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_notify_server(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 nfs4_label *a_label;
 916         struct nfs_fattr f_attr;
 917         struct nfs4_label *f_label;
 918         struct dentry *dir;
 919         struct dentry *dentry;
 920         struct nfs4_state_owner *owner;
 921         struct nfs4_state *state;
 922         struct iattr attrs;
 923         unsigned long timestamp;
 924         unsigned int rpc_done : 1;
 925         unsigned int file_created : 1;
 926         unsigned int is_recover : 1;
 927         int rpc_status;
 928         int cancelled;
 929 };
 930
 931 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
 932                 int err, struct nfs4_exception *exception)
 933 {
 934         if (err != -EINVAL)
 935                 return false;
 936         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 937                 return false;
 938         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
 939         exception->retry = 1;
 940         return true;
 941 }
 942
 943 static u32
 944 nfs4_map_atomic_open_share(struct nfs_server *server,
 945                 fmode_t fmode, int openflags)
 946 {
 947         u32 res = 0;
 948
 949         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
 950         case FMODE_READ:
 951                 res = NFS4_SHARE_ACCESS_READ;
 952                 break;
 953         case FMODE_WRITE:
 954                 res = NFS4_SHARE_ACCESS_WRITE;
 955                 break;
 956         case FMODE_READ|FMODE_WRITE:
 957                 res = NFS4_SHARE_ACCESS_BOTH;
 958         }
 959         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
 960                 goto out;
 961         /* Want no delegation if we're using O_DIRECT */
 962         if (openflags & O_DIRECT)
 963                 res |= NFS4_SHARE_WANT_NO_DELEG;
 964 out:
 965         return res;
 966 }
 967
 968 static enum open_claim_type4
 969 nfs4_map_atomic_open_claim(struct nfs_server *server,
 970                 enum open_claim_type4 claim)
 971 {
 972         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
 973                 return claim;
 974         switch (claim) {
 975         default:
 976                 return claim;
 977         case NFS4_OPEN_CLAIM_FH:
 978                 return NFS4_OPEN_CLAIM_NULL;
 979         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 980                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
 981         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 982                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
 983         }
 984 }
 985
 986 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
 987 {
 988         p->o_res.f_attr = &p->f_attr;
 989         p->o_res.f_label = p->f_label;
 990         p->o_res.seqid = p->o_arg.seqid;
 991         p->c_res.seqid = p->c_arg.seqid;
 992         p->o_res.server = p->o_arg.server;
 993         p->o_res.access_request = p->o_arg.access;
 994         nfs_fattr_init(&p->f_attr);
 995         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
 996 }
 997
 998 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
 999                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1000                 const struct iattr *attrs,
1001                 struct nfs4_label *label,
1002                 enum open_claim_type4 claim,
1003                 gfp_t gfp_mask)
1004 {
1005         struct dentry *parent = dget_parent(dentry);
1006         struct inode *dir = d_inode(parent);
1007         struct nfs_server *server = NFS_SERVER(dir);
1008         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1009         struct nfs4_opendata *p;
1010
1011         p = kzalloc(sizeof(*p), gfp_mask);
1012         if (p == NULL)
1013                 goto err;
1014
1015         p->f_label = nfs4_label_alloc(server, gfp_mask);
1016         if (IS_ERR(p->f_label))
1017                 goto err_free_p;
1018
1019         p->a_label = nfs4_label_alloc(server, gfp_mask);
1020         if (IS_ERR(p->a_label))
1021                 goto err_free_f;
1022
1023         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1024         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1025         if (IS_ERR(p->o_arg.seqid))
1026                 goto err_free_label;
1027         nfs_sb_active(dentry->d_sb);
1028         p->dentry = dget(dentry);
1029         p->dir = parent;
1030         p->owner = sp;
1031         atomic_inc(&sp->so_count);
1032         p->o_arg.open_flags = flags;
1033         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1034         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1035                         fmode, flags);
1036         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1037          * will return permission denied for all bits until close */
1038         if (!(flags & O_EXCL)) {
1039                 /* ask server to check for all possible rights as results
1040                  * are cached */
1041                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1042                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1043         }
1044         p->o_arg.clientid = server->nfs_client->cl_clientid;
1045         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1046         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1047         p->o_arg.name = &dentry->d_name;
1048         p->o_arg.server = server;
1049         p->o_arg.bitmask = nfs4_bitmask(server, label);
1050         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1051         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1052         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1053         switch (p->o_arg.claim) {
1054         case NFS4_OPEN_CLAIM_NULL:
1055         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1056         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1057                 p->o_arg.fh = NFS_FH(dir);
1058                 break;
1059         case NFS4_OPEN_CLAIM_PREVIOUS:
1060         case NFS4_OPEN_CLAIM_FH:
1061         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1062         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1063                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1064         }
1065         if (attrs != NULL && attrs->ia_valid != 0) {
1066                 __u32 verf[2];
1067
1068                 p->o_arg.u.attrs = &p->attrs;
1069                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1070
1071                 verf[0] = jiffies;
1072                 verf[1] = current->pid;
1073                 memcpy(p->o_arg.u.verifier.data, verf,
1074                                 sizeof(p->o_arg.u.verifier.data));
1075         }
1076         p->c_arg.fh = &p->o_res.fh;
1077         p->c_arg.stateid = &p->o_res.stateid;
1078         p->c_arg.seqid = p->o_arg.seqid;
1079         nfs4_init_opendata_res(p);
1080         kref_init(&p->kref);
1081         return p;
1082
1083 err_free_label:
1084         nfs4_label_free(p->a_label);
1085 err_free_f:
1086         nfs4_label_free(p->f_label);
1087 err_free_p:
1088         kfree(p);
1089 err:
1090         dput(parent);
1091         return NULL;
1092 }
1093
1094 static void nfs4_opendata_free(struct kref *kref)
1095 {
1096         struct nfs4_opendata *p = container_of(kref,
1097                         struct nfs4_opendata, kref);
1098         struct super_block *sb = p->dentry->d_sb;
1099
1100         nfs_free_seqid(p->o_arg.seqid);
1101         if (p->state != NULL)
1102                 nfs4_put_open_state(p->state);
1103         nfs4_put_state_owner(p->owner);
1104
1105         nfs4_label_free(p->a_label);
1106         nfs4_label_free(p->f_label);
1107
1108         dput(p->dir);
1109         dput(p->dentry);
1110         nfs_sb_deactive(sb);
1111         nfs_fattr_free_names(&p->f_attr);
1112         kfree(p->f_attr.mdsthreshold);
1113         kfree(p);
1114 }
1115
1116 static void nfs4_opendata_put(struct nfs4_opendata *p)
1117 {
1118         if (p != NULL)
1119                 kref_put(&p->kref, nfs4_opendata_free);
1120 }
1121
1122 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1123 {
1124         int ret;
1125
1126         ret = rpc_wait_for_completion_task(task);
1127         return ret;
1128 }
1129
1130 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1131                 fmode_t fmode)
1132 {
1133         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1134         case FMODE_READ|FMODE_WRITE:
1135                 return state->n_rdwr != 0;
1136         case FMODE_WRITE:
1137                 return state->n_wronly != 0;
1138         case FMODE_READ:
1139                 return state->n_rdonly != 0;
1140         }
1141         WARN_ON_ONCE(1);
1142         return false;
1143 }
1144
1145 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1146 {
1147         int ret = 0;
1148
1149         if (open_mode & (O_EXCL|O_TRUNC))
1150                 goto out;
1151         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1152                 case FMODE_READ:
1153                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1154                                 && state->n_rdonly != 0;
1155                         break;
1156                 case FMODE_WRITE:
1157                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1158                                 && state->n_wronly != 0;
1159                         break;
1160                 case FMODE_READ|FMODE_WRITE:
1161                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1162                                 && state->n_rdwr != 0;
1163         }
1164 out:
1165         return ret;
1166 }
1167
1168 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1169                 enum open_claim_type4 claim)
1170 {
1171         if (delegation == NULL)
1172                 return 0;
1173         if ((delegation->type & fmode) != fmode)
1174                 return 0;
1175         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1176                 return 0;
1177         switch (claim) {
1178         case NFS4_OPEN_CLAIM_NULL:
1179         case NFS4_OPEN_CLAIM_FH:
1180                 break;
1181         case NFS4_OPEN_CLAIM_PREVIOUS:
1182                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1183                         break;
1184         default:
1185                 return 0;
1186         }
1187         nfs_mark_delegation_referenced(delegation);
1188         return 1;
1189 }
1190
1191 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1192 {
1193         switch (fmode) {
1194                 case FMODE_WRITE:
1195                         state->n_wronly++;
1196                         break;
1197                 case FMODE_READ:
1198                         state->n_rdonly++;
1199                         break;
1200                 case FMODE_READ|FMODE_WRITE:
1201                         state->n_rdwr++;
1202         }
1203         nfs4_state_set_mode_locked(state, state->state | fmode);
1204 }
1205
1206 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1207 {
1208         struct nfs_client *clp = state->owner->so_server->nfs_client;
1209         bool need_recover = false;
1210
1211         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1212                 need_recover = true;
1213         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1214                 need_recover = true;
1215         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1216                 need_recover = true;
1217         if (need_recover)
1218                 nfs4_state_mark_reclaim_nograce(clp, state);
1219 }
1220
1221 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1222                 nfs4_stateid *stateid)
1223 {
1224         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1225                 return true;
1226         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1227                 nfs_test_and_clear_all_open_stateid(state);
1228                 return true;
1229         }
1230         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1231                 return true;
1232         return false;
1233 }
1234
1235 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1236 {
1237         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1238                 return;
1239         if (state->n_wronly)
1240                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1241         if (state->n_rdonly)
1242                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1243         if (state->n_rdwr)
1244                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1245         set_bit(NFS_OPEN_STATE, &state->flags);
1246 }
1247
1248 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1249                 nfs4_stateid *arg_stateid,
1250                 nfs4_stateid *stateid, fmode_t fmode)
1251 {
1252         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1253         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1254         case FMODE_WRITE:
1255                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1256                 break;
1257         case FMODE_READ:
1258                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1259                 break;
1260         case 0:
1261                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1262                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1263                 clear_bit(NFS_OPEN_STATE, &state->flags);
1264         }
1265         if (stateid == NULL)
1266                 return;
1267         /* Handle races with OPEN */
1268         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1269             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1270             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1271                 nfs_resync_open_stateid_locked(state);
1272                 return;
1273         }
1274         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1275                 nfs4_stateid_copy(&state->stateid, stateid);
1276         nfs4_stateid_copy(&state->open_stateid, stateid);
1277 }
1278
1279 static void nfs_clear_open_stateid(struct nfs4_state *state,
1280         nfs4_stateid *arg_stateid,
1281         nfs4_stateid *stateid, fmode_t fmode)
1282 {
1283         write_seqlock(&state->seqlock);
1284         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1285         write_sequnlock(&state->seqlock);
