:
 151                 return -EREMOTEIO;
 152         case -NFS4ERR_WRONGSEC:
 153         case -NFS4ERR_WRONG_CRED:
 154                 return -EPERM;
 155         case -NFS4ERR_BADOWNER:
 156         case -NFS4ERR_BADNAME:
 157                 return -EINVAL;
 158         case -NFS4ERR_SHARE_DENIED:
 159                 return -EACCES;
 160         case -NFS4ERR_MINOR_VERS_MISMATCH:
 161                 return -EPROTONOSUPPORT;
 162         case -NFS4ERR_FILE_OPEN:
 163                 return -EBUSY;
 164         default:
 165                 dprintk("%s could not handle NFSv4 error %d\n",
 166                                 __func__, -err);
 167                 break;
 168         }
 169         return -EIO;
 170 }
 171
 172 /*
 173  * This is our standard bitmap for GETATTR requests.
 174  */
 175 const u32 nfs4_fattr_bitmap[3] = {
 176         FATTR4_WORD0_TYPE
 177         | FATTR4_WORD0_CHANGE
 178         | FATTR4_WORD0_SIZE
 179         | FATTR4_WORD0_FSID
 180         | FATTR4_WORD0_FILEID,
 181         FATTR4_WORD1_MODE
 182         | FATTR4_WORD1_NUMLINKS
 183         | FATTR4_WORD1_OWNER
 184         | FATTR4_WORD1_OWNER_GROUP
 185         | FATTR4_WORD1_RAWDEV
 186         | FATTR4_WORD1_SPACE_USED
 187         | FATTR4_WORD1_TIME_ACCESS
 188         | FATTR4_WORD1_TIME_METADATA
 189         | FATTR4_WORD1_TIME_MODIFY
 190         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 191 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 192         FATTR4_WORD2_SECURITY_LABEL
 193 #endif
 194 };
 195
 196 static const u32 nfs4_pnfs_open_bitmap[3] = {
 197         FATTR4_WORD0_TYPE
 198         | FATTR4_WORD0_CHANGE
 199         | FATTR4_WORD0_SIZE
 200         | FATTR4_WORD0_FSID
 201         | FATTR4_WORD0_FILEID,
 202         FATTR4_WORD1_MODE
 203         | FATTR4_WORD1_NUMLINKS
 204         | FATTR4_WORD1_OWNER
 205         | FATTR4_WORD1_OWNER_GROUP
 206         | FATTR4_WORD1_RAWDEV
 207         | FATTR4_WORD1_SPACE_USED
 208         | FATTR4_WORD1_TIME_ACCESS
 209         | FATTR4_WORD1_TIME_METADATA
 210         | FATTR4_WORD1_TIME_MODIFY,
 211         FATTR4_WORD2_MDSTHRESHOLD
 212 };
 213
 214 static const u32 nfs4_open_noattr_bitmap[3] = {
 215         FATTR4_WORD0_TYPE
 216         | FATTR4_WORD0_CHANGE
 217         | FATTR4_WORD0_FILEID,
 218 };
 219
 220 const u32 nfs4_statfs_bitmap[3] = {
 221         FATTR4_WORD0_FILES_AVAIL
 222         | FATTR4_WORD0_FILES_FREE
 223         | FATTR4_WORD0_FILES_TOTAL,
 224         FATTR4_WORD1_SPACE_AVAIL
 225         | FATTR4_WORD1_SPACE_FREE
 226         | FATTR4_WORD1_SPACE_TOTAL
 227 };
 228
 229 const u32 nfs4_pathconf_bitmap[3] = {
 230         FATTR4_WORD0_MAXLINK
 231         | FATTR4_WORD0_MAXNAME,
 232         0
 233 };
 234
 235 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 236                         | FATTR4_WORD0_MAXREAD
 237                         | FATTR4_WORD0_MAXWRITE
 238                         | FATTR4_WORD0_LEASE_TIME,
 239                         FATTR4_WORD1_TIME_DELTA
 240                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 241                         FATTR4_WORD2_LAYOUT_BLKSIZE
 242 };
 243
 244 const u32 nfs4_fs_locations_bitmap[3] = {
 245         FATTR4_WORD0_TYPE
 246         | FATTR4_WORD0_CHANGE
 247         | FATTR4_WORD0_SIZE
 248         | FATTR4_WORD0_FSID
 249         | FATTR4_WORD0_FILEID
 250         | FATTR4_WORD0_FS_LOCATIONS,
 251         FATTR4_WORD1_MODE
 252         | FATTR4_WORD1_NUMLINKS
 253         | FATTR4_WORD1_OWNER
 254         | FATTR4_WORD1_OWNER_GROUP
 255         | FATTR4_WORD1_RAWDEV
 256         | FATTR4_WORD1_SPACE_USED
 257         | FATTR4_WORD1_TIME_ACCESS
 258         | FATTR4_WORD1_TIME_METADATA
 259         | FATTR4_WORD1_TIME_MODIFY
 260         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 261 };
 262
 263 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 264                 struct nfs4_readdir_arg *readdir)
 265 {
 266         __be32 *start, *p;
 267
 268         if (cookie > 2) {
 269                 readdir->cookie = cookie;
 270                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 271                 return;
 272         }
 273
 274         readdir->cookie = 0;
 275         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 276         if (cookie == 2)
 277                 return;
 278
 279         /*
 280          * NFSv4 servers do not return entries for '.' and '..'
 281          * Therefore, we fake these entries here.  We let '.'
 282          * have cookie 0 and '..' have cookie 1.  Note that
 283          * when talking to the server, we always send cookie 0
 284          * instead of 1 or 2.
 285          */
 286         start = p = kmap_atomic(*readdir->pages);
 287
 288         if (cookie == 0) {
 289                 *p++ = xdr_one;                                  /* next */
 290                 *p++ = xdr_zero;                   /* cookie, first word */
 291                 *p++ = xdr_one;                   /* cookie, second word */
 292                 *p++ = xdr_one;                             /* entry len */
 293                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 294                 p++;
 295                 *p++ = xdr_one;                         /* bitmap length */
 296                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 297                 *p++ = htonl(8);              /* attribute buffer length */
 298                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 299         }
 300
 301         *p++ = xdr_one;                                  /* next */
 302         *p++ = xdr_zero;                   /* cookie, first word */
 303         *p++ = xdr_two;                   /* cookie, second word */
 304         *p++ = xdr_two;                             /* entry len */
 305         memcpy(p, "..\0\0", 4);                         /* entry */
 306         p++;
 307         *p++ = xdr_one;                         /* bitmap length */
 308         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 309         *p++ = htonl(8);              /* attribute buffer length */
 310         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 311
 312         readdir->pgbase = (char *)p - (char *)start;
 313         readdir->count -= readdir->pgbase;
 314         kunmap_atomic(start);
 315 }
 316
 317 static long nfs4_update_delay(long *timeout)
 318 {
 319         long ret;
 320         if (!timeout)
 321                 return NFS4_POLL_RETRY_MAX;
 322         if (*timeout <= 0)
 323                 *timeout = NFS4_POLL_RETRY_MIN;
 324         if (*timeout > NFS4_POLL_RETRY_MAX)
 325                 *timeout = NFS4_POLL_RETRY_MAX;
 326         ret = *timeout;
 327         *timeout <<= 1;
 328         return ret;
 329 }
 330
 331 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 332 {
 333         int res = 0;
 334
 335         might_sleep();
 336
 337         freezable_schedule_timeout_killable_unsafe(
 338                 nfs4_update_delay(timeout));
 339         if (fatal_signal_pending(current))
 340                 res = -ERESTARTSYS;
 341         return res;
 342 }
 343
 344 /* This is the error handling routine for processes that are allowed
 345  * to sleep.
 346  */
 347 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 348 {
 349         struct nfs_client *clp = server->nfs_client;
 350         struct nfs4_state *state = exception->state;
 351         struct inode *inode = exception->inode;
 352         int ret = errorcode;
 353
 354         exception->retry = 0;
 355         switch(errorcode) {
 356                 case 0:
 357                         return 0;
 358                 case -NFS4ERR_OPENMODE:
 359                 case -NFS4ERR_DELEG_REVOKED:
 360                 case -NFS4ERR_ADMIN_REVOKED:
 361                 case -NFS4ERR_BAD_STATEID:
 362                         if (inode && nfs4_have_delegation(inode, FMODE_READ)) {
 363                                 nfs4_inode_return_delegation(inode);
 364                                 exception->retry = 1;
 365                                 return 0;
 366                         }
 367                         if (state == NULL)
 368                                 break;
 369                         ret = nfs4_schedule_stateid_recovery(server, state);
 370                         if (ret < 0)
 371                                 break;
 372                         goto wait_on_recovery;
 373                 case -NFS4ERR_EXPIRED:
 374                         if (state != NULL) {
 375                                 ret = nfs4_schedule_stateid_recovery(server, state);
 376                                 if (ret < 0)
 377                                         break;
 378                         }
 379                 case -NFS4ERR_STALE_STATEID:
 380                 case -NFS4ERR_STALE_CLIENTID:
 381                         nfs4_schedule_lease_recovery(clp);
 382                         goto wait_on_recovery;
 383                 case -NFS4ERR_MOVED:
 384                         ret = nfs4_schedule_migration_recovery(server);
 385                         if (ret < 0)
 386                                 break;
 387                         goto wait_on_recovery;
 388                 case -NFS4ERR_LEASE_MOVED:
 389                         nfs4_schedule_lease_moved_recovery(clp);
 390                         goto wait_on_recovery;
 391 #if defined(CONFIG_NFS_V4_1)
 392                 case -NFS4ERR_BADSESSION:
 393                 case -NFS4ERR_BADSLOT:
 394                 case -NFS4ERR_BAD_HIGH_SLOT:
 395                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 396                 case -NFS4ERR_DEADSESSION:
 397                 case -NFS4ERR_SEQ_FALSE_RETRY:
 398                 case -NFS4ERR_SEQ_MISORDERED:
 399                         dprintk("%s ERROR: %d Reset session\n", __func__,
 400                                 errorcode);
 401                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 402                         goto wait_on_recovery;
 403 #endif /* defined(CONFIG_NFS_V4_1) */
 404                 case -NFS4ERR_FILE_OPEN:
 405                         if (exception->timeout > HZ) {
 406                                 /* We have retried a decent amount, time to
 407                                  * fail
 408                                  */
 409                                 ret = -EBUSY;
 410                                 break;
 411                         }
 412                 case -NFS4ERR_GRACE:
 413                 case -NFS4ERR_DELAY:
 414                         ret = nfs4_delay(server->client, &exception->timeout);
 415                         if (ret != 0)
 416                                 break;
 417                 case -NFS4ERR_RETRY_UNCACHED_REP:
 418                 case -NFS4ERR_OLD_STATEID:
 419                         exception->retry = 1;
 420                         break;
 421                 case -NFS4ERR_BADOWNER:
 422                         /* The following works around a Linux server bug! */
 423                 case -NFS4ERR_BADNAME:
 424                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 425                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 426                                 exception->retry = 1;
 427                                 printk(KERN_WARNING "NFS: v4 server %s "
 428                                                 "does not accept raw "
 429                                                 "uid/gids. "
 430                                                 "Reenabling the idmapper.\n",
 431                                                 server->nfs_client->cl_hostname);
 432                         }
 433         }
 434         /* We failed to handle the error */
 435         return nfs4_map_errors(ret);
 436 wait_on_recovery:
 437         ret = nfs4_wait_clnt_recover(clp);
 438         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 439                 return -EIO;
 440         if (ret == 0)
 441                 exception->retry = 1;
 442         return ret;
 443 }
 444
 445 /*
 446  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 447  * or 'false' otherwise.
 448  */
 449 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 450 {
 451         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 452
 453         if (flavor == RPC_AUTH_GSS_KRB5I ||
 454             flavor == RPC_AUTH_GSS_KRB5P)
 455                 return true;
 456
 457         return false;
 458 }
 459
 460 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 461 {
 462         spin_lock(&clp->cl_lock);
 463         if (time_before(clp->cl_last_renewal,timestamp))
 464                 clp->cl_last_renewal = timestamp;
 465         spin_unlock(&clp->cl_lock);
 466 }
 467
 468 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 469 {
 470         struct nfs_client *clp = server->nfs_client;
 471
 472         if (!nfs4_has_session(clp))
 473                 do_renew_lease(clp, timestamp);
 474 }
 475
 476 struct nfs4_call_sync_data {
 477         const struct nfs_server *seq_server;
 478         struct nfs4_sequence_args *seq_args;
 479         struct nfs4_sequence_res *seq_res;
 480 };
 481
 482 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 483                         struct nfs4_sequence_res *res, int cache_reply)
 484 {
 485         args->sa_slot = NULL;
 486         args->sa_cache_this = cache_reply;
 487         args->sa_privileged = 0;
 488
 489         res->sr_slot = NULL;
 490 }
 491
 492 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 493 {
 494         args->sa_privileged = 1;
 495 }
 496
 497 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 498                          struct nfs4_sequence_args *args,
 499                          struct nfs4_sequence_res *res,
 500                          struct rpc_task *task)
 501 {
 502         struct nfs4_slot *slot;
 503
 504         /* slot already allocated? */
 505         if (res->sr_slot != NULL)
 506                 goto out_start;
 507
 508         spin_lock(&tbl->slot_tbl_lock);
 509         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 510                 goto out_sleep;
 511
 512         slot = nfs4_alloc_slot(tbl);
 513         if (IS_ERR(slot)) {
 514                 if (slot == ERR_PTR(-ENOMEM))
 515                         task->tk_timeout = HZ >> 2;
 516                 goto out_sleep;
 517         }
 518         spin_unlock(&tbl->slot_tbl_lock);
 519
 520         args->sa_slot = slot;
 521         res->sr_slot = slot;
 522
 523 out_start:
 524         rpc_call_start(task);
 525         return 0;
 526
 527 out_sleep:
 528         if (args->sa_privileged)
 529                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 530                                 NULL, RPC_PRIORITY_PRIVILEGED);
 531         else
 532                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 533         spin_unlock(&tbl->slot_tbl_lock);
 534         return -EAGAIN;
 535 }
 536 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 537
 538 static int nfs40_sequence_done(struct rpc_task *task,
 539                                struct nfs4_sequence_res *res)
 540 {
 541         struct nfs4_slot *slot = res->sr_slot;
 542         struct nfs4_slot_table *tbl;
 543
 544         if (slot == NULL)
 545                 goto out;
 546
 547         tbl = slot->table;
 548         spin_lock(&tbl->slot_tbl_lock);
 549         if (!nfs41_wake_and_assign_slot(tbl, slot))
 550                 nfs4_free_slot(tbl, slot);
 551         spin_unlock(&tbl->slot_tbl_lock);
 552
 553         res->sr_slot = NULL;
 554 out:
 555         return 1;
 556 }
 557
 558 #if defined(CONFIG_NFS_V4_1)
 559
 560 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 561 {
 562         struct nfs4_session *session;
 563         struct nfs4_slot_table *tbl;
 564         struct nfs4_slot *slot = res->sr_slot;
 565         bool send_new_highest_used_slotid = false;
 566
 567         tbl = slot->table;
 568         session = tbl->session;
 569
 570         spin_lock(&tbl->slot_tbl_lock);
 571         /* Be nice to the server: try to ensure that the last transmitted
 572          * value for highest_user_slotid <= target_highest_slotid
 573          */
 574         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 575                 send_new_highest_used_slotid = true;
 576
 577         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 578                 send_new_highest_used_slotid = false;
 579                 goto out_unlock;
 580         }
 581         nfs4_free_slot(tbl, slot);
 582
 583         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 584                 send_new_highest_used_slotid = false;
 585 out_unlock:
 586         spin_unlock(&tbl->slot_tbl_lock);
 587         res->sr_slot = NULL;
 588         if (send_new_highest_used_slotid)
 589                 nfs41_server_notify_highest_slotid_update(session->clp);
 590 }
 591
 592 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 593 {
 594         struct nfs4_session *session;
 595         struct nfs4_slot *slot = res->sr_slot;
 596         struct nfs_client *clp;
 597         bool interrupted = false;
 598         int ret = 1;
 599
 600         if (slot == NULL)
 601                 goto out_noaction;
 602         /* don't increment the sequence number if the task wasn't sent */
 603         if (!RPC_WAS_SENT(task))
 604                 goto out;
 605
 606         session = slot->table->session;
 607
 608         if (slot->interrupted) {
 609                 slot->interrupted = 0;
 610                 interrupted = true;
 611         }
 612
 613         trace_nfs4_sequence_done(session, res);
 614         /* Check the SEQUENCE operation status */
 615         switch (res->sr_status) {
 616         case 0:
 617                 /* Update the slot's sequence and clientid lease timer */
 618                 ++slot->seq_nr;
 619                 clp = session->clp;
 620                 do_renew_lease(clp, res->sr_timestamp);
 621                 /* Check sequence flags */
 622                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
 623                 nfs41_update_target_slotid(slot->table, slot, res);
 624                 break;
 625         case 1:
 626                 /*
 627                  * sr_status remains 1 if an RPC level error occurred.
 628                  * The server may or may not have processed the sequence
 629                  * operation..
 630                  * Mark the slot as having hosted an interrupted RPC call.
 631                  */
 632                 slot->interrupted = 1;
 633                 goto out;
 634         case -NFS4ERR_DELAY:
 635                 /* The server detected a resend of the RPC call and
 636                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 637                  * of RFC5661.
 638                  */
 639                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 640                         __func__,
 641                         slot->slot_nr,
 642                         slot->seq_nr);
 643                 goto out_retry;
 644         case -NFS4ERR_BADSLOT:
 645                 /*
 646                  * The slot id we used was probably retired. Try again
 647                  * using a different slot id.
 648                  */
 649                 goto retry_nowait;
 650         case -NFS4ERR_SEQ_MISORDERED:
 651                 /*
 652                  * Was the last operation on this sequence interrupted?
 653                  * If so, retry after bumping the sequence number.
 654                  */
 655                 if (interrupted) {
 656                         ++slot->seq_nr;
 657                         goto retry_nowait;
 658                 }
 659                 /*
 660                  * Could this slot have been previously retired?
 661                  * If so, then the server may be expecting seq_nr = 1!
 662                  */
 663                 if (slot->seq_nr != 1) {
 664                         slot->seq_nr = 1;
 665                         goto retry_nowait;
 666                 }
 667                 break;
 668         case -NFS4ERR_SEQ_FALSE_RETRY:
 669                 ++slot->seq_nr;
 670                 goto retry_nowait;
 671         default:
 672                 /* Just update the slot sequence no. */
 673                 ++slot->seq_nr;
 674         }
 675 out:
 676         /* The session may be reset by one of the error handlers. */
 677         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 678         nfs41_sequence_free_slot(res);
 679 out_noaction:
 680         return ret;
 681 retry_nowait:
 682         if (rpc_restart_call_prepare(task)) {
 683                 task->tk_status = 0;
 684                 ret = 0;
 685         }
 686         goto out;
 687 out_retry:
 688         if (!rpc_restart_call(task))
 689                 goto out;
 690         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 691         return 0;
 692 }
 693 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 694
 695 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 696 {
 697         if (res->sr_slot == NULL)
 698                 return 1;
 699         if (!res->sr_slot->table->session)
 700                 return nfs40_sequence_done(task, res);
 701         return nfs41_sequence_done(task, res);
 702 }
 703 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 704
 705 int nfs41_setup_sequence(struct nfs4_session *session,
 706                                 struct nfs4_sequence_args *args,
 707                                 struct nfs4_sequence_res *res,
 708                                 struct rpc_task *task)
 709 {
 710         struct nfs4_slot *slot;
 711         struct nfs4_slot_table *tbl;
 712
 713         dprintk("--> %s\n", __func__);
 714         /* slot already allocated? */
 715         if (res->sr_slot != NULL)
 716                 goto out_success;
 717
 718         tbl = &session->fc_slot_table;
 719
 720         task->tk_timeout = 0;
 721
 722         spin_lock(&tbl->slot_tbl_lock);
 723         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 724             !args->sa_privileged) {
 725                 /* The state manager will wait until the slot table is empty */
 726                 dprintk("%s session is draining\n", __func__);
 727                 goto out_sleep;
 728         }
 729
 730         slot = nfs4_alloc_slot(tbl);
 731         if (IS_ERR(slot)) {
 732                 /* If out of memory, try again in 1/4 second */
 733                 if (slot == ERR_PTR(-ENOMEM))
 734                         task->tk_timeout = HZ >> 2;
 735                 dprintk("<-- %s: no free slots\n", __func__);
 736                 goto out_sleep;
 737         }
 738         spin_unlock(&tbl->slot_tbl_lock);
 739
 740         args->sa_slot = slot;
 741
 742         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 743                         slot->slot_nr, slot->seq_nr);
 744
 745         res->sr_slot = slot;
 746         res->sr_timestamp = jiffies;
 747         res->sr_status_flags = 0;
 748         /*
 749          * sr_status is only set in decode_sequence, and so will remain
 750          * set to 1 if an rpc level failure occurs.
 751          */
 752         res->sr_status = 1;
 753         trace_nfs4_setup_sequence(session, args);
 754 out_success:
 755         rpc_call_start(task);
 756         return 0;
 757 out_sleep:
 758         /* Privileged tasks are queued with top priority */
 759         if (args->sa_privileged)
 760                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 761                                 NULL, RPC_PRIORITY_PRIVILEGED);
 762         else
 763                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 764         spin_unlock(&tbl->slot_tbl_lock);
 765         return -EAGAIN;
 766 }
 767 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 768
 769 static int nfs4_setup_sequence(const struct nfs_server *server,
 770                                struct nfs4_sequence_args *args,
 771                                struct nfs4_sequence_res *res,
 772                                struct rpc_task *task)
 773 {
 774         struct nfs4_session *session = nfs4_get_session(server);
 775         int ret = 0;
 776
 777         if (!session)
 778                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 779                                             args, res, task);
 780
 781         dprintk("--> %s clp %p session %p sr_slot %u\n",
 782                 __func__, session->clp, session, res->sr_slot ?
 783                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 784
 785         ret = nfs41_setup_sequence(session, args, res, task);
 786
 787         dprintk("<-- %s status=%d\n", __func__, ret);
 788         return ret;
 789 }
 790
 791 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 792 {
 793         struct nfs4_call_sync_data *data = calldata;
 794         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 795
 796         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 797
 798         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 799 }
 800
 801 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 802 {
 803         struct nfs4_call_sync_data *data = calldata;
 804
 805         nfs41_sequence_done(task, data->seq_res);
 806 }
 807
 808 static const struct rpc_call_ops nfs41_call_sync_ops = {
 809         .rpc_call_prepare = nfs41_call_sync_prepare,
 810         .rpc_call_done = nfs41_call_sync_done,
 811 };
 812
 813 #else   /* !CONFIG_NFS_V4_1 */
 814
 815 static int nfs4_setup_sequence(const struct nfs_server *server,
 816                                struct nfs4_sequence_args *args,
 817                                struct nfs4_sequence_res *res,
 818                                struct rpc_task *task)
 819 {
 820         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 821                                     args, res, task);
 822 }
 823
 824 int nfs4_sequence_done(struct rpc_task *task,
 825                        struct nfs4_sequence_res *res)
 826 {
 827         return nfs40_sequence_done(task, res);
 828 }
 829 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 830
 831 #endif  /* !CONFIG_NFS_V4_1 */
 832
 833 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 834 {
 835         struct nfs4_call_sync_data *data = calldata;
 836         nfs4_setup_sequence(data->seq_server,
 837                                 data->seq_args, data->seq_res, task);
 838 }
 839
 840 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 841 {
 842         struct nfs4_call_sync_data *data = calldata;
 843         nfs4_sequence_done(task, data->seq_res);
 844 }
 845
 846 static const struct rpc_call_ops nfs40_call_sync_ops = {
 847         .rpc_call_prepare = nfs40_call_sync_prepare,
 848         .rpc_call_done = nfs40_call_sync_done,
 849 };
 850
 851 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 852                                    struct nfs_server *server,
 853                                    struct rpc_message *msg,
 854                                    struct nfs4_sequence_args *args,
 855                                    struct nfs4_sequence_res *res)
 856 {
 857         int ret;
 858         struct rpc_task *task;
 859         struct nfs_client *clp = server->nfs_client;
 860         struct nfs4_call_sync_data data = {
 861                 .seq_server = server,
 862                 .seq_args = args,
 863                 .seq_res = res,
 864         };
 865         struct rpc_task_setup task_setup = {
 866                 .rpc_client = clnt,
 867                 .rpc_message = msg,
 868                 .callback_ops = clp->cl_mvops->call_sync_ops,
 869                 .callback_data = &data
 870         };
 871
 872         task = rpc_run_task(&task_setup);
 873         if (IS_ERR(task))
 874                 ret = PTR_ERR(task);
 875         else {
 876                 ret = task->tk_status;
 877                 rpc_put_task(task);
 878         }
 879         return ret;
 880 }
 881
 882 int nfs4_call_sync(struct rpc_clnt *clnt,
 883                    struct nfs_server *server,
 884                    struct rpc_message *msg,
 885                    struct nfs4_sequence_args *args,
 886                    struct nfs4_sequence_res *res,
 887                    int cache_reply)
 888 {
 889         nfs4_init_sequence(args, res, cache_reply);
 890         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 891 }
 892
 893 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
 894 {
 895         struct nfs_inode *nfsi = NFS_I(dir);
 896
 897         spin_lock(&dir->i_lock);
 898         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
 899         if (!cinfo->atomic || cinfo->before != dir->i_version)
 900                 nfs_force_lookup_revalidate(dir);
 901         dir->i_version = cinfo->after;
 902         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 903         nfs_fscache_invalidate(dir);
 904         spin_unlock(&dir->i_lock);
 905 }
 906
 907 struct nfs4_opendata {
 908         struct kref kref;
 909         struct nfs_openargs o_arg;
 910         struct nfs_openres o_res;
 911         struct nfs_open_confirmargs c_arg;
 912         struct nfs_open_confirmres c_res;
 913         struct nfs4_string owner_name;
 914         struct nfs4_string group_name;
 915         struct nfs_fattr f_attr;
 916         struct nfs4_label *f_label;
 917         struct dentry *dir;
 918         struct dentry *dentry;
 919         struct nfs4_state_owner *owner;
 920         struct nfs4_state *state;
 921         struct iattr attrs;
 922         unsigned long timestamp;
 923         unsigned int rpc_done : 1;
 924         unsigned int file_created : 1;
 925         unsigned int is_recover : 1;
 926         int rpc_status;
 927         int cancelled;
 928 };
 929
 930 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 931                 int err, struct nfs4_exception *exception)
 932 {
 933         if (err != -EINVAL)
 934                 return false;
 935         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 936                 return false;
 937         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 938         exception->retry = 1;
 939         return true;
 940 }
 941
 942 static u32
 943 nfs4_map_atomic_open_share(struct nfs_server *server,
 944                 fmode_t fmode, int openflags)
 945 {
 946         u32 res = 0;
 947
 948         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
 949         case FMODE_READ:
 950                 res = NFS4_SHARE_ACCESS_READ;
 951                 break;
 952         case FMODE_WRITE:
 953                 res = NFS4_SHARE_ACCESS_WRITE;
 954                 break;
 955         case FMODE_READ|FMODE_WRITE:
 956                 res = NFS4_SHARE_ACCESS_BOTH;
 957         }
 958         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 959                 goto out;
 960         /* Want no delegation if we're using O_DIRECT */
 961         if (openflags & O_DIRECT)
 962                 res |= NFS4_SHARE_WANT_NO_DELEG;
 963 out:
 964         return res;
 965 }
 966
 967 static enum open_claim_type4
 968 nfs4_map_atomic_open_claim(struct nfs_server *server,
 969                 enum open_claim_type4 claim)
 970 {
 971         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 972                 return claim;
 973         switch (claim) {
 974         default:
 975                 return claim;
 976         case NFS4_OPEN_CLAIM_FH:
 977                 return NFS4_OPEN_CLAIM_NULL;
 978         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 979                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 980         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 981                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 982         }
 983 }
 984
 985 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 986 {
 987         p->o_res.f_attr = &p->f_attr;
 988         p->o_res.f_label = p->f_label;
 989         p->o_res.seqid = p->o_arg.seqid;
 990         p->c_res.seqid = p->c_arg.seqid;
 991         p->o_res.server = p->o_arg.server;
 992         p->o_res.access_request = p->o_arg.access;
 993         nfs_fattr_init(&p->f_attr);
 994         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 995 }
 996
 997 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 998                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
 999                 const struct iattr *attrs,
1000                 struct nfs4_label *label,
1001                 enum open_claim_type4 claim,
1002                 gfp_t gfp_mask)
1003 {
1004         struct dentry *parent = dget_parent(dentry);
1005         struct inode *dir = d_inode(parent);
1006         struct nfs_server *server = NFS_SERVER(dir);
1007         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1008         struct nfs4_opendata *p;
1009
1010         p = kzalloc(sizeof(*p), gfp_mask);
1011         if (p == NULL)
1012                 goto err;
1013
1014         p->f_label = nfs4_label_alloc(server, gfp_mask);
1015         if (IS_ERR(p->f_label))
1016                 goto err_free_p;
1017
1018         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1019         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1020         if (IS_ERR(p->o_arg.seqid))
1021                 goto err_free_label;
1022         nfs_sb_active(dentry->d_sb);
1023         p->dentry = dget(dentry);
1024         p->dir = parent;
1025         p->owner = sp;
1026         atomic_inc(&sp->so_count);
1027         p->o_arg.open_flags = flags;
1028         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1029         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1030                         fmode, flags);
1031         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1032          * will return permission denied for all bits until close */
1033         if (!(flags & O_EXCL)) {
1034                 /* ask server to check for all possible rights as results
1035                  * are cached */
1036                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1037                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1038         }
1039         p->o_arg.clientid = server->nfs_client->cl_clientid;
1040         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1041         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1042         p->o_arg.name = &dentry->d_name;
1043         p->o_arg.server = server;
1044         p->o_arg.bitmask = nfs4_bitmask(server, label);
1045         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1046         p->o_arg.label = label;
1047         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1048         switch (p->o_arg.claim) {
1049         case NFS4_OPEN_CLAIM_NULL:
1050         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1051         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1052                 p->o_arg.fh = NFS_FH(dir);
1053                 break;
1054         case NFS4_OPEN_CLAIM_PREVIOUS:
1055         case NFS4_OPEN_CLAIM_FH:
1056         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1057         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1058                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1059         }
1060         if (attrs != NULL && attrs->ia_valid != 0) {
1061                 __u32 verf[2];
1062
1063                 p->o_arg.u.attrs = &p->attrs;
1064                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1065
1066                 verf[0] = jiffies;
1067                 verf[1] = current->pid;
1068                 memcpy(p->o_arg.u.verifier.data, verf,
1069                                 sizeof(p->o_arg.u.verifier.data));
1070         }
1071         p->c_arg.fh = &p->o_res.fh;
1072         p->c_arg.stateid = &p->o_res.stateid;
1073         p->c_arg.seqid = p->o_arg.seqid;
1074         nfs4_init_opendata_res(p);
1075         kref_init(&p->kref);
1076         return p;
1077
1078 err_free_label:
1079         nfs4_label_free(p->f_label);
1080 err_free_p:
1081         kfree(p);
1082 err:
1083         dput(parent);
1084         return NULL;
1085 }
1086
1087 static void nfs4_opendata_free(struct kref *kref)
1088 {
1089         struct nfs4_opendata *p = container_of(kref,
1090                         struct nfs4_opendata, kref);
1091         struct super_block *sb = p->dentry->d_sb;
1092
1093         nfs_free_seqid(p->o_arg.seqid);
1094         if (p->state != NULL)
1095                 nfs4_put_open_state(p->state);
1096         nfs4_put_state_owner(p->owner);
1097
1098         nfs4_label_free(p->f_label);
1099
1100         dput(p->dir);
1101         dput(p->dentry);
1102         nfs_sb_deactive(sb);
1103         nfs_fattr_free_names(&p->f_attr);
1104         kfree(p->f_attr.mdsthreshold);
1105         kfree(p);
1106 }
1107
1108 static void nfs4_opendata_put(struct nfs4_opendata *p)
1109 {
1110         if (p != NULL)
1111                 kref_put(&p->kref, nfs4_opendata_free);
1112 }
1113
1114 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1115 {
1116         int ret;
1117
1118         ret = rpc_wait_for_completion_task(task);
1119         return ret;
1120 }
1121
1122 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1123 {
1124         int ret = 0;
1125
1126         if (open_mode & (O_EXCL|O_TRUNC))
1127                 goto out;
1128         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1129                 case FMODE_READ:
1130                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1131                                 && state->n_rdonly != 0;
1132                         break;
1133                 case FMODE_WRITE:
1134                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1135                                 && state->n_wronly != 0;
1136                         break;
1137                 case FMODE_READ|FMODE_WRITE:
1138                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1139                                 && state->n_rdwr != 0;
1140         }
1141 out:
1142         return ret;
1143 }
1144
1145 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode)
1146 {
1147         if (delegation == NULL)
1148                 return 0;
1149         if ((delegation->type & fmode) != fmode)
1150                 return 0;
1151         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1152                 return 0;
1153         nfs_mark_delegation_referenced(delegation);
1154         return 1;
1155 }
1156
1157 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1158 {
1159         switch (fmode) {
1160                 case FMODE_WRITE:
1161                         state->n_wronly++;
1162                         break;
1163                 case FMODE_READ:
1164                         state->n_rdonly++;
1165                         break;
1166                 case FMODE_READ|FMODE_WRITE:
1167                         state->n_rdwr++;
1168         }
1169         nfs4_state_set_mode_locked(state, state->state | fmode);
1170 }
1171
1172 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1173 {
1174         struct nfs_client *clp = state->owner->so_server->nfs_client;
1175         bool need_recover = false;
1176
1177         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1178                 need_recover = true;
1179         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1180                 need_recover = true;
1181         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1182                 need_recover = true;
1183         if (need_recover)
1184                 nfs4_state_mark_reclaim_nograce(clp, state);
1185 }
1186
1187 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1188                 nfs4_stateid *stateid)
1189 {
1190         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1191                 return true;
1192         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1193                 nfs_test_and_clear_all_open_stateid(state);
1194                 return true;
1195         }
1196         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1197                 return true;
1198         return false;
1199 }
1200
1201 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1202 {
1203         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1204                 return;
1205         if (state->n_wronly)
1206                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1207         if (state->n_rdonly)
1208                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1209         if (state->n_rdwr)
1210                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1211         set_bit(NFS_OPEN_STATE, &state->flags);
1212 }
1213
1214 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1215                 nfs4_stateid *stateid, fmode_t fmode)
1216 {
1217         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1218         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1219         case FMODE_WRITE:
1220                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1221                 break;
1222         case FMODE_READ:
1223                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1224                 break;
1225         case 0:
1226                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1227                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1228                 clear_bit(NFS_OPEN_STATE, &state->flags);
1229         }
1230         if (stateid == NULL)
1231                 return;
1232         /* Handle races with OPEN */
1233         if (!nfs4_stateid_match_other(stateid, &state->open_stateid) ||
1234             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1235                 nfs_resync_open_stateid_locked(state);
1236                 return;
1237         }
1238         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1239                 nfs4_stateid_copy(&state->stateid, stateid);
1240         nfs4_stateid_copy(&state->open_stateid, stateid);
1241 }
1242
1243 static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1244 {
1245         write_seqlock(&state->seqlock);
1246         nfs_clear_open_stateid_locked(state, stateid, fmode);
1247         write_sequnlock(&state->seqlock);
1248         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1249                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1250 }
1251
1252 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1253 {
1254         switch (fmode) {
1255                 case FMODE_READ:
1256                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1257                         break;
1258                 case FMODE_WRITE:
1259                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1260                         break;
1261                 case FMODE_READ|FMODE_WRITE:
1262                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1263         }
1264         if (!nfs_need_update_open_stateid(state, stateid))
1265                 return;
1266         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1267                 nfs4_stateid_copy(&state->stateid, stateid);
1268         nfs4_stateid_copy(&state->open_stateid, stateid);
1269 }
1270
1271 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1272 {
1273         /*
1274          * Protect the call to nfs4_state_set_mode_locked and
1275          * serialise the stateid update
1276          */
1277         write_seqlock(&state->seqlock);
1278         if (deleg_stateid != NULL) {
1279                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1280                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1281         }
1282         if (open_stateid != NULL)
1283                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1284         write_sequnlock(&state->seqlock);
1285         spin_lock(&state->owner->so_lock);
1286         update_open_stateflags(state, fmode);
1287         spin_unlock(&state->owner->so_lock);
1288 }
1289
1290 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1291 {
1292         struct nfs_inode *nfsi = NFS_I(state->inode);
1293         struct nfs_delegation *deleg_cur;
1294         int ret = 0;
1295
1296         fmode &= (FMODE_READ|FMODE_WRITE);
1297
1298         rcu_read_lock();
1299         deleg_cur = rcu_dereference(nfsi->delegation);
1300         if (deleg_cur == NULL)
1301                 goto no_delegation;
1302
1303         spin_lock(&deleg_cur->lock);
1304         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1305            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1306             (deleg_cur->type & fmode) != fmode)
1307                 goto no_delegation_unlock;
1308
1309         if (delegation == NULL)
1310                 delegation = &deleg_cur->stateid;
1311         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1312                 goto no_delegation_unlock;
1313
1314         nfs_mark_delegation_referenced(deleg_cur);
1315         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1316         ret = 1;
1317 no_delegation_unlock:
1318         spin_unlock(&deleg_cur->lock);
1319 no_delegation:
1320         rcu_read_unlock();
1321
1322         if (!ret && open_stateid != NULL) {
1323                 __update_open_stateid(state, open_stateid, NULL, fmode);
1324                 ret = 1;
1325         }
1326         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1327                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1328
1329         return ret;
1330 }
1331
1332 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1333                 const nfs4_stateid *stateid)
1334 {
1335         struct nfs4_state *state = lsp->ls_state;
1336         bool ret = false;
1337
1338         spin_lock(&state->state_lock);
1339         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1340                 goto out_noupdate;
1341         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1342                 goto out_noupdate;
1343         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1344         ret = true;
1345 out_noupdate:
1346         spin_unlock(&state->state_lock);
1347         return ret;
1348 }
1349
1350 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1351 {
1352         struct nfs_delegation *delegation;
1353
1354         rcu_read_lock();
1355         delegation = rcu_dereference(NFS_I(inode)->delegation);
1356         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1357                 rcu_read_unlock();
1358                 return;
1359         }
1360         rcu_read_unlock();
1361         nfs4_inode_return_delegation(inode);
1362 }
1363
1364 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1365 {
1366         struct nfs4_state *state = opendata->state;
1367         struct nfs_inode *nfsi = NFS_I(state->inode);
1368         struct nfs_delegation *delegation;
1369         int open_mode = opendata->o_arg.open_flags;
1370         fmode_t fmode = opendata->o_arg.fmode;
1371         nfs4_stateid stateid;
1372         int ret = -EAGAIN;
1373
1374         for (;;) {
1375                 spin_lock(&state->owner->so_lock);
1376                 if (can_open_cached(state, fmode, open_mode)) {
1377                         update_open_stateflags(state, fmode);
1378                         spin_unlock(&state->owner->so_lock);
1379                         goto out_return_state;
1380                 }
1381                 spin_unlock(&state->owner->so_lock);
1382                 rcu_read_lock();
1383                 delegation = rcu_dereference(nfsi->delegation);
1384                 if (!can_open_delegated(delegation, fmode)) {
1385                         rcu_read_unlock();
1386                         break;
1387                 }
1388                 /* Save the delegation */
1389                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1390                 rcu_read_unlock();
1391                 nfs_release_seqid(opendata->o_arg.seqid);
1392                 if (!opendata->is_recover) {
1393                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1394                         if (ret != 0)
1395                                 goto out;
1396                 }
1397                 ret = -EAGAIN;
1398
1399                 /* Try to update the stateid using the delegation */
1400                 if (update_open_stateid(state, NULL, &stateid, fmode))
1401                         goto out_return_state;
1402         }
1403 out:
1404         return ERR_PTR(ret);
1405 out_return_state:
1406         atomic_inc(&state->count);
1407         return state;
1408 }
1409
1410 static void
1411 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1412 {
1413         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1414         struct nfs_delegation *delegation;
1415         int delegation_flags = 0;
1416
1417         rcu_read_lock();
1418         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1419         if (delegation)
1420                 delegation_flags = delegation->flags;
1421         rcu_read_unlock();
1422         if (data->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR) {
1423                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1424                                    "returning a delegation for "
1425                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1426                                    clp->cl_hostname);
1427         } else if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1428                 nfs_inode_set_delegation(state->inode,
1429                                          data->owner->so_cred,
1430                                          &data->o_res);
1431         else
1432                 nfs_inode_reclaim_delegation(state->inode,
1433                                              data->owner->so_cred,
1434                                              &data->o_res);
1435 }
1436
1437 /*
1438  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1439  * and update the nfs4_state.