1286         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1287                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1288 }
1289
1290 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1291 {
1292         switch (fmode) {
1293                 case FMODE_READ:
1294                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1295                         break;
1296                 case FMODE_WRITE:
1297                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1298                         break;
1299                 case FMODE_READ|FMODE_WRITE:
1300                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1301         }
1302         if (!nfs_need_update_open_stateid(state, stateid))
1303                 return;
1304         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1305                 nfs4_stateid_copy(&state->stateid, stateid);
1306         nfs4_stateid_copy(&state->open_stateid, stateid);
1307 }
1308
1309 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1310 {
1311         /*
1312          * Protect the call to nfs4_state_set_mode_locked and
1313          * serialise the stateid update
1314          */
1315         write_seqlock(&state->seqlock);
1316         if (deleg_stateid != NULL) {
1317                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1318                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1319         }
1320         if (open_stateid != NULL)
1321                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1322         write_sequnlock(&state->seqlock);
1323         spin_lock(&state->owner->so_lock);
1324         update_open_stateflags(state, fmode);
1325         spin_unlock(&state->owner->so_lock);
1326 }
1327
1328 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1329 {
1330         struct nfs_inode *nfsi = NFS_I(state->inode);
1331         struct nfs_delegation *deleg_cur;
1332         int ret = 0;
1333
1334         fmode &= (FMODE_READ|FMODE_WRITE);
1335
1336         rcu_read_lock();
1337         deleg_cur = rcu_dereference(nfsi->delegation);
1338         if (deleg_cur == NULL)
1339                 goto no_delegation;
1340
1341         spin_lock(&deleg_cur->lock);
1342         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1343            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1344             (deleg_cur->type & fmode) != fmode)
1345                 goto no_delegation_unlock;
1346
1347         if (delegation == NULL)
1348                 delegation = &deleg_cur->stateid;
1349         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1350                 goto no_delegation_unlock;
1351
1352         nfs_mark_delegation_referenced(deleg_cur);
1353         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1354         ret = 1;
1355 no_delegation_unlock:
1356         spin_unlock(&deleg_cur->lock);
1357 no_delegation:
1358         rcu_read_unlock();
1359
1360         if (!ret && open_stateid != NULL) {
1361                 __update_open_stateid(state, open_stateid, NULL, fmode);
1362                 ret = 1;
1363         }
1364         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1365                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1366
1367         return ret;
1368 }
1369
1370 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1371                 const nfs4_stateid *stateid)
1372 {
1373         struct nfs4_state *state = lsp->ls_state;
1374         bool ret = false;
1375
1376         spin_lock(&state->state_lock);
1377         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1378                 goto out_noupdate;
1379         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1380                 goto out_noupdate;
1381         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1382         ret = true;
1383 out_noupdate:
1384         spin_unlock(&state->state_lock);
1385         return ret;
1386 }
1387
1388 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1389 {
1390         struct nfs_delegation *delegation;
1391
1392         rcu_read_lock();
1393         delegation = rcu_dereference(NFS_I(inode)->delegation);
1394         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1395                 rcu_read_unlock();
1396                 return;
1397         }
1398         rcu_read_unlock();
1399         nfs4_inode_return_delegation(inode);
1400 }
1401
1402 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1403 {
1404         struct nfs4_state *state = opendata->state;
1405         struct nfs_inode *nfsi = NFS_I(state->inode);
1406         struct nfs_delegation *delegation;
1407         int open_mode = opendata->o_arg.open_flags;
1408         fmode_t fmode = opendata->o_arg.fmode;
1409         enum open_claim_type4 claim = opendata->o_arg.claim;
1410         nfs4_stateid stateid;
1411         int ret = -EAGAIN;
1412
1413         for (;;) {
1414                 spin_lock(&state->owner->so_lock);
1415                 if (can_open_cached(state, fmode, open_mode)) {
1416                         update_open_stateflags(state, fmode);
1417                         spin_unlock(&state->owner->so_lock);
1418                         goto out_return_state;
1419                 }
1420                 spin_unlock(&state->owner->so_lock);
1421                 rcu_read_lock();
1422                 delegation = rcu_dereference(nfsi->delegation);
1423                 if (!can_open_delegated(delegation, fmode, claim)) {
1424                         rcu_read_unlock();
1425                         break;
1426                 }
1427                 /* Save the delegation */
1428                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1429                 rcu_read_unlock();
1430                 nfs_release_seqid(opendata->o_arg.seqid);
1431                 if (!opendata->is_recover) {
1432                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1433                         if (ret != 0)
1434                                 goto out;
1435                 }
1436                 ret = -EAGAIN;
1437
1438                 /* Try to update the stateid using the delegation */
1439                 if (update_open_stateid(state, NULL, &stateid, fmode))
1440                         goto out_return_state;
1441         }
1442 out:
1443         return ERR_PTR(ret);
1444 out_return_state:
1445         atomic_inc(&state->count);
1446         return state;
1447 }
1448
1449 static void
1450 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1451 {
1452         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1453         struct nfs_delegation *delegation;
1454         int delegation_flags = 0;
1455
1456         rcu_read_lock();
1457         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1458         if (delegation)
1459                 delegation_flags = delegation->flags;
1460         rcu_read_unlock();
1461         switch (data->o_arg.claim) {
1462         default:
1463                 break;
1464         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1465         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1466                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1467                                    "returning a delegation for "
1468                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1469                                    clp->cl_hostname);
1470                 return;
1471         }
1472         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1473                 nfs_inode_set_delegation(state->inode,
1474                                          data->owner->so_cred,
1475                                          &data->o_res);
1476         else
1477                 nfs_inode_reclaim_delegation(state->inode,
1478                                              data->owner->so_cred,
1479                                              &data->o_res);
1480 }
1481
1482 /*
1483  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1484  * and update the nfs4_state.
1485  */
1486 static struct nfs4_state *
1487 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1488 {
1489         struct inode *inode = data->state->inode;
1490         struct nfs4_state *state = data->state;
1491         int ret;
1492
1493         if (!data->rpc_done) {
1494                 if (data->rpc_status) {
1495                         ret = data->rpc_status;
1496                         goto err;
1497                 }
1498                 /* cached opens have already been processed */
1499                 goto update;
1500         }
1501
1502         ret = nfs_refresh_inode(inode, &data->f_attr);
1503         if (ret)
1504                 goto err;
1505
1506         if (data->o_res.delegation_type != 0)
1507                 nfs4_opendata_check_deleg(data, state);
1508 update:
1509         update_open_stateid(state, &data->o_res.stateid, NULL,
1510                             data->o_arg.fmode);
1511         atomic_inc(&state->count);
1512
1513         return state;
1514 err:
1515         return ERR_PTR(ret);
1516
1517 }
1518
1519 static struct nfs4_state *
1520 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1521 {
1522         struct inode *inode;
1523         struct nfs4_state *state = NULL;
1524         int ret;
1525
1526         if (!data->rpc_done) {
1527                 state = nfs4_try_open_cached(data);
1528                 goto out;
1529         }
1530
1531         ret = -EAGAIN;
1532         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1533                 goto err;
1534         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1535         ret = PTR_ERR(inode);
1536         if (IS_ERR(inode))
1537                 goto err;
1538         ret = -ENOMEM;
1539         state = nfs4_get_open_state(inode, data->owner);
1540         if (state == NULL)
1541                 goto err_put_inode;
1542         if (data->o_res.delegation_type != 0)
1543                 nfs4_opendata_check_deleg(data, state);
1544         update_open_stateid(state, &data->o_res.stateid, NULL,
1545                         data->o_arg.fmode);
1546         iput(inode);
1547 out:
1548         nfs_release_seqid(data->o_arg.seqid);
1549         return state;
1550 err_put_inode:
1551         iput(inode);
1552 err:
1553         return ERR_PTR(ret);
1554 }
1555
1556 static struct nfs4_state *
1557 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1558 {
1559         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1560                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1561         return _nfs4_opendata_to_nfs4_state(data);
1562 }
1563
1564 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1565 {
1566         struct nfs_inode *nfsi = NFS_I(state->inode);
1567         struct nfs_open_context *ctx;
1568
1569         spin_lock(&state->inode->i_lock);
1570         list_for_each_entry(ctx, &nfsi->open_files, list) {
1571                 if (ctx->state != state)
1572                         continue;
1573                 get_nfs_open_context(ctx);
1574                 spin_unlock(&state->inode->i_lock);
1575                 return ctx;
1576         }
1577         spin_unlock(&state->inode->i_lock);
1578         return ERR_PTR(-ENOENT);
1579 }
1580
1581 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1582                 struct nfs4_state *state, enum open_claim_type4 claim)
1583 {
1584         struct nfs4_opendata *opendata;
1585
1586         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1587                         NULL, NULL, claim, GFP_NOFS);
1588         if (opendata == NULL)
1589                 return ERR_PTR(-ENOMEM);
1590         opendata->state = state;
1591         atomic_inc(&state->count);
1592         return opendata;
1593 }
1594
1595 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1596                 fmode_t fmode)
1597 {
1598         struct nfs4_state *newstate;
1599         int ret;
1600
1601         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1602                 return 0;
1603         opendata->o_arg.open_flags = 0;
1604         opendata->o_arg.fmode = fmode;
1605         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1606                         NFS_SB(opendata->dentry->d_sb),
1607                         fmode, 0);
1608         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1609         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1610         nfs4_init_opendata_res(opendata);
1611         ret = _nfs4_recover_proc_open(opendata);
1612         if (ret != 0)
1613                 return ret;
1614         newstate = nfs4_opendata_to_nfs4_state(opendata);
1615         if (IS_ERR(newstate))
1616                 return PTR_ERR(newstate);
1617         if (newstate != opendata->state)
1618                 ret = -ESTALE;
1619         nfs4_close_state(newstate, fmode);
1620         return ret;
1621 }
1622
1623 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1624 {
1625         int ret;
1626
1627         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1628         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1629         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1630         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1631         /* memory barrier prior to reading state->n_* */
1632         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1633         clear_bit(NFS_OPEN_STATE, &state->flags);
1634         smp_rmb();
1635         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1636         if (ret != 0)
1637                 return ret;
1638         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1639         if (ret != 0)
1640                 return ret;
1641         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1642         if (ret != 0)
1643                 return ret;
1644         /*
1645          * We may have performed cached opens for all three recoveries.
1646          * Check if we need to update the current stateid.
1647          */
1648         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1649             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1650                 write_seqlock(&state->seqlock);
1651                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1652                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1653                 write_sequnlock(&state->seqlock);
1654         }
1655         return 0;
1656 }
1657
1658 /*
1659  * OPEN_RECLAIM:
1660  *      reclaim state on the server after a reboot.
1661  */
1662 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1663 {
1664         struct nfs_delegation *delegation;
1665         struct nfs4_opendata *opendata;
1666         fmode_t delegation_type = 0;
1667         int status;
1668
1669         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1670                         NFS4_OPEN_CLAIM_PREVIOUS);
1671         if (IS_ERR(opendata))
1672                 return PTR_ERR(opendata);
1673         rcu_read_lock();
1674         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1675         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1676                 delegation_type = delegation->type;
1677         rcu_read_unlock();
1678         opendata->o_arg.u.delegation_type = delegation_type;
1679         status = nfs4_open_recover(opendata, state);
1680         nfs4_opendata_put(opendata);
1681         return status;
1682 }
1683
1684 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1685 {
1686         struct nfs_server *server = NFS_SERVER(state->inode);
1687         struct nfs4_exception exception = { };
1688         int err;
1689         do {
1690                 err = _nfs4_do_open_reclaim(ctx, state);
1691                 trace_nfs4_open_reclaim(ctx, 0, err);
1692                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1693                         continue;
1694                 if (err != -NFS4ERR_DELAY)
1695                         break;
1696                 nfs4_handle_exception(server, err, &exception);
1697         } while (exception.retry);
1698         return err;
1699 }
1700
1701 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1702 {
1703         struct nfs_open_context *ctx;
1704         int ret;
1705
1706         ctx = nfs4_state_find_open_context(state);
1707         if (IS_ERR(ctx))
1708                 return -EAGAIN;
1709         ret = nfs4_do_open_reclaim(ctx, state);
1710         put_nfs_open_context(ctx);
1711         return ret;
1712 }
1713
1714 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1715 {
1716         switch (err) {
1717                 default:
1718                         printk(KERN_ERR "NFS: %s: unhandled error "
1719                                         "%d.\n", __func__, err);
1720                 case 0:
1721                 case -ENOENT:
1722                 case -EAGAIN:
1723                 case -ESTALE:
1724                         break;
1725                 case -NFS4ERR_BADSESSION:
1726                 case -NFS4ERR_BADSLOT:
1727                 case -NFS4ERR_BAD_HIGH_SLOT:
1728                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1729                 case -NFS4ERR_DEADSESSION:
1730                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1731                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1732                         return -EAGAIN;
1733                 case -NFS4ERR_STALE_CLIENTID:
1734                 case -NFS4ERR_STALE_STATEID:
1735                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1736                 case -NFS4ERR_EXPIRED:
1737                         /* Don't recall a delegation if it was lost */
1738                         nfs4_schedule_lease_recovery(server->nfs_client);
1739                         return -EAGAIN;
1740                 case -NFS4ERR_MOVED:
1741                         nfs4_schedule_migration_recovery(server);
1742                         return -EAGAIN;
1743                 case -NFS4ERR_LEASE_MOVED:
1744                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1745                         return -EAGAIN;
1746                 case -NFS4ERR_DELEG_REVOKED:
1747                 case -NFS4ERR_ADMIN_REVOKED:
1748                 case -NFS4ERR_BAD_STATEID:
1749                 case -NFS4ERR_OPENMODE:
1750                         nfs_inode_find_state_and_recover(state->inode,
1751                                         stateid);
1752                         nfs4_schedule_stateid_recovery(server, state);
1753                         return -EAGAIN;
1754                 case -NFS4ERR_DELAY:
1755                 case -NFS4ERR_GRACE:
1756                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1757                         ssleep(1);
1758                         return -EAGAIN;
1759                 case -ENOMEM:
1760                 case -NFS4ERR_DENIED:
1761                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1762                         return 0;
1763         }
1764         return err;
1765 }
1766
1767 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1768                 struct nfs4_state *state, const nfs4_stateid *stateid,
1769                 fmode_t type)
1770 {
1771         struct nfs_server *server = NFS_SERVER(state->inode);
1772         struct nfs4_opendata *opendata;
1773         int err = 0;
1774
1775         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1776                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1777         if (IS_ERR(opendata))
1778                 return PTR_ERR(opendata);
1779         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1780         write_seqlock(&state->seqlock);
1781         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1782         write_sequnlock(&state->seqlock);
1783         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1784         switch (type & (FMODE_READ|FMODE_WRITE)) {
1785         case FMODE_READ|FMODE_WRITE:
1786         case FMODE_WRITE:
1787                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1788                 if (err)
1789                         break;
1790                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1791                 if (err)
1792                         break;
1793         case FMODE_READ:
1794                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1795         }
1796         nfs4_opendata_put(opendata);
1797         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1798 }
1799
1800 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1801 {
1802         struct nfs4_opendata *data = calldata;
1803
1804         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1805                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1806 }
1807
1808 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1809 {
1810         struct nfs4_opendata *data = calldata;
1811
1812         nfs40_sequence_done(task, &data->c_res.seq_res);
1813
1814         data->rpc_status = task->tk_status;
1815         if (data->rpc_status == 0) {
1816                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1817                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1818                 renew_lease(data->o_res.server, data->timestamp);
1819                 data->rpc_done = 1;
1820         }
1821 }
1822
1823 static void nfs4_open_confirm_release(void *calldata)
1824 {
1825         struct nfs4_opendata *data = calldata;
1826         struct nfs4_state *state = NULL;
1827
1828         /* If this request hasn't been cancelled, do nothing */
1829         if (data->cancelled == 0)
1830                 goto out_free;
1831         /* In case of error, no cleanup! */
1832         if (!data->rpc_done)
1833                 goto out_free;
1834         state = nfs4_opendata_to_nfs4_state(data);
1835         if (!IS_ERR(state))
1836                 nfs4_close_state(state, data->o_arg.fmode);
1837 out_free:
1838         nfs4_opendata_put(data);
1839 }
1840
1841 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1842         .rpc_call_prepare = nfs4_open_confirm_prepare,
1843         .rpc_call_done = nfs4_open_confirm_done,
1844         .rpc_release = nfs4_open_confirm_release,
1845 };
1846
1847 /*
1848  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1849  */
1850 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1851 {
1852         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1853         struct rpc_task *task;
1854         struct  rpc_message msg = {
1855                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1856                 .rpc_argp = &data->c_arg,
1857                 .rpc_resp = &data->c_res,
1858                 .rpc_cred = data->owner->so_cred,
1859         };
1860         struct rpc_task_setup task_setup_data = {
1861                 .rpc_client = server->client,
1862                 .rpc_message = &msg,
1863                 .callback_ops = &nfs4_open_confirm_ops,
1864                 .callback_data = data,
1865                 .workqueue = nfsiod_workqueue,
1866                 .flags = RPC_TASK_ASYNC,
1867         };
1868         int status;
1869
1870         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1871         kref_get(&data->kref);
1872         data->rpc_done = 0;
1873         data->rpc_status = 0;
1874         data->timestamp = jiffies;
1875         if (data->is_recover)
1876                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1877         task = rpc_run_task(&task_setup_data);
1878         if (IS_ERR(task))
1879                 return PTR_ERR(task);
1880         status = nfs4_wait_for_completion_rpc_task(task);
1881         if (status != 0) {
1882                 data->cancelled = 1;
1883                 smp_wmb();
1884         } else
1885                 status = data->rpc_status;
1886         rpc_put_task(task);
1887         return status;
1888 }
1889
1890 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1891 {
1892         struct nfs4_opendata *data = calldata;
1893         struct nfs4_state_owner *sp = data->owner;
1894         struct nfs_client *clp = sp->so_server->nfs_client;
1895         enum open_claim_type4 claim = data->o_arg.claim;
1896
1897         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1898                 goto out_wait;
1899         /*
1900          * Check if we still need to send an OPEN call, or if we can use
1901          * a delegation instead.
1902          */
1903         if (data->state != NULL) {
1904                 struct nfs_delegation *delegation;
1905
1906                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1907                         goto out_no_action;
1908                 rcu_read_lock();
1909                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1910                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1911                         goto unlock_no_action;
1912                 rcu_read_unlock();
1913         }
1914         /* Update client id. */
1915         data->o_arg.clientid = clp->cl_clientid;
1916         switch (claim) {
1917         default:
1918                 break;
1919         case NFS4_OPEN_CLAIM_PREVIOUS:
1920         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1921         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1922                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
1923         case NFS4_OPEN_CLAIM_FH:
1924                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
1925                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
1926         }
1927         data->timestamp = jiffies;
1928         if (nfs4_setup_sequence(data->o_arg.server,
1929                                 &data->o_arg.seq_args,
1930                                 &data->o_res.seq_res,
1931                                 task) != 0)
1932                 nfs_release_seqid(data->o_arg.seqid);
1933
1934         /* Set the create mode (note dependency on the session type) */
1935         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
1936         if (data->o_arg.open_flags & O_EXCL) {
1937                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
1938                 if (nfs4_has_persistent_session(clp))
1939                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
1940                 else if (clp->cl_mvops->minor_version > 0)
1941                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
1942         }
1943         return;
1944 unlock_no_action:
1945         rcu_read_unlock();
1946 out_no_action:
1947         task->tk_action = NULL;
1948 out_wait:
1949         nfs4_sequence_done(task, &data->o_res.seq_res);
1950 }
1951
1952 static void nfs4_open_done(struct rpc_task *task, void *calldata)
1953 {
1954         struct nfs4_opendata *data = calldata;
1955
1956         data->rpc_status = task->tk_status;
1957
1958         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
1959                 return;
1960
1961         if (task->tk_status == 0) {
1962                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
1963                         switch (data->o_res.f_attr->mode & S_IFMT) {
1964                         case S_IFREG:
1965                                 break;
1966                         case S_IFLNK:
1967                                 data->rpc_status = -ELOOP;
1968                                 break;
1969                         case S_IFDIR:
1970                                 data->rpc_status = -EISDIR;
1971                                 break;
1972                         default:
1973                                 data->rpc_status = -ENOTDIR;
1974                         }
1975                 }
1976                 renew_lease(data->o_res.server, data->timestamp);
1977                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
1978                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
1979         }
1980         data->rpc_done = 1;
1981 }
1982
1983 static void nfs4_open_release(void *calldata)
1984 {
1985         struct nfs4_opendata *data = calldata;
1986         struct nfs4_state *state = NULL;
1987
1988         /* If this request hasn't been cancelled, do nothing */
1989         if (data->cancelled == 0)
1990                 goto out_free;
1991         /* In case of error, no cleanup! */
1992         if (data->rpc_status != 0 || !data->rpc_done)
1993                 goto out_free;
1994         /* In case we need an open_confirm, no cleanup! */
1995         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
1996                 goto out_free;
1997         state = nfs4_opendata_to_nfs4_state(data);
1998         if (!IS_ERR(state))
1999                 nfs4_close_state(state, data->o_arg.fmode);
2000 out_free:
2001         nfs4_opendata_put(data);
2002 }
2003
2004 static const struct rpc_call_ops nfs4_open_ops = {
2005         .rpc_call_prepare = nfs4_open_prepare,
2006         .rpc_call_done = nfs4_open_done,
2007         .rpc_release = nfs4_open_release,
2008 };
2009
2010 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2011 {
2012         struct inode *dir = d_inode(data->dir);
2013         struct nfs_server *server = NFS_SERVER(dir);
2014         struct nfs_openargs *o_arg = &data->o_arg;
2015         struct nfs_openres *o_res = &data->o_res;
2016         struct rpc_task *task;
2017         struct rpc_message msg = {
2018                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2019                 .rpc_argp = o_arg,
2020                 .rpc_resp = o_res,
2021                 .rpc_cred = data->owner->so_cred,
2022         };
2023         struct rpc_task_setup task_setup_data = {
2024                 .rpc_client = server->client,
2025                 .rpc_message = &msg,
2026                 .callback_ops = &nfs4_open_ops,
2027                 .callback_data = data,
2028                 .workqueue = nfsiod_workqueue,
2029                 .flags = RPC_TASK_ASYNC,
2030         };
2031         int status;
2032
2033         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2034         kref_get(&data->kref);
2035         data->rpc_done = 0;
2036         data->rpc_status = 0;
2037         data->cancelled = 0;
2038         data->is_recover = 0;
2039         if (isrecover) {
2040                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2041                 data->is_recover = 1;
2042         }
2043         task = rpc_run_task(&task_setup_data);
2044         if (IS_ERR(task))
2045                 return PTR_ERR(task);
2046         status = nfs4_wait_for_completion_rpc_task(task);
2047         if (status != 0) {
2048                 data->cancelled = 1;
2049                 smp_wmb();
2050         } else
2051                 status = data->rpc_status;
2052         rpc_put_task(task);
2053
2054         return status;
2055 }
2056
2057 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2058 {
2059         struct inode *dir = d_inode(data->dir);
2060         struct nfs_openres *o_res = &data->o_res;
2061         int status;
2062
2063         status = nfs4_run_open_task(data, 1);
2064         if (status != 0 || !data->rpc_done)
2065                 return status;
2066
2067         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2068
2069         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2070                 status = _nfs4_proc_open_confirm(data);
2071                 if (status != 0)
2072                         return status;
2073         }
2074
2075         return status;
2076 }
2077
2078 /*
2079  * Additional permission checks in order to distinguish between an
2080  * open for read, and an open for execute. This works around the
2081  * fact that NFSv4 OPEN treats read and execute permissions as being
2082  * the same.
2083  * Note that in the non-execute case, we want to turn off permission
2084  * checking if we just created a new file (POSIX open() semantics).
2085  */
2086 static int nfs4_opendata_access(struct rpc_cred *cred,
2087                                 struct nfs4_opendata *opendata,
2088                                 struct nfs4_state *state, fmode_t fmode,
2089                                 int openflags)
2090 {
2091         struct nfs_access_entry cache;
2092         u32 mask;
2093
2094         /* access call failed or for some reason the server doesn't
2095          * support any access modes -- defer access call until later */
2096         if (opendata->o_res.access_supported == 0)
2097                 return 0;
2098
2099         mask = 0;
2100         /*
2101          * Use openflags to check for exec, because fmode won't
2102          * always have FMODE_EXEC set when file open for exec.
2103          */
2104         if (openflags & __FMODE_EXEC) {
2105                 /* ONLY check for exec rights */
2106                 mask = MAY_EXEC;
2107         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2108                 mask = MAY_READ;
2109
2110         cache.cred = cred;
2111         cache.jiffies = jiffies;
2112         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2113         nfs_access_add_cache(state->inode, &cache);
2114
2115         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2116                 return 0;
2117
2118         /* even though OPEN succeeded, access is denied. Close the file */
2119         nfs4_close_state(state, fmode);
2120         return -EACCES;
2121 }
2122
2123 /*
2124  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2125  */
2126 static int _nfs4_proc_open(struct nfs4_opendata *data)
2127 {
2128         struct inode *dir = d_inode(data->dir);
2129         struct nfs_server *server = NFS_SERVER(dir);
2130         struct nfs_openargs *o_arg = &data->o_arg;
2131         struct nfs_openres *o_res = &data->o_res;
2132         int status;
2133
2134         status = nfs4_run_open_task(data, 0);
2135         if (!data->rpc_done)
2136                 return status;
2137         if (status != 0) {
2138                 if (status == -NFS4ERR_BADNAME &&
2139                                 !(o_arg->open_flags & O_CREAT))
2140                         return -ENOENT;
2141                 return status;
2142         }
2143
2144         nfs_fattr_map_and_free_names(server, &data->f_attr);
2145
2146         if (o_arg->open_flags & O_CREAT) {
2147                 update_changeattr(dir, &o_res->cinfo);
2148                 if (o_arg->open_flags & O_EXCL)
2149                         data->file_created = 1;
2150                 else if (o_res->cinfo.before != o_res->cinfo.after)
2151                         data->file_created = 1;
2152         }
2153         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2154                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2155         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2156                 status = _nfs4_proc_open_confirm(data);
2157                 if (status != 0)
2158                         return status;
2159         }
2160         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2161                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2162         return 0;
2163 }
2164
2165 static int nfs4_recover_expired_lease(struct nfs_server *server)
2166 {
2167         return nfs4_client_recover_expired_lease(server->nfs_client);
2168 }
2169
2170 /*
2171  * OPEN_EXPIRED:
2172  *      reclaim state on the server after a network partition.