1440  */
1441 static struct nfs4_state *
1442 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1443 {
1444         struct inode *inode = data->state->inode;
1445         struct nfs4_state *state = data->state;
1446         int ret;
1447
1448         if (!data->rpc_done) {
1449                 if (data->rpc_status) {
1450                         ret = data->rpc_status;
1451                         goto err;
1452                 }
1453                 /* cached opens have already been processed */
1454                 goto update;
1455         }
1456
1457         ret = nfs_refresh_inode(inode, &data->f_attr);
1458         if (ret)
1459                 goto err;
1460
1461         if (data->o_res.delegation_type != 0)
1462                 nfs4_opendata_check_deleg(data, state);
1463 update:
1464         update_open_stateid(state, &data->o_res.stateid, NULL,
1465                             data->o_arg.fmode);
1466         atomic_inc(&state->count);
1467
1468         return state;
1469 err:
1470         return ERR_PTR(ret);
1471
1472 }
1473
1474 static struct nfs4_state *
1475 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1476 {
1477         struct inode *inode;
1478         struct nfs4_state *state = NULL;
1479         int ret;
1480
1481         if (!data->rpc_done) {
1482                 state = nfs4_try_open_cached(data);
1483                 goto out;
1484         }
1485
1486         ret = -EAGAIN;
1487         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1488                 goto err;
1489         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1490         ret = PTR_ERR(inode);
1491         if (IS_ERR(inode))
1492                 goto err;
1493         ret = -ENOMEM;
1494         state = nfs4_get_open_state(inode, data->owner);
1495         if (state == NULL)
1496                 goto err_put_inode;
1497         if (data->o_res.delegation_type != 0)
1498                 nfs4_opendata_check_deleg(data, state);
1499         update_open_stateid(state, &data->o_res.stateid, NULL,
1500                         data->o_arg.fmode);
1501         iput(inode);
1502 out:
1503         nfs_release_seqid(data->o_arg.seqid);
1504         return state;
1505 err_put_inode:
1506         iput(inode);
1507 err:
1508         return ERR_PTR(ret);
1509 }
1510
1511 static struct nfs4_state *
1512 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1513 {
1514         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1515                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1516         return _nfs4_opendata_to_nfs4_state(data);
1517 }
1518
1519 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1520 {
1521         struct nfs_inode *nfsi = NFS_I(state->inode);
1522         struct nfs_open_context *ctx;
1523
1524         spin_lock(&state->inode->i_lock);
1525         list_for_each_entry(ctx, &nfsi->open_files, list) {
1526                 if (ctx->state != state)
1527                         continue;
1528                 get_nfs_open_context(ctx);
1529                 spin_unlock(&state->inode->i_lock);
1530                 return ctx;
1531         }
1532         spin_unlock(&state->inode->i_lock);
1533         return ERR_PTR(-ENOENT);
1534 }
1535
1536 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1537                 struct nfs4_state *state, enum open_claim_type4 claim)
1538 {
1539         struct nfs4_opendata *opendata;
1540
1541         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1542                         NULL, NULL, claim, GFP_NOFS);
1543         if (opendata == NULL)
1544                 return ERR_PTR(-ENOMEM);
1545         opendata->state = state;
1546         atomic_inc(&state->count);
1547         return opendata;
1548 }
1549
1550 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, fmode_t fmode, struct nfs4_state **res)
1551 {
1552         struct nfs4_state *newstate;
1553         int ret;
1554
1555         if ((opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
1556              opendata->o_arg.claim == NFS4_OPEN_CLAIM_DELEG_CUR_FH) &&
1557             (opendata->o_arg.u.delegation_type & fmode) != fmode)
1558                 /* This mode can't have been delegated, so we must have
1559                  * a valid open_stateid to cover it - not need to reclaim.
1560                  */
1561                 return 0;
1562         opendata->o_arg.open_flags = 0;
1563         opendata->o_arg.fmode = fmode;
1564         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1565                         NFS_SB(opendata->dentry->d_sb),
1566                         fmode, 0);
1567         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1568         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1569         nfs4_init_opendata_res(opendata);
1570         ret = _nfs4_recover_proc_open(opendata);
1571         if (ret != 0)
1572                 return ret;
1573         newstate = nfs4_opendata_to_nfs4_state(opendata);
1574         if (IS_ERR(newstate))
1575                 return PTR_ERR(newstate);
1576         nfs4_close_state(newstate, fmode);
1577         *res = newstate;
1578         return 0;
1579 }
1580
1581 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1582 {
1583         struct nfs4_state *newstate;
1584         int ret;
1585
1586         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1587         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1588         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1589         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1590         /* memory barrier prior to reading state->n_* */
1591         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1592         clear_bit(NFS_OPEN_STATE, &state->flags);
1593         smp_rmb();
1594         if (state->n_rdwr != 0) {
1595                 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
1596                 if (ret != 0)
1597                         return ret;
1598                 if (newstate != state)
1599                         return -ESTALE;
1600         }
1601         if (state->n_wronly != 0) {
1602                 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
1603                 if (ret != 0)
1604                         return ret;
1605                 if (newstate != state)
1606                         return -ESTALE;
1607         }
1608         if (state->n_rdonly != 0) {
1609                 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
1610                 if (ret != 0)
1611                         return ret;
1612                 if (newstate != state)
1613                         return -ESTALE;
1614         }
1615         /*
1616          * We may have performed cached opens for all three recoveries.
1617          * Check if we need to update the current stateid.
1618          */
1619         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1620             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1621                 write_seqlock(&state->seqlock);
1622                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1623                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1624                 write_sequnlock(&state->seqlock);
1625         }
1626         return 0;
1627 }
1628
1629 /*
1630  * OPEN_RECLAIM:
1631  *      reclaim state on the server after a reboot.
1632  */
1633 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1634 {
1635         struct nfs_delegation *delegation;
1636         struct nfs4_opendata *opendata;
1637         fmode_t delegation_type = 0;
1638         int status;
1639
1640         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1641                         NFS4_OPEN_CLAIM_PREVIOUS);
1642         if (IS_ERR(opendata))
1643                 return PTR_ERR(opendata);
1644         rcu_read_lock();
1645         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1646         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1647                 delegation_type = delegation->type;
1648         rcu_read_unlock();
1649         opendata->o_arg.u.delegation_type = delegation_type;
1650         status = nfs4_open_recover(opendata, state);
1651         nfs4_opendata_put(opendata);
1652         return status;
1653 }
1654
1655 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1656 {
1657         struct nfs_server *server = NFS_SERVER(state->inode);
1658         struct nfs4_exception exception = { };
1659         int err;
1660         do {
1661                 err = _nfs4_do_open_reclaim(ctx, state);
1662                 trace_nfs4_open_reclaim(ctx, 0, err);
1663                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1664                         continue;
1665                 if (err != -NFS4ERR_DELAY)
1666                         break;
1667                 nfs4_handle_exception(server, err, &exception);
1668         } while (exception.retry);
1669         return err;
1670 }
1671
1672 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1673 {
1674         struct nfs_open_context *ctx;
1675         int ret;
1676
1677         ctx = nfs4_state_find_open_context(state);
1678         if (IS_ERR(ctx))
1679                 return -EAGAIN;
1680         ret = nfs4_do_open_reclaim(ctx, state);
1681         put_nfs_open_context(ctx);
1682         return ret;
1683 }
1684
1685 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1686 {
1687         switch (err) {
1688                 default:
1689                         printk(KERN_ERR "NFS: %s: unhandled error "
1690                                         "%d.\n", __func__, err);
1691                 case 0:
1692                 case -ENOENT:
1693                 case -EAGAIN:
1694                 case -ESTALE:
1695                         break;
1696                 case -NFS4ERR_BADSESSION:
1697                 case -NFS4ERR_BADSLOT:
1698                 case -NFS4ERR_BAD_HIGH_SLOT:
1699                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1700                 case -NFS4ERR_DEADSESSION:
1701                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1702                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1703                         return -EAGAIN;
1704                 case -NFS4ERR_STALE_CLIENTID:
1705                 case -NFS4ERR_STALE_STATEID:
1706                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1707                 case -NFS4ERR_EXPIRED:
1708                         /* Don't recall a delegation if it was lost */
1709                         nfs4_schedule_lease_recovery(server->nfs_client);
1710                         return -EAGAIN;
1711                 case -NFS4ERR_MOVED:
1712                         nfs4_schedule_migration_recovery(server);
1713                         return -EAGAIN;
1714                 case -NFS4ERR_LEASE_MOVED:
1715                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1716                         return -EAGAIN;
1717                 case -NFS4ERR_DELEG_REVOKED:
1718                 case -NFS4ERR_ADMIN_REVOKED:
1719                 case -NFS4ERR_BAD_STATEID:
1720                 case -NFS4ERR_OPENMODE:
1721                         nfs_inode_find_state_and_recover(state->inode,
1722                                         stateid);
1723                         nfs4_schedule_stateid_recovery(server, state);
1724                         return -EAGAIN;
1725                 case -NFS4ERR_DELAY:
1726                 case -NFS4ERR_GRACE:
1727                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1728                         ssleep(1);
1729                         return -EAGAIN;
1730                 case -ENOMEM:
1731                 case -NFS4ERR_DENIED:
1732                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1733                         return 0;
1734         }
1735         return err;
1736 }
1737
1738 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
1739 {
1740         struct nfs_server *server = NFS_SERVER(state->inode);
1741         struct nfs4_opendata *opendata;
1742         int err;
1743
1744         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1745                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1746         if (IS_ERR(opendata))
1747                 return PTR_ERR(opendata);
1748         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1749         err = nfs4_open_recover(opendata, state);
1750         nfs4_opendata_put(opendata);
1751         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1752 }
1753
1754 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1755 {
1756         struct nfs4_opendata *data = calldata;
1757
1758         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1759                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1760 }
1761
1762 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1763 {
1764         struct nfs4_opendata *data = calldata;
1765
1766         nfs40_sequence_done(task, &data->c_res.seq_res);
1767
1768         data->rpc_status = task->tk_status;
1769         if (data->rpc_status == 0) {
1770                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1771                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1772                 renew_lease(data->o_res.server, data->timestamp);
1773                 data->rpc_done = 1;
1774         }
1775 }
1776
1777 static void nfs4_open_confirm_release(void *calldata)
1778 {
1779         struct nfs4_opendata *data = calldata;
1780         struct nfs4_state *state = NULL;
1781
1782         /* If this request hasn't been cancelled, do nothing */
1783         if (data->cancelled == 0)
1784                 goto out_free;
1785         /* In case of error, no cleanup! */
1786         if (!data->rpc_done)
1787                 goto out_free;
1788         state = nfs4_opendata_to_nfs4_state(data);
1789         if (!IS_ERR(state))
1790                 nfs4_close_state(state, data->o_arg.fmode);
1791 out_free:
1792         nfs4_opendata_put(data);
1793 }
1794
1795 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1796         .rpc_call_prepare = nfs4_open_confirm_prepare,
1797         .rpc_call_done = nfs4_open_confirm_done,
1798         .rpc_release = nfs4_open_confirm_release,
1799 };
1800
1801 /*
1802  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1803  */
1804 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1805 {
1806         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1807         struct rpc_task *task;
1808         struct  rpc_message msg = {
1809                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1810                 .rpc_argp = &data->c_arg,
1811                 .rpc_resp = &data->c_res,
1812                 .rpc_cred = data->owner->so_cred,
1813         };
1814         struct rpc_task_setup task_setup_data = {
1815                 .rpc_client = server->client,
1816                 .rpc_message = &msg,
1817                 .callback_ops = &nfs4_open_confirm_ops,
1818                 .callback_data = data,
1819                 .workqueue = nfsiod_workqueue,
1820                 .flags = RPC_TASK_ASYNC,
1821         };
1822         int status;
1823
1824         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1825         kref_get(&data->kref);
1826         data->rpc_done = 0;
1827         data->rpc_status = 0;
1828         data->timestamp = jiffies;
1829         task = rpc_run_task(&task_setup_data);
1830         if (IS_ERR(task))
1831                 return PTR_ERR(task);
1832         status = nfs4_wait_for_completion_rpc_task(task);
1833         if (status != 0) {
1834                 data->cancelled = 1;
1835                 smp_wmb();
1836         } else
1837                 status = data->rpc_status;
1838         rpc_put_task(task);
1839         return status;
1840 }
1841
1842 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1843 {
1844         struct nfs4_opendata *data = calldata;
1845         struct nfs4_state_owner *sp = data->owner;
1846         struct nfs_client *clp = sp->so_server->nfs_client;
1847
1848         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1849                 goto out_wait;
1850         /*
1851          * Check if we still need to send an OPEN call, or if we can use
1852          * a delegation instead.
1853          */
1854         if (data->state != NULL) {
1855                 struct nfs_delegation *delegation;
1856
1857                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1858                         goto out_no_action;
1859                 rcu_read_lock();
1860                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1861                 if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR &&
1862                     data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH &&
1863                     can_open_delegated(delegation, data->o_arg.fmode))
1864                         goto unlock_no_action;
1865                 rcu_read_unlock();
1866         }
1867         /* Update client id. */
1868         data->o_arg.clientid = clp->cl_clientid;
1869         switch (data->o_arg.claim) {
1870         case NFS4_OPEN_CLAIM_PREVIOUS:
1871         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1872         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1873                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1874         case NFS4_OPEN_CLAIM_FH:
1875                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1876                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1877         }
1878         data->timestamp = jiffies;
1879         if (nfs4_setup_sequence(data->o_arg.server,
1880                                 &data->o_arg.seq_args,
1881                                 &data->o_res.seq_res,
1882                                 task) != 0)
1883                 nfs_release_seqid(data->o_arg.seqid);
1884
1885         /* Set the create mode (note dependency on the session type) */
1886         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1887         if (data->o_arg.open_flags & O_EXCL) {
1888                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1889                 if (nfs4_has_persistent_session(clp))
1890                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1891                 else if (clp->cl_mvops->minor_version > 0)
1892                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1893         }
1894         return;
1895 unlock_no_action:
1896         rcu_read_unlock();
1897 out_no_action:
1898         task->tk_action = NULL;
1899 out_wait:
1900         nfs4_sequence_done(task, &data->o_res.seq_res);
1901 }
1902
1903 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1904 {
1905         struct nfs4_opendata *data = calldata;
1906
1907         data->rpc_status = task->tk_status;
1908
1909         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1910                 return;
1911
1912         if (task->tk_status == 0) {
1913                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1914                         switch (data->o_res.f_attr->mode & S_IFMT) {
1915                         case S_IFREG:
1916                                 break;
1917                         case S_IFLNK:
1918                                 data->rpc_status = -ELOOP;
1919                                 break;
1920                         case S_IFDIR:
1921                                 data->rpc_status = -EISDIR;
1922                                 break;
1923                         default:
1924                                 data->rpc_status = -ENOTDIR;
1925                         }
1926                 }
1927                 renew_lease(data->o_res.server, data->timestamp);
1928                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1929                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1930         }
1931         data->rpc_done = 1;
1932 }
1933
1934 static void nfs4_open_release(void *calldata)
1935 {
1936         struct nfs4_opendata *data = calldata;
1937         struct nfs4_state *state = NULL;
1938
1939         /* If this request hasn't been cancelled, do nothing */
1940         if (data->cancelled == 0)
1941                 goto out_free;
1942         /* In case of error, no cleanup! */
1943         if (data->rpc_status != 0 || !data->rpc_done)
1944                 goto out_free;
1945         /* In case we need an open_confirm, no cleanup! */
1946         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1947                 goto out_free;
1948         state = nfs4_opendata_to_nfs4_state(data);
1949         if (!IS_ERR(state))
1950                 nfs4_close_state(state, data->o_arg.fmode);
1951 out_free:
1952         nfs4_opendata_put(data);
1953 }
1954
1955 static const struct rpc_call_ops nfs4_open_ops = {
1956         .rpc_call_prepare = nfs4_open_prepare,
1957         .rpc_call_done = nfs4_open_done,
1958         .rpc_release = nfs4_open_release,
1959 };
1960
1961 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
1962 {
1963         struct inode *dir = d_inode(data->dir);
1964         struct nfs_server *server = NFS_SERVER(dir);
1965         struct nfs_openargs *o_arg = &data->o_arg;
1966         struct nfs_openres *o_res = &data->o_res;
1967         struct rpc_task *task;
1968         struct rpc_message msg = {
1969                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
1970                 .rpc_argp = o_arg,
1971                 .rpc_resp = o_res,
1972                 .rpc_cred = data->owner->so_cred,
1973         };
1974         struct rpc_task_setup task_setup_data = {
1975                 .rpc_client = server->client,
1976                 .rpc_message = &msg,
1977                 .callback_ops = &nfs4_open_ops,
1978                 .callback_data = data,
1979                 .workqueue = nfsiod_workqueue,
1980                 .flags = RPC_TASK_ASYNC,
1981         };
1982         int status;
1983
1984         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
1985         kref_get(&data->kref);
1986         data->rpc_done = 0;
1987         data->rpc_status = 0;
1988         data->cancelled = 0;
1989         data->is_recover = 0;
1990         if (isrecover) {
1991                 nfs4_set_sequence_privileged(&o_arg->seq_args);
1992                 data->is_recover = 1;
1993         }
1994         task = rpc_run_task(&task_setup_data);
1995         if (IS_ERR(task))
1996                 return PTR_ERR(task);
1997         status = nfs4_wait_for_completion_rpc_task(task);
1998         if (status != 0) {
1999                 data->cancelled = 1;
2000                 smp_wmb();
2001         } else
2002                 status = data->rpc_status;
2003         rpc_put_task(task);
2004
2005         return status;
2006 }
2007
2008 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2009 {
2010         struct inode *dir = d_inode(data->dir);
2011         struct nfs_openres *o_res = &data->o_res;
2012         int status;
2013
2014         status = nfs4_run_open_task(data, 1);
2015         if (status != 0 || !data->rpc_done)
2016                 return status;
2017
2018         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2019
2020         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2021                 status = _nfs4_proc_open_confirm(data);
2022                 if (status != 0)
2023                         return status;
2024         }
2025
2026         return status;
2027 }
2028
2029 /*
2030  * Additional permission checks in order to distinguish between an
2031  * open for read, and an open for execute. This works around the
2032  * fact that NFSv4 OPEN treats read and execute permissions as being
2033  * the same.
2034  * Note that in the non-execute case, we want to turn off permission
2035  * checking if we just created a new file (POSIX open() semantics).
2036  */
2037 static int nfs4_opendata_access(struct rpc_cred *cred,
2038                                 struct nfs4_opendata *opendata,
2039                                 struct nfs4_state *state, fmode_t fmode,
2040                                 int openflags)
2041 {
2042         struct nfs_access_entry cache;
2043         u32 mask;
2044
2045         /* access call failed or for some reason the server doesn't
2046          * support any access modes -- defer access call until later */
2047         if (opendata->o_res.access_supported == 0)
2048                 return 0;
2049
2050         mask = 0;
2051         /*
2052          * Use openflags to check for exec, because fmode won't
2053          * always have FMODE_EXEC set when file open for exec.
2054          */
2055         if (openflags & __FMODE_EXEC) {
2056                 /* ONLY check for exec rights */
2057                 mask = MAY_EXEC;
2058         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2059                 mask = MAY_READ;
2060
2061         cache.cred = cred;
2062         cache.jiffies = jiffies;
2063         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2064         nfs_access_add_cache(state->inode, &cache);
2065
2066         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2067                 return 0;
2068
2069         /* even though OPEN succeeded, access is denied. Close the file */
2070         nfs4_close_state(state, fmode);
2071         return -EACCES;
2072 }
2073
2074 /*
2075  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2076  */
2077 static int _nfs4_proc_open(struct nfs4_opendata *data)
2078 {
2079         struct inode *dir = d_inode(data->dir);
2080         struct nfs_server *server = NFS_SERVER(dir);
2081         struct nfs_openargs *o_arg = &data->o_arg;
2082         struct nfs_openres *o_res = &data->o_res;
2083         int status;
2084
2085         status = nfs4_run_open_task(data, 0);
2086         if (!data->rpc_done)
2087                 return status;
2088         if (status != 0) {
2089                 if (status == -NFS4ERR_BADNAME &&
2090                                 !(o_arg->open_flags & O_CREAT))
2091                         return -ENOENT;
2092                 return status;
2093         }
2094
2095         nfs_fattr_map_and_free_names(server, &data->f_attr);
2096
2097         if (o_arg->open_flags & O_CREAT) {
2098                 update_changeattr(dir, &o_res->cinfo);
2099                 if (o_arg->open_flags & O_EXCL)
2100                         data->file_created = 1;
2101                 else if (o_res->cinfo.before != o_res->cinfo.after)
2102                         data->file_created = 1;
2103         }
2104         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2105                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2106         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2107                 status = _nfs4_proc_open_confirm(data);
2108                 if (status != 0)
2109                         return status;
2110         }
2111         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2112                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2113         return 0;
2114 }
2115
2116 static int nfs4_recover_expired_lease(struct nfs_server *server)
2117 {
2118         return nfs4_client_recover_expired_lease(server->nfs_client);
2119 }
2120
2121 /*
2122  * OPEN_EXPIRED:
2123  *      reclaim state on the server after a network partition.
2124  *      Assumes caller holds the appropriate lock
2125  */
2126 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2127 {
2128         struct nfs4_opendata *opendata;
2129         int ret;
2130
2131         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2132                         NFS4_OPEN_CLAIM_FH);
2133         if (IS_ERR(opendata))
2134                 return PTR_ERR(opendata);
2135         ret = nfs4_open_recover(opendata, state);
2136         if (ret == -ESTALE)
2137                 d_drop(ctx->dentry);
2138         nfs4_opendata_put(opendata);
2139         return ret;
2140 }
2141
2142 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2143 {
2144         struct nfs_server *server = NFS_SERVER(state->inode);
2145         struct nfs4_exception exception = { };
2146         int err;
2147
2148         do {
2149                 err = _nfs4_open_expired(ctx, state);
2150                 trace_nfs4_open_expired(ctx, 0, err);
2151                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2152                         continue;
2153                 switch (err) {
2154                 default:
2155                         goto out;
2156                 case -NFS4ERR_GRACE:
2157                 case -NFS4ERR_DELAY:
2158                         nfs4_handle_exception(server, err, &exception);
2159                         err = 0;
2160                 }
2161         } while (exception.retry);
2162 out:
2163         return err;
2164 }
2165
2166 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2167 {
2168         struct nfs_open_context *ctx;
2169         int ret;
2170
2171         ctx = nfs4_state_find_open_context(state);
2172         if (IS_ERR(ctx))
2173                 return -EAGAIN;
2174         ret = nfs4_do_open_expired(ctx, state);
2175         put_nfs_open_context(ctx);
2176         return ret;
2177 }
2178
2179 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2180 {
2181         nfs_remove_bad_delegation(state->inode);
2182         write_seqlock(&state->seqlock);
2183         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2184         write_sequnlock(&state->seqlock);
2185         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2186 }
2187
2188 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2189 {
2190         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2191                 nfs_finish_clear_delegation_stateid(state);
2192 }
2193
2194 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2195 {
2196         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2197         nfs40_clear_delegation_stateid(state);
2198         return nfs4_open_expired(sp, state);
2199 }
2200
2201 #if defined(CONFIG_NFS_V4_1)
2202 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2203 {
2204         struct nfs_server *server = NFS_SERVER(state->inode);
2205         nfs4_stateid stateid;
2206         struct nfs_delegation *delegation;
2207         struct rpc_cred *cred;
2208         int status;
2209
2210         /* Get the delegation credential for use by test/free_stateid */
2211         rcu_read_lock();
2212         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2213         if (delegation == NULL) {
2214                 rcu_read_unlock();
2215                 return;
2216         }
2217
2218         nfs4_stateid_copy(&stateid, &delegation->stateid);
2219         cred = get_rpccred(delegation->cred);
2220         rcu_read_unlock();
2221         status = nfs41_test_stateid(server, &stateid, cred);
2222         trace_nfs4_test_delegation_stateid(state, NULL, status);
2223
2224         if (status != NFS_OK) {
2225                 /* Free the stateid unless the server explicitly
2226                  * informs us the stateid is unrecognized. */
2227                 if (status != -NFS4ERR_BAD_STATEID)
2228                         nfs41_free_stateid(server, &stateid, cred);
2229                 nfs_finish_clear_delegation_stateid(state);
2230         }
2231
2232         put_rpccred(cred);
2233 }
2234
2235 /**
2236  * nfs41_check_open_stateid - possibly free an open stateid
2237  *
2238  * @state: NFSv4 state for an inode
2239  *
2240  * Returns NFS_OK if recovery for this stateid is now finished.