2173  *      Assumes caller holds the appropriate lock
2174  */
2175 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2176 {
2177         struct nfs4_opendata *opendata;
2178         int ret;
2179
2180         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2181                         NFS4_OPEN_CLAIM_FH);
2182         if (IS_ERR(opendata))
2183                 return PTR_ERR(opendata);
2184         ret = nfs4_open_recover(opendata, state);
2185         if (ret == -ESTALE)
2186                 d_drop(ctx->dentry);
2187         nfs4_opendata_put(opendata);
2188         return ret;
2189 }
2190
2191 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2192 {
2193         struct nfs_server *server = NFS_SERVER(state->inode);
2194         struct nfs4_exception exception = { };
2195         int err;
2196
2197         do {
2198                 err = _nfs4_open_expired(ctx, state);
2199                 trace_nfs4_open_expired(ctx, 0, err);
2200                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2201                         continue;
2202                 switch (err) {
2203                 default:
2204                         goto out;
2205                 case -NFS4ERR_GRACE:
2206                 case -NFS4ERR_DELAY:
2207                         nfs4_handle_exception(server, err, &exception);
2208                         err = 0;
2209                 }
2210         } while (exception.retry);
2211 out:
2212         return err;
2213 }
2214
2215 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2216 {
2217         struct nfs_open_context *ctx;
2218         int ret;
2219
2220         ctx = nfs4_state_find_open_context(state);
2221         if (IS_ERR(ctx))
2222                 return -EAGAIN;
2223         ret = nfs4_do_open_expired(ctx, state);
2224         put_nfs_open_context(ctx);
2225         return ret;
2226 }
2227
2228 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2229 {
2230         nfs_remove_bad_delegation(state->inode);
2231         write_seqlock(&state->seqlock);
2232         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2233         write_sequnlock(&state->seqlock);
2234         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2235 }
2236
2237 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2238 {
2239         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2240                 nfs_finish_clear_delegation_stateid(state);
2241 }
2242
2243 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2244 {
2245         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2246         nfs40_clear_delegation_stateid(state);
2247         return nfs4_open_expired(sp, state);
2248 }
2249
2250 #if defined(CONFIG_NFS_V4_1)
2251 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2252 {
2253         struct nfs_server *server = NFS_SERVER(state->inode);
2254         nfs4_stateid stateid;
2255         struct nfs_delegation *delegation;
2256         struct rpc_cred *cred;
2257         int status;
2258
2259         /* Get the delegation credential for use by test/free_stateid */
2260         rcu_read_lock();
2261         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2262         if (delegation == NULL) {
2263                 rcu_read_unlock();
2264                 return;
2265         }
2266
2267         nfs4_stateid_copy(&stateid, &delegation->stateid);
2268         cred = get_rpccred(delegation->cred);
2269         rcu_read_unlock();
2270         status = nfs41_test_stateid(server, &stateid, cred);
2271         trace_nfs4_test_delegation_stateid(state, NULL, status);
2272
2273         if (status != NFS_OK) {
2274                 /* Free the stateid unless the server explicitly
2275                  * informs us the stateid is unrecognized. */
2276                 if (status != -NFS4ERR_BAD_STATEID)
2277                         nfs41_free_stateid(server, &stateid, cred);
2278                 nfs_finish_clear_delegation_stateid(state);
2279         }
2280
2281         put_rpccred(cred);
2282 }
2283
2284 /**
2285  * nfs41_check_open_stateid - possibly free an open stateid
2286  *
2287  * @state: NFSv4 state for an inode
2288  *
2289  * Returns NFS_OK if recovery for this stateid is now finished.
2290  * Otherwise a negative NFS4ERR value is returned.
2291  */
2292 static int nfs41_check_open_stateid(struct nfs4_state *state)
2293 {
2294         struct nfs_server *server = NFS_SERVER(state->inode);
2295         nfs4_stateid *stateid = &state->open_stateid;
2296         struct rpc_cred *cred = state->owner->so_cred;
2297         int status;
2298
2299         /* If a state reset has been done, test_stateid is unneeded */
2300         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2301             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2302             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2303                 return -NFS4ERR_BAD_STATEID;
2304
2305         status = nfs41_test_stateid(server, stateid, cred);
2306         trace_nfs4_test_open_stateid(state, NULL, status);
2307         if (status != NFS_OK) {
2308                 /* Free the stateid unless the server explicitly
2309                  * informs us the stateid is unrecognized. */
2310                 if (status != -NFS4ERR_BAD_STATEID)
2311                         nfs41_free_stateid(server, stateid, cred);
2312
2313                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2314                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2315                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2316                 clear_bit(NFS_OPEN_STATE, &state->flags);
2317         }
2318         return status;
2319 }
2320
2321 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2322 {
2323         int status;
2324
2325         nfs41_check_delegation_stateid(state);
2326         status = nfs41_check_open_stateid(state);
2327         if (status != NFS_OK)
2328                 status = nfs4_open_expired(sp, state);
2329         return status;
2330 }
2331 #endif
2332
2333 /*
2334  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2335  * fields corresponding to attributes that were used to store the verifier.
2336  * Make sure we clobber those fields in the later setattr call
2337  */
2338 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2339                                 struct iattr *sattr, struct nfs4_label **label)
2340 {
2341         const u32 *attrset = opendata->o_res.attrset;
2342
2343         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2344             !(sattr->ia_valid & ATTR_ATIME_SET))
2345                 sattr->ia_valid |= ATTR_ATIME;
2346
2347         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2348             !(sattr->ia_valid & ATTR_MTIME_SET))
2349                 sattr->ia_valid |= ATTR_MTIME;
2350
2351         /* Except MODE, it seems harmless of setting twice. */
2352         if ((attrset[1] & FATTR4_WORD1_MODE))
2353                 sattr->ia_valid &= ~ATTR_MODE;
2354
2355         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2356                 *label = NULL;
2357 }
2358
2359 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2360                 fmode_t fmode,
2361                 int flags,
2362                 struct nfs_open_context *ctx)
2363 {
2364         struct nfs4_state_owner *sp = opendata->owner;
2365         struct nfs_server *server = sp->so_server;
2366         struct dentry *dentry;
2367         struct nfs4_state *state;
2368         unsigned int seq;
2369         int ret;
2370
2371         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2372
2373         ret = _nfs4_proc_open(opendata);
2374         if (ret != 0)
2375                 goto out;
2376
2377         state = nfs4_opendata_to_nfs4_state(opendata);
2378         ret = PTR_ERR(state);
2379         if (IS_ERR(state))
2380                 goto out;
2381         if (server->caps & NFS_CAP_POSIX_LOCK)
2382                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2383
2384         dentry = opendata->dentry;
2385         if (d_really_is_negative(dentry)) {
2386                 /* FIXME: Is this d_drop() ever needed? */
2387                 d_drop(dentry);
2388                 dentry = d_add_unique(dentry, igrab(state->inode));
2389                 if (dentry == NULL) {
2390                         dentry = opendata->dentry;
2391                 } else if (dentry != ctx->dentry) {
2392                         dput(ctx->dentry);
2393                         ctx->dentry = dget(dentry);
2394                 }
2395                 nfs_set_verifier(dentry,
2396                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2397         }
2398
2399         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2400         if (ret != 0)
2401                 goto out;
2402
2403         ctx->state = state;
2404         if (d_inode(dentry) == state->inode) {
2405                 nfs_inode_attach_open_context(ctx);
2406                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2407                         nfs4_schedule_stateid_recovery(server, state);
2408         }
2409 out:
2410         return ret;
2411 }
2412
2413 /*
2414  * Returns a referenced nfs4_state
2415  */
2416 static int _nfs4_do_open(struct inode *dir,
2417                         struct nfs_open_context *ctx,
2418                         int flags,
2419                         struct iattr *sattr,
2420                         struct nfs4_label *label,
2421                         int *opened)
2422 {
2423         struct nfs4_state_owner  *sp;
2424         struct nfs4_state     *state = NULL;
2425         struct nfs_server       *server = NFS_SERVER(dir);
2426         struct nfs4_opendata *opendata;
2427         struct dentry *dentry = ctx->dentry;
2428         struct rpc_cred *cred = ctx->cred;
2429         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2430         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2431         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2432         struct nfs4_label *olabel = NULL;
2433         int status;
2434
2435         /* Protect against reboot recovery conflicts */
2436         status = -ENOMEM;
2437         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2438         if (sp == NULL) {
2439                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2440                 goto out_err;
2441         }
2442         status = nfs4_recover_expired_lease(server);
2443         if (status != 0)
2444                 goto err_put_state_owner;
2445         if (d_really_is_positive(dentry))
2446                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2447         status = -ENOMEM;
2448         if (d_really_is_positive(dentry))
2449                 claim = NFS4_OPEN_CLAIM_FH;
2450         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2451                         label, claim, GFP_KERNEL);
2452         if (opendata == NULL)
2453                 goto err_put_state_owner;
2454
2455         if (label) {
2456                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2457                 if (IS_ERR(olabel)) {
2458                         status = PTR_ERR(olabel);
2459                         goto err_opendata_put;
2460                 }
2461         }
2462
2463         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2464                 if (!opendata->f_attr.mdsthreshold) {
2465                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2466                         if (!opendata->f_attr.mdsthreshold)
2467                                 goto err_free_label;
2468                 }
2469                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2470         }
2471         if (d_really_is_positive(dentry))
2472                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2473
2474         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2475         if (status != 0)
2476                 goto err_free_label;
2477         state = ctx->state;
2478
2479         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2480             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2481                 nfs4_exclusive_attrset(opendata, sattr, &label);
2482
2483                 nfs_fattr_init(opendata->o_res.f_attr);
2484                 status = nfs4_do_setattr(state->inode, cred,
2485                                 opendata->o_res.f_attr, sattr,
2486                                 state, label, olabel);
2487                 if (status == 0) {
2488                         nfs_setattr_update_inode(state->inode, sattr,
2489                                         opendata->o_res.f_attr);
2490                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2491                 }
2492         }
2493         if (opened && opendata->file_created)
2494                 *opened |= FILE_CREATED;
2495
2496         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2497                 *ctx_th = opendata->f_attr.mdsthreshold;
2498                 opendata->f_attr.mdsthreshold = NULL;
2499         }
2500
2501         nfs4_label_free(olabel);
2502
2503         nfs4_opendata_put(opendata);
2504         nfs4_put_state_owner(sp);
2505         return 0;
2506 err_free_label:
2507         nfs4_label_free(olabel);
2508 err_opendata_put:
2509         nfs4_opendata_put(opendata);
2510 err_put_state_owner:
2511         nfs4_put_state_owner(sp);
2512 out_err:
2513         return status;
2514 }
2515
2516
2517 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2518                                         struct nfs_open_context *ctx,
2519                                         int flags,
2520                                         struct iattr *sattr,
2521                                         struct nfs4_label *label,
2522                                         int *opened)
2523 {
2524         struct nfs_server *server = NFS_SERVER(dir);
2525         struct nfs4_exception exception = { };
2526         struct nfs4_state *res;
2527         int status;
2528
2529         do {
2530                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2531                 res = ctx->state;
2532                 trace_nfs4_open_file(ctx, flags, status);
2533                 if (status == 0)
2534                         break;
2535                 /* NOTE: BAD_SEQID means the server and client disagree about the
2536                  * book-keeping w.r.t. state-changing operations
2537                  * (OPEN/CLOSE/LOCK/LOCKU...)
2538                  * It is actually a sign of a bug on the client or on the server.
2539                  *
2540                  * If we receive a BAD_SEQID error in the particular case of
2541                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2542                  * have unhashed the old state_owner for us, and that we can
2543                  * therefore safely retry using a new one. We should still warn
2544                  * the user though...
2545                  */
2546                 if (status == -NFS4ERR_BAD_SEQID) {
2547                         pr_warn_ratelimited("NFS: v4 server %s "
2548                                         " returned a bad sequence-id error!\n",
2549                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2550                         exception.retry = 1;
2551                         continue;
2552                 }
2553                 /*
2554                  * BAD_STATEID on OPEN means that the server cancelled our
2555                  * state before it received the OPEN_CONFIRM.
2556                  * Recover by retrying the request as per the discussion
2557                  * on Page 181 of RFC3530.
2558                  */
2559                 if (status == -NFS4ERR_BAD_STATEID) {
2560                         exception.retry = 1;
2561                         continue;
2562                 }
2563                 if (status == -EAGAIN) {
2564                         /* We must have found a delegation */
2565                         exception.retry = 1;
2566                         continue;
2567                 }
2568                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2569                         continue;
2570                 res = ERR_PTR(nfs4_handle_exception(server,
2571                                         status, &exception));
2572         } while (exception.retry);
2573         return res;
2574 }
2575
2576 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2577                             struct nfs_fattr *fattr, struct iattr *sattr,
2578                             struct nfs4_state *state, struct nfs4_label *ilabel,
2579                             struct nfs4_label *olabel)
2580 {
2581         struct nfs_server *server = NFS_SERVER(inode);
2582         struct nfs_setattrargs  arg = {
2583                 .fh             = NFS_FH(inode),
2584                 .iap            = sattr,
2585                 .server         = server,
2586                 .bitmask = server->attr_bitmask,
2587                 .label          = ilabel,
2588         };
2589         struct nfs_setattrres  res = {
2590                 .fattr          = fattr,
2591                 .label          = olabel,
2592                 .server         = server,
2593         };
2594         struct rpc_message msg = {
2595                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2596                 .rpc_argp       = &arg,
2597                 .rpc_resp       = &res,
2598                 .rpc_cred       = cred,
2599         };
2600         unsigned long timestamp = jiffies;
2601         fmode_t fmode;
2602         bool truncate;
2603         int status;
2604
2605         arg.bitmask = nfs4_bitmask(server, ilabel);
2606         if (ilabel)
2607                 arg.bitmask = nfs4_bitmask(server, olabel);
2608
2609         nfs_fattr_init(fattr);
2610
2611         /* Servers should only apply open mode checks for file size changes */
2612         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2613         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2614
2615         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2616                 /* Use that stateid */
2617         } else if (truncate && state != NULL) {
2618                 struct nfs_lockowner lockowner = {
2619                         .l_owner = current->files,
2620                         .l_pid = current->tgid,
2621                 };
2622                 if (!nfs4_valid_open_stateid(state))
2623                         return -EBADF;
2624                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2625                                 &lockowner) == -EIO)
2626                         return -EBADF;
2627         } else
2628                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2629
2630         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2631         if (status == 0 && state != NULL)
2632                 renew_lease(server, timestamp);
2633         return status;
2634 }
2635
2636 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2637                            struct nfs_fattr *fattr, struct iattr *sattr,
2638                            struct nfs4_state *state, struct nfs4_label *ilabel,
2639                            struct nfs4_label *olabel)
2640 {
2641         struct nfs_server *server = NFS_SERVER(inode);
2642         struct nfs4_exception exception = {
2643                 .state = state,
2644                 .inode = inode,
2645         };
2646         int err;
2647         do {
2648                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2649                 trace_nfs4_setattr(inode, err);
2650                 switch (err) {
2651                 case -NFS4ERR_OPENMODE:
2652                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2653                                 pr_warn_once("NFSv4: server %s is incorrectly "
2654                                                 "applying open mode checks to "
2655                                                 "a SETATTR that is not "
2656                                                 "changing file size.\n",
2657                                                 server->nfs_client->cl_hostname);
2658                         }
2659                         if (state && !(state->state & FMODE_WRITE)) {
2660                                 err = -EBADF;
2661                                 if (sattr->ia_valid & ATTR_OPEN)
2662                                         err = -EACCES;
2663                                 goto out;
2664                         }
2665                 }
2666                 err = nfs4_handle_exception(server, err, &exception);
2667         } while (exception.retry);
2668 out:
2669         return err;
2670 }
2671
2672 static bool
2673 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2674 {
2675         if (inode == NULL || !nfs_have_layout(inode))
2676                 return false;
2677
2678         return pnfs_wait_on_layoutreturn(inode, task);
2679 }
2680
2681 struct nfs4_closedata {
2682         struct inode *inode;
2683         struct nfs4_state *state;
2684         struct nfs_closeargs arg;
2685         struct nfs_closeres res;
2686         struct nfs_fattr fattr;
2687         unsigned long timestamp;
2688         bool roc;
2689         u32 roc_barrier;
2690 };
2691
2692 static void nfs4_free_closedata(void *data)
2693 {
2694         struct nfs4_closedata *calldata = data;
2695         struct nfs4_state_owner *sp = calldata->state->owner;
2696         struct super_block *sb = calldata->state->inode->i_sb;
2697
2698         if (calldata->roc)
2699                 pnfs_roc_release(calldata->state->inode);
2700         nfs4_put_open_state(calldata->state);
2701         nfs_free_seqid(calldata->arg.seqid);
2702         nfs4_put_state_owner(sp);
2703         nfs_sb_deactive(sb);
2704         kfree(calldata);
2705 }
2706
2707 static void nfs4_close_done(struct rpc_task *task, void *data)
2708 {
2709         struct nfs4_closedata *calldata = data;
2710         struct nfs4_state *state = calldata->state;
2711         struct nfs_server *server = NFS_SERVER(calldata->inode);
2712         nfs4_stateid *res_stateid = NULL;
2713
2714         dprintk("%s: begin!\n", __func__);
2715         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2716                 return;
2717         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2718         /* hmm. we are done with the inode, and in the process of freeing
2719          * the state_owner. we keep this around to process errors
2720          */
2721         switch (task->tk_status) {
2722                 case 0:
2723                         res_stateid = &calldata->res.stateid;
2724                         if (calldata->roc)
2725                                 pnfs_roc_set_barrier(state->inode,
2726                                                      calldata->roc_barrier);
2727                         renew_lease(server, calldata->timestamp);
2728                         break;
2729                 case -NFS4ERR_ADMIN_REVOKED:
2730                 case -NFS4ERR_STALE_STATEID:
2731                 case -NFS4ERR_OLD_STATEID:
2732                 case -NFS4ERR_BAD_STATEID:
2733                 case -NFS4ERR_EXPIRED:
2734                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2735                                                 &state->open_stateid)) {
2736                                 rpc_restart_call_prepare(task);
2737                                 goto out_release;
2738                         }
2739                         if (calldata->arg.fmode == 0)
2740                                 break;
2741                 default:
2742                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2743                                 rpc_restart_call_prepare(task);
2744                                 goto out_release;
2745                         }
2746         }
2747         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2748                         res_stateid, calldata->arg.fmode);
2749 out_release:
2750         nfs_release_seqid(calldata->arg.seqid);
2751         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2752         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2753 }
2754
2755 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2756 {
2757         struct nfs4_closedata *calldata = data;
2758         struct nfs4_state *state = calldata->state;
2759         struct inode *inode = calldata->inode;
2760         bool is_rdonly, is_wronly, is_rdwr;
2761         int call_close = 0;
2762
2763         dprintk("%s: begin!\n", __func__);
2764         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2765                 goto out_wait;
2766
2767         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2768         spin_lock(&state->owner->so_lock);
2769         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2770         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2771         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2772         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2773         /* Calculate the change in open mode */
2774         calldata->arg.fmode = 0;
2775         if (state->n_rdwr == 0) {
2776                 if (state->n_rdonly == 0)
2777                         call_close |= is_rdonly;
2778                 else if (is_rdonly)
2779                         calldata->arg.fmode |= FMODE_READ;
2780                 if (state->n_wronly == 0)
2781                         call_close |= is_wronly;
2782                 else if (is_wronly)
2783                         calldata->arg.fmode |= FMODE_WRITE;
2784         } else if (is_rdwr)
2785                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2786
2787         if (calldata->arg.fmode == 0)
2788                 call_close |= is_rdwr;
2789
2790         if (!nfs4_valid_open_stateid(state))
2791                 call_close = 0;
2792         spin_unlock(&state->owner->so_lock);
2793
2794         if (!call_close) {
2795                 /* Note: exit _without_ calling nfs4_close_done */
2796                 goto out_no_action;
2797         }
2798
2799         if (nfs4_wait_on_layoutreturn(inode, task)) {
2800                 nfs_release_seqid(calldata->arg.seqid);
2801                 goto out_wait;
2802         }
2803
2804         if (calldata->arg.fmode == 0)
2805                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2806         if (calldata->roc)
2807                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2808
2809         calldata->arg.share_access =
2810                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2811                                 calldata->arg.fmode, 0);
2812
2813         nfs_fattr_init(calldata->res.fattr);
2814         calldata->timestamp = jiffies;
2815         if (nfs4_setup_sequence(NFS_SERVER(inode),
2816                                 &calldata->arg.seq_args,
2817                                 &calldata->res.seq_res,
2818                                 task) != 0)
2819                 nfs_release_seqid(calldata->arg.seqid);
2820         dprintk("%s: done!\n", __func__);
2821         return;
2822 out_no_action:
2823         task->tk_action = NULL;
2824 out_wait:
2825         nfs4_sequence_done(task, &calldata->res.seq_res);
2826 }
2827
2828 static const struct rpc_call_ops nfs4_close_ops = {
2829         .rpc_call_prepare = nfs4_close_prepare,
2830         .rpc_call_done = nfs4_close_done,
2831         .rpc_release = nfs4_free_closedata,
2832 };
2833
2834 static bool nfs4_roc(struct inode *inode)
2835 {
2836         if (!nfs_have_layout(inode))
2837                 return false;
2838         return pnfs_roc(inode);
2839 }
2840
2841 /*
2842  * It is possible for data to be read/written from a mem-mapped file
2843  * after the sys_close call (which hits the vfs layer as a flush).
2844  * This means that we can't safely call nfsv4 close on a file until
2845  * the inode is cleared. This in turn means that we are not good
2846  * NFSv4 citizens - we do not indicate to the server to update the file's
2847  * share state even when we are done with one of the three share
2848  * stateid's in the inode.
2849  *
2850  * NOTE: Caller must be holding the sp->so_owner semaphore!