2241  * Otherwise a negative NFS4ERR value is returned.
2242  */
2243 static int nfs41_check_open_stateid(struct nfs4_state *state)
2244 {
2245         struct nfs_server *server = NFS_SERVER(state->inode);
2246         nfs4_stateid *stateid = &state->open_stateid;
2247         struct rpc_cred *cred = state->owner->so_cred;
2248         int status;
2249
2250         /* If a state reset has been done, test_stateid is unneeded */
2251         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2252             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2253             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2254                 return -NFS4ERR_BAD_STATEID;
2255
2256         status = nfs41_test_stateid(server, stateid, cred);
2257         trace_nfs4_test_open_stateid(state, NULL, status);
2258         if (status != NFS_OK) {
2259                 /* Free the stateid unless the server explicitly
2260                  * informs us the stateid is unrecognized. */
2261                 if (status != -NFS4ERR_BAD_STATEID)
2262                         nfs41_free_stateid(server, stateid, cred);
2263
2264                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2265                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2266                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2267                 clear_bit(NFS_OPEN_STATE, &state->flags);
2268         }
2269         return status;
2270 }
2271
2272 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2273 {
2274         int status;
2275
2276         nfs41_check_delegation_stateid(state);
2277         status = nfs41_check_open_stateid(state);
2278         if (status != NFS_OK)
2279                 status = nfs4_open_expired(sp, state);
2280         return status;
2281 }
2282 #endif
2283
2284 /*
2285  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2286  * fields corresponding to attributes that were used to store the verifier.
2287  * Make sure we clobber those fields in the later setattr call
2288  */
2289 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
2290 {
2291         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2292             !(sattr->ia_valid & ATTR_ATIME_SET))
2293                 sattr->ia_valid |= ATTR_ATIME;
2294
2295         if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2296             !(sattr->ia_valid & ATTR_MTIME_SET))
2297                 sattr->ia_valid |= ATTR_MTIME;
2298 }
2299
2300 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2301                 fmode_t fmode,
2302                 int flags,
2303                 struct nfs_open_context *ctx)
2304 {
2305         struct nfs4_state_owner *sp = opendata->owner;
2306         struct nfs_server *server = sp->so_server;
2307         struct dentry *dentry;
2308         struct nfs4_state *state;
2309         unsigned int seq;
2310         int ret;
2311
2312         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2313
2314         ret = _nfs4_proc_open(opendata);
2315         if (ret != 0)
2316                 goto out;
2317
2318         state = nfs4_opendata_to_nfs4_state(opendata);
2319         ret = PTR_ERR(state);
2320         if (IS_ERR(state))
2321                 goto out;
2322         if (server->caps & NFS_CAP_POSIX_LOCK)
2323                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2324
2325         dentry = opendata->dentry;
2326         if (d_really_is_negative(dentry)) {
2327                 /* FIXME: Is this d_drop() ever needed? */
2328                 d_drop(dentry);
2329                 dentry = d_add_unique(dentry, igrab(state->inode));
2330                 if (dentry == NULL) {
2331                         dentry = opendata->dentry;
2332                 } else if (dentry != ctx->dentry) {
2333                         dput(ctx->dentry);
2334                         ctx->dentry = dget(dentry);
2335                 }
2336                 nfs_set_verifier(dentry,
2337                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2338         }
2339
2340         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2341         if (ret != 0)
2342                 goto out;
2343
2344         ctx->state = state;
2345         if (d_inode(dentry) == state->inode) {
2346                 nfs_inode_attach_open_context(ctx);
2347                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2348                         nfs4_schedule_stateid_recovery(server, state);
2349         }
2350 out:
2351         return ret;
2352 }
2353
2354 /*
2355  * Returns a referenced nfs4_state
2356  */
2357 static int _nfs4_do_open(struct inode *dir,
2358                         struct nfs_open_context *ctx,
2359                         int flags,
2360                         struct iattr *sattr,
2361                         struct nfs4_label *label,
2362                         int *opened)
2363 {
2364         struct nfs4_state_owner  *sp;
2365         struct nfs4_state     *state = NULL;
2366         struct nfs_server       *server = NFS_SERVER(dir);
2367         struct nfs4_opendata *opendata;
2368         struct dentry *dentry = ctx->dentry;
2369         struct rpc_cred *cred = ctx->cred;
2370         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2371         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2372         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2373         struct nfs4_label *olabel = NULL;
2374         int status;
2375
2376         /* Protect against reboot recovery conflicts */
2377         status = -ENOMEM;
2378         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2379         if (sp == NULL) {
2380                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2381                 goto out_err;
2382         }
2383         status = nfs4_recover_expired_lease(server);
2384         if (status != 0)
2385                 goto err_put_state_owner;
2386         if (d_really_is_positive(dentry))
2387                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2388         status = -ENOMEM;
2389         if (d_really_is_positive(dentry))
2390                 claim = NFS4_OPEN_CLAIM_FH;
2391         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2392                         label, claim, GFP_KERNEL);
2393         if (opendata == NULL)
2394                 goto err_put_state_owner;
2395
2396         if (label) {
2397                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2398                 if (IS_ERR(olabel)) {
2399                         status = PTR_ERR(olabel);
2400                         goto err_opendata_put;
2401                 }
2402         }
2403
2404         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2405                 if (!opendata->f_attr.mdsthreshold) {
2406                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2407                         if (!opendata->f_attr.mdsthreshold)
2408                                 goto err_free_label;
2409                 }
2410                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2411         }
2412         if (d_really_is_positive(dentry))
2413                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2414
2415         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2416         if (status != 0)
2417                 goto err_free_label;
2418         state = ctx->state;
2419
2420         if ((opendata->o_arg.open_flags & O_EXCL) &&
2421             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2422                 nfs4_exclusive_attrset(opendata, sattr);
2423
2424                 nfs_fattr_init(opendata->o_res.f_attr);
2425                 status = nfs4_do_setattr(state->inode, cred,
2426                                 opendata->o_res.f_attr, sattr,
2427                                 state, label, olabel);
2428                 if (status == 0) {
2429                         nfs_setattr_update_inode(state->inode, sattr,
2430                                         opendata->o_res.f_attr);
2431                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2432                 }
2433         }
2434         if (opendata->file_created)
2435                 *opened |= FILE_CREATED;
2436
2437         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2438                 *ctx_th = opendata->f_attr.mdsthreshold;
2439                 opendata->f_attr.mdsthreshold = NULL;
2440         }
2441
2442         nfs4_label_free(olabel);
2443
2444         nfs4_opendata_put(opendata);
2445         nfs4_put_state_owner(sp);
2446         return 0;
2447 err_free_label:
2448         nfs4_label_free(olabel);
2449 err_opendata_put:
2450         nfs4_opendata_put(opendata);
2451 err_put_state_owner:
2452         nfs4_put_state_owner(sp);
2453 out_err:
2454         return status;
2455 }
2456
2457
2458 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2459                                         struct nfs_open_context *ctx,
2460                                         int flags,
2461                                         struct iattr *sattr,
2462                                         struct nfs4_label *label,
2463                                         int *opened)
2464 {
2465         struct nfs_server *server = NFS_SERVER(dir);
2466         struct nfs4_exception exception = { };
2467         struct nfs4_state *res;
2468         int status;
2469
2470         do {
2471                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2472                 res = ctx->state;
2473                 trace_nfs4_open_file(ctx, flags, status);
2474                 if (status == 0)
2475                         break;
2476                 /* NOTE: BAD_SEQID means the server and client disagree about the
2477                  * book-keeping w.r.t. state-changing operations
2478                  * (OPEN/CLOSE/LOCK/LOCKU...)
2479                  * It is actually a sign of a bug on the client or on the server.
2480                  *
2481                  * If we receive a BAD_SEQID error in the particular case of
2482                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2483                  * have unhashed the old state_owner for us, and that we can
2484                  * therefore safely retry using a new one. We should still warn
2485                  * the user though...
2486                  */
2487                 if (status == -NFS4ERR_BAD_SEQID) {
2488                         pr_warn_ratelimited("NFS: v4 server %s "
2489                                         " returned a bad sequence-id error!\n",
2490                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2491                         exception.retry = 1;
2492                         continue;
2493                 }
2494                 /*
2495                  * BAD_STATEID on OPEN means that the server cancelled our
2496                  * state before it received the OPEN_CONFIRM.
2497                  * Recover by retrying the request as per the discussion
2498                  * on Page 181 of RFC3530.
2499                  */
2500                 if (status == -NFS4ERR_BAD_STATEID) {
2501                         exception.retry = 1;
2502                         continue;
2503                 }
2504                 if (status == -EAGAIN) {
2505                         /* We must have found a delegation */
2506                         exception.retry = 1;
2507                         continue;
2508                 }
2509                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2510                         continue;
2511                 res = ERR_PTR(nfs4_handle_exception(server,
2512                                         status, &exception));
2513         } while (exception.retry);
2514         return res;
2515 }
2516
2517 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2518                             struct nfs_fattr *fattr, struct iattr *sattr,
2519                             struct nfs4_state *state, struct nfs4_label *ilabel,
2520                             struct nfs4_label *olabel)
2521 {
2522         struct nfs_server *server = NFS_SERVER(inode);
2523         struct nfs_setattrargs  arg = {
2524                 .fh             = NFS_FH(inode),
2525                 .iap            = sattr,
2526                 .server         = server,
2527                 .bitmask = server->attr_bitmask,
2528                 .label          = ilabel,
2529         };
2530         struct nfs_setattrres  res = {
2531                 .fattr          = fattr,
2532                 .label          = olabel,
2533                 .server         = server,
2534         };
2535         struct rpc_message msg = {
2536                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2537                 .rpc_argp       = &arg,
2538                 .rpc_resp       = &res,
2539                 .rpc_cred       = cred,
2540         };
2541         unsigned long timestamp = jiffies;
2542         fmode_t fmode;
2543         bool truncate;
2544         int status;
2545
2546         arg.bitmask = nfs4_bitmask(server, ilabel);
2547         if (ilabel)
2548                 arg.bitmask = nfs4_bitmask(server, olabel);
2549
2550         nfs_fattr_init(fattr);
2551
2552         /* Servers should only apply open mode checks for file size changes */
2553         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2554         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2555
2556         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2557                 /* Use that stateid */
2558         } else if (truncate && state != NULL) {
2559                 struct nfs_lockowner lockowner = {
2560                         .l_owner = current->files,
2561                         .l_pid = current->tgid,
2562                 };
2563                 if (!nfs4_valid_open_stateid(state))
2564                         return -EBADF;
2565                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2566                                 &lockowner) == -EIO)
2567                         return -EBADF;
2568         } else
2569                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2570
2571         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2572         if (status == 0 && state != NULL)
2573                 renew_lease(server, timestamp);
2574         return status;
2575 }
2576
2577 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2578                            struct nfs_fattr *fattr, struct iattr *sattr,
2579                            struct nfs4_state *state, struct nfs4_label *ilabel,
2580                            struct nfs4_label *olabel)
2581 {
2582         struct nfs_server *server = NFS_SERVER(inode);
2583         struct nfs4_exception exception = {
2584                 .state = state,
2585                 .inode = inode,
2586         };
2587         int err;
2588         do {
2589                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2590                 trace_nfs4_setattr(inode, err);
2591                 switch (err) {
2592                 case -NFS4ERR_OPENMODE:
2593                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2594                                 pr_warn_once("NFSv4: server %s is incorrectly "
2595                                                 "applying open mode checks to "
2596                                                 "a SETATTR that is not "
2597                                                 "changing file size.\n",
2598                                                 server->nfs_client->cl_hostname);
2599                         }
2600                         if (state && !(state->state & FMODE_WRITE)) {
2601                                 err = -EBADF;
2602                                 if (sattr->ia_valid & ATTR_OPEN)
2603                                         err = -EACCES;
2604                                 goto out;
2605                         }
2606                 }
2607                 err = nfs4_handle_exception(server, err, &exception);
2608         } while (exception.retry);
2609 out:
2610         return err;
2611 }
2612
2613 struct nfs4_closedata {
2614         struct inode *inode;
2615         struct nfs4_state *state;
2616         struct nfs_closeargs arg;
2617         struct nfs_closeres res;
2618         struct nfs_fattr fattr;
2619         unsigned long timestamp;
2620         bool roc;
2621         u32 roc_barrier;
2622 };
2623
2624 static void nfs4_free_closedata(void *data)
2625 {
2626         struct nfs4_closedata *calldata = data;
2627         struct nfs4_state_owner *sp = calldata->state->owner;
2628         struct super_block *sb = calldata->state->inode->i_sb;
2629
2630         if (calldata->roc)
2631                 pnfs_roc_release(calldata->state->inode);
2632         nfs4_put_open_state(calldata->state);
2633         nfs_free_seqid(calldata->arg.seqid);
2634         nfs4_put_state_owner(sp);
2635         nfs_sb_deactive(sb);
2636         kfree(calldata);
2637 }
2638
2639 static void nfs4_close_done(struct rpc_task *task, void *data)
2640 {
2641         struct nfs4_closedata *calldata = data;
2642         struct nfs4_state *state = calldata->state;
2643         struct nfs_server *server = NFS_SERVER(calldata->inode);
2644         nfs4_stateid *res_stateid = NULL;
2645
2646         dprintk("%s: begin!\n", __func__);
2647         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2648                 return;
2649         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2650         /* hmm. we are done with the inode, and in the process of freeing
2651          * the state_owner. we keep this around to process errors
2652          */
2653         switch (task->tk_status) {
2654                 case 0:
2655                         res_stateid = &calldata->res.stateid;
2656                         if (calldata->arg.fmode == 0 && calldata->roc)
2657                                 pnfs_roc_set_barrier(state->inode,
2658                                                      calldata->roc_barrier);
2659                         renew_lease(server, calldata->timestamp);
2660                         break;
2661                 case -NFS4ERR_ADMIN_REVOKED:
2662                 case -NFS4ERR_STALE_STATEID:
2663                 case -NFS4ERR_OLD_STATEID:
2664                 case -NFS4ERR_BAD_STATEID:
2665                 case -NFS4ERR_EXPIRED:
2666                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2667                                                 &state->open_stateid)) {
2668                                 rpc_restart_call_prepare(task);
2669                                 goto out_release;
2670                         }
2671                         if (calldata->arg.fmode == 0)
2672                                 break;
2673                 default:
2674                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2675                                 rpc_restart_call_prepare(task);
2676                                 goto out_release;
2677                         }
2678         }
2679         nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode);
2680 out_release:
2681         nfs_release_seqid(calldata->arg.seqid);
2682         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2683         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2684 }
2685
2686 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2687 {
2688         struct nfs4_closedata *calldata = data;
2689         struct nfs4_state *state = calldata->state;
2690         struct inode *inode = calldata->inode;
2691         bool is_rdonly, is_wronly, is_rdwr;
2692         int call_close = 0;
2693
2694         dprintk("%s: begin!\n", __func__);
2695         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2696                 goto out_wait;
2697
2698         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2699         spin_lock(&state->owner->so_lock);
2700         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2701         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2702         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2703         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2704         /* Calculate the change in open mode */
2705         calldata->arg.fmode = 0;
2706         if (state->n_rdwr == 0) {
2707                 if (state->n_rdonly == 0)
2708                         call_close |= is_rdonly;
2709                 else if (is_rdonly)
2710                         calldata->arg.fmode |= FMODE_READ;
2711                 if (state->n_wronly == 0)
2712                         call_close |= is_wronly;
2713                 else if (is_wronly)
2714                         calldata->arg.fmode |= FMODE_WRITE;
2715         } else if (is_rdwr)
2716                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2717
2718         if (calldata->arg.fmode == 0)
2719                 call_close |= is_rdwr;
2720
2721         if (!nfs4_valid_open_stateid(state))
2722                 call_close = 0;
2723         spin_unlock(&state->owner->so_lock);
2724
2725         if (!call_close) {
2726                 /* Note: exit _without_ calling nfs4_close_done */
2727                 goto out_no_action;
2728         }
2729
2730         if (calldata->arg.fmode == 0) {
2731                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2732                 if (calldata->roc &&
2733                     pnfs_roc_drain(inode, &calldata->roc_barrier, task)) {
2734                         nfs_release_seqid(calldata->arg.seqid);
2735                         goto out_wait;
2736                     }
2737         }
2738         calldata->arg.share_access =
2739                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2740                                 calldata->arg.fmode, 0);
2741
2742         nfs_fattr_init(calldata->res.fattr);
2743         calldata->timestamp = jiffies;
2744         if (nfs4_setup_sequence(NFS_SERVER(inode),
2745                                 &calldata->arg.seq_args,
2746                                 &calldata->res.seq_res,
2747                                 task) != 0)
2748                 nfs_release_seqid(calldata->arg.seqid);
2749         dprintk("%s: done!\n", __func__);
2750         return;
2751 out_no_action:
2752         task->tk_action = NULL;
2753 out_wait:
2754         nfs4_sequence_done(task, &calldata->res.seq_res);
2755 }
2756
2757 static const struct rpc_call_ops nfs4_close_ops = {
2758         .rpc_call_prepare = nfs4_close_prepare,
2759         .rpc_call_done = nfs4_close_done,
2760         .rpc_release = nfs4_free_closedata,
2761 };
2762
2763 static bool nfs4_roc(struct inode *inode)
2764 {
2765         if (!nfs_have_layout(inode))
2766                 return false;
2767         return pnfs_roc(inode);
2768 }
2769
2770 /*
2771  * It is possible for data to be read/written from a mem-mapped file
2772  * after the sys_close call (which hits the vfs layer as a flush).
2773  * This means that we can't safely call nfsv4 close on a file until
2774  * the inode is cleared. This in turn means that we are not good
2775  * NFSv4 citizens - we do not indicate to the server to update the file's
2776  * share state even when we are done with one of the three share
2777  * stateid's in the inode.
2778  *
2779  * NOTE: Caller must be holding the sp->so_owner semaphore!
2780  */
2781 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2782 {
2783         struct nfs_server *server = NFS_SERVER(state->inode);
2784         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2785         struct nfs4_closedata *calldata;
2786         struct nfs4_state_owner *sp = state->owner;
2787         struct rpc_task *task;
2788         struct rpc_message msg = {
2789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2790                 .rpc_cred = state->owner->so_cred,
2791         };
2792         struct rpc_task_setup task_setup_data = {
2793                 .rpc_client = server->client,
2794                 .rpc_message = &msg,
2795                 .callback_ops = &nfs4_close_ops,
2796                 .workqueue = nfsiod_workqueue,
2797                 .flags = RPC_TASK_ASYNC,
2798         };
2799         int status = -ENOMEM;
2800
2801         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2802                 &task_setup_data.rpc_client, &msg);
2803
2804         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2805         if (calldata == NULL)
2806                 goto out;
2807         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2808         calldata->inode = state->inode;
2809         calldata->state = state;
2810         calldata->arg.fh = NFS_FH(state->inode);
2811         /* Serialization for the sequence id */
2812         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2813         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2814         if (IS_ERR(calldata->arg.seqid))
2815                 goto out_free_calldata;
2816         calldata->arg.fmode = 0;
2817         calldata->arg.bitmask = server->cache_consistency_bitmask;
2818         calldata->res.fattr = &calldata->fattr;
2819         calldata->res.seqid = calldata->arg.seqid;
2820         calldata->res.server = server;
2821         calldata->roc = nfs4_roc(state->inode);
2822         nfs_sb_active(calldata->inode->i_sb);
2823
2824         msg.rpc_argp = &calldata->arg;
2825         msg.rpc_resp = &calldata->res;
2826         task_setup_data.callback_data = calldata;
2827         task = rpc_run_task(&task_setup_data);
2828         if (IS_ERR(task))
2829                 return PTR_ERR(task);
2830         status = 0;
2831         if (wait)
2832                 status = rpc_wait_for_completion_task(task);
2833         rpc_put_task(task);
2834         return status;
2835 out_free_calldata:
2836         kfree(calldata);
2837 out:
2838         nfs4_put_open_state(state);
2839         nfs4_put_state_owner(sp);
2840         return status;
2841 }
2842
2843 static struct inode *
2844 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2845                 int open_flags, struct iattr *attr, int *opened)
2846 {
2847         struct nfs4_state *state;
2848         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2849
2850         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2851
2852         /* Protect against concurrent sillydeletes */
2853         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2854
2855         nfs4_label_release_security(label);
2856
2857         if (IS_ERR(state))
2858                 return ERR_CAST(state);
2859         return state->inode;
2860 }
2861
2862 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2863 {
2864         if (ctx->state == NULL)
2865                 return;
2866         if (is_sync)
2867                 nfs4_close_sync(ctx->state, ctx->mode);
2868         else
2869                 nfs4_close_state(ctx->state, ctx->mode);
2870 }
2871
2872 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2873 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2874 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2875
2876 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2877 {
2878         struct nfs4_server_caps_arg args = {
2879                 .fhandle = fhandle,
2880         };
2881         struct nfs4_server_caps_res res = {};
2882         struct rpc_message msg = {
2883                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2884                 .rpc_argp = &args,
2885                 .rpc_resp = &res,
2886         };
2887         int status;
2888
2889         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2890         if (status == 0) {
2891                 /* Sanity check the server answers */
2892                 switch (server->nfs_client->cl_minorversion) {
2893                 case 0:
2894                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2895                         res.attr_bitmask[2] = 0;
2896                         break;
2897                 case 1:
2898                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2899                         break;
2900                 case 2:
2901                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2902                 }
2903                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2904                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2905                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2906                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2907                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2908                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2909                                 NFS_CAP_SECURITY_LABEL);
2910                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2911                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2912                         server->caps |= NFS_CAP_ACLS;
2913                 if (res.has_links != 0)
2914                         server->caps |= NFS_CAP_HARDLINKS;
2915                 if (res.has_symlinks != 0)
2916                         server->caps |= NFS_CAP_SYMLINKS;
2917                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2918                         server->caps |= NFS_CAP_FILEID;
2919                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
2920                         server->caps |= NFS_CAP_MODE;
2921                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
2922                         server->caps |= NFS_CAP_NLINK;
2923                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
2924                         server->caps |= NFS_CAP_OWNER;
2925                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
2926                         server->caps |= NFS_CAP_OWNER_GROUP;
2927                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
2928                         server->caps |= NFS_CAP_ATIME;
2929                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
2930                         server->caps |= NFS_CAP_CTIME;
2931                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
2932                         server->caps |= NFS_CAP_MTIME;
2933 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
2934                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
2935                         server->caps |= NFS_CAP_SECURITY_LABEL;
2936 #endif
2937                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
2938                                 sizeof(server->attr_bitmask));
2939                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
2940
2941                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
2942                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
2943                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
2944                 server->cache_consistency_bitmask[2] = 0;
2945                 server->acl_bitmask = res.acl_bitmask;
2946                 server->fh_expire_type = res.fh_expire_type;
2947         }
2948
2949         return status;
2950 }
2951
2952 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2953 {
2954         struct nfs4_exception exception = { };
2955         int err;
2956         do {
2957                 err = nfs4_handle_exception(server,
2958                                 _nfs4_server_capabilities(server, fhandle),
2959                                 &exception);
2960         } while (exception.retry);
2961         return err;
2962 }
2963
2964 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2965                 struct nfs_fsinfo *info)
2966 {
2967         u32 bitmask[3];
2968         struct nfs4_lookup_root_arg args = {
2969                 .bitmask = bitmask,
2970         };
2971         struct nfs4_lookup_res res = {
2972                 .server = server,
2973                 .fattr = info->fattr,
2974                 .fh = fhandle,
2975         };
2976         struct rpc_message msg = {
2977                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
2978                 .rpc_argp = &args,
2979                 .rpc_resp = &res,
2980         };
2981
2982         bitmask[0] = nfs4_fattr_bitmap[0];
2983         bitmask[1] = nfs4_fattr_bitmap[1];
2984         /*
2985          * Process the label in the upcoming getfattr
2986          */
2987         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
2988
2989         nfs_fattr_init(info->fattr);
2990         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2991 }
2992
2993 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
2994                 struct nfs_fsinfo *info)
2995 {
2996         struct nfs4_exception exception = { };
2997         int err;
2998         do {
2999                 err = _nfs4_lookup_root(server, fhandle, info);
3000                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3001                 switch (err) {
3002                 case 0:
3003                 case -NFS4ERR_WRONGSEC:
3004                         goto out;
3005                 default:
3006                         err = nfs4_handle_exception(server, err, &exception);
3007                 }
3008         } while (exception.retry);
3009 out:
3010         return err;
3011 }
3012
3013 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3014                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3015 {
3016         struct rpc_auth_create_args auth_args = {
3017                 .pseudoflavor = flavor,
3018         };
3019         struct rpc_auth *auth;
3020         int ret;
3021
3022         auth = rpcauth_create(&auth_args, server->client);
3023         if (IS_ERR(auth)) {
3024                 ret = -EACCES;
3025                 goto out;
3026         }
3027         ret = nfs4_lookup_root(server, fhandle, info);
3028 out:
3029         return ret;
3030 }
3031
3032 /*
3033  * Retry pseudoroot lookup with various security flavors.  We do this when:
3034  *
3035  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3036  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3037  *
3038  * Returns zero on success, or a negative NFS4ERR value, or a
3039  * negative errno value.
3040  */
3041 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3042                               struct nfs_fsinfo *info)
3043 {
3044         /* Per 3530bis 15.33.5 */
3045         static const rpc_authflavor_t flav_array[] = {
3046                 RPC_AUTH_GSS_KRB5P,
3047                 RPC_AUTH_GSS_KRB5I,
3048                 RPC_AUTH_GSS_KRB5,
3049                 RPC_AUTH_UNIX,                  /* courtesy */
3050                 RPC_AUTH_NULL,
3051         };
3052         int status = -EPERM;
3053         size_t i;
3054
3055         if (server->auth_info.flavor_len > 0) {
3056                 /* try each flavor specified by user */
3057                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3058                         status = nfs4_lookup_root_sec(server, fhandle, info,
3059                                                 server->auth_info.flavors[i]);
3060                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3061                                 continue;
3062                         break;
3063                 }
3064         } else {
3065                 /* no flavors specified by user, try default list */
3066                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3067                         status = nfs4_lookup_root_sec(server, fhandle, info,
3068                                                       flav_array[i]);
3069                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3070                                 continue;
3071                         break;
3072                 }
3073         }
3074
3075         /*
3076          * -EACCESS could mean that the user doesn't have correct permissions
3077          * to access the mount.  It could also mean that we tried to mount
3078          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3079          * existing mount programs don't handle -EACCES very well so it should
3080          * be mapped to -EPERM instead.
3081          */
3082         if (status == -EACCES)
3083                 status = -EPERM;
3084         return status;
3085 }
3086
3087 static int nfs4_do_find_root_sec(struct nfs_server *server,
3088                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3089 {
3090         int mv = server->nfs_client->cl_minorversion;
3091         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3092 }
3093
3094 /**
3095  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3096  * @server: initialized nfs_server handle
3097  * @fhandle: we fill in the pseudo-fs root file handle
3098  * @info: we fill in an FSINFO struct
3099  * @auth_probe: probe the auth flavours
3100  *
3101  * Returns zero on success, or a negative errno.
3102  */
3103 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3104                          struct nfs_fsinfo *info,
3105                          bool auth_probe)
3106 {
3107         int status = 0;
3108
3109         if (!auth_probe)
3110                 status = nfs4_lookup_root(server, fhandle, info);
3111
3112         if (auth_probe || status == NFS4ERR_WRONGSEC)
3113                 status = nfs4_do_find_root_sec(server, fhandle, info);
3114
3115         if (status == 0)
3116                 status = nfs4_server_capabilities(server, fhandle);
3117         if (status == 0)
3118                 status = nfs4_do_fsinfo(server, fhandle, info);
3119
3120         return nfs4_map_errors(status);
3121 }
3122
3123 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3124                               struct nfs_fsinfo *info)
3125 {
3126         int error;
3127         struct nfs_fattr *fattr = info->fattr;
3128         struct nfs4_label *label = NULL;
3129
3130         error = nfs4_server_capabilities(server, mntfh);
3131         if (error < 0) {
3132                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3133                 return error;
3134         }
3135
3136         label = nfs4_label_alloc(server, GFP_KERNEL);
3137         if (IS_ERR(label))
3138                 return PTR_ERR(label);
3139
3140         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3141         if (error < 0) {
3142                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3143                 goto err_free_label;
3144         }
3145
3146         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3147             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3148                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3149
3150 err_free_label:
3151         nfs4_label_free(label);
3152
3153         return error;
3154 }
3155
3156 /*
3157  * Get locations and (maybe) other attributes of a referral.
3158  * Note that we'll actually follow the referral later when
3159  * we detect fsid mismatch in inode revalidation
3160  */
3161 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3162                              const struct qstr *name, struct nfs_fattr *fattr,
3163                              struct nfs_fh *fhandle)
3164 {
3165         int status = -ENOMEM;
3166         struct page *page = NULL;
3167         struct nfs4_fs_locations *locations = NULL;
3168
3169         page = alloc_page(GFP_KERNEL);
3170         if (page == NULL)
3171                 goto out;
3172         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3173         if (locations == NULL)
3174                 goto out;
3175
3176         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3177         if (status != 0)
3178                 goto out;
3179
3180         /*
3181          * If the fsid didn't change, this is a migration event, not a
3182          * referral.  Cause us to drop into the exception handler, which
3183          * will kick off migration recovery.
3184          */
3185         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3186                 dprintk("%s: server did not return a different fsid for"
3187                         " a referral at %s\n", __func__, name->name);
3188                 status = -NFS4ERR_MOVED;
3189                 goto out;
3190         }
3191         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3192         nfs_fixup_referral_attributes(&locations->fattr);
3193
3194         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3195         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3196         memset(fhandle, 0, sizeof(struct nfs_fh));
3197 out:
3198         if (page)
3199                 __free_page(page);
3200         kfree(locations);
3201         return status;
3202 }
3203
3204 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3205                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3206 {
3207         struct nfs4_getattr_arg args = {
3208                 .fh = fhandle,
3209                 .bitmask = server->attr_bitmask,
3210         };
3211         struct nfs4_getattr_res res = {
3212                 .fattr = fattr,
3213                 .label = label,
3214                 .server = server,
3215         };
3216         struct rpc_message msg = {
3217                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3218                 .rpc_argp = &args,
3219                 .rpc_resp = &res,
3220         };
3221
3222         args.bitmask = nfs4_bitmask(server, label);
3223
3224         nfs_fattr_init(fattr);
3225         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3226 }
3227
3228 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3229                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3230 {
3231         struct nfs4_exception exception = { };
3232         int err;
3233         do {
3234                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3235                 trace_nfs4_getattr(server, fhandle, fattr, err);
3236                 err = nfs4_handle_exception(server, err,
3237                                 &exception);
3238         } while (exception.retry);
3239         return err;
3240 }
3241
3242 /*
3243  * The file is not closed if it is opened due to the a request to change
3244  * the size of the file. The open call will not be needed once the
3245  * VFS layer lookup-intents are implemented.
3246  *
3247  * Close is called when the inode is destroyed.
3248  * If we haven't opened the file for O_WRONLY, we
3249  * need to in the size_change case to obtain a stateid.
3250  *
3251  * Got race?
3252  * Because OPEN is always done by name in nfsv4, it is
3253  * possible that we opened a different file by the same
3254  * name.  We can recognize this race condition, but we
3255  * can't do anything about it besides returning an error.
3256  *
3257  * This will be fixed with VFS changes (lookup-intent).