2851  */
2852 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2853 {
2854         struct nfs_server *server = NFS_SERVER(state->inode);
2855         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2856         struct nfs4_closedata *calldata;
2857         struct nfs4_state_owner *sp = state->owner;
2858         struct rpc_task *task;
2859         struct rpc_message msg = {
2860                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2861                 .rpc_cred = state->owner->so_cred,
2862         };
2863         struct rpc_task_setup task_setup_data = {
2864                 .rpc_client = server->client,
2865                 .rpc_message = &msg,
2866                 .callback_ops = &nfs4_close_ops,
2867                 .workqueue = nfsiod_workqueue,
2868                 .flags = RPC_TASK_ASYNC,
2869         };
2870         int status = -ENOMEM;
2871
2872         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2873                 &task_setup_data.rpc_client, &msg);
2874
2875         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2876         if (calldata == NULL)
2877                 goto out;
2878         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2879         calldata->inode = state->inode;
2880         calldata->state = state;
2881         calldata->arg.fh = NFS_FH(state->inode);
2882         /* Serialization for the sequence id */
2883         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2884         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2885         if (IS_ERR(calldata->arg.seqid))
2886                 goto out_free_calldata;
2887         calldata->arg.fmode = 0;
2888         calldata->arg.bitmask = server->cache_consistency_bitmask;
2889         calldata->res.fattr = &calldata->fattr;
2890         calldata->res.seqid = calldata->arg.seqid;
2891         calldata->res.server = server;
2892         calldata->roc = nfs4_roc(state->inode);
2893         nfs_sb_active(calldata->inode->i_sb);
2894
2895         msg.rpc_argp = &calldata->arg;
2896         msg.rpc_resp = &calldata->res;
2897         task_setup_data.callback_data = calldata;
2898         task = rpc_run_task(&task_setup_data);
2899         if (IS_ERR(task))
2900                 return PTR_ERR(task);
2901         status = 0;
2902         if (wait)
2903                 status = rpc_wait_for_completion_task(task);
2904         rpc_put_task(task);
2905         return status;
2906 out_free_calldata:
2907         kfree(calldata);
2908 out:
2909         nfs4_put_open_state(state);
2910         nfs4_put_state_owner(sp);
2911         return status;
2912 }
2913
2914 static struct inode *
2915 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2916                 int open_flags, struct iattr *attr, int *opened)
2917 {
2918         struct nfs4_state *state;
2919         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2920
2921         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
2922
2923         /* Protect against concurrent sillydeletes */
2924         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
2925
2926         nfs4_label_release_security(label);
2927
2928         if (IS_ERR(state))
2929                 return ERR_CAST(state);
2930         return state->inode;
2931 }
2932
2933 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
2934 {
2935         if (ctx->state == NULL)
2936                 return;
2937         if (is_sync)
2938                 nfs4_close_sync(ctx->state, ctx->mode);
2939         else
2940                 nfs4_close_state(ctx->state, ctx->mode);
2941 }
2942
2943 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
2944 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
2945 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
2946
2947 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
2948 {
2949         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
2950         struct nfs4_server_caps_arg args = {
2951                 .fhandle = fhandle,
2952                 .bitmask = bitmask,
2953         };
2954         struct nfs4_server_caps_res res = {};
2955         struct rpc_message msg = {
2956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
2957                 .rpc_argp = &args,
2958                 .rpc_resp = &res,
2959         };
2960         int status;
2961
2962         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
2963                      FATTR4_WORD0_FH_EXPIRE_TYPE |
2964                      FATTR4_WORD0_LINK_SUPPORT |
2965                      FATTR4_WORD0_SYMLINK_SUPPORT |
2966                      FATTR4_WORD0_ACLSUPPORT;
2967         if (minorversion)
2968                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
2969
2970         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
2971         if (status == 0) {
2972                 /* Sanity check the server answers */
2973                 switch (minorversion) {
2974                 case 0:
2975                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
2976                         res.attr_bitmask[2] = 0;
2977                         break;
2978                 case 1:
2979                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
2980                         break;
2981                 case 2:
2982                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
2983                 }
2984                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
2985                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
2986                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
2987                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
2988                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
2989                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
2990                                 NFS_CAP_SECURITY_LABEL);
2991                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
2992                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
2993                         server->caps |= NFS_CAP_ACLS;
2994                 if (res.has_links != 0)
2995                         server->caps |= NFS_CAP_HARDLINKS;
2996                 if (res.has_symlinks != 0)
2997                         server->caps |= NFS_CAP_SYMLINKS;
2998                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
2999                         server->caps |= NFS_CAP_FILEID;
3000                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3001                         server->caps |= NFS_CAP_MODE;
3002                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3003                         server->caps |= NFS_CAP_NLINK;
3004                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3005                         server->caps |= NFS_CAP_OWNER;
3006                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3007                         server->caps |= NFS_CAP_OWNER_GROUP;
3008                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3009                         server->caps |= NFS_CAP_ATIME;
3010                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3011                         server->caps |= NFS_CAP_CTIME;
3012                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3013                         server->caps |= NFS_CAP_MTIME;
3014 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3015                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3016                         server->caps |= NFS_CAP_SECURITY_LABEL;
3017 #endif
3018                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3019                                 sizeof(server->attr_bitmask));
3020                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3021
3022                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3023                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3024                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3025                 server->cache_consistency_bitmask[2] = 0;
3026                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3027                         sizeof(server->exclcreat_bitmask));
3028                 server->acl_bitmask = res.acl_bitmask;
3029                 server->fh_expire_type = res.fh_expire_type;
3030         }
3031
3032         return status;
3033 }
3034
3035 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3036 {
3037         struct nfs4_exception exception = { };
3038         int err;
3039         do {
3040                 err = nfs4_handle_exception(server,
3041                                 _nfs4_server_capabilities(server, fhandle),
3042                                 &exception);
3043         } while (exception.retry);
3044         return err;
3045 }
3046
3047 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3048                 struct nfs_fsinfo *info)
3049 {
3050         u32 bitmask[3];
3051         struct nfs4_lookup_root_arg args = {
3052                 .bitmask = bitmask,
3053         };
3054         struct nfs4_lookup_res res = {
3055                 .server = server,
3056                 .fattr = info->fattr,
3057                 .fh = fhandle,
3058         };
3059         struct rpc_message msg = {
3060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3061                 .rpc_argp = &args,
3062                 .rpc_resp = &res,
3063         };
3064
3065         bitmask[0] = nfs4_fattr_bitmap[0];
3066         bitmask[1] = nfs4_fattr_bitmap[1];
3067         /*
3068          * Process the label in the upcoming getfattr
3069          */
3070         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3071
3072         nfs_fattr_init(info->fattr);
3073         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3074 }
3075
3076 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3077                 struct nfs_fsinfo *info)
3078 {
3079         struct nfs4_exception exception = { };
3080         int err;
3081         do {
3082                 err = _nfs4_lookup_root(server, fhandle, info);
3083                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3084                 switch (err) {
3085                 case 0:
3086                 case -NFS4ERR_WRONGSEC:
3087                         goto out;
3088                 default:
3089                         err = nfs4_handle_exception(server, err, &exception);
3090                 }
3091         } while (exception.retry);
3092 out:
3093         return err;
3094 }
3095
3096 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3097                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3098 {
3099         struct rpc_auth_create_args auth_args = {
3100                 .pseudoflavor = flavor,
3101         };
3102         struct rpc_auth *auth;
3103         int ret;
3104
3105         auth = rpcauth_create(&auth_args, server->client);
3106         if (IS_ERR(auth)) {
3107                 ret = -EACCES;
3108                 goto out;
3109         }
3110         ret = nfs4_lookup_root(server, fhandle, info);
3111 out:
3112         return ret;
3113 }
3114
3115 /*
3116  * Retry pseudoroot lookup with various security flavors.  We do this when:
3117  *
3118  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3119  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3120  *
3121  * Returns zero on success, or a negative NFS4ERR value, or a
3122  * negative errno value.
3123  */
3124 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3125                               struct nfs_fsinfo *info)
3126 {
3127         /* Per 3530bis 15.33.5 */
3128         static const rpc_authflavor_t flav_array[] = {
3129                 RPC_AUTH_GSS_KRB5P,
3130                 RPC_AUTH_GSS_KRB5I,
3131                 RPC_AUTH_GSS_KRB5,
3132                 RPC_AUTH_UNIX,                  /* courtesy */
3133                 RPC_AUTH_NULL,
3134         };
3135         int status = -EPERM;
3136         size_t i;
3137
3138         if (server->auth_info.flavor_len > 0) {
3139                 /* try each flavor specified by user */
3140                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3141                         status = nfs4_lookup_root_sec(server, fhandle, info,
3142                                                 server->auth_info.flavors[i]);
3143                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3144                                 continue;
3145                         break;
3146                 }
3147         } else {
3148                 /* no flavors specified by user, try default list */
3149                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3150                         status = nfs4_lookup_root_sec(server, fhandle, info,
3151                                                       flav_array[i]);
3152                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3153                                 continue;
3154                         break;
3155                 }
3156         }
3157
3158         /*
3159          * -EACCESS could mean that the user doesn't have correct permissions
3160          * to access the mount.  It could also mean that we tried to mount
3161          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3162          * existing mount programs don't handle -EACCES very well so it should
3163          * be mapped to -EPERM instead.
3164          */
3165         if (status == -EACCES)
3166                 status = -EPERM;
3167         return status;
3168 }
3169
3170 static int nfs4_do_find_root_sec(struct nfs_server *server,
3171                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3172 {
3173         int mv = server->nfs_client->cl_minorversion;
3174         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3175 }
3176
3177 /**
3178  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3179  * @server: initialized nfs_server handle
3180  * @fhandle: we fill in the pseudo-fs root file handle
3181  * @info: we fill in an FSINFO struct
3182  * @auth_probe: probe the auth flavours
3183  *
3184  * Returns zero on success, or a negative errno.
3185  */
3186 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3187                          struct nfs_fsinfo *info,
3188                          bool auth_probe)
3189 {
3190         int status = 0;
3191
3192         if (!auth_probe)
3193                 status = nfs4_lookup_root(server, fhandle, info);
3194
3195         if (auth_probe || status == NFS4ERR_WRONGSEC)
3196                 status = nfs4_do_find_root_sec(server, fhandle, info);
3197
3198         if (status == 0)
3199                 status = nfs4_server_capabilities(server, fhandle);
3200         if (status == 0)
3201                 status = nfs4_do_fsinfo(server, fhandle, info);
3202
3203         return nfs4_map_errors(status);
3204 }
3205
3206 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3207                               struct nfs_fsinfo *info)
3208 {
3209         int error;
3210         struct nfs_fattr *fattr = info->fattr;
3211         struct nfs4_label *label = NULL;
3212
3213         error = nfs4_server_capabilities(server, mntfh);
3214         if (error < 0) {
3215                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3216                 return error;
3217         }
3218
3219         label = nfs4_label_alloc(server, GFP_KERNEL);
3220         if (IS_ERR(label))
3221                 return PTR_ERR(label);
3222
3223         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3224         if (error < 0) {
3225                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3226                 goto err_free_label;
3227         }
3228
3229         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3230             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3231                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3232
3233 err_free_label:
3234         nfs4_label_free(label);
3235
3236         return error;
3237 }
3238
3239 /*
3240  * Get locations and (maybe) other attributes of a referral.
3241  * Note that we'll actually follow the referral later when
3242  * we detect fsid mismatch in inode revalidation
3243  */
3244 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3245                              const struct qstr *name, struct nfs_fattr *fattr,
3246                              struct nfs_fh *fhandle)
3247 {
3248         int status = -ENOMEM;
3249         struct page *page = NULL;
3250         struct nfs4_fs_locations *locations = NULL;
3251
3252         page = alloc_page(GFP_KERNEL);
3253         if (page == NULL)
3254                 goto out;
3255         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3256         if (locations == NULL)
3257                 goto out;
3258
3259         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3260         if (status != 0)
3261                 goto out;
3262
3263         /*
3264          * If the fsid didn't change, this is a migration event, not a
3265          * referral.  Cause us to drop into the exception handler, which
3266          * will kick off migration recovery.
3267          */
3268         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3269                 dprintk("%s: server did not return a different fsid for"
3270                         " a referral at %s\n", __func__, name->name);
3271                 status = -NFS4ERR_MOVED;
3272                 goto out;
3273         }
3274         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3275         nfs_fixup_referral_attributes(&locations->fattr);
3276
3277         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3278         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3279         memset(fhandle, 0, sizeof(struct nfs_fh));
3280 out:
3281         if (page)
3282                 __free_page(page);
3283         kfree(locations);
3284         return status;
3285 }
3286
3287 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3288                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3289 {
3290         struct nfs4_getattr_arg args = {
3291                 .fh = fhandle,
3292                 .bitmask = server->attr_bitmask,
3293         };
3294         struct nfs4_getattr_res res = {
3295                 .fattr = fattr,
3296                 .label = label,
3297                 .server = server,
3298         };
3299         struct rpc_message msg = {
3300                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3301                 .rpc_argp = &args,
3302                 .rpc_resp = &res,
3303         };
3304
3305         args.bitmask = nfs4_bitmask(server, label);
3306
3307         nfs_fattr_init(fattr);
3308         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3309 }
3310
3311 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3312                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3313 {
3314         struct nfs4_exception exception = { };
3315         int err;
3316         do {
3317                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3318                 trace_nfs4_getattr(server, fhandle, fattr, err);
3319                 err = nfs4_handle_exception(server, err,
3320                                 &exception);
3321         } while (exception.retry);
3322         return err;
3323 }
3324
3325 /*
3326  * The file is not closed if it is opened due to the a request to change
3327  * the size of the file. The open call will not be needed once the
3328  * VFS layer lookup-intents are implemented.
3329  *
3330  * Close is called when the inode is destroyed.
3331  * If we haven't opened the file for O_WRONLY, we
3332  * need to in the size_change case to obtain a stateid.
3333  *
3334  * Got race?
3335  * Because OPEN is always done by name in nfsv4, it is
3336  * possible that we opened a different file by the same
3337  * name.  We can recognize this race condition, but we
3338  * can't do anything about it besides returning an error.
3339  *
3340  * This will be fixed with VFS changes (lookup-intent).
3341  */
3342 static int
3343 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3344                   struct iattr *sattr)
3345 {
3346         struct inode *inode = d_inode(dentry);
3347         struct rpc_cred *cred = NULL;
3348         struct nfs4_state *state = NULL;
3349         struct nfs4_label *label = NULL;
3350         int status;
3351
3352         if (pnfs_ld_layoutret_on_setattr(inode) &&
3353             sattr->ia_valid & ATTR_SIZE &&
3354             sattr->ia_size < i_size_read(inode))
3355                 pnfs_commit_and_return_layout(inode);
3356
3357         nfs_fattr_init(fattr);
3358
3359         /* Deal with open(O_TRUNC) */
3360         if (sattr->ia_valid & ATTR_OPEN)
3361                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3362
3363         /* Optimization: if the end result is no change, don't RPC */
3364         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3365                 return 0;
3366
3367         /* Search for an existing open(O_WRITE) file */
3368         if (sattr->ia_valid & ATTR_FILE) {
3369                 struct nfs_open_context *ctx;
3370
3371                 ctx = nfs_file_open_context(sattr->ia_file);
3372                 if (ctx) {
3373                         cred = ctx->cred;
3374                         state = ctx->state;
3375                 }
3376         }
3377
3378         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3379         if (IS_ERR(label))
3380                 return PTR_ERR(label);
3381
3382         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3383         if (status == 0) {
3384                 nfs_setattr_update_inode(inode, sattr, fattr);
3385                 nfs_setsecurity(inode, fattr, label);
3386         }
3387         nfs4_label_free(label);
3388         return status;
3389 }
3390
3391 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3392                 const struct qstr *name, struct nfs_fh *fhandle,
3393                 struct nfs_fattr *fattr, struct nfs4_label *label)
3394 {
3395         struct nfs_server *server = NFS_SERVER(dir);
3396         int                    status;
3397         struct nfs4_lookup_arg args = {
3398                 .bitmask = server->attr_bitmask,
3399                 .dir_fh = NFS_FH(dir),
3400                 .name = name,
3401         };
3402         struct nfs4_lookup_res res = {
3403                 .server = server,
3404                 .fattr = fattr,
3405                 .label = label,
3406                 .fh = fhandle,
3407         };
3408         struct rpc_message msg = {
3409                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3410                 .rpc_argp = &args,
3411                 .rpc_resp = &res,
3412         };
3413
3414         args.bitmask = nfs4_bitmask(server, label);
3415
3416         nfs_fattr_init(fattr);
3417
3418         dprintk("NFS call  lookup %s\n", name->name);
3419         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3420         dprintk("NFS reply lookup: %d\n", status);
3421         return status;
3422 }
3423
3424 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3425 {
3426         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3427                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3428         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3429         fattr->nlink = 2;
3430 }
3431
3432 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3433                                    struct qstr *name, struct nfs_fh *fhandle,
3434                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3435 {
3436         struct nfs4_exception exception = { };
3437         struct rpc_clnt *client = *clnt;
3438         int err;
3439         do {
3440                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3441                 trace_nfs4_lookup(dir, name, err);
3442                 switch (err) {
3443                 case -NFS4ERR_BADNAME:
3444                         err = -ENOENT;
3445                         goto out;
3446                 case -NFS4ERR_MOVED:
3447                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3448                         if (err == -NFS4ERR_MOVED)
3449                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3450                         goto out;
3451                 case -NFS4ERR_WRONGSEC:
3452                         err = -EPERM;
3453                         if (client != *clnt)
3454                                 goto out;
3455                         client = nfs4_negotiate_security(client, dir, name);
3456                         if (IS_ERR(client))
3457                                 return PTR_ERR(client);
3458
3459                         exception.retry = 1;
3460                         break;
3461                 default:
3462                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3463                 }
3464         } while (exception.retry);
3465
3466 out:
3467         if (err == 0)
3468                 *clnt = client;
3469         else if (client != *clnt)
3470                 rpc_shutdown_client(client);
3471
3472         return err;
3473 }
3474
3475 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3476                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3477                             struct nfs4_label *label)
3478 {
3479         int status;
3480         struct rpc_clnt *client = NFS_CLIENT(dir);
3481
3482         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3483         if (client != NFS_CLIENT(dir)) {
3484                 rpc_shutdown_client(client);
3485                 nfs_fixup_secinfo_attributes(fattr);
3486         }
3487         return status;
3488 }
3489
3490 struct rpc_clnt *
3491 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3492                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3493 {
3494         struct rpc_clnt *client = NFS_CLIENT(dir);
3495         int status;
3496
3497         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3498         if (status < 0)
3499                 return ERR_PTR(status);
3500         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3501 }
3502
3503 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3504 {
3505         struct nfs_server *server = NFS_SERVER(inode);
3506         struct nfs4_accessargs args = {
3507                 .fh = NFS_FH(inode),
3508                 .bitmask = server->cache_consistency_bitmask,
3509         };
3510         struct nfs4_accessres res = {
3511                 .server = server,
3512         };
3513         struct rpc_message msg = {
3514                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3515                 .rpc_argp = &args,
3516                 .rpc_resp = &res,
3517                 .rpc_cred = entry->cred,
3518         };
3519         int mode = entry->mask;
3520         int status = 0;
3521
3522         /*
3523          * Determine which access bits we want to ask for...
3524          */
3525         if (mode & MAY_READ)
3526                 args.access |= NFS4_ACCESS_READ;
3527         if (S_ISDIR(inode->i_mode)) {
3528                 if (mode & MAY_WRITE)
3529                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3530                 if (mode & MAY_EXEC)
3531                         args.access |= NFS4_ACCESS_LOOKUP;
3532         } else {
3533                 if (mode & MAY_WRITE)
3534                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3535                 if (mode & MAY_EXEC)
3536                         args.access |= NFS4_ACCESS_EXECUTE;
3537         }
3538
3539         res.fattr = nfs_alloc_fattr();
3540         if (res.fattr == NULL)
3541                 return -ENOMEM;
3542
3543         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3544         if (!status) {
3545                 nfs_access_set_mask(entry, res.access);
3546                 nfs_refresh_inode(inode, res.fattr);
3547         }
3548         nfs_free_fattr(res.fattr);
3549         return status;
3550 }
3551
3552 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3553 {
3554         struct nfs4_exception exception = { };
3555         int err;
3556         do {
3557                 err = _nfs4_proc_access(inode, entry);
3558                 trace_nfs4_access(inode, err);
3559                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3560                                 &exception);
3561         } while (exception.retry);
3562         return err;
3563 }
3564
3565 /*
3566  * TODO: For the time being, we don't try to get any attributes
3567  * along with any of the zero-copy operations READ, READDIR,
3568  * READLINK, WRITE.
3569  *
3570  * In the case of the first three, we want to put the GETATTR
3571  * after the read-type operation -- this is because it is hard
3572  * to predict the length of a GETATTR response in v4, and thus
3573  * align the READ data correctly.  This means that the GETATTR
3574  * may end up partially falling into the page cache, and we should
3575  * shift it into the 'tail' of the xdr_buf before processing.
3576  * To do this efficiently, we need to know the total length
3577  * of data received, which doesn't seem to be available outside
3578  * of the RPC layer.
3579  *
3580  * In the case of WRITE, we also want to put the GETATTR after
3581  * the operation -- in this case because we want to make sure
3582  * we get the post-operation mtime and size.
3583  *
3584  * Both of these changes to the XDR layer would in fact be quite
3585  * minor, but I decided to leave them for a subsequent patch.
3586  */
3587 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3588                 unsigned int pgbase, unsigned int pglen)
3589 {
3590         struct nfs4_readlink args = {
3591                 .fh       = NFS_FH(inode),
3592                 .pgbase   = pgbase,
3593                 .pglen    = pglen,
3594                 .pages    = &page,
3595         };
3596         struct nfs4_readlink_res res;
3597         struct rpc_message msg = {
3598                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3599                 .rpc_argp = &args,
3600                 .rpc_resp = &res,
3601         };
3602
3603         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3604 }
3605
3606 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3607                 unsigned int pgbase, unsigned int pglen)
3608 {
3609         struct nfs4_exception exception = { };
3610         int err;
3611         do {
3612                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3613                 trace_nfs4_readlink(inode, err);
3614                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3615                                 &exception);
3616         } while (exception.retry);
3617         return err;
3618 }
3619
3620 /*
3621  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3622  */
3623 static int
3624 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3625                  int flags)
3626 {
3627         struct nfs4_label l, *ilabel = NULL;
3628         struct nfs_open_context *ctx;
3629         struct nfs4_state *state;
3630         int status = 0;
3631
3632         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3633         if (IS_ERR(ctx))
3634                 return PTR_ERR(ctx);
3635
3636         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3637
3638         sattr->ia_mode &= ~current_umask();
3639         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3640         if (IS_ERR(state)) {
3641                 status = PTR_ERR(state);
3642                 goto out;
3643         }
3644 out:
3645         nfs4_label_release_security(ilabel);
3646         put_nfs_open_context(ctx);
3647         return status;
3648 }
3649
3650 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3651 {
3652         struct nfs_server *server = NFS_SERVER(dir);
3653         struct nfs_removeargs args = {
3654                 .fh = NFS_FH(dir),
3655                 .name = *name,
3656         };
3657         struct nfs_removeres res = {
3658                 .server = server,
3659         };
3660         struct rpc_message msg = {
3661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3662                 .rpc_argp = &args,
3663                 .rpc_resp = &res,
3664         };
3665         int status;
3666
3667         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3668         if (status == 0)
3669                 update_changeattr(dir, &res.cinfo);
3670         return status;
3671 }
3672
3673 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3674 {
3675         struct nfs4_exception exception = { };
3676         int err;
3677         do {
3678                 err = _nfs4_proc_remove(dir, name);
3679                 trace_nfs4_remove(dir, name, err);
3680                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3681                                 &exception);
3682         } while (exception.retry);
3683         return err;
3684 }
3685
3686 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3687 {
3688         struct nfs_server *server = NFS_SERVER(dir);
3689         struct nfs_removeargs *args = msg->rpc_argp;
3690         struct nfs_removeres *res = msg->rpc_resp;
3691
3692         res->server = server;
3693         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3694         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3695
3696         nfs_fattr_init(res->dir_attr);
3697 }
3698
3699 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3700 {
3701         nfs4_setup_sequence(NFS_SERVER(data->dir),
3702                         &data->args.seq_args,
3703                         &data->res.seq_res,
3704                         task);
3705 }
3706
3707 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3708 {
3709         struct nfs_unlinkdata *data = task->tk_calldata;
3710         struct nfs_removeres *res = &data->res;
3711
3712         if (!nfs4_sequence_done(task, &res->seq_res))
3713                 return 0;
3714         if (nfs4_async_handle_error(task, res->server, NULL,
3715                                     &data->timeout) == -EAGAIN)
3716                 return 0;
3717         update_changeattr(dir, &res->cinfo);
3718         return 1;
3719 }
3720
3721 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3722 {
3723         struct nfs_server *server = NFS_SERVER(dir);
3724         struct nfs_renameargs *arg = msg->rpc_argp;
3725         struct nfs_renameres *res = msg->rpc_resp;
3726
3727         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3728         res->server = server;
3729         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3730 }
3731
3732 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3733 {
3734         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3735                         &data->args.seq_args,
3736                         &data->res.seq_res,
3737                         task);
3738 }
3739
3740 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3741                                  struct inode *new_dir)
3742 {
3743         struct nfs_renamedata *data = task->tk_calldata;
3744         struct nfs_renameres *res = &data->res;
3745
3746         if (!nfs4_sequence_done(task, &res->seq_res))
3747                 return 0;
3748         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3749                 return 0;
3750
3751         update_changeattr(old_dir, &res->old_cinfo);
3752         update_changeattr(new_dir, &res->new_cinfo);
3753         return 1;
3754 }
3755
3756 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3757 {
3758         struct nfs_server *server = NFS_SERVER(inode);