3258  */
3259 static int
3260 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3261                   struct iattr *sattr)
3262 {
3263         struct inode *inode = d_inode(dentry);
3264         struct rpc_cred *cred = NULL;
3265         struct nfs4_state *state = NULL;
3266         struct nfs4_label *label = NULL;
3267         int status;
3268
3269         if (pnfs_ld_layoutret_on_setattr(inode) &&
3270             sattr->ia_valid & ATTR_SIZE &&
3271             sattr->ia_size < i_size_read(inode))
3272                 pnfs_commit_and_return_layout(inode);
3273
3274         nfs_fattr_init(fattr);
3275
3276         /* Deal with open(O_TRUNC) */
3277         if (sattr->ia_valid & ATTR_OPEN)
3278                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3279
3280         /* Optimization: if the end result is no change, don't RPC */
3281         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3282                 return 0;
3283
3284         /* Search for an existing open(O_WRITE) file */
3285         if (sattr->ia_valid & ATTR_FILE) {
3286                 struct nfs_open_context *ctx;
3287
3288                 ctx = nfs_file_open_context(sattr->ia_file);
3289                 if (ctx) {
3290                         cred = ctx->cred;
3291                         state = ctx->state;
3292                 }
3293         }
3294
3295         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3296         if (IS_ERR(label))
3297                 return PTR_ERR(label);
3298
3299         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3300         if (status == 0) {
3301                 nfs_setattr_update_inode(inode, sattr, fattr);
3302                 nfs_setsecurity(inode, fattr, label);
3303         }
3304         nfs4_label_free(label);
3305         return status;
3306 }
3307
3308 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3309                 const struct qstr *name, struct nfs_fh *fhandle,
3310                 struct nfs_fattr *fattr, struct nfs4_label *label)
3311 {
3312         struct nfs_server *server = NFS_SERVER(dir);
3313         int                    status;
3314         struct nfs4_lookup_arg args = {
3315                 .bitmask = server->attr_bitmask,
3316                 .dir_fh = NFS_FH(dir),
3317                 .name = name,
3318         };
3319         struct nfs4_lookup_res res = {
3320                 .server = server,
3321                 .fattr = fattr,
3322                 .label = label,
3323                 .fh = fhandle,
3324         };
3325         struct rpc_message msg = {
3326                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3327                 .rpc_argp = &args,
3328                 .rpc_resp = &res,
3329         };
3330
3331         args.bitmask = nfs4_bitmask(server, label);
3332
3333         nfs_fattr_init(fattr);
3334
3335         dprintk("NFS call  lookup %s\n", name->name);
3336         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3337         dprintk("NFS reply lookup: %d\n", status);
3338         return status;
3339 }
3340
3341 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3342 {
3343         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3344                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3345         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3346         fattr->nlink = 2;
3347 }
3348
3349 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3350                                    struct qstr *name, struct nfs_fh *fhandle,
3351                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3352 {
3353         struct nfs4_exception exception = { };
3354         struct rpc_clnt *client = *clnt;
3355         int err;
3356         do {
3357                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3358                 trace_nfs4_lookup(dir, name, err);
3359                 switch (err) {
3360                 case -NFS4ERR_BADNAME:
3361                         err = -ENOENT;
3362                         goto out;
3363                 case -NFS4ERR_MOVED:
3364                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3365                         if (err == -NFS4ERR_MOVED)
3366                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3367                         goto out;
3368                 case -NFS4ERR_WRONGSEC:
3369                         err = -EPERM;
3370                         if (client != *clnt)
3371                                 goto out;
3372                         client = nfs4_negotiate_security(client, dir, name);
3373                         if (IS_ERR(client))
3374                                 return PTR_ERR(client);
3375
3376                         exception.retry = 1;
3377                         break;
3378                 default:
3379                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3380                 }
3381         } while (exception.retry);
3382
3383 out:
3384         if (err == 0)
3385                 *clnt = client;
3386         else if (client != *clnt)
3387                 rpc_shutdown_client(client);
3388
3389         return err;
3390 }
3391
3392 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3393                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3394                             struct nfs4_label *label)
3395 {
3396         int status;
3397         struct rpc_clnt *client = NFS_CLIENT(dir);
3398
3399         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3400         if (client != NFS_CLIENT(dir)) {
3401                 rpc_shutdown_client(client);
3402                 nfs_fixup_secinfo_attributes(fattr);
3403         }
3404         return status;
3405 }
3406
3407 struct rpc_clnt *
3408 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3409                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3410 {
3411         struct rpc_clnt *client = NFS_CLIENT(dir);
3412         int status;
3413
3414         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3415         if (status < 0)
3416                 return ERR_PTR(status);
3417         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3418 }
3419
3420 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3421 {
3422         struct nfs_server *server = NFS_SERVER(inode);
3423         struct nfs4_accessargs args = {
3424                 .fh = NFS_FH(inode),
3425                 .bitmask = server->cache_consistency_bitmask,
3426         };
3427         struct nfs4_accessres res = {
3428                 .server = server,
3429         };
3430         struct rpc_message msg = {
3431                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3432                 .rpc_argp = &args,
3433                 .rpc_resp = &res,
3434                 .rpc_cred = entry->cred,
3435         };
3436         int mode = entry->mask;
3437         int status = 0;
3438
3439         /*
3440          * Determine which access bits we want to ask for...
3441          */
3442         if (mode & MAY_READ)
3443                 args.access |= NFS4_ACCESS_READ;
3444         if (S_ISDIR(inode->i_mode)) {
3445                 if (mode & MAY_WRITE)
3446                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3447                 if (mode & MAY_EXEC)
3448                         args.access |= NFS4_ACCESS_LOOKUP;
3449         } else {
3450                 if (mode & MAY_WRITE)
3451                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3452                 if (mode & MAY_EXEC)
3453                         args.access |= NFS4_ACCESS_EXECUTE;
3454         }
3455
3456         res.fattr = nfs_alloc_fattr();
3457         if (res.fattr == NULL)
3458                 return -ENOMEM;
3459
3460         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3461         if (!status) {
3462                 nfs_access_set_mask(entry, res.access);
3463                 nfs_refresh_inode(inode, res.fattr);
3464         }
3465         nfs_free_fattr(res.fattr);
3466         return status;
3467 }
3468
3469 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3470 {
3471         struct nfs4_exception exception = { };
3472         int err;
3473         do {
3474                 err = _nfs4_proc_access(inode, entry);
3475                 trace_nfs4_access(inode, err);
3476                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3477                                 &exception);
3478         } while (exception.retry);
3479         return err;
3480 }
3481
3482 /*
3483  * TODO: For the time being, we don't try to get any attributes
3484  * along with any of the zero-copy operations READ, READDIR,
3485  * READLINK, WRITE.
3486  *
3487  * In the case of the first three, we want to put the GETATTR
3488  * after the read-type operation -- this is because it is hard
3489  * to predict the length of a GETATTR response in v4, and thus
3490  * align the READ data correctly.  This means that the GETATTR
3491  * may end up partially falling into the page cache, and we should
3492  * shift it into the 'tail' of the xdr_buf before processing.
3493  * To do this efficiently, we need to know the total length
3494  * of data received, which doesn't seem to be available outside
3495  * of the RPC layer.
3496  *
3497  * In the case of WRITE, we also want to put the GETATTR after
3498  * the operation -- in this case because we want to make sure
3499  * we get the post-operation mtime and size.
3500  *
3501  * Both of these changes to the XDR layer would in fact be quite
3502  * minor, but I decided to leave them for a subsequent patch.
3503  */
3504 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3505                 unsigned int pgbase, unsigned int pglen)
3506 {
3507         struct nfs4_readlink args = {
3508                 .fh       = NFS_FH(inode),
3509                 .pgbase   = pgbase,
3510                 .pglen    = pglen,
3511                 .pages    = &page,
3512         };
3513         struct nfs4_readlink_res res;
3514         struct rpc_message msg = {
3515                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3516                 .rpc_argp = &args,
3517                 .rpc_resp = &res,
3518         };
3519
3520         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3521 }
3522
3523 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3524                 unsigned int pgbase, unsigned int pglen)
3525 {
3526         struct nfs4_exception exception = { };
3527         int err;
3528         do {
3529                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3530                 trace_nfs4_readlink(inode, err);
3531                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3532                                 &exception);
3533         } while (exception.retry);
3534         return err;
3535 }
3536
3537 /*
3538  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3539  */
3540 static int
3541 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3542                  int flags)
3543 {
3544         struct nfs4_label l, *ilabel = NULL;
3545         struct nfs_open_context *ctx;
3546         struct nfs4_state *state;
3547         int opened = 0;
3548         int status = 0;
3549
3550         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3551         if (IS_ERR(ctx))
3552                 return PTR_ERR(ctx);
3553
3554         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3555
3556         sattr->ia_mode &= ~current_umask();
3557         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened);
3558         if (IS_ERR(state)) {
3559                 status = PTR_ERR(state);
3560                 goto out;
3561         }
3562 out:
3563         nfs4_label_release_security(ilabel);
3564         put_nfs_open_context(ctx);
3565         return status;
3566 }
3567
3568 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3569 {
3570         struct nfs_server *server = NFS_SERVER(dir);
3571         struct nfs_removeargs args = {
3572                 .fh = NFS_FH(dir),
3573                 .name = *name,
3574         };
3575         struct nfs_removeres res = {
3576                 .server = server,
3577         };
3578         struct rpc_message msg = {
3579                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3580                 .rpc_argp = &args,
3581                 .rpc_resp = &res,
3582         };
3583         int status;
3584
3585         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3586         if (status == 0)
3587                 update_changeattr(dir, &res.cinfo);
3588         return status;
3589 }
3590
3591 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3592 {
3593         struct nfs4_exception exception = { };
3594         int err;
3595         do {
3596                 err = _nfs4_proc_remove(dir, name);
3597                 trace_nfs4_remove(dir, name, err);
3598                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3599                                 &exception);
3600         } while (exception.retry);
3601         return err;
3602 }
3603
3604 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3605 {
3606         struct nfs_server *server = NFS_SERVER(dir);
3607         struct nfs_removeargs *args = msg->rpc_argp;
3608         struct nfs_removeres *res = msg->rpc_resp;
3609
3610         res->server = server;
3611         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3612         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3613
3614         nfs_fattr_init(res->dir_attr);
3615 }
3616
3617 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3618 {
3619         nfs4_setup_sequence(NFS_SERVER(data->dir),
3620                         &data->args.seq_args,
3621                         &data->res.seq_res,
3622                         task);
3623 }
3624
3625 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3626 {
3627         struct nfs_unlinkdata *data = task->tk_calldata;
3628         struct nfs_removeres *res = &data->res;
3629
3630         if (!nfs4_sequence_done(task, &res->seq_res))
3631                 return 0;
3632         if (nfs4_async_handle_error(task, res->server, NULL,
3633                                     &data->timeout) == -EAGAIN)
3634                 return 0;
3635         update_changeattr(dir, &res->cinfo);
3636         return 1;
3637 }
3638
3639 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3640 {
3641         struct nfs_server *server = NFS_SERVER(dir);
3642         struct nfs_renameargs *arg = msg->rpc_argp;
3643         struct nfs_renameres *res = msg->rpc_resp;
3644
3645         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3646         res->server = server;
3647         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3648 }
3649
3650 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3651 {
3652         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3653                         &data->args.seq_args,
3654                         &data->res.seq_res,
3655                         task);
3656 }
3657
3658 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3659                                  struct inode *new_dir)
3660 {
3661         struct nfs_renamedata *data = task->tk_calldata;
3662         struct nfs_renameres *res = &data->res;
3663
3664         if (!nfs4_sequence_done(task, &res->seq_res))
3665                 return 0;
3666         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3667                 return 0;
3668
3669         update_changeattr(old_dir, &res->old_cinfo);
3670         update_changeattr(new_dir, &res->new_cinfo);
3671         return 1;
3672 }
3673
3674 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3675 {
3676         struct nfs_server *server = NFS_SERVER(inode);
3677         struct nfs4_link_arg arg = {
3678                 .fh     = NFS_FH(inode),
3679                 .dir_fh = NFS_FH(dir),
3680                 .name   = name,
3681                 .bitmask = server->attr_bitmask,
3682         };
3683         struct nfs4_link_res res = {
3684                 .server = server,
3685                 .label = NULL,
3686         };
3687         struct rpc_message msg = {
3688                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3689                 .rpc_argp = &arg,
3690                 .rpc_resp = &res,
3691         };
3692         int status = -ENOMEM;
3693
3694         res.fattr = nfs_alloc_fattr();
3695         if (res.fattr == NULL)
3696                 goto out;
3697
3698         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3699         if (IS_ERR(res.label)) {
3700                 status = PTR_ERR(res.label);
3701                 goto out;
3702         }
3703         arg.bitmask = nfs4_bitmask(server, res.label);
3704
3705         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3706         if (!status) {
3707                 update_changeattr(dir, &res.cinfo);
3708                 status = nfs_post_op_update_inode(inode, res.fattr);
3709                 if (!status)
3710                         nfs_setsecurity(inode, res.fattr, res.label);
3711         }
3712
3713
3714         nfs4_label_free(res.label);
3715
3716 out:
3717         nfs_free_fattr(res.fattr);
3718         return status;
3719 }
3720
3721 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3722 {
3723         struct nfs4_exception exception = { };
3724         int err;
3725         do {
3726                 err = nfs4_handle_exception(NFS_SERVER(inode),
3727                                 _nfs4_proc_link(inode, dir, name),
3728                                 &exception);
3729         } while (exception.retry);
3730         return err;
3731 }
3732
3733 struct nfs4_createdata {
3734         struct rpc_message msg;
3735         struct nfs4_create_arg arg;
3736         struct nfs4_create_res res;
3737         struct nfs_fh fh;
3738         struct nfs_fattr fattr;
3739         struct nfs4_label *label;
3740 };
3741
3742 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3743                 struct qstr *name, struct iattr *sattr, u32 ftype)
3744 {
3745         struct nfs4_createdata *data;
3746
3747         data = kzalloc(sizeof(*data), GFP_KERNEL);
3748         if (data != NULL) {
3749                 struct nfs_server *server = NFS_SERVER(dir);
3750
3751                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3752                 if (IS_ERR(data->label))
3753                         goto out_free;
3754
3755                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3756                 data->msg.rpc_argp = &data->arg;
3757                 data->msg.rpc_resp = &data->res;
3758                 data->arg.dir_fh = NFS_FH(dir);
3759                 data->arg.server = server;
3760                 data->arg.name = name;
3761                 data->arg.attrs = sattr;
3762                 data->arg.ftype = ftype;
3763                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3764                 data->res.server = server;
3765                 data->res.fh = &data->fh;
3766                 data->res.fattr = &data->fattr;
3767                 data->res.label = data->label;
3768                 nfs_fattr_init(data->res.fattr);
3769         }
3770         return data;
3771 out_free:
3772         kfree(data);
3773         return NULL;
3774 }
3775
3776 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3777 {
3778         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3779                                     &data->arg.seq_args, &data->res.seq_res, 1);
3780         if (status == 0) {
3781                 update_changeattr(dir, &data->res.dir_cinfo);
3782                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3783         }
3784         return status;
3785 }
3786
3787 static void nfs4_free_createdata(struct nfs4_createdata *data)
3788 {
3789         nfs4_label_free(data->label);
3790         kfree(data);
3791 }
3792
3793 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3794                 struct page *page, unsigned int len, struct iattr *sattr,
3795                 struct nfs4_label *label)
3796 {
3797         struct nfs4_createdata *data;
3798         int status = -ENAMETOOLONG;
3799
3800         if (len > NFS4_MAXPATHLEN)
3801                 goto out;
3802
3803         status = -ENOMEM;
3804         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3805         if (data == NULL)
3806                 goto out;
3807
3808         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3809         data->arg.u.symlink.pages = &page;
3810         data->arg.u.symlink.len = len;
3811         data->arg.label = label;
3812
3813         status = nfs4_do_create(dir, dentry, data);
3814
3815         nfs4_free_createdata(data);
3816 out:
3817         return status;
3818 }
3819
3820 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3821                 struct page *page, unsigned int len, struct iattr *sattr)
3822 {
3823         struct nfs4_exception exception = { };
3824         struct nfs4_label l, *label = NULL;
3825         int err;
3826
3827         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3828
3829         do {
3830                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3831                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3832                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3833                                 &exception);
3834         } while (exception.retry);
3835
3836         nfs4_label_release_security(label);
3837         return err;
3838 }
3839
3840 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3841                 struct iattr *sattr, struct nfs4_label *label)
3842 {
3843         struct nfs4_createdata *data;
3844         int status = -ENOMEM;
3845
3846         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3847         if (data == NULL)
3848                 goto out;
3849
3850         data->arg.label = label;
3851         status = nfs4_do_create(dir, dentry, data);
3852
3853         nfs4_free_createdata(data);
3854 out:
3855         return status;
3856 }
3857
3858 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3859                 struct iattr *sattr)
3860 {
3861         struct nfs4_exception exception = { };
3862         struct nfs4_label l, *label = NULL;
3863         int err;
3864
3865         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3866
3867         sattr->ia_mode &= ~current_umask();
3868         do {
3869                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3870                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3871                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3872                                 &exception);
3873         } while (exception.retry);
3874         nfs4_label_release_security(label);
3875
3876         return err;
3877 }
3878
3879 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3880                 u64 cookie, struct page **pages, unsigned int count, int plus)
3881 {
3882         struct inode            *dir = d_inode(dentry);
3883         struct nfs4_readdir_arg args = {
3884                 .fh = NFS_FH(dir),
3885                 .pages = pages,
3886                 .pgbase = 0,
3887                 .count = count,
3888                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3889                 .plus = plus,
3890         };
3891         struct nfs4_readdir_res res;
3892         struct rpc_message msg = {
3893                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3894                 .rpc_argp = &args,
3895                 .rpc_resp = &res,
3896                 .rpc_cred = cred,
3897         };
3898         int                     status;
3899
3900         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3901                         dentry,
3902                         (unsigned long long)cookie);
3903         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3904         res.pgbase = args.pgbase;
3905         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3906         if (status >= 0) {
3907                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3908                 status += args.pgbase;
3909         }
3910
3911         nfs_invalidate_atime(dir);
3912
3913         dprintk("%s: returns %d\n", __func__, status);
3914         return status;
3915 }
3916
3917 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3918                 u64 cookie, struct page **pages, unsigned int count, int plus)
3919 {
3920         struct nfs4_exception exception = { };
3921         int err;
3922         do {
3923                 err = _nfs4_proc_readdir(dentry, cred, cookie,
3924                                 pages, count, plus);
3925                 trace_nfs4_readdir(d_inode(dentry), err);
3926                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
3927                                 &exception);
3928         } while (exception.retry);
3929         return err;
3930 }
3931
3932 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3933                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
3934 {
3935         struct nfs4_createdata *data;
3936         int mode = sattr->ia_mode;
3937         int status = -ENOMEM;
3938
3939         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
3940         if (data == NULL)
3941                 goto out;
3942
3943         if (S_ISFIFO(mode))
3944                 data->arg.ftype = NF4FIFO;
3945         else if (S_ISBLK(mode)) {
3946                 data->arg.ftype = NF4BLK;
3947                 data->arg.u.device.specdata1 = MAJOR(rdev);
3948                 data->arg.u.device.specdata2 = MINOR(rdev);
3949         }
3950         else if (S_ISCHR(mode)) {
3951                 data->arg.ftype = NF4CHR;
3952                 data->arg.u.device.specdata1 = MAJOR(rdev);
3953                 data->arg.u.device.specdata2 = MINOR(rdev);
3954         } else if (!S_ISSOCK(mode)) {
3955                 status = -EINVAL;
3956                 goto out_free;
3957         }
3958
3959         data->arg.label = label;
3960         status = nfs4_do_create(dir, dentry, data);
3961 out_free:
3962         nfs4_free_createdata(data);
3963 out:
3964         return status;
3965 }
3966
3967 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
3968                 struct iattr *sattr, dev_t rdev)
3969 {
3970         struct nfs4_exception exception = { };
3971         struct nfs4_label l, *label = NULL;
3972         int err;
3973
3974         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3975
3976         sattr->ia_mode &= ~current_umask();
3977         do {
3978                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
3979                 trace_nfs4_mknod(dir, &dentry->d_name, err);
3980                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3981                                 &exception);
3982         } while (exception.retry);
3983
3984         nfs4_label_release_security(label);
3985
3986         return err;
3987 }
3988
3989 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
3990                  struct nfs_fsstat *fsstat)
3991 {
3992         struct nfs4_statfs_arg args = {
3993                 .fh = fhandle,
3994                 .bitmask = server->attr_bitmask,
3995         };
3996         struct nfs4_statfs_res res = {
3997                 .fsstat = fsstat,
3998         };
3999         struct rpc_message msg = {
4000                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4001                 .rpc_argp = &args,
4002                 .rpc_resp = &res,
4003         };
4004
4005         nfs_fattr_init(fsstat->fattr);
4006         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4007 }
4008
4009 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4010 {
4011         struct nfs4_exception exception = { };
4012         int err;
4013         do {
4014                 err = nfs4_handle_exception(server,
4015                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4016                                 &exception);
4017         } while (exception.retry);
4018         return err;
4019 }
4020
4021 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4022                 struct nfs_fsinfo *fsinfo)
4023 {
4024         struct nfs4_fsinfo_arg args = {
4025                 .fh = fhandle,
4026                 .bitmask = server->attr_bitmask,
4027         };
4028         struct nfs4_fsinfo_res res = {
4029                 .fsinfo = fsinfo,
4030         };
4031         struct rpc_message msg = {
4032                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4033                 .rpc_argp = &args,
4034                 .rpc_resp = &res,
4035         };
4036
4037         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4038 }
4039
4040 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4041 {
4042         struct nfs4_exception exception = { };
4043         unsigned long now = jiffies;
4044         int err;
4045
4046         do {
4047                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4048                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4049                 if (err == 0) {
4050                         struct nfs_client *clp = server->nfs_client;
4051
4052                         spin_lock(&clp->cl_lock);
4053                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4054                         clp->cl_last_renewal = now;
4055                         spin_unlock(&clp->cl_lock);
4056                         break;
4057                 }
4058                 err = nfs4_handle_exception(server, err, &exception);
4059         } while (exception.retry);
4060         return err;
4061 }
4062
4063 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4064 {
4065         int error;
4066
4067         nfs_fattr_init(fsinfo->fattr);
4068         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4069         if (error == 0) {
4070                 /* block layout checks this! */
4071                 server->pnfs_blksize = fsinfo->blksize;
4072                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4073         }
4074
4075         return error;
4076 }
4077
4078 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4079                 struct nfs_pathconf *pathconf)
4080 {
4081         struct nfs4_pathconf_arg args = {
4082                 .fh = fhandle,
4083                 .bitmask = server->attr_bitmask,
4084         };
4085         struct nfs4_pathconf_res res = {
4086                 .pathconf = pathconf,
4087         };
4088         struct rpc_message msg = {
4089                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4090                 .rpc_argp = &args,
4091                 .rpc_resp = &res,
4092         };
4093
4094         /* None of the pathconf attributes are mandatory to implement */
4095         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4096                 memset(pathconf, 0, sizeof(*pathconf));
4097                 return 0;
4098         }
4099
4100         nfs_fattr_init(pathconf->fattr);
4101         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4102 }
4103
4104 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4105                 struct nfs_pathconf *pathconf)
4106 {
4107         struct nfs4_exception exception = { };
4108         int err;
4109
4110         do {
4111                 err = nfs4_handle_exception(server,
4112                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4113                                 &exception);
4114         } while (exception.retry);
4115         return err;
4116 }
4117
4118 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4119                 const struct nfs_open_context *ctx,
4120                 const struct nfs_lock_context *l_ctx,
4121                 fmode_t fmode)
4122 {
4123         const struct nfs_lockowner *lockowner = NULL;
4124
4125         if (l_ctx != NULL)
4126                 lockowner = &l_ctx->lockowner;
4127         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4128 }
4129 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4130
4131 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4132                 const struct nfs_open_context *ctx,
4133                 const struct nfs_lock_context *l_ctx,
4134                 fmode_t fmode)
4135 {
4136         nfs4_stateid current_stateid;
4137
4138         /* If the current stateid represents a lost lock, then exit */
4139         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4140                 return true;
4141         return nfs4_stateid_match(stateid, &current_stateid);
4142 }
4143
4144 static bool nfs4_error_stateid_expired(int err)
4145 {
4146         switch (err) {
4147         case -NFS4ERR_DELEG_REVOKED:
4148         case -NFS4ERR_ADMIN_REVOKED:
4149         case -NFS4ERR_BAD_STATEID:
4150         case -NFS4ERR_STALE_STATEID:
4151         case -NFS4ERR_OLD_STATEID:
4152         case -NFS4ERR_OPENMODE:
4153         case -NFS4ERR_EXPIRED:
4154                 return true;
4155         }
4156         return false;
4157 }
4158
4159 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4160 {
4161         nfs_invalidate_atime(hdr->inode);
4162 }
4163
4164 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4165 {
4166         struct nfs_server *server = NFS_SERVER(hdr->inode);
4167
4168         trace_nfs4_read(hdr, task->tk_status);
4169         if (nfs4_async_handle_error(task, server,
4170                                     hdr->args.context->state,
4171                                     NULL) == -EAGAIN) {
4172                 rpc_restart_call_prepare(task);
4173                 return -EAGAIN;
4174         }
4175
4176         __nfs4_read_done_cb(hdr);
4177         if (task->tk_status > 0)
4178                 renew_lease(server, hdr->timestamp);
4179         return 0;
4180 }
4181
4182 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4183                 struct nfs_pgio_args *args)
4184 {
4185
4186         if (!nfs4_error_stateid_expired(task->tk_status) ||
4187                 nfs4_stateid_is_current(&args->stateid,
4188                                 args->context,
4189                                 args->lock_context,
4190                                 FMODE_READ))
4191                 return false;
4192         rpc_restart_call_prepare(task);
4193         return true;
4194 }
4195
4196 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4197 {
4198
4199         dprintk("--> %s\n", __func__);
4200
4201         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4202                 return -EAGAIN;
4203         if (nfs4_read_stateid_changed(task, &hdr->args))
4204                 return -EAGAIN;
4205         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4206                                     nfs4_read_done_cb(task, hdr);
4207 }
4208
4209 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4210                                  struct rpc_message *msg)
4211 {
4212         hdr->timestamp   = jiffies;
4213         hdr->pgio_done_cb = nfs4_read_done_cb;
4214         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4215         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4216 }
4217
4218 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4219                                       struct nfs_pgio_header *hdr)
4220 {
4221         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4222                         &hdr->args.seq_args,
4223                         &hdr->res.seq_res,
4224                         task))
4225                 return 0;
4226         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4227                                 hdr->args.lock_context,
4228                                 hdr->rw_ops->rw_mode) == -EIO)
4229                 return -EIO;
4230         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4231                 return -EIO;
4232         return 0;
4233 }
4234
4235 static int nfs4_write_done_cb(struct rpc_task *task,
4236                               struct nfs_pgio_header *hdr)
4237 {
4238         struct inode *inode = hdr->inode;
4239
4240         trace_nfs4_write(hdr, task->tk_status);
4241         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4242                                     hdr->args.context->state,
4243                                     NULL) == -EAGAIN) {
4244                 rpc_restart_call_prepare(task);
4245                 return -EAGAIN;
4246         }
4247         if (task->tk_status >= 0) {
4248                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4249                 nfs_writeback_update_inode(hdr);
4250         }
4251         return 0;
4252 }
4253
4254 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4255                 struct nfs_pgio_args *args)
4256 {
4257
4258         if (!nfs4_error_stateid_expired(task->tk_status) ||
4259                 nfs4_stateid_is_current(&args->stateid,
4260                                 args->context,
4261                                 args->lock_context,
4262                                 FMODE_WRITE))
4263                 return false;
4264         rpc_restart_call_prepare(task);
4265         return true;
4266 }
4267
4268 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4269 {
4270         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4271                 return -EAGAIN;
4272         if (nfs4_write_stateid_changed(task, &hdr->args))
4273                 return -EAGAIN;
4274         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4275                 nfs4_write_done_cb(task, hdr);
4276 }
4277
4278 static
4279 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4280 {
4281         /* Don't request attributes for pNFS or O_DIRECT writes */
4282         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4283                 return false;
4284         /* Otherwise, request attributes if and only if we don't hold
4285          * a delegation
4286          */
4287         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4288 }
4289
4290 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4291                                   struct rpc_message *msg)
4292 {
4293         struct nfs_server *server = NFS_SERVER(hdr->inode);
4294
4295         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4296                 hdr->args.bitmask = NULL;
4297                 hdr->res.fattr = NULL;
4298         } else
4299                 hdr->args.bitmask = server->cache_consistency_bitmask;
4300
4301         if (!hdr->pgio_done_cb)
4302                 hdr->pgio_done_cb = nfs4_write_done_cb;
4303         hdr->res.server = server;
4304         hdr->timestamp   = jiffies;
4305
4306         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4307         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4308 }
4309
4310 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4311 {
4312         nfs4_setup_sequence(NFS_SERVER(data->inode),
4313                         &data->args.seq_args,
4314                         &data->res.seq_res,
4315                         task);
4316 }
4317
4318 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4319 {
4320         struct inode *inode = data->inode;
4321
4322         trace_nfs4_commit(data, task->tk_status);
4323         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4324                                     NULL, NULL) == -EAGAIN) {
4325                 rpc_restart_call_prepare(task);
4326                 return -EAGAIN;
4327         }
4328         return 0;
4329 }
4330
4331 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4332 {
4333         if (!nfs4_sequence_done(task, &data->res.seq_res))
4334                 return -EAGAIN;
4335         return data->commit_done_cb(task, data);
4336 }
4337
4338 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4339 {
4340         struct nfs_server *server = NFS_SERVER(data->inode);
4341
4342         if (data->commit_done_cb == NULL)
4343                 data->commit_done_cb = nfs4_commit_done_cb;
4344         data->res.server = server;
4345         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4346         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4347 }
4348
4349 struct nfs4_renewdata {
4350         struct nfs_client       *client;
4351         unsigned long           timestamp;
4352 };
4353
4354 /*
4355  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4356  * standalone procedure for queueing an asynchronous RENEW.
4357  */
4358 static void nfs4_renew_release(void *calldata)
4359 {
4360         struct nfs4_renewdata *data = calldata;
4361         struct nfs_client *clp = data->client;
4362
4363         if (atomic_read(&clp->cl_count) > 1)
4364                 nfs4_schedule_state_renewal(clp);
4365         nfs_put_client(clp);
4366         kfree(data);
4367 }
4368
4369 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4370 {
4371         struct nfs4_renewdata *data = calldata;
4372         struct nfs_client *clp = data->client;
4373         unsigned long timestamp = data->timestamp;
4374
4375         trace_nfs4_renew_async(clp, task->tk_status);
4376         switch (task->tk_status) {
4377         case 0:
4378                 break;
4379         case -NFS4ERR_LEASE_MOVED:
4380                 nfs4_schedule_lease_moved_recovery(clp);
4381                 break;
4382         default:
4383                 /* Unless we're shutting down, schedule state recovery! */
4384                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4385                         return;
4386                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4387                         nfs4_schedule_lease_recovery(clp);
4388                         return;
4389                 }
4390                 nfs4_schedule_path_down_recovery(clp);
4391         }
4392         do_renew_lease(clp, timestamp);
4393 }
4394
4395 static const struct rpc_call_ops nfs4_renew_ops = {
4396         .rpc_call_done = nfs4_renew_done,
4397         .rpc_release = nfs4_renew_release,
4398 };
4399
4400 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4401 {
4402         struct rpc_message msg = {
4403                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4404                 .rpc_argp       = clp,
4405                 .rpc_cred       = cred,
4406         };
4407         struct nfs4_renewdata *data;
4408
4409         if (renew_flags == 0)
4410                 return 0;
4411         if (!atomic_inc_not_zero(&clp->cl_count))
4412                 return -EIO;
4413         data = kmalloc(sizeof(*data), GFP_NOFS);
4414         if (data == NULL)
4415                 return -ENOMEM;
4416         data->client = clp;
4417         data->timestamp = jiffies;
4418         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4419                         &nfs4_renew_ops, data);
4420 }
4421
4422 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4423 {
4424         struct rpc_message msg = {
4425                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4426                 .rpc_argp       = clp,
4427                 .rpc_cred       = cred,
4428         };
4429         unsigned long now = jiffies;
4430         int status;
4431
4432         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4433         if (status < 0)
4434                 return status;
4435         do_renew_lease(clp, now);
4436         return 0;
4437 }
4438
4439 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4440 {
4441         return server->caps & NFS_CAP_ACLS;
4442 }
4443
4444 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4445  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4446  * the stack.
4447  */
4448 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4449
4450 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4451                 struct page **pages, unsigned int *pgbase)
4452 {
4453         struct page *newpage, **spages;
4454         int rc = 0;
4455         size_t len;
4456         spages = pages;
4457
4458         do {
4459                 len = min_t(size_t, PAGE_SIZE, buflen);
4460                 newpage = alloc_page(GFP_KERNEL);
4461
4462                 if (newpage == NULL)
4463                         goto unwind;
4464                 memcpy(page_address(newpage), buf, len);
4465                 buf += len;
4466                 buflen -= len;
4467                 *pages++ = newpage;
4468                 rc++;
4469         } while (buflen != 0);
4470
4471         return rc;
4472
4473 unwind:
4474         for(; rc > 0; rc--)
4475                 __free_page(spages[rc-1]);
4476         return -ENOMEM;
4477 }
4478
4479 struct nfs4_cached_acl {
4480         int cached;
4481         size_t len;
4482         char data[0];
4483 };
4484
4485 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4486 {
4487         struct nfs_inode *nfsi = NFS_I(inode);
4488
4489         spin_lock(&inode->i_lock);
4490         kfree(nfsi->nfs4_acl);
4491         nfsi->nfs4_acl = acl;
4492         spin_unlock(&inode->i_lock);
4493 }
4494
4495 static void nfs4_zap_acl_attr(struct inode *inode)
4496 {
4497         nfs4_set_cached_acl(inode, NULL);
4498 }
4499
4500 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4501 {
4502         struct nfs_inode *nfsi = NFS_I(inode);
4503         struct nfs4_cached_acl *acl;
4504         int ret = -ENOENT;
4505
4506         spin_lock(&inode->i_lock);
4507         acl = nfsi->nfs4_acl;
4508         if (acl == NULL)
4509                 goto out;
4510         if (buf == NULL) /* user is just asking for length */
4511                 goto out_len;
4512         if (acl->cached == 0)
4513                 goto out;
4514         ret = -ERANGE; /* see getxattr(2) man page */
4515         if (acl->len > buflen)
4516                 goto out;
4517         memcpy(buf, acl->data, acl->len);
4518 out_len:
4519         ret = acl->len;
4520 out:
4521         spin_unlock(&inode->i_lock);
4522         return ret;
4523 }
4524
4525 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4526 {
4527         struct nfs4_cached_acl *acl;
4528         size_t buflen = sizeof(*acl) + acl_len;
4529
4530         if (buflen <= PAGE_SIZE) {
4531                 acl = kmalloc(buflen, GFP_KERNEL);
4532                 if (acl == NULL)
4533                         goto out;
4534                 acl->cached = 1;
4535                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4536         } else {
4537                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4538                 if (acl == NULL)
4539                         goto out;
4540                 acl->cached = 0;
4541         }
4542         acl->len = acl_len;
4543 out:
4544         nfs4_set_cached_acl(inode, acl);
4545 }
4546
4547 /*
4548  * The getxattr API returns the required buffer length when called with a
4549  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4550  * the required buf.  On a NULL buf, we send a page of data to the server
4551  * guessing that the ACL request can be serviced by a page. If so, we cache
4552  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4553  * the cache. If not so, we throw away the page, and cache the required
4554  * length. The next getxattr call will then produce another round trip to
4555  * the server, this time with the input buf of the required size.
4556  */
4557 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4558 {
4559         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4560         struct nfs_getaclargs args = {
4561                 .fh = NFS_FH(inode),
4562                 .acl_pages = pages,
4563                 .acl_len = buflen,
4564         };
4565         struct nfs_getaclres res = {
4566                 .acl_len = buflen,
4567         };
4568         struct rpc_message msg = {
4569                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4570                 .rpc_argp = &args,
4571                 .rpc_resp = &res,
4572         };
4573         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4574         int ret = -ENOMEM, i;
4575
4576         /* As long as we're doing a round trip to the server anyway,
4577          * let's be prepared for a page of acl data. */
4578         if (npages == 0)
4579                 npages = 1;
4580         if (npages > ARRAY_SIZE(pages))
4581                 return -ERANGE;
4582
4583         for (i = 0; i < npages; i++) {
4584                 pages[i] = alloc_page(GFP_KERNEL);
4585                 if (!pages[i])
4586                         goto out_free;
4587         }
4588
4589         /* for decoding across pages */
4590         res.acl_scratch = alloc_page(GFP_KERNEL);
4591         if (!res.acl_scratch)
4592                 goto out_free;
4593
4594         args.acl_len = npages * PAGE_SIZE;
4595         args.acl_pgbase = 0;
4596
4597         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4598                 __func__, buf, buflen, npages, args.acl_len);
4599         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4600                              &msg, &args.seq_args, &res.seq_res, 0);
4601         if (ret)
4602                 goto out_free;
4603
4604         /* Handle the case where the passed-in buffer is too short */
4605         if (res.acl_flags & NFS4_ACL_TRUNC) {
4606                 /* Did the user only issue a request for the acl length? */
4607                 if (buf == NULL)
4608                         goto out_ok;
4609                 ret = -ERANGE;
4610                 goto out_free;
4611         }
4612         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4613         if (buf) {
4614                 if (res.acl_len > buflen) {
4615                         ret = -ERANGE;
4616                         goto out_free;
4617                 }
4618                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4619         }
4620 out_ok:
4621         ret = res.acl_len;
4622 out_free:
4623         for (i = 0; i < npages; i++)
4624                 if (pages[i])
4625                         __free_page(pages[i]);
4626         if (res.acl_scratch)
4627                 __free_page(res.acl_scratch);
4628         return ret;
4629 }
4630
4631 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4632 {
4633         struct nfs4_exception exception = { };
4634         ssize_t ret;
4635         do {
4636                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4637                 trace_nfs4_get_acl(inode, ret);
4638                 if (ret >= 0)
4639                         break;
4640                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4641         } while (exception.retry);
4642         return ret;
4643 }
4644
4645 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4646 {
4647         struct nfs_server *server = NFS_SERVER(inode);
4648         int ret;
4649
4650         if (!nfs4_server_supports_acls(server))
4651                 return -EOPNOTSUPP;
4652         ret = nfs_revalidate_inode(server, inode);
4653         if (ret < 0)
4654                 return ret;
4655         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4656                 nfs_zap_acl_cache(inode);
4657         ret = nfs4_read_cached_acl(inode, buf, buflen);
4658         if (ret != -ENOENT)
4659                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4660                  * but no cached acl data, just the acl length */
4661                 return ret;
4662         return nfs4_get_acl_uncached(inode, buf, buflen);
4663 }
4664
4665 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4666 {
4667         struct nfs_server *server = NFS_SERVER(inode);
4668         struct page *pages[NFS4ACL_MAXPAGES];
4669         struct nfs_setaclargs arg = {
4670                 .fh             = NFS_FH(inode),
4671                 .acl_pages      = pages,
4672                 .acl_len        = buflen,
4673         };
4674         struct nfs_setaclres res;
4675         struct rpc_message msg = {
4676                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4677                 .rpc_argp       = &arg,
4678                 .rpc_resp       = &res,
4679         };
4680         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4681         int ret, i;
4682
4683         if (!nfs4_server_supports_acls(server))
4684                 return -EOPNOTSUPP;
4685         if (npages > ARRAY_SIZE(pages))
4686                 return -ERANGE;
4687         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4688         if (i < 0)
4689                 return i;
4690         nfs4_inode_return_delegation(inode);
4691         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4692
4693         /*
4694          * Free each page after tx, so the only ref left is
4695          * held by the network stack
4696          */
4697         for (; i > 0; i--)
4698                 put_page(pages[i-1]);
4699
4700         /*
4701          * Acl update can result in inode attribute update.
4702          * so mark the attribute cache invalid.
4703          */
4704         spin_lock(&inode->i_lock);
4705         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4706         spin_unlock(&inode->i_lock);
4707         nfs_access_zap_cache(inode);
4708         nfs_zap_acl_cache(inode);
4709         return ret;
4710 }
4711
4712 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4713 {
4714         struct nfs4_exception exception = { };
4715         int err;
4716         do {
4717                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4718                 trace_nfs4_set_acl(inode, err);
4719                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4720                                 &exception);
4721         } while (exception.retry);
4722         return err;
4723 }
4724
4725 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4726 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4727                                         size_t buflen)
4728 {
4729         struct nfs_server *server = NFS_SERVER(inode);
4730         struct nfs_fattr fattr;
4731         struct nfs4_label label = {0, 0, buflen, buf};
4732
4733         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4734         struct nfs4_getattr_arg arg = {
4735                 .fh             = NFS_FH(inode),
4736                 .bitmask        = bitmask,
4737         };
4738         struct nfs4_getattr_res res = {
4739                 .fattr          = &fattr,
4740                 .label          = &label,
4741                 .server         = server,
4742         };
4743         struct rpc_message msg = {
4744                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4745                 .rpc_argp       = &arg,
4746                 .rpc_resp       = &res,
4747         };
4748         int ret;
4749
4750         nfs_fattr_init(&fattr);
4751
4752         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4753         if (ret)
4754                 return ret;
4755         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4756                 return -ENOENT;
4757         if (buflen < label.len)
4758                 return -ERANGE;
4759         return 0;
4760 }
4761
4762 static int nfs4_get_security_label(struct inode *inode, void *buf,
4763                                         size_t buflen)
4764 {
4765         struct nfs4_exception exception = { };
4766         int err;
4767
4768         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4769                 return -EOPNOTSUPP;
4770
4771         do {
4772                 err = _nfs4_get_security_label(inode, buf, buflen);
4773                 trace_nfs4_get_security_label(inode, err);
4774                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4775                                 &exception);
4776         } while (exception.retry);
4777         return err;
4778 }
4779
4780 static int _nfs4_do_set_security_label(struct inode *inode,
4781                 struct nfs4_label *ilabel,
4782                 struct nfs_fattr *fattr,
4783                 struct nfs4_label *olabel)
4784 {
4785
4786         struct iattr sattr = {0};
4787         struct nfs_server *server = NFS_SERVER(inode);
4788         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4789         struct nfs_setattrargs arg = {
4790                 .fh             = NFS_FH(inode),
4791                 .iap            = &sattr,
4792                 .server         = server,
4793                 .bitmask        = bitmask,
4794                 .label          = ilabel,
4795         };
4796         struct nfs_setattrres res = {
4797                 .fattr          = fattr,
4798                 .label          = olabel,
4799                 .server         = server,
4800         };
4801         struct rpc_message msg = {
4802                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4803                 .rpc_argp       = &arg,
4804                 .rpc_resp       = &res,
4805         };
4806         int status;
4807
4808         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4809
4810         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4811         if (status)
4812                 dprintk("%s failed: %d\n", __func__, status);
4813
4814         return status;
4815 }
4816
4817 static int nfs4_do_set_security_label(struct inode *inode,
4818                 struct nfs4_label *ilabel,
4819                 struct nfs_fattr *fattr,
4820                 struct nfs4_label *olabel)
4821 {
4822         struct nfs4_exception exception = { };
4823         int err;
4824
4825         do {
4826                 err = _nfs4_do_set_security_label(inode, ilabel,
4827                                 fattr, olabel);
4828                 trace_nfs4_set_security_label(inode, err);
4829                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4830                                 &exception);
4831         } while (exception.retry);
4832         return err;
4833 }
4834
4835 static int
4836 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4837 {
4838         struct nfs4_label ilabel, *olabel = NULL;
4839         struct nfs_fattr fattr;
4840         struct rpc_cred *cred;
4841         struct inode *inode = d_inode(dentry);
4842         int status;
4843
4844         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4845                 return -EOPNOTSUPP;
4846
4847         nfs_fattr_init(&fattr);
4848
4849         ilabel.pi = 0;
4850         ilabel.lfs = 0;
4851         ilabel.label = (char *)buf;
4852         ilabel.len = buflen;
4853
4854         cred = rpc_lookup_cred();
4855         if (IS_ERR(cred))
4856                 return PTR_ERR(cred);
4857
4858         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4859         if (IS_ERR(olabel)) {
4860                 status = -PTR_ERR(olabel);
4861                 goto out;
4862         }
4863
4864         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4865         if (status == 0)
4866                 nfs_setsecurity(inode, &fattr, olabel);
4867
4868         nfs4_label_free(olabel);
4869 out:
4870         put_rpccred(cred);
4871         return status;
4872 }
4873 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4874
4875
4876 static int
4877 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4878                         struct nfs4_state *state, long *timeout)
4879 {
4880         struct nfs_client *clp = server->nfs_client;
4881
4882         if (task->tk_status >= 0)
4883                 return 0;
4884         switch(task->tk_status) {
4885                 case -NFS4ERR_DELEG_REVOKED:
4886                 case -NFS4ERR_ADMIN_REVOKED:
4887                 case -NFS4ERR_BAD_STATEID:
4888                 case -NFS4ERR_OPENMODE:
4889                         if (state == NULL)
4890                                 break;
4891                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4892                                 goto recovery_failed;
4893                         goto wait_on_recovery;
4894                 case -NFS4ERR_EXPIRED:
4895                         if (state != NULL) {
4896                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4897                                         goto recovery_failed;
4898                         }
4899                 case -NFS4ERR_STALE_STATEID:
4900                 case -NFS4ERR_STALE_CLIENTID:
4901                         nfs4_schedule_lease_recovery(clp);
4902                         goto wait_on_recovery;
4903                 case -NFS4ERR_MOVED:
4904                         if (nfs4_schedule_migration_recovery(server) < 0)
4905                                 goto recovery_failed;
4906                         goto wait_on_recovery;
4907                 case -NFS4ERR_LEASE_MOVED:
4908                         nfs4_schedule_lease_moved_recovery(clp);
4909                         goto wait_on_recovery;
4910 #if defined(CONFIG_NFS_V4_1)
4911                 case -NFS4ERR_BADSESSION:
4912                 case -NFS4ERR_BADSLOT:
4913                 case -NFS4ERR_BAD_HIGH_SLOT:
4914                 case -NFS4ERR_DEADSESSION:
4915                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4916                 case -NFS4ERR_SEQ_FALSE_RETRY:
4917                 case -NFS4ERR_SEQ_MISORDERED:
4918                         dprintk("%s ERROR %d, Reset session\n", __func__,
4919                                 task->tk_status);
4920                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
4921                         goto wait_on_recovery;
4922 #endif /* CONFIG_NFS_V4_1 */
4923                 case -NFS4ERR_DELAY:
4924                         nfs_inc_server_stats(server, NFSIOS_DELAY);
4925                         rpc_delay(task, nfs4_update_delay(timeout));
4926                         goto restart_call;
4927                 case -NFS4ERR_GRACE:
4928                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
4929                 case -NFS4ERR_RETRY_UNCACHED_REP:
4930                 case -NFS4ERR_OLD_STATEID:
4931                         goto restart_call;
4932         }
4933         task->tk_status = nfs4_map_errors(task->tk_status);
4934         return 0;
4935 recovery_failed:
4936         task->tk_status = -EIO;
4937         return 0;
4938 wait_on_recovery:
4939         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
4940         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
4941                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
4942         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
4943                 goto recovery_failed;
4944 restart_call:
4945         task->tk_status = 0;
4946         return -EAGAIN;
4947 }
4948
4949 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
4950                                     nfs4_verifier *bootverf)
4951 {
4952         __be32 verf[2];
4953
4954         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
4955                 /* An impossible timestamp guarantees this value
4956                  * will never match a generated boot time. */
4957                 verf[0] = 0;
4958                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
4959         } else {
4960                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
4961                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
4962                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
4963         }
4964         memcpy(bootverf->data, verf, sizeof(bootverf->data));
4965 }
4966
4967 static int
4968 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
4969 {
4970         int result;
4971         size_t len;
4972         char *str;
4973         bool retried = false;
4974
4975         if (clp->cl_owner_id != NULL)
4976                 return 0;
4977 retry:
4978         rcu_read_lock();
4979         len = 10 + strlen(clp->cl_ipaddr) + 1 +
4980                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
4981                 1 +
4982                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
4983                 1;
4984         rcu_read_unlock();
4985
4986         if (len > NFS4_OPAQUE_LIMIT + 1)
4987                 return -EINVAL;
4988
4989         /*
4990          * Since this string is allocated at mount time, and held until the
4991          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
4992          * about a memory-reclaim deadlock.
4993          */
4994         str = kmalloc(len, GFP_KERNEL);
4995         if (!str)
4996                 return -ENOMEM;
4997
4998         rcu_read_lock();
4999         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5000                         clp->cl_ipaddr,
5001                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5002                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5003         rcu_read_unlock();
5004
5005         /* Did something change? */
5006         if (result >= len) {
5007                 kfree(str);
5008                 if (retried)
5009                         return -EINVAL;
5010                 retried = true;
5011                 goto retry;
5012         }
5013         clp->cl_owner_id = str;
5014         return 0;
5015 }
5016
5017 static int
5018 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5019 {
5020         int result;
5021         size_t len;
5022         char *str;
5023
5024         len = 10 + 10 + 1 + 10 + 1 +
5025                 strlen(nfs4_client_id_uniquifier) + 1 +
5026                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5027
5028         if (len > NFS4_OPAQUE_LIMIT + 1)
5029                 return -EINVAL;
5030
5031         /*
5032          * Since this string is allocated at mount time, and held until the
5033          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5034          * about a memory-reclaim deadlock.
5035          */
5036         str = kmalloc(len, GFP_KERNEL);
5037         if (!str)
5038                 return -ENOMEM;
5039
5040         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5041                         clp->rpc_ops->version, clp->cl_minorversion,
5042                         nfs4_client_id_uniquifier,
5043                         clp->cl_rpcclient->cl_nodename);
5044         if (result >= len) {
5045                 kfree(str);
5046                 return -EINVAL;
5047         }
5048         clp->cl_owner_id = str;
5049         return 0;
5050 }
5051
5052 static int
5053 nfs4_init_uniform_client_string(struct nfs_client *clp)
5054 {
5055         int result;
5056         size_t len;
5057         char *str;
5058
5059         if (clp->cl_owner_id != NULL)
5060                 return 0;
5061
5062         if (nfs4_client_id_uniquifier[0] != '\0')
5063                 return nfs4_init_uniquifier_client_string(clp);
5064
5065         len = 10 + 10 + 1 + 10 + 1 +
5066                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5067
5068         if (len > NFS4_OPAQUE_LIMIT + 1)
5069                 return -EINVAL;
5070
5071         /*
5072          * Since this string is allocated at mount time, and held until the
5073          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5074          * about a memory-reclaim deadlock.
5075          */
5076         str = kmalloc(len, GFP_KERNEL);
5077         if (!str)
5078                 return -ENOMEM;
5079
5080         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5081                         clp->rpc_ops->version, clp->cl_minorversion,
5082                         clp->cl_rpcclient->cl_nodename);
5083         if (result >= len) {
5084                 kfree(str);
5085                 return -EINVAL;
5086         }
5087         clp->cl_owner_id = str;
5088         return 0;
5089 }
5090
5091 /*
5092  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5093  * services.  Advertise one based on the address family of the
5094  * clientaddr.
5095  */
5096 static unsigned int
5097 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5098 {
5099         if (strchr(clp->cl_ipaddr, ':') != NULL)
5100                 return scnprintf(buf, len, "tcp6");
5101         else
5102                 return scnprintf(buf, len, "tcp");
5103 }
5104
5105 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5106 {
5107         struct nfs4_setclientid *sc = calldata;
5108
5109         if (task->tk_status == 0)
5110                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5111 }
5112
5113 static const struct rpc_call_ops nfs4_setclientid_ops = {
5114         .rpc_call_done = nfs4_setclientid_done,
5115 };
5116
5117 /**
5118  * nfs4_proc_setclientid - Negotiate client ID
5119  * @clp: state data structure
5120  * @program: RPC program for NFSv4 callback service
5121  * @port: IP port number for NFS4 callback service
5122  * @cred: RPC credential to use for this call
5123  * @res: where to place the result
5124  *
5125  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5126  */
5127 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5128                 unsigned short port, struct rpc_cred *cred,
5129                 struct nfs4_setclientid_res *res)
5130 {
5131         nfs4_verifier sc_verifier;
5132         struct nfs4_setclientid setclientid = {
5133                 .sc_verifier = &sc_verifier,
5134                 .sc_prog = program,
5135                 .sc_clnt = clp,
5136         };
5137         struct rpc_message msg = {
5138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5139                 .rpc_argp = &setclientid,
5140                 .rpc_resp = res,
5141                 .rpc_cred = cred,
5142         };
5143         struct rpc_task *task;
5144         struct rpc_task_setup task_setup_data = {
5145                 .rpc_client = clp->cl_rpcclient,
5146                 .rpc_message = &msg,
5147                 .callback_ops = &nfs4_setclientid_ops,
5148                 .callback_data = &setclientid,
5149                 .flags = RPC_TASK_TIMEOUT,
5150         };
5151         int status;
5152
5153         /* nfs_client_id4 */
5154         nfs4_init_boot_verifier(clp, &sc_verifier);
5155
5156         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5157                 status = nfs4_init_uniform_client_string(clp);
5158         else
5159                 status = nfs4_init_nonuniform_client_string(clp);
5160
5161         if (status)
5162                 goto out;
5163
5164         /* cb_client4 */
5165         setclientid.sc_netid_len =
5166                                 nfs4_init_callback_netid(clp,
5167                                                 setclientid.sc_netid,
5168                                                 sizeof(setclientid.sc_netid));
5169         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5170                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5171                                 clp->cl_ipaddr, port >> 8, port & 255);
5172
5173         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5174                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5175                 clp->cl_owner_id);
5176         task = rpc_run_task(&task_setup_data);
5177         if (IS_ERR(task)) {
5178                 status = PTR_ERR(task);
5179                 goto out;
5180         }
5181         status = task->tk_status;
5182         if (setclientid.sc_cred) {
5183                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5184                 put_rpccred(setclientid.sc_cred);
5185         }
5186         rpc_put_task(task);
5187 out:
5188         trace_nfs4_setclientid(clp, status);
5189         dprintk("NFS reply setclientid: %d\n", status);
5190         return status;
5191 }
5192
5193 /**
5194  * nfs4_proc_setclientid_confirm - Confirm client ID
5195  * @clp: state data structure
5196  * @res: result of a previous SETCLIENTID
5197  * @cred: RPC credential to use for this call
5198  *
5199  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5200  */
5201 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5202                 struct nfs4_setclientid_res *arg,
5203                 struct rpc_cred *cred)
5204 {
5205         struct rpc_message msg = {
5206                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5207                 .rpc_argp = arg,
5208                 .rpc_cred = cred,
5209         };
5210         int status;
5211
5212         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5213                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5214                 clp->cl_clientid);
5215         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5216         trace_nfs4_setclientid_confirm(clp, status);
5217         dprintk("NFS reply setclientid_confirm: %d\n", status);
5218         return status;
5219 }
5220
5221 struct nfs4_delegreturndata {
5222         struct nfs4_delegreturnargs args;
5223         struct nfs4_delegreturnres res;
5224         struct nfs_fh fh;
5225         nfs4_stateid stateid;
5226         unsigned long timestamp;
5227         struct nfs_fattr fattr;
5228         int rpc_status;
5229         struct inode *inode;
5230         bool roc;
5231         u32 roc_barrier;
5232 };
5233
5234 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5235 {
5236         struct nfs4_delegreturndata *data = calldata;
5237
5238         if (!nfs4_sequence_done(task, &data->res.seq_res))
5239                 return;
5240
5241         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5242         switch (task->tk_status) {
5243         case 0:
5244                 renew_lease(data->res.server, data->timestamp);
5245         case -NFS4ERR_ADMIN_REVOKED:
5246         case -NFS4ERR_DELEG_REVOKED:
5247         case -NFS4ERR_BAD_STATEID:
5248         case -NFS4ERR_OLD_STATEID:
5249         case -NFS4ERR_STALE_STATEID:
5250         case -NFS4ERR_EXPIRED:
5251                 task->tk_status = 0;
5252                 if (data->roc)
5253                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5254                 break;
5255         default:
5256                 if (nfs4_async_handle_error(task, data->res.server,
5257                                             NULL, NULL) == -EAGAIN) {
5258                         rpc_restart_call_prepare(task);
5259                         return;
5260                 }
5261         }
5262         data->rpc_status = task->tk_status;
5263 }
5264
5265 static void nfs4_delegreturn_release(void *calldata)
5266 {
5267         struct nfs4_delegreturndata *data = calldata;
5268         struct inode *inode = data->inode;
5269
5270         if (inode) {
5271                 if (data->roc)
5272                         pnfs_roc_release(inode);
5273                 nfs_iput_and_deactive(inode);
5274         }
5275         kfree(calldata);
5276 }
5277
5278 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5279 {
5280         struct nfs4_delegreturndata *d_data;
5281
5282         d_data = (struct nfs4_delegreturndata *)data;
5283
5284         if (d_data->roc &&
5285             pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task))
5286                 return;
5287
5288         nfs4_setup_sequence(d_data->res.server,
5289                         &d_data->args.seq_args,
5290                         &d_data->res.seq_res,
5291                         task);
5292 }
5293
5294 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5295         .rpc_call_prepare = nfs4_delegreturn_prepare,
5296         .rpc_call_done = nfs4_delegreturn_done,
5297         .rpc_release = nfs4_delegreturn_release,
5298 };
5299
5300 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5301 {
5302         struct nfs4_delegreturndata *data;
5303         struct nfs_server *server = NFS_SERVER(inode);
5304         struct rpc_task *task;
5305         struct rpc_message msg = {
5306                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5307                 .rpc_cred = cred,
5308         };
5309         struct rpc_task_setup task_setup_data = {
5310                 .rpc_client = server->client,
5311                 .rpc_message = &msg,
5312                 .callback_ops = &nfs4_delegreturn_ops,
5313                 .flags = RPC_TASK_ASYNC,
5314         };
5315         int status = 0;
5316
5317         data = kzalloc(sizeof(*data), GFP_NOFS);
5318         if (data == NULL)
5319                 return -ENOMEM;
5320         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5321         data->args.fhandle = &data->fh;
5322         data->args.stateid = &data->stateid;
5323         data->args.bitmask = server->cache_consistency_bitmask;
5324         nfs_copy_fh(&data->fh, NFS_FH(inode));
5325         nfs4_stateid_copy(&data->stateid, stateid);
5326         data->res.fattr = &data->fattr;
5327         data->res.server = server;
5328         nfs_fattr_init(data->res.fattr);
5329         data->timestamp = jiffies;
5330         data->rpc_status = 0;
5331         data->inode = nfs_igrab_and_active(inode);
5332         if (data->inode)
5333                 data->roc = nfs4_roc(inode);
5334
5335         task_setup_data.callback_data = data;
5336         msg.rpc_argp = &data->args;
5337         msg.rpc_resp = &data->res;
5338         task = rpc_run_task(&task_setup_data);
5339         if (IS_ERR(task))
5340                 return PTR_ERR(task);
5341         if (!issync)
5342                 goto out;
5343         status = nfs4_wait_for_completion_rpc_task(task);
5344         if (status != 0)
5345                 goto out;
5346         status = data->rpc_status;
5347         if (status == 0)
5348                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5349         else
5350                 nfs_refresh_inode(inode, &data->fattr);
5351 out:
5352         rpc_put_task(task);
5353         return status;
5354 }
5355
5356 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5357 {
5358         struct nfs_server *server = NFS_SERVER(inode);
5359         struct nfs4_exception exception = { };
5360         int err;
5361         do {
5362                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5363                 trace_nfs4_delegreturn(inode, err);
5364                 switch (err) {
5365                         case -NFS4ERR_STALE_STATEID:
5366                         case -NFS4ERR_EXPIRED:
5367                         case 0:
5368                                 return 0;
5369                 }
5370                 err = nfs4_handle_exception(server, err, &exception);
5371         } while (exception.retry);
5372         return err;
5373 }
5374
5375 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5376 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5377
5378 /*
5379  * sleep, with exponential backoff, and retry the LOCK operation.
5380  */
5381 static unsigned long
5382 nfs4_set_lock_task_retry(unsigned long timeout)
5383 {
5384         freezable_schedule_timeout_killable_unsafe(timeout);
5385         timeout <<= 1;
5386         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5387                 return NFS4_LOCK_MAXTIMEOUT;
5388         return timeout;
5389 }
5390
5391 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5392 {
5393         struct inode *inode = state->inode;
5394         struct nfs_server *server = NFS_SERVER(inode);
5395         struct nfs_client *clp = server->nfs_client;
5396         struct nfs_lockt_args arg = {
5397                 .fh = NFS_FH(inode),
5398                 .fl = request,
5399         };
5400         struct nfs_lockt_res res = {
5401                 .denied = request,
5402         };
5403         struct rpc_message msg = {
5404                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5405                 .rpc_argp       = &arg,
5406                 .rpc_resp       = &res,
5407                 .rpc_cred       = state->owner->so_cred,
5408         };
5409         struct nfs4_lock_state *lsp;
5410         int status;
5411
5412         arg.lock_owner.clientid = clp->cl_clientid;
5413         status = nfs4_set_lock_state(state, request);
5414         if (status != 0)
5415                 goto out;
5416         lsp = request->fl_u.nfs4_fl.owner;
5417         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5418         arg.lock_owner.s_dev = server->s_dev;
5419         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5420         switch (status) {
5421                 case 0:
5422                         request->fl_type = F_UNLCK;
5423                         break;
5424                 case -NFS4ERR_DENIED:
5425                         status = 0;
5426         }
5427         request->fl_ops->fl_release_private(request);
5428         request->fl_ops = NULL;
5429 out:
5430         return status;
5431 }
5432
5433 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5434 {
5435         struct nfs4_exception exception = { };
5436         int err;
5437
5438         do {
5439                 err = _nfs4_proc_getlk(state, cmd, request);
5440                 trace_nfs4_get_lock(request, state, cmd, err);
5441                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5442                                 &exception);
5443         } while (exception.retry);
5444         return err;
5445 }
5446
5447 static int do_vfs_lock(struct file *file, struct file_lock *fl)
5448 {
5449         int res = 0;
5450         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5451                 case FL_POSIX:
5452                         res = posix_lock_file_wait(file, fl);
5453                         break;
5454                 case FL_FLOCK:
5455                         res = flock_lock_file_wait(file, fl);
5456                         break;
5457                 default:
5458                         BUG();
5459         }
5460         return res;
5461 }
5462
5463 struct nfs4_unlockdata {
5464         struct nfs_locku_args arg;
5465         struct nfs_locku_res res;
5466         struct nfs4_lock_state *lsp;
5467         struct nfs_open_context *ctx;
5468         struct file_lock fl;
5469         const struct nfs_server *server;
5470         unsigned long timestamp;
5471 };
5472
5473 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5474                 struct nfs_open_context *ctx,
5475                 struct nfs4_lock_state *lsp,
5476                 struct nfs_seqid *seqid)
5477 {
5478         struct nfs4_unlockdata *p;
5479         struct inode *inode = lsp->ls_state->inode;
5480
5481         p = kzalloc(sizeof(*p), GFP_NOFS);
5482         if (p == NULL)
5483                 return NULL;
5484         p->arg.fh = NFS_FH(inode);
5485         p->arg.fl = &p->fl;
5486         p->arg.seqid = seqid;
5487         p->res.seqid = seqid;
5488         p->lsp = lsp;
5489         atomic_inc(&lsp->ls_count);
5490         /* Ensure we don't close file until we're done freeing locks! */
5491         p->ctx = get_nfs_open_context(ctx);
5492         get_file(fl->fl_file);
5493         memcpy(&p->fl, fl, sizeof(p->fl));
5494         p->server = NFS_SERVER(inode);
5495         return p;
5496 }
5497
5498 static void nfs4_locku_release_calldata(void *data)
5499 {
5500         struct nfs4_unlockdata *calldata = data;
5501         nfs_free_seqid(calldata->arg.seqid);
5502         nfs4_put_lock_state(calldata->lsp);
5503         put_nfs_open_context(calldata->ctx);
5504         fput(calldata->fl.fl_file);
5505         kfree(calldata);
5506 }
5507
5508 static void nfs4_locku_done(struct rpc_task *task, void *data)
5509 {
5510         struct nfs4_unlockdata *calldata = data;
5511
5512         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5513                 return;
5514         switch (task->tk_status) {
5515                 case 0:
5516                         renew_lease(calldata->server, calldata->timestamp);
5517                         do_vfs_lock(calldata->fl.fl_file, &calldata->fl);
5518                         if (nfs4_update_lock_stateid(calldata->lsp,
5519                                         &calldata->res.stateid))
5520                                 break;
5521                 case -NFS4ERR_BAD_STATEID:
5522                 case -NFS4ERR_OLD_STATEID:
5523                 case -NFS4ERR_STALE_STATEID:
5524                 case -NFS4ERR_EXPIRED:
5525                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5526                                                 &calldata->lsp->ls_stateid))
5527                                 rpc_restart_call_prepare(task);
5528                         break;
5529                 default:
5530                         if (nfs4_async_handle_error(task, calldata->server,
5531                                                     NULL, NULL) == -EAGAIN)
5532                                 rpc_restart_call_prepare(task);
5533         }
5534         nfs_release_seqid(calldata->arg.seqid);
5535 }
5536
5537 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5538 {
5539         struct nfs4_unlockdata *calldata = data;
5540
5541         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5542                 goto out_wait;
5543         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5544         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5545                 /* Note: exit _without_ running nfs4_locku_done */
5546                 goto out_no_action;
5547         }
5548         calldata->timestamp = jiffies;
5549         if (nfs4_setup_sequence(calldata->server,
5550                                 &calldata->arg.seq_args,
5551                                 &calldata->res.seq_res,
5552                                 task) != 0)
5553                 nfs_release_seqid(calldata->arg.seqid);
5554         return;
5555 out_no_action:
5556         task->tk_action = NULL;
5557 out_wait:
5558         nfs4_sequence_done(task, &calldata->res.seq_res);
5559 }
5560
5561 static const struct rpc_call_ops nfs4_locku_ops = {
5562         .rpc_call_prepare = nfs4_locku_prepare,
5563         .rpc_call_done = nfs4_locku_done,
5564         .rpc_release = nfs4_locku_release_calldata,
5565 };
5566
5567 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5568                 struct nfs_open_context *ctx,
5569                 struct nfs4_lock_state *lsp,
5570                 struct nfs_seqid *seqid)
5571 {
5572         struct nfs4_unlockdata *data;
5573         struct rpc_message msg = {
5574                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5575                 .rpc_cred = ctx->cred,
5576         };
5577         struct rpc_task_setup task_setup_data = {
5578                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5579                 .rpc_message = &msg,
5580                 .callback_ops = &nfs4_locku_ops,
5581                 .workqueue = nfsiod_workqueue,
5582                 .flags = RPC_TASK_ASYNC,
5583         };
5584
5585         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5586                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5587
5588         /* Ensure this is an unlock - when canceling a lock, the
5589          * canceled lock is passed in, and it won't be an unlock.
5590          */
5591         fl->fl_type = F_UNLCK;
5592
5593         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5594         if (data == NULL) {
5595                 nfs_free_seqid(seqid);
5596                 return ERR_PTR(-ENOMEM);
5597         }
5598
5599         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5600         msg.rpc_argp = &data->arg;
5601         msg.rpc_resp = &data->res;
5602         task_setup_data.callback_data = data;
5603         return rpc_run_task(&task_setup_data);
5604 }
5605
5606 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5607 {
5608         struct inode *inode = state->inode;
5609         struct nfs4_state_owner *sp = state->owner;
5610         struct nfs_inode *nfsi = NFS_I(inode);
5611         struct nfs_seqid *seqid;
5612         struct nfs4_lock_state *lsp;
5613         struct rpc_task *task;
5614         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5615         int status = 0;
5616         unsigned char fl_flags = request->fl_flags;
5617
5618         status = nfs4_set_lock_state(state, request);
5619         /* Unlock _before_ we do the RPC call */
5620         request->fl_flags |= FL_EXISTS;
5621         /* Exclude nfs_delegation_claim_locks() */
5622         mutex_lock(&sp->so_delegreturn_mutex);
5623         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5624         down_read(&nfsi->rwsem);
5625         if (do_vfs_lock(request->fl_file, request) == -ENOENT) {
5626                 up_read(&nfsi->rwsem);
5627                 mutex_unlock(&sp->so_delegreturn_mutex);
5628                 goto out;
5629         }
5630         up_read(&nfsi->rwsem);
5631         mutex_unlock(&sp->so_delegreturn_mutex);
5632         if (status != 0)
5633                 goto out;
5634         /* Is this a delegated lock? */
5635         lsp = request->fl_u.nfs4_fl.owner;
5636         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5637                 goto out;
5638         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5639         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5640         status = -ENOMEM;
5641         if (IS_ERR(seqid))
5642                 goto out;
5643         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5644         status = PTR_ERR(task);
5645         if (IS_ERR(task))
5646                 goto out;
5647         status = nfs4_wait_for_completion_rpc_task(task);
5648         rpc_put_task(task);
5649 out:
5650         request->fl_flags = fl_flags;
5651         trace_nfs4_unlock(request, state, F_SETLK, status);
5652         return status;
5653 }
5654
5655 struct nfs4_lockdata {
5656         struct nfs_lock_args arg;
5657         struct nfs_lock_res res;
5658         struct nfs4_lock_state *lsp;
5659         struct nfs_open_context *ctx;
5660         struct file_lock fl;
5661         unsigned long timestamp;
5662         int rpc_status;
5663         int cancelled;
5664         struct nfs_server *server;
5665 };
5666
5667 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5668                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5669                 gfp_t gfp_mask)
5670 {
5671         struct nfs4_lockdata *p;
5672         struct inode *inode = lsp->ls_state->inode;
5673         struct nfs_server *server = NFS_SERVER(inode);
5674         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5675
5676         p = kzalloc(sizeof(*p), gfp_mask);
5677         if (p == NULL)
5678                 return NULL;
5679
5680         p->arg.fh = NFS_FH(inode);
5681         p->arg.fl = &p->fl;
5682         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5683         if (IS_ERR(p->arg.open_seqid))
5684                 goto out_free;
5685         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5686         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5687         if (IS_ERR(p->arg.lock_seqid))
5688                 goto out_free_seqid;
5689         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5690         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5691         p->arg.lock_owner.s_dev = server->s_dev;
5692         p->res.lock_seqid = p->arg.lock_seqid;
5693         p->lsp = lsp;
5694         p->server = server;
5695         atomic_inc(&lsp->ls_count);
5696         p->ctx = get_nfs_open_context(ctx);
5697         get_file(fl->fl_file);
5698         memcpy(&p->fl, fl, sizeof(p->fl));
5699         return p;
5700 out_free_seqid:
5701         nfs_free_seqid(p->arg.open_seqid);
5702 out_free:
5703         kfree(p);
5704         return NULL;
5705 }
5706
5707 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5708 {
5709         struct nfs4_lockdata *data = calldata;
5710         struct nfs4_state *state = data->lsp->ls_state;
5711
5712         dprintk("%s: begin!\n", __func__);
5713         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5714                 goto out_wait;
5715         /* Do we need to do an open_to_lock_owner? */
5716         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5717                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5718                         goto out_release_lock_seqid;
5719                 }
5720                 nfs4_stateid_copy(&data->arg.open_stateid,
5721                                 &state->open_stateid);
5722                 data->arg.new_lock_owner = 1;
5723                 data->res.open_seqid = data->arg.open_seqid;
5724         } else {
5725                 data->arg.new_lock_owner = 0;
5726                 nfs4_stateid_copy(&data->arg.lock_stateid,
5727                                 &data->lsp->ls_stateid);
5728         }
5729         if (!nfs4_valid_open_stateid(state)) {
5730                 data->rpc_status = -EBADF;
5731                 task->tk_action = NULL;
5732                 goto out_release_open_seqid;
5733         }
5734         data->timestamp = jiffies;
5735         if (nfs4_setup_sequence(data->server,
5736                                 &data->arg.seq_args,
5737                                 &data->res.seq_res,
5738                                 task) == 0)
5739                 return;
5740 out_release_open_seqid:
5741         nfs_release_seqid(data->arg.open_seqid);
5742 out_release_lock_seqid:
5743         nfs_release_seqid(data->arg.lock_seqid);
5744 out_wait:
5745         nfs4_sequence_done(task, &data->res.seq_res);
5746         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5747 }
5748
5749 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5750 {
5751         struct nfs4_lockdata *data = calldata;
5752         struct nfs4_lock_state *lsp = data->lsp;
5753
5754         dprintk("%s: begin!\n", __func__);
5755
5756         if (!nfs4_sequence_done(task, &data->res.seq_res))
5757                 return;
5758
5759         data->rpc_status = task->tk_status;
5760         switch (task->tk_status) {
5761         case 0:
5762                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5763                                 data->timestamp);
5764                 if (data->arg.new_lock) {
5765                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5766                         if (do_vfs_lock(data->fl.fl_file, &data->fl) < 0) {
5767                                 rpc_restart_call_prepare(task);
5768                                 break;
5769                         }
5770                 }
5771                 if (data->arg.new_lock_owner != 0) {
5772                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5773                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5774                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5775                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5776                         rpc_restart_call_prepare(task);
5777                 break;
5778         case -NFS4ERR_BAD_STATEID:
5779         case -NFS4ERR_OLD_STATEID:
5780         case -NFS4ERR_STALE_STATEID:
5781         case -NFS4ERR_EXPIRED:
5782                 if (data->arg.new_lock_owner != 0) {
5783                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5784                                                 &lsp->ls_state->open_stateid))
5785                                 rpc_restart_call_prepare(task);
5786                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5787                                                 &lsp->ls_stateid))
5788                                 rpc_restart_call_prepare(task);
5789         }
5790         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5791 }
5792
5793 static void nfs4_lock_release(void *calldata)
5794 {
5795         struct nfs4_lockdata *data = calldata;
5796
5797         dprintk("%s: begin!\n", __func__);
5798         nfs_free_seqid(data->arg.open_seqid);
5799         if (data->cancelled != 0) {
5800                 struct rpc_task *task;
5801                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5802                                 data->arg.lock_seqid);
5803                 if (!IS_ERR(task))
5804                         rpc_put_task_async(task);
5805                 dprintk("%s: cancelling lock!\n", __func__);
5806         } else
5807                 nfs_free_seqid(data->arg.lock_seqid);
5808         nfs4_put_lock_state(data->lsp);
5809         put_nfs_open_context(data->ctx);
5810         fput(data->fl.fl_file);
5811         kfree(data);
5812         dprintk("%s: done!\n", __func__);
5813 }
5814
5815 static const struct rpc_call_ops nfs4_lock_ops = {
5816         .rpc_call_prepare = nfs4_lock_prepare,
5817         .rpc_call_done = nfs4_lock_done,
5818         .rpc_release = nfs4_lock_release,
5819 };
5820
5821 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5822 {
5823         switch (error) {
5824         case -NFS4ERR_ADMIN_REVOKED:
5825         case -NFS4ERR_BAD_STATEID:
5826                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5827                 if (new_lock_owner != 0 ||
5828                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5829                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5830                 break;
5831         case -NFS4ERR_STALE_STATEID:
5832                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5833         case -NFS4ERR_EXPIRED:
5834                 nfs4_schedule_lease_recovery(server->nfs_client);
5835         };
5836 }
5837
5838 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5839 {
5840         struct nfs4_lockdata *data;
5841         struct rpc_task *task;
5842         struct rpc_message msg = {
5843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5844                 .rpc_cred = state->owner->so_cred,
5845         };
5846         struct rpc_task_setup task_setup_data = {
5847                 .rpc_client = NFS_CLIENT(state->inode),
5848                 .rpc_message = &msg,
5849                 .callback_ops = &nfs4_lock_ops,
5850                 .workqueue = nfsiod_workqueue,
5851                 .flags = RPC_TASK_ASYNC,
5852         };
5853         int ret;
5854
5855         dprintk("%s: begin!\n", __func__);
5856         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5857                         fl->fl_u.nfs4_fl.owner,
5858                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5859         if (data == NULL)
5860                 return -ENOMEM;
5861         if (IS_SETLKW(cmd))
5862                 data->arg.block = 1;
5863         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5864         msg.rpc_argp = &data->arg;
5865         msg.rpc_resp = &data->res;
5866         task_setup_data.callback_data = data;
5867         if (recovery_type > NFS_LOCK_NEW) {
5868                 if (recovery_type == NFS_LOCK_RECLAIM)
5869                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5870                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5871         } else
5872                 data->arg.new_lock = 1;
5873         task = rpc_run_task(&task_setup_data);
5874         if (IS_ERR(task))
5875                 return PTR_ERR(task);
5876         ret = nfs4_wait_for_completion_rpc_task(task);
5877         if (ret == 0) {
5878                 ret = data->rpc_status;
5879                 if (ret)
5880                         nfs4_handle_setlk_error(data->server, data->lsp,
5881                                         data->arg.new_lock_owner, ret);
5882         } else
5883                 data->cancelled = 1;
5884         rpc_put_task(task);
5885         dprintk("%s: done, ret = %d!\n", __func__, ret);
5886         return ret;
5887 }
5888
5889 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5890 {
5891         struct nfs_server *server = NFS_SERVER(state->inode);
5892         struct nfs4_exception exception = {
5893                 .inode = state->inode,
5894         };
5895         int err;
5896
5897         do {
5898                 /* Cache the lock if possible... */
5899                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5900                         return 0;
5901                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5902                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5903                 if (err != -NFS4ERR_DELAY)
5904                         break;
5905                 nfs4_handle_exception(server, err, &exception);
5906         } while (exception.retry);
5907         return err;
5908 }
5909
5910 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5911 {
5912         struct nfs_server *server = NFS_SERVER(state->inode);
5913         struct nfs4_exception exception = {
5914                 .inode = state->inode,
5915         };
5916         int err;
5917
5918         err = nfs4_set_lock_state(state, request);
5919         if (err != 0)
5920                 return err;
5921         if (!recover_lost_locks) {
5922                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5923                 return 0;
5924         }
5925         do {
5926                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5927                         return 0;
5928                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5929                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5930                 switch (err) {
5931                 default:
5932                         goto out;
5933                 case -NFS4ERR_GRACE:
5934                 case -NFS4ERR_DELAY:
5935                         nfs4_handle_exception(server, err, &exception);
5936                         err = 0;
5937                 }
5938         } while (exception.retry);
5939 out:
5940         return err;
5941 }
5942
5943 #if defined(CONFIG_NFS_V4_1)
5944 /**
5945  * nfs41_check_expired_locks - possibly free a lock stateid
5946  *
5947  * @state: NFSv4 state for an inode
5948  *
5949  * Returns NFS_OK if recovery for this stateid is now finished.
5950  * Otherwise a negative NFS4ERR value is returned.
5951  */
5952 static int nfs41_check_expired_locks(struct nfs4_state *state)
5953 {
5954         int status, ret = -NFS4ERR_BAD_STATEID;
5955         struct nfs4_lock_state *lsp;
5956         struct nfs_server *server = NFS_SERVER(state->inode);
5957
5958         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
5959                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
5960                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
5961
5962                         status = nfs41_test_stateid(server,
5963                                         &lsp->ls_stateid,
5964                                         cred);
5965                         trace_nfs4_test_lock_stateid(state, lsp, status);
5966                         if (status != NFS_OK) {
5967                                 /* Free the stateid unless the server
5968                                  * informs us the stateid is unrecognized. */
5969                                 if (status != -NFS4ERR_BAD_STATEID)
5970                                         nfs41_free_stateid(server,
5971                                                         &lsp->ls_stateid,
5972                                                         cred);
5973                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5974                                 ret = status;
5975                         }
5976                 }
5977         };
5978
5979         return ret;
5980 }
5981
5982 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
5983 {
5984         int status = NFS_OK;
5985
5986         if (test_bit(LK_STATE_IN_USE, &state->flags))
5987                 status = nfs41_check_expired_locks(state);
5988         if (status != NFS_OK)
5989                 status = nfs4_lock_expired(state, request);
5990         return status;
5991 }
5992 #endif
5993
5994 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5995 {
5996         struct nfs_inode *nfsi = NFS_I(state->inode);
5997         unsigned char fl_flags = request->fl_flags;
5998         int status = -ENOLCK;
5999
6000         if ((fl_flags & FL_POSIX) &&
6001                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6002                 goto out;
6003         /* Is this a delegated open? */
6004         status = nfs4_set_lock_state(state, request);
6005         if (status != 0)
6006                 goto out;
6007         request->fl_flags |= FL_ACCESS;
6008         status = do_vfs_lock(request->fl_file, request);
6009         if (status < 0)
6010                 goto out;
6011         down_read(&nfsi->rwsem);
6012         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6013                 /* Yes: cache locks! */
6014                 /* ...but avoid races with delegation recall... */
6015                 request->fl_flags = fl_flags & ~FL_SLEEP;
6016                 status = do_vfs_lock(request->fl_file, request);
6017                 up_read(&nfsi->rwsem);
6018                 goto out;
6019         }
6020         up_read(&nfsi->rwsem);
6021         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6022 out:
6023         request->fl_flags = fl_flags;
6024         return status;
6025 }
6026
6027 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6028 {
6029         struct nfs4_exception exception = {
6030                 .state = state,
6031                 .inode = state->inode,
6032         };
6033         int err;
6034
6035         do {
6036                 err = _nfs4_proc_setlk(state, cmd, request);
6037                 trace_nfs4_set_lock(request, state, cmd, err);
6038                 if (err == -NFS4ERR_DENIED)
6039                         err = -EAGAIN;
6040                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6041                                 err, &exception);
6042         } while (exception.retry);
6043         return err;
6044 }
6045
6046 static int
6047 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6048 {
6049         struct nfs_open_context *ctx;
6050         struct nfs4_state *state;
6051         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6052         int status;
6053
6054         /* verify open state */
6055         ctx = nfs_file_open_context(filp);
6056         state = ctx->state;
6057
6058         if (request->fl_start < 0 || request->fl_end < 0)
6059                 return -EINVAL;
6060
6061         if (IS_GETLK(cmd)) {
6062                 if (state != NULL)
6063                         return nfs4_proc_getlk(state, F_GETLK, request);
6064                 return 0;
6065         }
6066
6067         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6068                 return -EINVAL;
6069
6070         if (request->fl_type == F_UNLCK) {
6071                 if (state != NULL)
6072                         return nfs4_proc_unlck(state, cmd, request);
6073                 return 0;
6074         }
6075
6076         if (state == NULL)
6077                 return -ENOLCK;
6078         /*
6079          * Don't rely on the VFS having checked the file open mode,
6080          * since it won't do this for flock() locks.
6081          */
6082         switch (request->fl_type) {
6083         case F_RDLCK:
6084                 if (!(filp->f_mode & FMODE_READ))
6085                         return -EBADF;
6086                 break;
6087         case F_WRLCK:
6088                 if (!(filp->f_mode & FMODE_WRITE))
6089                         return -EBADF;
6090         }
6091
6092         do {
6093                 status = nfs4_proc_setlk(state, cmd, request);
6094                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6095                         break;
6096                 timeout = nfs4_set_lock_task_retry(timeout);
6097                 status = -ERESTARTSYS;
6098                 if (signalled())
6099                         break;
6100         } while(status < 0);
6101         return status;
6102 }
6103
6104 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6105 {
6106         struct nfs_server *server = NFS_SERVER(state->inode);
6107         int err;
6108
6109         err = nfs4_set_lock_state(state, fl);
6110         if (err != 0)
6111                 return err;
6112         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6113         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6114 }
6115
6116 struct nfs_release_lockowner_data {
6117         struct nfs4_lock_state *lsp;
6118         struct nfs_server *server;
6119         struct nfs_release_lockowner_args args;
6120         struct nfs_release_lockowner_res res;
6121         unsigned long timestamp;
6122 };
6123
6124 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6125 {
6126         struct nfs_release_lockowner_data *data = calldata;
6127         struct nfs_server *server = data->server;
6128         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6129                              &data->args.seq_args, &data->res.seq_res, task);
6130         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6131         data->timestamp = jiffies;
6132 }
6133
6134 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6135 {
6136         struct nfs_release_lockowner_data *data = calldata;
6137         struct nfs_server *server = data->server;
6138
6139         nfs40_sequence_done(task, &data->res.seq_res);
6140
6141         switch (task->tk_status) {
6142         case 0:
6143                 renew_lease(server, data->timestamp);
6144                 break;
6145         case -NFS4ERR_STALE_CLIENTID:
6146         case -NFS4ERR_EXPIRED:
6147                 nfs4_schedule_lease_recovery(server->nfs_client);
6148                 break;
6149         case -NFS4ERR_LEASE_MOVED:
6150         case -NFS4ERR_DELAY:
6151                 if (nfs4_async_handle_error(task, server,
6152                                             NULL, NULL) == -EAGAIN)
6153                         rpc_restart_call_prepare(task);
6154         }
6155 }
6156
6157 static void nfs4_release_lockowner_release(void *calldata)
6158 {
6159         struct nfs_release_lockowner_data *data = calldata;
6160         nfs4_free_lock_state(data->server, data->lsp);
6161         kfree(calldata);
6162 }
6163
6164 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6165         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6166         .rpc_call_done = nfs4_release_lockowner_done,
6167         .rpc_release = nfs4_release_lockowner_release,
6168 };
6169
6170 static void
6171 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6172 {
6173         struct nfs_release_lockowner_data *data;
6174         struct rpc_message msg = {
6175                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6176         };
6177
6178         if (server->nfs_client->cl_mvops->minor_version != 0)
6179                 return;
6180
6181         data = kmalloc(sizeof(*data), GFP_NOFS);
6182         if (!data)
6183                 return;
6184         data->lsp = lsp;
6185         data->server = server;
6186         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6187         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6188         data->args.lock_owner.s_dev = server->s_dev;
6189
6190         msg.rpc_argp = &data->args;
6191         msg.rpc_resp = &data->res;
6192         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6193         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6194 }
6195
6196 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6197
6198 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6199                                    const void *buf, size_t buflen,
6200                                    int flags, int type)
6201 {
6202         if (strcmp(key, "") != 0)
6203                 return -EINVAL;
6204
6205         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6206 }
6207
6208 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6209                                    void *buf, size_t buflen, int type)
6210 {
6211         if (strcmp(key, "") != 0)
6212                 return -EINVAL;
6213
6214         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6215 }
6216
6217 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6218                                        size_t list_len, const char *name,
6219                                        size_t name_len, int type)
6220 {
6221         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6222
6223         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6224                 return 0;
6225
6226         if (list && len <= list_len)
6227                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6228         return len;
6229 }
6230
6231 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6232 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6233 {
6234         return server->caps & NFS_CAP_SECURITY_LABEL;
6235 }
6236
6237 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6238                                    const void *buf, size_t buflen,
6239                                    int flags, int type)
6240 {
6241         if (security_ismaclabel(key))
6242                 return nfs4_set_security_label(dentry, buf, buflen);
6243
6244         return -EOPNOTSUPP;
6245 }
6246
6247 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6248                                    void *buf, size_t buflen, int type)
6249 {
6250         if (security_ismaclabel(key))
6251                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6252         return -EOPNOTSUPP;
6253 }
6254
6255 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6256                                        size_t list_len, const char *name,
6257                                        size_t name_len, int type)
6258 {
6259         size_t len = 0;
6260
6261         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6262                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6263                 if (list && len <= list_len)
6264                         security_inode_listsecurity(d_inode(dentry), list, len);
6265         }
6266         return len;
6267 }
6268
6269 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6270         .prefix = XATTR_SECURITY_PREFIX,
6271         .list   = nfs4_xattr_list_nfs4_label,
6272         .get    = nfs4_xattr_get_nfs4_label,
6273         .set    = nfs4_xattr_set_nfs4_label,
6274 };
6275 #endif
6276
6277
6278 /*
6279  * nfs_fhget will use either the mounted_on_fileid or the fileid
6280  */
6281 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6282 {
6283         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6284                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6285               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6286               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6287                 return;
6288
6289         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6290                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6291         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6292         fattr->nlink = 2;
6293 }
6294
6295 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6296                                    const struct qstr *name,
6297                                    struct nfs4_fs_locations *fs_locations,
6298                                    struct page *page)
6299 {
6300         struct nfs_server *server = NFS_SERVER(dir);
6301         u32 bitmask[3] = {
6302                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6303         };
6304         struct nfs4_fs_locations_arg args = {
6305                 .dir_fh = NFS_FH(dir),
6306                 .name = name,
6307                 .page = page,
6308                 .bitmask = bitmask,
6309         };
6310         struct nfs4_fs_locations_res res = {
6311                 .fs_locations = fs_locations,
6312         };
6313         struct rpc_message msg = {
6314                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6315                 .rpc_argp = &args,
6316                 .rpc_resp = &res,
6317         };
6318         int status;
6319
6320         dprintk("%s: start\n", __func__);
6321
6322         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6323          * is not supported */
6324         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6325                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6326         else
6327                 bitmask[0] |= FATTR4_WORD0_FILEID;
6328
6329         nfs_fattr_init(&fs_locations->fattr);
6330         fs_locations->server = server;
6331         fs_locations->nlocations = 0;
6332         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6333         dprintk("%s: returned status = %d\n", __func__, status);
6334         return status;
6335 }
6336
6337 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6338                            const struct qstr *name,
6339                            struct nfs4_fs_locations *fs_locations,
6340                            struct page *page)
6341 {
6342         struct nfs4_exception exception = { };
6343         int err;
6344         do {
6345                 err = _nfs4_proc_fs_locations(client, dir, name,
6346                                 fs_locations, page);
6347                 trace_nfs4_get_fs_locations(dir, name, err);
6348                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6349                                 &exception);
6350         } while (exception.retry);
6351         return err;
6352 }
6353
6354 /*
6355  * This operation also signals the server that this client is
6356  * performing migration recovery.  The server can stop returning
6357  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6358  * appended to this compound to identify the client ID which is
6359  * performing recovery.
6360  */
6361 static int _nfs40_proc_get_locations(struct inode *inode,
6362                                      struct nfs4_fs_locations *locations,
6363                                      struct page *page, struct rpc_cred *cred)
6364 {
6365         struct nfs_server *server = NFS_SERVER(inode);
6366         struct rpc_clnt *clnt = server->client;
6367         u32 bitmask[2] = {
6368                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6369         };
6370         struct nfs4_fs_locations_arg args = {
6371                 .clientid       = server->nfs_client->cl_clientid,
6372                 .fh             = NFS_FH(inode),
6373                 .page           = page,
6374                 .bitmask        = bitmask,
6375                 .migration      = 1,            /* skip LOOKUP */
6376                 .renew          = 1,            /* append RENEW */
6377         };
6378         struct nfs4_fs_locations_res res = {
6379                 .fs_locations   = locations,
6380                 .migration      = 1,
6381                 .renew          = 1,
6382         };
6383         struct rpc_message msg = {
6384                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6385                 .rpc_argp       = &args,
6386                 .rpc_resp       = &res,
6387                 .rpc_cred       = cred,
6388         };
6389         unsigned long now = jiffies;
6390         int status;
6391
6392         nfs_fattr_init(&locations->fattr);
6393         locations->server = server;
6394         locations->nlocations = 0;
6395
6396         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6397         nfs4_set_sequence_privileged(&args.seq_args);
6398         status = nfs4_call_sync_sequence(clnt, server, &msg,
6399                                         &args.seq_args, &res.seq_res);
6400         if (status)
6401                 return status;
6402
6403         renew_lease(server, now);
6404         return 0;
6405 }
6406
6407 #ifdef CONFIG_NFS_V4_1
6408
6409 /*
6410  * This operation also signals the server that this client is
6411  * performing migration recovery.  The server can stop asserting
6412  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6413  * performing this operation is identified in the SEQUENCE
6414  * operation in this compound.
6415  *
6416  * When the client supports GETATTR(fs_locations_info), it can
6417  * be plumbed in here.
6418  */
6419 static int _nfs41_proc_get_locations(struct inode *inode,
6420                                      struct nfs4_fs_locations *locations,
6421                                      struct page *page, struct rpc_cred *cred)
6422 {
6423         struct nfs_server *server = NFS_SERVER(inode);
6424         struct rpc_clnt *clnt = server->client;
6425         u32 bitmask[2] = {
6426                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6427         };
6428         struct nfs4_fs_locations_arg args = {
6429                 .fh             = NFS_FH(inode),
6430                 .page           = page,
6431                 .bitmask        = bitmask,
6432                 .migration      = 1,            /* skip LOOKUP */
6433         };
6434         struct nfs4_fs_locations_res res = {
6435                 .fs_locations   = locations,
6436                 .migration      = 1,
6437         };
6438         struct rpc_message msg = {
6439                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6440                 .rpc_argp       = &args,
6441                 .rpc_resp       = &res,
6442                 .rpc_cred       = cred,
6443         };
6444         int status;
6445
6446         nfs_fattr_init(&locations->fattr);
6447         locations->server = server;
6448         locations->nlocations = 0;
6449
6450         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6451         nfs4_set_sequence_privileged(&args.seq_args);
6452         status = nfs4_call_sync_sequence(clnt, server, &msg,
6453                                         &args.seq_args, &res.seq_res);
6454         if (status == NFS4_OK &&
6455             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6456                 status = -NFS4ERR_LEASE_MOVED;
6457         return status;
6458 }
6459
6460 #endif  /* CONFIG_NFS_V4_1 */
6461
6462 /**
6463  * nfs4_proc_get_locations - discover locations for a migrated FSID
6464  * @inode: inode on FSID that is migrating
6465  * @locations: result of query
6466  * @page: buffer
6467  * @cred: credential to use for this operation
6468  *
6469  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6470  * operation failed, or a negative errno if a local error occurred.
6471  *
6472  * On success, "locations" is filled in, but if the server has
6473  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6474  * asserted.
6475  *
6476  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6477  * from this client that require migration recovery.
6478  */
6479 int nfs4_proc_get_locations(struct inode *inode,
6480                             struct nfs4_fs_locations *locations,
6481                             struct page *page, struct rpc_cred *cred)
6482 {
6483         struct nfs_server *server = NFS_SERVER(inode);
6484         struct nfs_client *clp = server->nfs_client;
6485         const struct nfs4_mig_recovery_ops *ops =
6486                                         clp->cl_mvops->mig_recovery_ops;
6487         struct nfs4_exception exception = { };
6488         int status;
6489
6490         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6491                 (unsigned long long)server->fsid.major,
6492                 (unsigned long long)server->fsid.minor,
6493                 clp->cl_hostname);
6494         nfs_display_fhandle(NFS_FH(inode), __func__);
6495
6496         do {
6497                 status = ops->get_locations(inode, locations, page, cred);
6498                 if (status != -NFS4ERR_DELAY)
6499                         break;
6500                 nfs4_handle_exception(server, status, &exception);
6501         } while (exception.retry);
6502         return status;
6503 }
6504
6505 /*
6506  * This operation also signals the server that this client is
6507  * performing "lease moved" recovery.  The server can stop
6508  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6509  * is appended to this compound to identify the client ID which is
6510  * performing recovery.
6511  */
6512 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6513 {
6514         struct nfs_server *server = NFS_SERVER(inode);
6515         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6516         struct rpc_clnt *clnt = server->client;
6517         struct nfs4_fsid_present_arg args = {
6518                 .fh             = NFS_FH(inode),
6519                 .clientid       = clp->cl_clientid,
6520                 .renew          = 1,            /* append RENEW */
6521         };
6522         struct nfs4_fsid_present_res res = {
6523                 .renew          = 1,
6524         };
6525         struct rpc_message msg = {
6526                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6527                 .rpc_argp       = &args,
6528                 .rpc_resp       = &res,
6529                 .rpc_cred       = cred,
6530         };
6531         unsigned long now = jiffies;
6532         int status;
6533
6534         res.fh = nfs_alloc_fhandle();
6535         if (res.fh == NULL)
6536                 return -ENOMEM;
6537
6538         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6539         nfs4_set_sequence_privileged(&args.seq_args);
6540         status = nfs4_call_sync_sequence(clnt, server, &msg,
6541                                                 &args.seq_args, &res.seq_res);
6542         nfs_free_fhandle(res.fh);
6543         if (status)
6544                 return status;
6545
6546         do_renew_lease(clp, now);
6547         return 0;
6548 }
6549
6550 #ifdef CONFIG_NFS_V4_1
6551
6552 /*
6553  * This operation also signals the server that this client is
6554  * performing "lease moved" recovery.  The server can stop asserting
6555  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6556  * this operation is identified in the SEQUENCE operation in this
6557  * compound.
6558  */
6559 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6560 {
6561         struct nfs_server *server = NFS_SERVER(inode);
6562         struct rpc_clnt *clnt = server->client;
6563         struct nfs4_fsid_present_arg args = {
6564                 .fh             = NFS_FH(inode),
6565         };
6566         struct nfs4_fsid_present_res res = {
6567         };
6568         struct rpc_message msg = {
6569                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6570                 .rpc_argp       = &args,
6571                 .rpc_resp       = &res,
6572                 .rpc_cred       = cred,
6573         };
6574         int status;
6575
6576         res.fh = nfs_alloc_fhandle();
6577         if (res.fh == NULL)
6578                 return -ENOMEM;
6579
6580         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6581         nfs4_set_sequence_privileged(&args.seq_args);
6582         status = nfs4_call_sync_sequence(clnt, server, &msg,
6583                                                 &args.seq_args, &res.seq_res);
6584         nfs_free_fhandle(res.fh);
6585         if (status == NFS4_OK &&
6586             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6587                 status = -NFS4ERR_LEASE_MOVED;
6588         return status;
6589 }
6590
6591 #endif  /* CONFIG_NFS_V4_1 */
6592
6593 /**
6594  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6595  * @inode: inode on FSID to check
6596  * @cred: credential to use for this operation
6597  *
6598  * Server indicates whether the FSID is present, moved, or not
6599  * recognized.  This operation is necessary to clear a LEASE_MOVED
6600  * condition for this client ID.
6601  *
6602  * Returns NFS4_OK if the FSID is present on this server,
6603  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6604  *  NFS4ERR code if some error occurred on the server, or a
6605  *  negative errno if a local failure occurred.
6606  */
6607 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6608 {
6609         struct nfs_server *server = NFS_SERVER(inode);
6610         struct nfs_client *clp = server->nfs_client;
6611         const struct nfs4_mig_recovery_ops *ops =
6612                                         clp->cl_mvops->mig_recovery_ops;
6613         struct nfs4_exception exception = { };
6614         int status;
6615
6616         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6617                 (unsigned long long)server->fsid.major,
6618                 (unsigned long long)server->fsid.minor,
6619                 clp->cl_hostname);
6620         nfs_display_fhandle(NFS_FH(inode), __func__);
6621
6622         do {
6623                 status = ops->fsid_present(inode, cred);
6624                 if (status != -NFS4ERR_DELAY)
6625                         break;
6626                 nfs4_handle_exception(server, status, &exception);
6627         } while (exception.retry);
6628         return status;
6629 }
6630
6631 /**
6632  * If 'use_integrity' is true and the state managment nfs_client
6633  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6634  * and the machine credential as per RFC3530bis and RFC5661 Security
6635  * Considerations sections. Otherwise, just use the user cred with the
6636  * filesystem's rpc_client.
6637  */
6638 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6639 {
6640         int status;
6641         struct nfs4_secinfo_arg args = {
6642                 .dir_fh = NFS_FH(dir),
6643                 .name   = name,
6644         };
6645         struct nfs4_secinfo_res res = {
6646                 .flavors     = flavors,
6647         };
6648         struct rpc_message msg = {
6649                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6650                 .rpc_argp = &args,
6651                 .rpc_resp = &res,
6652         };
6653         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6654         struct rpc_cred *cred = NULL;
6655
6656         if (use_integrity) {
6657                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6658                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6659                 msg.rpc_cred = cred;
6660         }
6661
6662         dprintk("NFS call  secinfo %s\n", name->name);
6663
6664         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6665                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6666
6667         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6668                                 &res.seq_res, 0);
6669         dprintk("NFS reply  secinfo: %d\n", status);
6670
6671         if (cred)
6672                 put_rpccred(cred);
6673
6674         return status;
6675 }
6676
6677 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6678                       struct nfs4_secinfo_flavors *flavors)
6679 {
6680         struct nfs4_exception exception = { };
6681         int err;
6682         do {
6683                 err = -NFS4ERR_WRONGSEC;
6684
6685                 /* try to use integrity protection with machine cred */
6686                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6687                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6688
6689                 /*
6690                  * if unable to use integrity protection, or SECINFO with
6691                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6692                  * disallowed by spec, but exists in deployed servers) use
6693                  * the current filesystem's rpc_client and the user cred.
6694                  */
6695                 if (err == -NFS4ERR_WRONGSEC)
6696                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6697
6698                 trace_nfs4_secinfo(dir, name, err);
6699                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6700                                 &exception);
6701         } while (exception.retry);
6702         return err;
6703 }
6704
6705 #ifdef CONFIG_NFS_V4_1
6706 /*
6707  * Check the exchange flags returned by the server for invalid flags, having
6708  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6709  * DS flags set.
6710  */
6711 static int nfs4_check_cl_exchange_flags(u32 flags)
6712 {
6713         if (flags & ~EXCHGID4_FLAG_MASK_R)
6714                 goto out_inval;
6715         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6716             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6717                 goto out_inval;
6718         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6719                 goto out_inval;
6720         return NFS_OK;
6721 out_inval:
6722         return -NFS4ERR_INVAL;
6723 }
6724
6725 static bool
6726 nfs41_same_server_scope(struct nfs41_server_scope *a,
6727                         struct nfs41_server_scope *b)
6728 {
6729         if (a->server_scope_sz == b->server_scope_sz &&
6730             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6731                 return true;
6732
6733         return false;
6734 }
6735
6736 /*
6737  * nfs4_proc_bind_conn_to_session()
6738  *
6739  * The 4.1 client currently uses the same TCP connection for the
6740  * fore and backchannel.
6741  */
6742 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6743 {
6744         int status;
6745         struct nfs41_bind_conn_to_session_args args = {
6746                 .client = clp,
6747                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6748         };
6749         struct nfs41_bind_conn_to_session_res res;
6750         struct rpc_message msg = {
6751                 .rpc_proc =
6752                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6753                 .rpc_argp = &args,
6754                 .rpc_resp = &res,
6755                 .rpc_cred = cred,
6756         };
6757
6758         dprintk("--> %s\n", __func__);
6759
6760         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6761         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6762                 args.dir = NFS4_CDFC4_FORE;
6763
6764         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6765         trace_nfs4_bind_conn_to_session(clp, status);
6766         if (status == 0) {
6767                 if (memcmp(res.sessionid.data,
6768                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6769                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6770                         status = -EIO;
6771                         goto out;
6772                 }
6773                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6774                         dprintk("NFS: %s: Unexpected direction from server\n",
6775                                 __func__);
6776                         status = -EIO;
6777                         goto out;
6778                 }
6779                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6780                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6781                                 __func__);
6782                         status = -EIO;
6783                         goto out;
6784                 }
6785         }
6786 out:
6787         dprintk("<-- %s status= %d\n", __func__, status);
6788         return status;
6789 }
6790
6791 /*
6792  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6793  * and operations we'd like to see to enable certain features in the allow map
6794  */
6795 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6796         .how = SP4_MACH_CRED,
6797         .enforce.u.words = {
6798                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6799                       1 << (OP_EXCHANGE_ID - 32) |
6800                       1 << (OP_CREATE_SESSION - 32) |
6801                       1 << (OP_DESTROY_SESSION - 32) |
6802                       1 << (OP_DESTROY_CLIENTID - 32)
6803         },
6804         .allow.u.words = {
6805                 [0] = 1 << (OP_CLOSE) |
6806                       1 << (OP_LOCKU) |
6807                       1 << (OP_COMMIT),
6808                 [1] = 1 << (OP_SECINFO - 32) |
6809                       1 << (OP_SECINFO_NO_NAME - 32) |
6810                       1 << (OP_TEST_STATEID - 32) |
6811                       1 << (OP_FREE_STATEID - 32) |
6812                       1 << (OP_WRITE - 32)
6813         }
6814 };
6815
6816 /*
6817  * Select the state protection mode for client `clp' given the server results
6818  * from exchange_id in `sp'.
6819  *
6820  * Returns 0 on success, negative errno otherwise.
6821  */
6822 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6823                                  struct nfs41_state_protection *sp)
6824 {
6825         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6826                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6827                       1 << (OP_EXCHANGE_ID - 32) |
6828                       1 << (OP_CREATE_SESSION - 32) |
6829                       1 << (OP_DESTROY_SESSION - 32) |
6830                       1 << (OP_DESTROY_CLIENTID - 32)
6831         };
6832         unsigned int i;
6833
6834         if (sp->how == SP4_MACH_CRED) {
6835                 /* Print state protect result */
6836                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6837                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6838                         if (test_bit(i, sp->enforce.u.longs))
6839                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6840                         if (test_bit(i, sp->allow.u.longs))
6841                                 dfprintk(MOUNT, "  allow op %d\n", i);
6842                 }
6843
6844                 /* make sure nothing is on enforce list that isn't supported */
6845                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6846                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6847                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6848                                 return -EINVAL;
6849                         }
6850                 }
6851
6852                 /*
6853                  * Minimal mode - state operations are allowed to use machine
6854                  * credential.  Note this already happens by default, so the
6855                  * client doesn't have to do anything more than the negotiation.
6856                  *
6857                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6858                  *       we're already using the machine cred for exchange_id
6859                  *       and will never use a different cred.
6860                  */
6861                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6862                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6863                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6864                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6865                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6866                         dfprintk(MOUNT, "  minimal mode enabled\n");
6867                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6868                 } else {
6869                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6870                         return -EINVAL;
6871                 }
6872
6873                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6874                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6875                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6876                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6877                 }
6878
6879                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6880                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6881                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6882                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6883                 }
6884
6885                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6886                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6887                         dfprintk(MOUNT, "  stateid mode enabled\n");
6888                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6889                 }
6890
6891                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6892                         dfprintk(MOUNT, "  write mode enabled\n");
6893                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6894                 }
6895
6896                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6897                         dfprintk(MOUNT, "  commit mode enabled\n");
6898                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6899                 }
6900         }
6901
6902         return 0;
6903 }
6904
6905 /*
6906  * _nfs4_proc_exchange_id()
6907  *
6908  * Wrapper for EXCHANGE_ID operation.
6909  */
6910 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6911         u32 sp4_how)
6912 {
6913         nfs4_verifier verifier;
6914         struct nfs41_exchange_id_args args = {
6915                 .verifier = &verifier,
6916                 .client = clp,
6917 #ifdef CONFIG_NFS_V4_1_MIGRATION
6918                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6919                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6920                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6921 #else
6922                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6923                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6924 #endif
6925         };
6926         struct nfs41_exchange_id_res res = {
6927                 0
6928         };
6929         int status;
6930         struct rpc_message msg = {
6931                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6932                 .rpc_argp = &args,
6933                 .rpc_resp = &res,
6934                 .rpc_cred = cred,
6935         };
6936
6937         nfs4_init_boot_verifier(clp, &verifier);
6938
6939         status = nfs4_init_uniform_client_string(clp);
6940         if (status)
6941                 goto out;
6942
6943         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
6944                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6945                 clp->cl_owner_id);
6946
6947         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
6948                                         GFP_NOFS);
6949         if (unlikely(res.server_owner == NULL)) {
6950                 status = -ENOMEM;
6951                 goto out;
6952         }
6953
6954         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
6955                                         GFP_NOFS);
6956         if (unlikely(res.server_scope == NULL)) {
6957                 status = -ENOMEM;
6958                 goto out_server_owner;
6959         }
6960
6961         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
6962         if (unlikely(res.impl_id == NULL)) {
6963                 status = -ENOMEM;
6964                 goto out_server_scope;
6965         }
6966
6967         switch (sp4_how) {
6968         case SP4_NONE:
6969                 args.state_protect.how = SP4_NONE;
6970                 break;
6971
6972         case SP4_MACH_CRED:
6973                 args.state_protect = nfs4_sp4_mach_cred_request;
6974                 break;
6975
6976         default:
6977                 /* unsupported! */
6978                 WARN_ON_ONCE(1);
6979                 status = -EINVAL;
6980                 goto out_impl_id;
6981         }
6982
6983         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6984         trace_nfs4_exchange_id(clp, status);
6985         if (status == 0)
6986                 status = nfs4_check_cl_exchange_flags(res.flags);
6987
6988         if (status == 0)
6989                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
6990
6991         if (status == 0) {
6992                 clp->cl_clientid = res.clientid;
6993                 clp->cl_exchange_flags = res.flags;
6994                 /* Client ID is not confirmed */
6995                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
6996                         clear_bit(NFS4_SESSION_ESTABLISHED,
6997                                         &clp->cl_session->session_state);
6998                         clp->cl_seqid = res.seqid;
6999                 }
7000
7001                 kfree(clp->cl_serverowner);
7002                 clp->cl_serverowner = res.server_owner;
7003                 res.server_owner = NULL;
7004
7005                 /* use the most recent implementation id */
7006                 kfree(clp->cl_implid);
7007                 clp->cl_implid = res.impl_id;
7008                 res.impl_id = NULL;
7009
7010                 if (clp->cl_serverscope != NULL &&
7011                     !nfs41_same_server_scope(clp->cl_serverscope,
7012                                              res.server_scope)) {
7013                         dprintk("%s: server_scope mismatch detected\n",
7014                                 __func__);
7015                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7016                         kfree(clp->cl_serverscope);
7017                         clp->cl_serverscope = NULL;
7018                 }
7019
7020                 if (clp->cl_serverscope == NULL) {
7021                         clp->cl_serverscope = res.server_scope;
7022                         res.server_scope = NULL;
7023                 }
7024         }
7025
7026 out_impl_id:
7027         kfree(res.impl_id);
7028 out_server_scope:
7029         kfree(res.server_scope);
7030 out_server_owner:
7031         kfree(res.server_owner);
7032 out:
7033         if (clp->cl_implid != NULL)
7034                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7035                         "domain: %s, name: %s, date: %llu,%u\n",
7036                         clp->cl_implid->domain, clp->cl_implid->name,
7037                         clp->cl_implid->date.seconds,
7038                         clp->cl_implid->date.nseconds);
7039         dprintk("NFS reply exchange_id: %d\n", status);
7040         return status;
7041 }
7042
7043 /*
7044  * nfs4_proc_exchange_id()
7045  *
7046  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7047  *
7048  * Since the clientid has expired, all compounds using sessions
7049  * associated with the stale clientid will be returning
7050  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7051  * be in some phase of session reset.
7052  *
7053  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7054  */
7055 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7056 {
7057         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7058         int status;
7059
7060         /* try SP4_MACH_CRED if krb5i/p */
7061         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7062             authflavor == RPC_AUTH_GSS_KRB5P) {
7063                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7064                 if (!status)
7065                         return 0;
7066         }
7067
7068         /* try SP4_NONE */
7069         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7070 }
7071
7072 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7073                 struct rpc_cred *cred)
7074 {
7075         struct rpc_message msg = {
7076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7077                 .rpc_argp = clp,
7078                 .rpc_cred = cred,
7079         };
7080         int status;
7081
7082         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7083         trace_nfs4_destroy_clientid(clp, status);
7084         if (status)
7085                 dprintk("NFS: Got error %d from the server %s on "
7086                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7087         return status;
7088 }
7089
7090 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7091                 struct rpc_cred *cred)
7092 {
7093         unsigned int loop;
7094         int ret;
7095
7096         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7097                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7098                 switch (ret) {
7099                 case -NFS4ERR_DELAY:
7100                 case -NFS4ERR_CLIENTID_BUSY:
7101                         ssleep(1);
7102                         break;
7103                 default:
7104                         return ret;
7105                 }
7106         }
7107         return 0;
7108 }
7109
7110 int nfs4_destroy_clientid(struct nfs_client *clp)
7111 {
7112         struct rpc_cred *cred;
7113         int ret = 0;
7114
7115         if (clp->cl_mvops->minor_version < 1)
7116                 goto out;
7117         if (clp->cl_exchange_flags == 0)
7118                 goto out;
7119         if (clp->cl_preserve_clid)
7120                 goto out;
7121         cred = nfs4_get_clid_cred(clp);
7122         ret = nfs4_proc_destroy_clientid(clp, cred);
7123         if (cred)
7124                 put_rpccred(cred);
7125         switch (ret) {
7126         case 0:
7127         case -NFS4ERR_STALE_CLIENTID:
7128                 clp->cl_exchange_flags = 0;
7129         }
7130 out:
7131         return ret;
7132 }
7133
7134 struct nfs4_get_lease_time_data {
7135         struct nfs4_get_lease_time_args *args;
7136         struct nfs4_get_lease_time_res *res;
7137         struct nfs_client *clp;
7138 };
7139
7140 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7141                                         void *calldata)
7142 {
7143         struct nfs4_get_lease_time_data *data =
7144                         (struct nfs4_get_lease_time_data *)calldata;
7145
7146         dprintk("--> %s\n", __func__);
7147         /* just setup sequence, do not trigger session recovery
7148            since we're invoked within one */
7149         nfs41_setup_sequence(data->clp->cl_session,
7150                         &data->args->la_seq_args,
7151                         &data->res->lr_seq_res,
7152                         task);
7153         dprintk("<-- %s\n", __func__);
7154 }
7155
7156 /*
7157  * Called from nfs4_state_manager thread for session setup, so don't recover
7158  * from sequence operation or clientid errors.
7159  */
7160 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7161 {
7162         struct nfs4_get_lease_time_data *data =
7163                         (struct nfs4_get_lease_time_data *)calldata;
7164
7165         dprintk("--> %s\n", __func__);
7166         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7167                 return;
7168         switch (task->tk_status) {
7169         case -NFS4ERR_DELAY:
7170         case -NFS4ERR_GRACE:
7171                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7172                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7173                 task->tk_status = 0;
7174                 /* fall through */
7175         case -NFS4ERR_RETRY_UNCACHED_REP:
7176                 rpc_restart_call_prepare(task);
7177                 return;
7178         }
7179         dprintk("<-- %s\n", __func__);
7180 }
7181
7182 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7183         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7184         .rpc_call_done = nfs4_get_lease_time_done,
7185 };
7186
7187 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7188 {
7189         struct rpc_task *task;
7190         struct nfs4_get_lease_time_args args;
7191         struct nfs4_get_lease_time_res res = {
7192                 .lr_fsinfo = fsinfo,
7193         };
7194         struct nfs4_get_lease_time_data data = {
7195                 .args = &args,
7196                 .res = &res,
7197                 .clp = clp,
7198         };
7199         struct rpc_message msg = {
7200                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7201                 .rpc_argp = &args,
7202                 .rpc_resp = &res,
7203         };
7204         struct rpc_task_setup task_setup = {
7205                 .rpc_client = clp->cl_rpcclient,
7206                 .rpc_message = &msg,
7207                 .callback_ops = &nfs4_get_lease_time_ops,
7208                 .callback_data = &data,
7209                 .flags = RPC_TASK_TIMEOUT,
7210         };
7211         int status;
7212
7213         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7214         nfs4_set_sequence_privileged(&args.la_seq_args);
7215         dprintk("--> %s\n", __func__);
7216         task = rpc_run_task(&task_setup);
7217
7218         if (IS_ERR(task))
7219                 status = PTR_ERR(task);
7220         else {
7221                 status = task->tk_status;
7222                 rpc_put_task(task);
7223         }
7224         dprintk("<-- %s return %d\n", __func__, status);
7225
7226         return status;
7227 }
7228
7229 /*
7230  * Initialize the values to be used by the client in CREATE_SESSION
7231  * If nfs4_init_session set the fore channel request and response sizes,
7232  * use them.
7233  *
7234  * Set the back channel max_resp_sz_cached to zero to force the client to
7235  * always set csa_cachethis to FALSE because the current implementation
7236  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7237  */
7238 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7239 {
7240         unsigned int max_rqst_sz, max_resp_sz;
7241
7242         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7243         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7244
7245         /* Fore channel attributes */
7246         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7247         args->fc_attrs.max_resp_sz = max_resp_sz;
7248         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7249         args->fc_attrs.max_reqs = max_session_slots;
7250
7251         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7252                 "max_ops=%u max_reqs=%u\n",
7253                 __func__,
7254                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7255                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7256
7257         /* Back channel attributes */
7258         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7259         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7260         args->bc_attrs.max_resp_sz_cached = 0;
7261         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7262         args->bc_attrs.max_reqs = 1;
7263
7264         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7265                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7266                 __func__,
7267                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7268                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7269                 args->bc_attrs.max_reqs);
7270 }
7271
7272 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7273                 struct nfs41_create_session_res *res)
7274 {
7275         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7276         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7277
7278         if (rcvd->max_resp_sz > sent->max_resp_sz)
7279                 return -EINVAL;
7280         /*
7281          * Our requested max_ops is the minimum we need; we're not
7282          * prepared to break up compounds into smaller pieces than that.
7283          * So, no point even trying to continue if the server won't
7284          * cooperate:
7285          */
7286         if (rcvd->max_ops < sent->max_ops)
7287                 return -EINVAL;
7288         if (rcvd->max_reqs == 0)
7289                 return -EINVAL;
7290         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7291                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7292         return 0;
7293 }
7294
7295 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7296                 struct nfs41_create_session_res *res)
7297 {
7298         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7299         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7300
7301         if (!(res->flags & SESSION4_BACK_CHAN))
7302                 goto out;
7303         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7304                 return -EINVAL;
7305         if (rcvd->max_resp_sz < sent->max_resp_sz)
7306                 return -EINVAL;
7307         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7308                 return -EINVAL;
7309         /* These would render the backchannel useless: */
7310         if (rcvd->max_ops != sent->max_ops)
7311                 return -EINVAL;
7312         if (rcvd->max_reqs != sent->max_reqs)
7313                 return -EINVAL;
7314 out:
7315         return 0;
7316 }
7317
7318 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7319                                      struct nfs41_create_session_res *res)
7320 {
7321         int ret;
7322
7323         ret = nfs4_verify_fore_channel_attrs(args, res);
7324         if (ret)
7325                 return ret;
7326         return nfs4_verify_back_channel_attrs(args, res);
7327 }
7328
7329 static void nfs4_update_session(struct nfs4_session *session,
7330                 struct nfs41_create_session_res *res)
7331 {
7332         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7333         /* Mark client id and session as being confirmed */
7334         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7335         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7336         session->flags = res->flags;
7337         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7338         if (res->flags & SESSION4_BACK_CHAN)
7339                 memcpy(&session->bc_attrs, &res->bc_attrs,
7340                                 sizeof(session->bc_attrs));
7341 }
7342
7343 static int _nfs4_proc_create_session(struct nfs_client *clp,
7344                 struct rpc_cred *cred)
7345 {
7346         struct nfs4_session *session = clp->cl_session;
7347         struct nfs41_create_session_args args = {
7348                 .client = clp,
7349                 .clientid = clp->cl_clientid,
7350                 .seqid = clp->cl_seqid,
7351                 .cb_program = NFS4_CALLBACK,
7352         };
7353         struct nfs41_create_session_res res;
7354
7355         struct rpc_message msg = {
7356                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7357                 .rpc_argp = &args,
7358                 .rpc_resp = &res,
7359                 .rpc_cred = cred,
7360         };
7361         int status;
7362
7363         nfs4_init_channel_attrs(&args);
7364         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7365
7366         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7367         trace_nfs4_create_session(clp, status);
7368
7369         if (!status) {
7370                 /* Verify the session's negotiated channel_attrs values */
7371                 status = nfs4_verify_channel_attrs(&args, &res);
7372                 /* Increment the clientid slot sequence id */
7373                 if (clp->cl_seqid == res.seqid)
7374                         clp->cl_seqid++;
7375                 if (status)
7376                         goto out;
7377                 nfs4_update_session(session, &res);
7378         }
7379 out:
7380         return status;
7381 }
7382
7383 /*
7384  * Issues a CREATE_SESSION operation to the server.
7385  * It is the responsibility of the caller to verify the session is
7386  * expired before calling this routine.
7387  */
7388 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7389 {
7390         int status;
7391         unsigned *ptr;
7392         struct nfs4_session *session = clp->cl_session;
7393
7394         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7395
7396         status = _nfs4_proc_create_session(clp, cred);
7397         if (status)
7398                 goto out;
7399
7400         /* Init or reset the session slot tables */
7401         status = nfs4_setup_session_slot_tables(session);
7402         dprintk("slot table setup returned %d\n", status);
7403         if (status)
7404                 goto out;
7405
7406         ptr = (unsigned *)&session->sess_id.data[0];
7407         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7408                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7409 out:
7410         dprintk("<-- %s\n", __func__);
7411         return status;
7412 }
7413
7414 /*
7415  * Issue the over-the-wire RPC DESTROY_SESSION.
7416  * The caller must serialize access to this routine.
7417  */
7418 int nfs4_proc_destroy_session(struct nfs4_session *session,
7419                 struct rpc_cred *cred)
7420 {
7421         struct rpc_message msg = {
7422                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7423                 .rpc_argp = session,
7424                 .rpc_cred = cred,
7425         };
7426         int status = 0;
7427
7428         dprintk("--> nfs4_proc_destroy_session\n");
7429
7430         /* session is still being setup */
7431         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7432                 return 0;
7433
7434         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7435         trace_nfs4_destroy_session(session->clp, status);
7436
7437         if (status)
7438                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7439                         "Session has been destroyed regardless...\n", status);
7440
7441         dprintk("<-- nfs4_proc_destroy_session\n");
7442         return status;
7443 }
7444
7445 /*
7446  * Renew the cl_session lease.
7447  */
7448 struct nfs4_sequence_data {
7449         struct nfs_client *clp;
7450         struct nfs4_sequence_args args;
7451         struct nfs4_sequence_res res;
7452 };
7453
7454 static void nfs41_sequence_release(void *data)
7455 {
7456         struct nfs4_sequence_data *calldata = data;
7457         struct nfs_client *clp = calldata->clp;
7458
7459         if (atomic_read(&clp->cl_count) > 1)
7460                 nfs4_schedule_state_renewal(clp);
7461         nfs_put_client(clp);
7462         kfree(calldata);
7463 }
7464
7465 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7466 {
7467         switch(task->tk_status) {
7468         case -NFS4ERR_DELAY:
7469                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7470                 return -EAGAIN;
7471         default:
7472                 nfs4_schedule_lease_recovery(clp);
7473         }
7474         return 0;
7475 }
7476
7477 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7478 {
7479         struct nfs4_sequence_data *calldata = data;
7480         struct nfs_client *clp = calldata->clp;
7481
7482         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7483                 return;
7484
7485         trace_nfs4_sequence(clp, task->tk_status);
7486         if (task->tk_status < 0) {
7487                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7488                 if (atomic_read(&clp->cl_count) == 1)
7489                         goto out;
7490
7491                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7492                         rpc_restart_call_prepare(task);
7493                         return;
7494                 }
7495         }
7496         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7497 out:
7498         dprintk("<-- %s\n", __func__);
7499 }
7500
7501 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7502 {
7503         struct nfs4_sequence_data *calldata = data;
7504         struct nfs_client *clp = calldata->clp;
7505         struct nfs4_sequence_args *args;
7506         struct nfs4_sequence_res *res;
7507
7508         args = task->tk_msg.rpc_argp;
7509         res = task->tk_msg.rpc_resp;
7510
7511         nfs41_setup_sequence(clp->cl_session, args, res, task);
7512 }
7513
7514 static const struct rpc_call_ops nfs41_sequence_ops = {
7515         .rpc_call_done = nfs41_sequence_call_done,
7516         .rpc_call_prepare = nfs41_sequence_prepare,
7517         .rpc_release = nfs41_sequence_release,
7518 };
7519
7520 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7521                 struct rpc_cred *cred,
7522                 bool is_privileged)
7523 {
7524         struct nfs4_sequence_data *calldata;
7525         struct rpc_message msg = {
7526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7527                 .rpc_cred = cred,
7528         };
7529         struct rpc_task_setup task_setup_data = {
7530                 .rpc_client = clp->cl_rpcclient,
7531                 .rpc_message = &msg,
7532                 .callback_ops = &nfs41_sequence_ops,
7533                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7534         };
7535
7536         if (!atomic_inc_not_zero(&clp->cl_count))
7537                 return ERR_PTR(-EIO);
7538         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7539         if (calldata == NULL) {
7540                 nfs_put_client(clp);
7541                 return ERR_PTR(-ENOMEM);
7542         }
7543         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7544         if (is_privileged)
7545                 nfs4_set_sequence_privileged(&calldata->args);
7546         msg.rpc_argp = &calldata->args;
7547         msg.rpc_resp = &calldata->res;
7548         calldata->clp = clp;
7549         task_setup_data.callback_data = calldata;
7550
7551         return rpc_run_task(&task_setup_data);
7552 }
7553
7554 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7555 {
7556         struct rpc_task *task;
7557         int ret = 0;
7558
7559         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7560                 return -EAGAIN;
7561         task = _nfs41_proc_sequence(clp, cred, false);
7562         if (IS_ERR(task))
7563                 ret = PTR_ERR(task);
7564         else
7565                 rpc_put_task_async(task);
7566         dprintk("<-- %s status=%d\n", __func__, ret);
7567         return ret;
7568 }
7569
7570 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7571 {
7572         struct rpc_task *task;
7573         int ret;
7574
7575         task = _nfs41_proc_sequence(clp, cred, true);
7576         if (IS_ERR(task)) {
7577                 ret = PTR_ERR(task);
7578                 goto out;
7579         }
7580         ret = rpc_wait_for_completion_task(task);
7581         if (!ret)
7582                 ret = task->tk_status;
7583         rpc_put_task(task);
7584 out:
7585         dprintk("<-- %s status=%d\n", __func__, ret);
7586         return ret;
7587 }
7588
7589 struct nfs4_reclaim_complete_data {
7590         struct nfs_client *clp;
7591         struct nfs41_reclaim_complete_args arg;
7592         struct nfs41_reclaim_complete_res res;
7593 };
7594
7595 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7596 {
7597         struct nfs4_reclaim_complete_data *calldata = data;
7598
7599         nfs41_setup_sequence(calldata->clp->cl_session,
7600                         &calldata->arg.seq_args,
7601                         &calldata->res.seq_res,
7602                         task);
7603 }
7604
7605 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7606 {
7607         switch(task->tk_status) {
7608         case 0:
7609         case -NFS4ERR_COMPLETE_ALREADY:
7610         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7611                 break;
7612         case -NFS4ERR_DELAY:
7613                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7614                 /* fall through */
7615         case -NFS4ERR_RETRY_UNCACHED_REP:
7616                 return -EAGAIN;
7617         default:
7618                 nfs4_schedule_lease_recovery(clp);
7619         }
7620         return 0;
7621 }
7622
7623 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7624 {
7625         struct nfs4_reclaim_complete_data *calldata = data;
7626         struct nfs_client *clp = calldata->clp;
7627         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7628
7629         dprintk("--> %s\n", __func__);
7630         if (!nfs41_sequence_done(task, res))
7631                 return;
7632
7633         trace_nfs4_reclaim_complete(clp, task->tk_status);
7634         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7635                 rpc_restart_call_prepare(task);
7636                 return;
7637         }
7638         dprintk("<-- %s\n", __func__);
7639 }
7640
7641 static void nfs4_free_reclaim_complete_data(void *data)
7642 {
7643         struct nfs4_reclaim_complete_data *calldata = data;
7644
7645         kfree(calldata);
7646 }
7647
7648 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7649         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7650         .rpc_call_done = nfs4_reclaim_complete_done,
7651         .rpc_release = nfs4_free_reclaim_complete_data,
7652 };
7653
7654 /*
7655  * Issue a global reclaim complete.
7656  */
7657 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7658                 struct rpc_cred *cred)
7659 {
7660         struct nfs4_reclaim_complete_data *calldata;
7661         struct rpc_task *task;
7662         struct rpc_message msg = {
7663                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7664                 .rpc_cred = cred,
7665         };
7666         struct rpc_task_setup task_setup_data = {
7667                 .rpc_client = clp->cl_rpcclient,
7668                 .rpc_message = &msg,
7669                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7670                 .flags = RPC_TASK_ASYNC,
7671         };
7672         int status = -ENOMEM;
7673
7674         dprintk("--> %s\n", __func__);
7675         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7676         if (calldata == NULL)
7677                 goto out;
7678         calldata->clp = clp;
7679         calldata->arg.one_fs = 0;
7680
7681         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7682         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7683         msg.rpc_argp = &calldata->arg;
7684         msg.rpc_resp = &calldata->res;
7685         task_setup_data.callback_data = calldata;
7686         task = rpc_run_task(&task_setup_data);
7687         if (IS_ERR(task)) {
7688                 status = PTR_ERR(task);
7689                 goto out;
7690         }
7691         status = nfs4_wait_for_completion_rpc_task(task);
7692         if (status == 0)
7693                 status = task->tk_status;
7694         rpc_put_task(task);
7695         return 0;
7696 out:
7697         dprintk("<-- %s status=%d\n", __func__, status);
7698         return status;
7699 }
7700
7701 static void
7702 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7703 {
7704         struct nfs4_layoutget *lgp = calldata;
7705         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7706         struct nfs4_session *session = nfs4_get_session(server);
7707
7708         dprintk("--> %s\n", __func__);
7709         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7710          * right now covering the LAYOUTGET we are about to send.
7711          * However, that is not so catastrophic, and there seems
7712          * to be no way to prevent it completely.
7713          */
7714         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7715                                 &lgp->res.seq_res, task))
7716                 return;
7717         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7718                                           NFS_I(lgp->args.inode)->layout,
7719                                           &lgp->args.range,
7720                                           lgp->args.ctx->state)) {
7721                 rpc_exit(task, NFS4_OK);
7722         }
7723 }
7724
7725 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7726 {
7727         struct nfs4_layoutget *lgp = calldata;
7728         struct inode *inode = lgp->args.inode;
7729         struct nfs_server *server = NFS_SERVER(inode);
7730         struct pnfs_layout_hdr *lo;
7731         struct nfs4_state *state = NULL;
7732         unsigned long timeo, now, giveup;
7733
7734         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7735
7736         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7737                 goto out;
7738
7739         switch (task->tk_status) {
7740         case 0:
7741                 goto out;
7742         /*
7743          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7744          * (or clients) writing to the same RAID stripe
7745          */
7746         case -NFS4ERR_LAYOUTTRYLATER:
7747         /*
7748          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7749          * existing layout before getting a new one).
7750          */
7751         case -NFS4ERR_RECALLCONFLICT:
7752                 timeo = rpc_get_timeout(task->tk_client);
7753                 giveup = lgp->args.timestamp + timeo;
7754                 now = jiffies;
7755                 if (time_after(giveup, now)) {
7756                         unsigned long delay;
7757
7758                         /* Delay for:
7759                          * - Not less then NFS4_POLL_RETRY_MIN.
7760                          * - One last time a jiffie before we give up
7761                          * - exponential backoff (time_now minus start_attempt)
7762                          */
7763                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7764                                     min((giveup - now - 1),
7765                                         now - lgp->args.timestamp));
7766
7767                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7768                                 __func__, delay);
7769                         rpc_delay(task, delay);
7770                         task->tk_status = 0;
7771                         rpc_restart_call_prepare(task);
7772                         goto out; /* Do not call nfs4_async_handle_error() */
7773                 }
7774                 break;
7775         case -NFS4ERR_EXPIRED:
7776         case -NFS4ERR_BAD_STATEID:
7777                 spin_lock(&inode->i_lock);
7778                 lo = NFS_I(inode)->layout;
7779                 if (!lo || list_empty(&lo->plh_segs)) {
7780                         spin_unlock(&inode->i_lock);
7781                         /* If the open stateid was bad, then recover it. */
7782                         state = lgp->args.ctx->state;
7783                 } else {
7784                         LIST_HEAD(head);
7785
7786                         /*
7787                          * Mark the bad layout state as invalid, then retry
7788                          * with the current stateid.
7789                          */
7790                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7791                         spin_unlock(&inode->i_lock);
7792                         pnfs_free_lseg_list(&head);
7793
7794                         task->tk_status = 0;
7795                         rpc_restart_call_prepare(task);
7796                 }
7797         }
7798         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7799                 rpc_restart_call_prepare(task);
7800 out:
7801         dprintk("<-- %s\n", __func__);
7802 }
7803
7804 static size_t max_response_pages(struct nfs_server *server)
7805 {
7806         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7807         return nfs_page_array_len(0, max_resp_sz);
7808 }
7809
7810 static void nfs4_free_pages(struct page **pages, size_t size)
7811 {
7812         int i;
7813
7814         if (!pages)
7815                 return;
7816
7817         for (i = 0; i < size; i++) {
7818                 if (!pages[i])
7819                         break;
7820                 __free_page(pages[i]);
7821         }
7822         kfree(pages);
7823 }
7824
7825 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7826 {
7827         struct page **pages;
7828         int i;
7829
7830         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7831         if (!pages) {
7832                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7833                 return NULL;
7834         }
7835
7836         for (i = 0; i < size; i++) {
7837                 pages[i] = alloc_page(gfp_flags);
7838                 if (!pages[i]) {
7839                         dprintk("%s: failed to allocate page\n", __func__);
7840                         nfs4_free_pages(pages, size);
7841                         return NULL;
7842                 }
7843         }
7844
7845         return pages;
7846 }
7847
7848 static void nfs4_layoutget_release(void *calldata)
7849 {
7850         struct nfs4_layoutget *lgp = calldata;
7851         struct inode *inode = lgp->args.inode;
7852         struct nfs_server *server = NFS_SERVER(inode);
7853         size_t max_pages = max_response_pages(server);
7854
7855         dprintk("--> %s\n", __func__);
7856         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7857         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7858         put_nfs_open_context(lgp->args.ctx);
7859         kfree(calldata);
7860         dprintk("<-- %s\n", __func__);
7861 }
7862
7863 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7864         .rpc_call_prepare = nfs4_layoutget_prepare,
7865         .rpc_call_done = nfs4_layoutget_done,
7866         .rpc_release = nfs4_layoutget_release,
7867 };
7868
7869 struct pnfs_layout_segment *
7870 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7871 {
7872         struct inode *inode = lgp->args.inode;
7873         struct nfs_server *server = NFS_SERVER(inode);
7874         size_t max_pages = max_response_pages(server);
7875         struct rpc_task *task;
7876         struct rpc_message msg = {
7877                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7878                 .rpc_argp = &lgp->args,
7879                 .rpc_resp = &lgp->res,
7880                 .rpc_cred = lgp->cred,
7881         };
7882         struct rpc_task_setup task_setup_data = {
7883                 .rpc_client = server->client,
7884                 .rpc_message = &msg,
7885                 .callback_ops = &nfs4_layoutget_call_ops,
7886                 .callback_data = lgp,
7887                 .flags = RPC_TASK_ASYNC,
7888         };
7889         struct pnfs_layout_segment *lseg = NULL;
7890         int status = 0;
7891
7892         dprintk("--> %s\n", __func__);
7893
7894         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7895         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7896
7897         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7898         if (!lgp->args.layout.pages) {
7899                 nfs4_layoutget_release(lgp);
7900                 return ERR_PTR(-ENOMEM);
7901         }
7902         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7903         lgp->args.timestamp = jiffies;
7904
7905         lgp->res.layoutp = &lgp->args.layout;
7906         lgp->res.seq_res.sr_slot = NULL;
7907         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7908
7909         task = rpc_run_task(&task_setup_data);
7910         if (IS_ERR(task))
7911                 return ERR_CAST(task);
7912         status = nfs4_wait_for_completion_rpc_task(task);
7913         if (status == 0)
7914                 status = task->tk_status;
7915         trace_nfs4_layoutget(lgp->args.ctx,
7916                         &lgp->args.range,
7917                         &lgp->res.range,
7918                         status);
7919         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
7920         if (status == 0 && lgp->res.layoutp->len)
7921                 lseg = pnfs_layout_process(lgp);
7922         rpc_put_task(task);
7923         dprintk("<-- %s status=%d\n", __func__, status);
7924         if (status)
7925                 return ERR_PTR(status);
7926         return lseg;
7927 }
7928
7929 static void
7930 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
7931 {
7932         struct nfs4_layoutreturn *lrp = calldata;
7933
7934         dprintk("--> %s\n", __func__);
7935         nfs41_setup_sequence(lrp->clp->cl_session,
7936                         &lrp->args.seq_args,
7937                         &lrp->res.seq_res,
7938                         task);
7939 }
7940
7941 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
7942 {
7943         struct nfs4_layoutreturn *lrp = calldata;
7944         struct nfs_server *server;
7945
7946         dprintk("--> %s\n", __func__);
7947
7948         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
7949                 return;
7950
7951         server = NFS_SERVER(lrp->args.inode);
7952         switch (task->tk_status) {
7953         default:
7954                 task->tk_status = 0;
7955         case 0:
7956                 break;
7957         case -NFS4ERR_DELAY:
7958                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
7959                         break;
7960                 rpc_restart_call_prepare(task);
7961                 return;
7962         }
7963         dprintk("<-- %s\n", __func__);
7964 }
7965
7966 static void nfs4_layoutreturn_release(void *calldata)
7967 {
7968         struct nfs4_layoutreturn *lrp = calldata;
7969         struct pnfs_layout_hdr *lo = lrp->args.layout;
7970         LIST_HEAD(freeme);
7971
7972         dprintk("--> %s\n", __func__);
7973         spin_lock(&lo->plh_inode->i_lock);
7974         if (lrp->res.lrs_present)
7975                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
7976         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
7977         pnfs_clear_layoutreturn_waitbit(lo);
7978         lo->plh_block_lgets--;
7979         spin_unlock(&lo->plh_inode->i_lock);
7980         pnfs_free_lseg_list(&freeme);
7981         pnfs_put_layout_hdr(lrp->args.layout);
7982         nfs_iput_and_deactive(lrp->inode);
7983         kfree(calldata);
7984         dprintk("<-- %s\n", __func__);
7985 }
7986
7987 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
7988         .rpc_call_prepare = nfs4_layoutreturn_prepare,
7989         .rpc_call_done = nfs4_layoutreturn_done,
7990         .rpc_release = nfs4_layoutreturn_release,
7991 };
7992
7993 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
7994 {
7995         struct rpc_task *task;
7996         struct rpc_message msg = {
7997                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
7998                 .rpc_argp = &lrp->args,
7999                 .rpc_resp = &lrp->res,
8000                 .rpc_cred = lrp->cred,
8001         };
8002         struct rpc_task_setup task_setup_data = {
8003                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8004                 .rpc_message = &msg,
8005                 .callback_ops = &nfs4_layoutreturn_call_ops,
8006                 .callback_data = lrp,
8007         };
8008         int status = 0;
8009
8010         dprintk("--> %s\n", __func__);
8011         if (!sync) {
8012                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8013                 if (!lrp->inode) {
8014                         nfs4_layoutreturn_release(lrp);
8015                         return -EAGAIN;
8016                 }
8017                 task_setup_data.flags |= RPC_TASK_ASYNC;
8018         }
8019         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8020         task = rpc_run_task(&task_setup_data);
8021         if (IS_ERR(task))
8022                 return PTR_ERR(task);
8023         if (sync)
8024                 status = task->tk_status;
8025         trace_nfs4_layoutreturn(lrp->args.inode, status);
8026         dprintk("<-- %s status=%d\n", __func__, status);
8027         rpc_put_task(task);
8028         return status;
8029 }
8030
8031 static int
8032 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8033                 struct pnfs_device *pdev,
8034                 struct rpc_cred *cred)
8035 {
8036         struct nfs4_getdeviceinfo_args args = {
8037                 .pdev = pdev,
8038                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8039                         NOTIFY_DEVICEID4_DELETE,
8040         };
8041         struct nfs4_getdeviceinfo_res res = {
8042                 .pdev = pdev,
8043         };
8044         struct rpc_message msg = {
8045                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8046                 .rpc_argp = &args,
8047                 .rpc_resp = &res,
8048                 .rpc_cred = cred,
8049         };
8050         int status;
8051
8052         dprintk("--> %s\n", __func__);
8053         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8054         if (res.notification & ~args.notify_types)
8055                 dprintk("%s: unsupported notification\n", __func__);
8056         if (res.notification != args.notify_types)
8057                 pdev->nocache = 1;
8058
8059         dprintk("<-- %s status=%d\n", __func__, status);
8060
8061         return status;
8062 }
8063
8064 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8065                 struct pnfs_device *pdev,
8066                 struct rpc_cred *cred)
8067 {
8068         struct nfs4_exception exception = { };
8069         int err;
8070
8071         do {
8072                 err = nfs4_handle_exception(server,
8073                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8074                                         &exception);
8075         } while (exception.retry);
8076         return err;
8077 }
8078 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8079
8080 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8081 {
8082         struct nfs4_layoutcommit_data *data = calldata;
8083         struct nfs_server *server = NFS_SERVER(data->args.inode);
8084         struct nfs4_session *session = nfs4_get_session(server);
8085
8086         nfs41_setup_sequence(session,
8087                         &data->args.seq_args,
8088                         &data->res.seq_res,
8089                         task);
8090 }
8091
8092 static void
8093 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8094 {
8095         struct nfs4_layoutcommit_data *data = calldata;
8096         struct nfs_server *server = NFS_SERVER(data->args.inode);
8097
8098         if (!nfs41_sequence_done(task, &data->res.seq_res))
8099                 return;
8100
8101         switch (task->tk_status) { /* Just ignore these failures */
8102         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8103         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8104         case -NFS4ERR_BADLAYOUT:     /* no layout */
8105         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8106                 task->tk_status = 0;
8107         case 0:
8108                 break;
8109         default:
8110                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8111                         rpc_restart_call_prepare(task);
8112                         return;
8113                 }
8114         }
8115 }
8116
8117 static void nfs4_layoutcommit_release(void *calldata)
8118 {
8119         struct nfs4_layoutcommit_data *data = calldata;
8120
8121         pnfs_cleanup_layoutcommit(data);
8122         nfs_post_op_update_inode_force_wcc(data->args.inode,
8123                                            data->res.fattr);
8124         put_rpccred(data->cred);
8125         nfs_iput_and_deactive(data->inode);
8126         kfree(data);
8127 }
8128
8129 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8130         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8131         .rpc_call_done = nfs4_layoutcommit_done,
8132         .rpc_release = nfs4_layoutcommit_release,
8133 };
8134
8135 int
8136 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8137 {
8138         struct rpc_message msg = {
8139                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8140                 .rpc_argp = &data->args,
8141                 .rpc_resp = &data->res,
8142                 .rpc_cred = data->cred,
8143         };
8144         struct rpc_task_setup task_setup_data = {
8145                 .task = &data->task,
8146                 .rpc_client = NFS_CLIENT(data->args.inode),
8147                 .rpc_message = &msg,
8148                 .callback_ops = &nfs4_layoutcommit_ops,
8149                 .callback_data = data,
8150         };
8151         struct rpc_task *task;
8152         int status = 0;
8153
8154         dprintk("NFS: initiating layoutcommit call. sync %d "
8155                 "lbw: %llu inode %lu\n", sync,
8156                 data->args.lastbytewritten,
8157                 data->args.inode->i_ino);
8158
8159         if (!sync) {
8160                 data->inode = nfs_igrab_and_active(data->args.inode);
8161                 if (data->inode == NULL) {
8162                         nfs4_layoutcommit_release(data);
8163                         return -EAGAIN;
8164                 }
8165                 task_setup_data.flags = RPC_TASK_ASYNC;
8166         }
8167         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8168         task = rpc_run_task(&task_setup_data);
8169         if (IS_ERR(task))
8170                 return PTR_ERR(task);
8171         if (sync)
8172                 status = task->tk_status;
8173         trace_nfs4_layoutcommit(data->args.inode, status);
8174         dprintk("%s: status %d\n", __func__, status);
8175         rpc_put_task(task);
8176         return status;
8177 }
8178
8179 /**
8180  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8181  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8182  */
8183 static int
8184 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8185                     struct nfs_fsinfo *info,
8186                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8187 {
8188         struct nfs41_secinfo_no_name_args args = {
8189                 .style = SECINFO_STYLE_CURRENT_FH,
8190         };
8191         struct nfs4_secinfo_res res = {
8192                 .flavors = flavors,
8193         };
8194         struct rpc_message msg = {
8195                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8196                 .rpc_argp = &args,
8197                 .rpc_resp = &res,
8198         };
8199         struct rpc_clnt *clnt = server->client;
8200         struct rpc_cred *cred = NULL;
8201         int status;
8202
8203         if (use_integrity) {
8204                 clnt = server->nfs_client->cl_rpcclient;
8205                 cred = nfs4_get_clid_cred(server->nfs_client);
8206                 msg.rpc_cred = cred;
8207         }
8208
8209         dprintk("--> %s\n", __func__);
8210         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8211                                 &res.seq_res, 0);
8212         dprintk("<-- %s status=%d\n", __func__, status);
8213
8214         if (cred)
8215                 put_rpccred(cred);
8216
8217         return status;
8218 }
8219
8220 static int
8221 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8222                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8223 {
8224         struct nfs4_exception exception = { };
8225         int err;
8226         do {
8227                 /* first try using integrity protection */
8228                 err = -NFS4ERR_WRONGSEC;
8229
8230                 /* try to use integrity protection with machine cred */
8231                 if (_nfs4_is_integrity_protected(server->nfs_client))
8232                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8233                                                           flavors, true);
8234
8235                 /*
8236                  * if unable to use integrity protection, or SECINFO with
8237                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8238                  * disallowed by spec, but exists in deployed servers) use
8239                  * the current filesystem's rpc_client and the user cred.
8240                  */
8241                 if (err == -NFS4ERR_WRONGSEC)
8242                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8243                                                           flavors, false);
8244
8245                 switch (err) {
8246                 case 0:
8247                 case -NFS4ERR_WRONGSEC:
8248                 case -ENOTSUPP:
8249                         goto out;
8250                 default:
8251                         err = nfs4_handle_exception(server, err, &exception);
8252                 }
8253         } while (exception.retry);
8254 out:
8255         return err;
8256 }
8257
8258 static int
8259 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8260                     struct nfs_fsinfo *info)
8261 {
8262         int err;
8263         struct page *page;
8264         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8265         struct nfs4_secinfo_flavors *flavors;
8266         struct nfs4_secinfo4 *secinfo;
8267         int i;
8268
8269         page = alloc_page(GFP_KERNEL);
8270         if (!page) {
8271                 err = -ENOMEM;
8272                 goto out;
8273         }
8274
8275         flavors = page_address(page);
8276         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8277
8278         /*
8279          * Fall back on "guess and check" method if
8280          * the server doesn't support SECINFO_NO_NAME
8281          */
8282         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8283                 err = nfs4_find_root_sec(server, fhandle, info);
8284                 goto out_freepage;
8285         }
8286         if (err)
8287                 goto out_freepage;
8288
8289         for (i = 0; i < flavors->num_flavors; i++) {
8290                 secinfo = &flavors->flavors[i];
8291
8292                 switch (secinfo->flavor) {
8293                 case RPC_AUTH_NULL:
8294                 case RPC_AUTH_UNIX:
8295                 case RPC_AUTH_GSS:
8296                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8297                                         &secinfo->flavor_info);
8298                         break;
8299                 default:
8300                         flavor = RPC_AUTH_MAXFLAVOR;
8301                         break;
8302                 }
8303
8304                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8305                         flavor = RPC_AUTH_MAXFLAVOR;
8306
8307                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8308                         err = nfs4_lookup_root_sec(server, fhandle,
8309                                                    info, flavor);
8310                         if (!err)
8311                                 break;
8312                 }
8313         }
8314
8315         if (flavor == RPC_AUTH_MAXFLAVOR)
8316                 err = -EPERM;
8317
8318 out_freepage:
8319         put_page(page);
8320         if (err == -EACCES)
8321                 return -EPERM;
8322 out:
8323         return err;
8324 }
8325
8326 static int _nfs41_test_stateid(struct nfs_server *server,
8327                 nfs4_stateid *stateid,
8328                 struct rpc_cred *cred)
8329 {
8330         int status;
8331         struct nfs41_test_stateid_args args = {
8332                 .stateid = stateid,
8333         };
8334         struct nfs41_test_stateid_res res;
8335         struct rpc_message msg = {
8336                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8337                 .rpc_argp = &args,
8338                 .rpc_resp = &res,
8339                 .rpc_cred = cred,
8340         };
8341         struct rpc_clnt *rpc_client = server->client;
8342
8343         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8344                 &rpc_client, &msg);
8345
8346         dprintk("NFS call  test_stateid %p\n", stateid);
8347         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8348         nfs4_set_sequence_privileged(&args.seq_args);
8349         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8350                         &args.seq_args, &res.seq_res);
8351         if (status != NFS_OK) {
8352                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8353                 return status;
8354         }
8355         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8356         return -res.status;
8357 }
8358
8359 /**
8360  * nfs41_test_stateid - perform a TEST_STATEID operation
8361  *
8362  * @server: server / transport on which to perform the operation
8363  * @stateid: state ID to test
8364  * @cred: credential
8365  *
8366  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8367  * Otherwise a negative NFS4ERR value is returned if the operation
8368  * failed or the state ID is not currently valid.
8369  */
8370 static int nfs41_test_stateid(struct nfs_server *server,
8371                 nfs4_stateid *stateid,
8372                 struct rpc_cred *cred)
8373 {
8374         struct nfs4_exception exception = { };
8375         int err;
8376         do {
8377                 err = _nfs41_test_stateid(server, stateid, cred);
8378                 if (err != -NFS4ERR_DELAY)
8379                         break;
8380                 nfs4_handle_exception(server, err, &exception);
8381         } while (exception.retry);
8382         return err;
8383 }
8384
8385 struct nfs_free_stateid_data {
8386         struct nfs_server *server;
8387         struct nfs41_free_stateid_args args;
8388         struct nfs41_free_stateid_res res;
8389 };
8390
8391 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8392 {
8393         struct nfs_free_stateid_data *data = calldata;
8394         nfs41_setup_sequence(nfs4_get_session(data->server),
8395                         &data->args.seq_args,
8396                         &data->res.seq_res,
8397                         task);
8398 }
8399
8400 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8401 {
8402         struct nfs_free_stateid_data *data = calldata;
8403
8404         nfs41_sequence_done(task, &data->res.seq_res);
8405
8406         switch (task->tk_status) {
8407         case -NFS4ERR_DELAY:
8408                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8409                         rpc_restart_call_prepare(task);
8410         }
8411 }
8412
8413 static void nfs41_free_stateid_release(void *calldata)
8414 {
8415         kfree(calldata);
8416 }
8417
8418 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8419         .rpc_call_prepare = nfs41_free_stateid_prepare,
8420         .rpc_call_done = nfs41_free_stateid_done,
8421         .rpc_release = nfs41_free_stateid_release,
8422 };
8423
8424 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8425                 nfs4_stateid *stateid,
8426                 struct rpc_cred *cred,
8427                 bool privileged)
8428 {
8429         struct rpc_message msg = {
8430                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8431                 .rpc_cred = cred,
8432         };
8433         struct rpc_task_setup task_setup = {
8434                 .rpc_client = server->client,
8435                 .rpc_message = &msg,
8436                 .callback_ops = &nfs41_free_stateid_ops,
8437                 .flags = RPC_TASK_ASYNC,
8438         };
8439         struct nfs_free_stateid_data *data;
8440
8441         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8442                 &task_setup.rpc_client, &msg);
8443
8444         dprintk("NFS call  free_stateid %p\n", stateid);
8445         data = kmalloc(sizeof(*data), GFP_NOFS);
8446         if (!data)
8447                 return ERR_PTR(-ENOMEM);
8448         data->server = server;
8449         nfs4_stateid_copy(&data->args.stateid, stateid);
8450
8451         task_setup.callback_data = data;
8452
8453         msg.rpc_argp = &data->args;
8454         msg.rpc_resp = &data->res;
8455         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8456         if (privileged)
8457                 nfs4_set_sequence_privileged(&data->args.seq_args);
8458
8459         return rpc_run_task(&task_setup);
8460 }
8461
8462 /**
8463  * nfs41_free_stateid - perform a FREE_STATEID operation
8464  *
8465  * @server: server / transport on which to perform the operation
8466  * @stateid: state ID to release
8467  * @cred: credential
8468  *
8469  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8470  * negative NFS4ERR value is returned.
8471  */
8472 static int nfs41_free_stateid(struct nfs_server *server,
8473                 nfs4_stateid *stateid,
8474                 struct rpc_cred *cred)
8475 {
8476         struct rpc_task *task;
8477         int ret;
8478
8479         task = _nfs41_free_stateid(server, stateid, cred, true);
8480         if (IS_ERR(task))
8481                 return PTR_ERR(task);
8482         ret = rpc_wait_for_completion_task(task);
8483         if (!ret)
8484                 ret = task->tk_status;
8485         rpc_put_task(task);
8486         return ret;
8487 }
8488
8489 static void
8490 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8491 {
8492         struct rpc_task *task;
8493         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8494
8495         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8496         nfs4_free_lock_state(server, lsp);
8497         if (IS_ERR(task))
8498                 return;
8499         rpc_put_task(task);
8500 }
8501
8502 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8503                 const nfs4_stateid *s2)
8504 {
8505         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8506                 return false;
8507
8508         if (s1->seqid == s2->seqid)
8509                 return true;
8510         if (s1->seqid == 0 || s2->seqid == 0)
8511                 return true;
8512
8513         return false;
8514 }
8515
8516 #endif /* CONFIG_NFS_V4_1 */
8517
8518 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8519                 const nfs4_stateid *s2)
8520 {
8521         return nfs4_stateid_match(s1, s2);
8522 }
8523
8524
8525 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8526         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8527         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8528         .recover_open   = nfs4_open_reclaim,
8529         .recover_lock   = nfs4_lock_reclaim,
8530         .establish_clid = nfs4_init_clientid,
8531         .detect_trunking = nfs40_discover_server_trunking,
8532 };
8533
8534 #if defined(CONFIG_NFS_V4_1)
8535 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8536         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8537         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8538         .recover_open   = nfs4_open_reclaim,
8539         .recover_lock   = nfs4_lock_reclaim,
8540         .establish_clid = nfs41_init_clientid,
8541         .reclaim_complete = nfs41_proc_reclaim_complete,
8542         .detect_trunking = nfs41_discover_server_trunking,
8543 };
8544 #endif /* CONFIG_NFS_V4_1 */
8545
8546 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8547         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8548         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8549         .recover_open   = nfs40_open_expired,
8550         .recover_lock   = nfs4_lock_expired,
8551         .establish_clid = nfs4_init_clientid,
8552 };
8553
8554 #if defined(CONFIG_NFS_V4_1)
8555 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8556         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8557         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8558         .recover_open   = nfs41_open_expired,
8559         .recover_lock   = nfs41_lock_expired,
8560         .establish_clid = nfs41_init_clientid,
8561 };
8562 #endif /* CONFIG_NFS_V4_1 */
8563
8564 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8565         .sched_state_renewal = nfs4_proc_async_renew,
8566         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8567         .renew_lease = nfs4_proc_renew,
8568 };
8569
8570 #if defined(CONFIG_NFS_V4_1)
8571 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8572         .sched_state_renewal = nfs41_proc_async_sequence,
8573         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8574         .renew_lease = nfs4_proc_sequence,
8575 };
8576 #endif
8577
8578 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8579         .get_locations = _nfs40_proc_get_locations,
8580         .fsid_present = _nfs40_proc_fsid_present,
8581 };
8582
8583 #if defined(CONFIG_NFS_V4_1)
8584 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8585         .get_locations = _nfs41_proc_get_locations,
8586         .fsid_present = _nfs41_proc_fsid_present,
8587 };
8588 #endif  /* CONFIG_NFS_V4_1 */
8589
8590 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8591         .minor_version = 0,
8592         .init_caps = NFS_CAP_READDIRPLUS
8593                 | NFS_CAP_ATOMIC_OPEN
8594                 | NFS_CAP_CHANGE_ATTR
8595                 | NFS_CAP_POSIX_LOCK,
8596         .init_client = nfs40_init_client,
8597         .shutdown_client = nfs40_shutdown_client,
8598         .match_stateid = nfs4_match_stateid,
8599         .find_root_sec = nfs4_find_root_sec,
8600         .free_lock_state = nfs4_release_lockowner,
8601         .alloc_seqid = nfs_alloc_seqid,
8602         .call_sync_ops = &nfs40_call_sync_ops,
8603         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8604         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8605         .state_renewal_ops = &nfs40_state_renewal_ops,
8606         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8607 };
8608
8609 #if defined(CONFIG_NFS_V4_1)
8610 static struct nfs_seqid *
8611 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8612 {
8613         return NULL;
8614 }
8615
8616 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8617         .minor_version = 1,
8618         .init_caps = NFS_CAP_READDIRPLUS
8619                 | NFS_CAP_ATOMIC_OPEN
8620                 | NFS_CAP_CHANGE_ATTR
8621                 | NFS_CAP_POSIX_LOCK
8622                 | NFS_CAP_STATEID_NFSV41
8623                 | NFS_CAP_ATOMIC_OPEN_V1,
8624         .init_client = nfs41_init_client,
8625         .shutdown_client = nfs41_shutdown_client,
8626         .match_stateid = nfs41_match_stateid,
8627         .find_root_sec = nfs41_find_root_sec,
8628         .free_lock_state = nfs41_free_lock_state,
8629         .alloc_seqid = nfs_alloc_no_seqid,
8630         .call_sync_ops = &nfs41_call_sync_ops,
8631         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8632         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8633         .state_renewal_ops = &nfs41_state_renewal_ops,
8634         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8635 };
8636 #endif
8637
8638 #if defined(CONFIG_NFS_V4_2)
8639 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8640         .minor_version = 2,
8641         .init_caps = NFS_CAP_READDIRPLUS
8642                 | NFS_CAP_ATOMIC_OPEN
8643                 | NFS_CAP_CHANGE_ATTR
8644                 | NFS_CAP_POSIX_LOCK
8645                 | NFS_CAP_STATEID_NFSV41
8646                 | NFS_CAP_ATOMIC_OPEN_V1
8647                 | NFS_CAP_ALLOCATE
8648                 | NFS_CAP_DEALLOCATE
8649                 | NFS_CAP_SEEK
8650                 | NFS_CAP_LAYOUTSTATS,
8651         .init_client = nfs41_init_client,
8652         .shutdown_client = nfs41_shutdown_client,
8653         .match_stateid = nfs41_match_stateid,
8654         .find_root_sec = nfs41_find_root_sec,
8655         .free_lock_state = nfs41_free_lock_state,
8656         .call_sync_ops = &nfs41_call_sync_ops,
8657         .alloc_seqid = nfs_alloc_no_seqid,
8658         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8659         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8660         .state_renewal_ops = &nfs41_state_renewal_ops,
8661 };
8662 #endif
8663
8664 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8665         [0] = &nfs_v4_0_minor_ops,
8666 #if defined(CONFIG_NFS_V4_1)
8667         [1] = &nfs_v4_1_minor_ops,
8668 #endif
8669 #if defined(CONFIG_NFS_V4_2)
8670         [2] = &nfs_v4_2_minor_ops,
8671 #endif
8672 };
8673
8674 static const struct inode_operations nfs4_dir_inode_operations = {
8675         .create         = nfs_create,
8676         .lookup         = nfs_lookup,
8677         .atomic_open    = nfs_atomic_open,
8678         .link           = nfs_link,
8679         .unlink         = nfs_unlink,
8680         .symlink        = nfs_symlink,
8681         .mkdir          = nfs_mkdir,
8682         .rmdir          = nfs_rmdir,
8683         .mknod          = nfs_mknod,
8684         .rename         = nfs_rename,
8685         .permission     = nfs_permission,
8686         .getattr        = nfs_getattr,
8687         .setattr        = nfs_setattr,
8688         .getxattr       = generic_getxattr,
8689         .setxattr       = generic_setxattr,
8690         .listxattr      = generic_listxattr,
8691         .removexattr    = generic_removexattr,
8692 };
8693
8694 static const struct inode_operations nfs4_file_inode_operations = {
8695         .permission     = nfs_permission,
8696         .getattr        = nfs_getattr,
8697         .setattr        = nfs_setattr,
8698         .getxattr       = generic_getxattr,
8699         .setxattr       = generic_setxattr,
8700         .listxattr      = generic_listxattr,
8701         .removexattr    = generic_removexattr,
8702 };
8703
8704 const struct nfs_rpc_ops nfs_v4_clientops = {
8705         .version        = 4,                    /* protocol version */
8706         .dentry_ops     = &nfs4_dentry_operations,
8707         .dir_inode_ops  = &nfs4_dir_inode_operations,
8708         .file_inode_ops = &nfs4_file_inode_operations,
8709         .file_ops       = &nfs4_file_operations,
8710         .getroot        = nfs4_proc_get_root,
8711         .submount       = nfs4_submount,
8712         .try_mount      = nfs4_try_mount,
8713         .getattr        = nfs4_proc_getattr,
8714         .setattr        = nfs4_proc_setattr,
8715         .lookup         = nfs4_proc_lookup,
8716         .access         = nfs4_proc_access,
8717         .readlink       = nfs4_proc_readlink,
8718         .create         = nfs4_proc_create,
8719         .remove         = nfs4_proc_remove,
8720         .unlink_setup   = nfs4_proc_unlink_setup,
8721         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8722         .unlink_done    = nfs4_proc_unlink_done,
8723         .rename_setup   = nfs4_proc_rename_setup,
8724         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8725         .rename_done    = nfs4_proc_rename_done,
8726         .link           = nfs4_proc_link,
8727         .symlink        = nfs4_proc_symlink,
8728         .mkdir          = nfs4_proc_mkdir,
8729         .rmdir          = nfs4_proc_remove,
8730         .readdir        = nfs4_proc_readdir,
8731         .mknod          = nfs4_proc_mknod,
8732         .statfs         = nfs4_proc_statfs,
8733         .fsinfo         = nfs4_proc_fsinfo,
8734         .pathconf       = nfs4_proc_pathconf,
8735         .set_capabilities = nfs4_server_capabilities,
8736         .decode_dirent  = nfs4_decode_dirent,
8737         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8738         .read_setup     = nfs4_proc_read_setup,
8739         .read_done      = nfs4_read_done,
8740         .write_setup    = nfs4_proc_write_setup,
8741         .write_done     = nfs4_write_done,
8742         .commit_setup   = nfs4_proc_commit_setup,
8743         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8744         .commit_done    = nfs4_commit_done,
8745         .lock           = nfs4_proc_lock,
8746         .clear_acl_cache = nfs4_zap_acl_attr,
8747         .close_context  = nfs4_close_context,
8748         .open_context   = nfs4_atomic_open,
8749         .have_delegation = nfs4_have_delegation,
8750         .return_delegation = nfs4_inode_return_delegation,
8751         .alloc_client   = nfs4_alloc_client,
8752         .init_client    = nfs4_init_client,
8753         .free_client    = nfs4_free_client,
8754         .create_server  = nfs4_create_server,
8755         .clone_server   = nfs_clone_server,
8756 };
8757
8758 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8759         .prefix = XATTR_NAME_NFSV4_ACL,
8760         .list   = nfs4_xattr_list_nfs4_acl,
8761         .get    = nfs4_xattr_get_nfs4_acl,
8762         .set    = nfs4_xattr_set_nfs4_acl,
8763 };
8764
8765 const struct xattr_handler *nfs4_xattr_handlers[] = {
8766         &nfs4_xattr_nfs4_acl_handler,
8767 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8768         &nfs4_xattr_nfs4_label_handler,
8769 #endif
8770         NULL
8771 };
8772
8773 /*
8774  * Local variables:
8775  *  c-basic-offset: 8
8776  * End:
8777  */