3759         struct nfs4_link_arg arg = {
3760                 .fh     = NFS_FH(inode),
3761                 .dir_fh = NFS_FH(dir),
3762                 .name   = name,
3763                 .bitmask = server->attr_bitmask,
3764         };
3765         struct nfs4_link_res res = {
3766                 .server = server,
3767                 .label = NULL,
3768         };
3769         struct rpc_message msg = {
3770                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3771                 .rpc_argp = &arg,
3772                 .rpc_resp = &res,
3773         };
3774         int status = -ENOMEM;
3775
3776         res.fattr = nfs_alloc_fattr();
3777         if (res.fattr == NULL)
3778                 goto out;
3779
3780         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3781         if (IS_ERR(res.label)) {
3782                 status = PTR_ERR(res.label);
3783                 goto out;
3784         }
3785         arg.bitmask = nfs4_bitmask(server, res.label);
3786
3787         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3788         if (!status) {
3789                 update_changeattr(dir, &res.cinfo);
3790                 status = nfs_post_op_update_inode(inode, res.fattr);
3791                 if (!status)
3792                         nfs_setsecurity(inode, res.fattr, res.label);
3793         }
3794
3795
3796         nfs4_label_free(res.label);
3797
3798 out:
3799         nfs_free_fattr(res.fattr);
3800         return status;
3801 }
3802
3803 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3804 {
3805         struct nfs4_exception exception = { };
3806         int err;
3807         do {
3808                 err = nfs4_handle_exception(NFS_SERVER(inode),
3809                                 _nfs4_proc_link(inode, dir, name),
3810                                 &exception);
3811         } while (exception.retry);
3812         return err;
3813 }
3814
3815 struct nfs4_createdata {
3816         struct rpc_message msg;
3817         struct nfs4_create_arg arg;
3818         struct nfs4_create_res res;
3819         struct nfs_fh fh;
3820         struct nfs_fattr fattr;
3821         struct nfs4_label *label;
3822 };
3823
3824 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3825                 struct qstr *name, struct iattr *sattr, u32 ftype)
3826 {
3827         struct nfs4_createdata *data;
3828
3829         data = kzalloc(sizeof(*data), GFP_KERNEL);
3830         if (data != NULL) {
3831                 struct nfs_server *server = NFS_SERVER(dir);
3832
3833                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3834                 if (IS_ERR(data->label))
3835                         goto out_free;
3836
3837                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3838                 data->msg.rpc_argp = &data->arg;
3839                 data->msg.rpc_resp = &data->res;
3840                 data->arg.dir_fh = NFS_FH(dir);
3841                 data->arg.server = server;
3842                 data->arg.name = name;
3843                 data->arg.attrs = sattr;
3844                 data->arg.ftype = ftype;
3845                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3846                 data->res.server = server;
3847                 data->res.fh = &data->fh;
3848                 data->res.fattr = &data->fattr;
3849                 data->res.label = data->label;
3850                 nfs_fattr_init(data->res.fattr);
3851         }
3852         return data;
3853 out_free:
3854         kfree(data);
3855         return NULL;
3856 }
3857
3858 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3859 {
3860         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3861                                     &data->arg.seq_args, &data->res.seq_res, 1);
3862         if (status == 0) {
3863                 update_changeattr(dir, &data->res.dir_cinfo);
3864                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3865         }
3866         return status;
3867 }
3868
3869 static void nfs4_free_createdata(struct nfs4_createdata *data)
3870 {
3871         nfs4_label_free(data->label);
3872         kfree(data);
3873 }
3874
3875 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3876                 struct page *page, unsigned int len, struct iattr *sattr,
3877                 struct nfs4_label *label)
3878 {
3879         struct nfs4_createdata *data;
3880         int status = -ENAMETOOLONG;
3881
3882         if (len > NFS4_MAXPATHLEN)
3883                 goto out;
3884
3885         status = -ENOMEM;
3886         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3887         if (data == NULL)
3888                 goto out;
3889
3890         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3891         data->arg.u.symlink.pages = &page;
3892         data->arg.u.symlink.len = len;
3893         data->arg.label = label;
3894
3895         status = nfs4_do_create(dir, dentry, data);
3896
3897         nfs4_free_createdata(data);
3898 out:
3899         return status;
3900 }
3901
3902 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3903                 struct page *page, unsigned int len, struct iattr *sattr)
3904 {
3905         struct nfs4_exception exception = { };
3906         struct nfs4_label l, *label = NULL;
3907         int err;
3908
3909         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3910
3911         do {
3912                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3913                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3914                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3915                                 &exception);
3916         } while (exception.retry);
3917
3918         nfs4_label_release_security(label);
3919         return err;
3920 }
3921
3922 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3923                 struct iattr *sattr, struct nfs4_label *label)
3924 {
3925         struct nfs4_createdata *data;
3926         int status = -ENOMEM;
3927
3928         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
3929         if (data == NULL)
3930                 goto out;
3931
3932         data->arg.label = label;
3933         status = nfs4_do_create(dir, dentry, data);
3934
3935         nfs4_free_createdata(data);
3936 out:
3937         return status;
3938 }
3939
3940 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
3941                 struct iattr *sattr)
3942 {
3943         struct nfs4_exception exception = { };
3944         struct nfs4_label l, *label = NULL;
3945         int err;
3946
3947         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3948
3949         sattr->ia_mode &= ~current_umask();
3950         do {
3951                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
3952                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
3953                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3954                                 &exception);
3955         } while (exception.retry);
3956         nfs4_label_release_security(label);
3957
3958         return err;
3959 }
3960
3961 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
3962                 u64 cookie, struct page **pages, unsigned int count, int plus)
3963 {
3964         struct inode            *dir = d_inode(dentry);
3965         struct nfs4_readdir_arg args = {
3966                 .fh = NFS_FH(dir),
3967                 .pages = pages,
3968                 .pgbase = 0,
3969                 .count = count,
3970                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
3971                 .plus = plus,
3972         };
3973         struct nfs4_readdir_res res;
3974         struct rpc_message msg = {
3975                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
3976                 .rpc_argp = &args,
3977                 .rpc_resp = &res,
3978                 .rpc_cred = cred,
3979         };
3980         int                     status;
3981
3982         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
3983                         dentry,
3984                         (unsigned long long)cookie);
3985         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
3986         res.pgbase = args.pgbase;
3987         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
3988         if (status >= 0) {
3989                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
3990                 status += args.pgbase;
3991         }
3992
3993         nfs_invalidate_atime(dir);
3994
3995         dprintk("%s: returns %d\n", __func__, status);
3996         return status;
3997 }
3998
3999 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4000                 u64 cookie, struct page **pages, unsigned int count, int plus)
4001 {
4002         struct nfs4_exception exception = { };
4003         int err;
4004         do {
4005                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4006                                 pages, count, plus);
4007                 trace_nfs4_readdir(d_inode(dentry), err);
4008                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4009                                 &exception);
4010         } while (exception.retry);
4011         return err;
4012 }
4013
4014 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4015                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4016 {
4017         struct nfs4_createdata *data;
4018         int mode = sattr->ia_mode;
4019         int status = -ENOMEM;
4020
4021         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4022         if (data == NULL)
4023                 goto out;
4024
4025         if (S_ISFIFO(mode))
4026                 data->arg.ftype = NF4FIFO;
4027         else if (S_ISBLK(mode)) {
4028                 data->arg.ftype = NF4BLK;
4029                 data->arg.u.device.specdata1 = MAJOR(rdev);
4030                 data->arg.u.device.specdata2 = MINOR(rdev);
4031         }
4032         else if (S_ISCHR(mode)) {
4033                 data->arg.ftype = NF4CHR;
4034                 data->arg.u.device.specdata1 = MAJOR(rdev);
4035                 data->arg.u.device.specdata2 = MINOR(rdev);
4036         } else if (!S_ISSOCK(mode)) {
4037                 status = -EINVAL;
4038                 goto out_free;
4039         }
4040
4041         data->arg.label = label;
4042         status = nfs4_do_create(dir, dentry, data);
4043 out_free:
4044         nfs4_free_createdata(data);
4045 out:
4046         return status;
4047 }
4048
4049 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4050                 struct iattr *sattr, dev_t rdev)
4051 {
4052         struct nfs4_exception exception = { };
4053         struct nfs4_label l, *label = NULL;
4054         int err;
4055
4056         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4057
4058         sattr->ia_mode &= ~current_umask();
4059         do {
4060                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4061                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4062                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4063                                 &exception);
4064         } while (exception.retry);
4065
4066         nfs4_label_release_security(label);
4067
4068         return err;
4069 }
4070
4071 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4072                  struct nfs_fsstat *fsstat)
4073 {
4074         struct nfs4_statfs_arg args = {
4075                 .fh = fhandle,
4076                 .bitmask = server->attr_bitmask,
4077         };
4078         struct nfs4_statfs_res res = {
4079                 .fsstat = fsstat,
4080         };
4081         struct rpc_message msg = {
4082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4083                 .rpc_argp = &args,
4084                 .rpc_resp = &res,
4085         };
4086
4087         nfs_fattr_init(fsstat->fattr);
4088         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4089 }
4090
4091 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4092 {
4093         struct nfs4_exception exception = { };
4094         int err;
4095         do {
4096                 err = nfs4_handle_exception(server,
4097                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4098                                 &exception);
4099         } while (exception.retry);
4100         return err;
4101 }
4102
4103 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4104                 struct nfs_fsinfo *fsinfo)
4105 {
4106         struct nfs4_fsinfo_arg args = {
4107                 .fh = fhandle,
4108                 .bitmask = server->attr_bitmask,
4109         };
4110         struct nfs4_fsinfo_res res = {
4111                 .fsinfo = fsinfo,
4112         };
4113         struct rpc_message msg = {
4114                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4115                 .rpc_argp = &args,
4116                 .rpc_resp = &res,
4117         };
4118
4119         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4120 }
4121
4122 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4123 {
4124         struct nfs4_exception exception = { };
4125         unsigned long now = jiffies;
4126         int err;
4127
4128         do {
4129                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4130                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4131                 if (err == 0) {
4132                         struct nfs_client *clp = server->nfs_client;
4133
4134                         spin_lock(&clp->cl_lock);
4135                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4136                         clp->cl_last_renewal = now;
4137                         spin_unlock(&clp->cl_lock);
4138                         break;
4139                 }
4140                 err = nfs4_handle_exception(server, err, &exception);
4141         } while (exception.retry);
4142         return err;
4143 }
4144
4145 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4146 {
4147         int error;
4148
4149         nfs_fattr_init(fsinfo->fattr);
4150         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4151         if (error == 0) {
4152                 /* block layout checks this! */
4153                 server->pnfs_blksize = fsinfo->blksize;
4154                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4155         }
4156
4157         return error;
4158 }
4159
4160 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4161                 struct nfs_pathconf *pathconf)
4162 {
4163         struct nfs4_pathconf_arg args = {
4164                 .fh = fhandle,
4165                 .bitmask = server->attr_bitmask,
4166         };
4167         struct nfs4_pathconf_res res = {
4168                 .pathconf = pathconf,
4169         };
4170         struct rpc_message msg = {
4171                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4172                 .rpc_argp = &args,
4173                 .rpc_resp = &res,
4174         };
4175
4176         /* None of the pathconf attributes are mandatory to implement */
4177         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4178                 memset(pathconf, 0, sizeof(*pathconf));
4179                 return 0;
4180         }
4181
4182         nfs_fattr_init(pathconf->fattr);
4183         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4184 }
4185
4186 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4187                 struct nfs_pathconf *pathconf)
4188 {
4189         struct nfs4_exception exception = { };
4190         int err;
4191
4192         do {
4193                 err = nfs4_handle_exception(server,
4194                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4195                                 &exception);
4196         } while (exception.retry);
4197         return err;
4198 }
4199
4200 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4201                 const struct nfs_open_context *ctx,
4202                 const struct nfs_lock_context *l_ctx,
4203                 fmode_t fmode)
4204 {
4205         const struct nfs_lockowner *lockowner = NULL;
4206
4207         if (l_ctx != NULL)
4208                 lockowner = &l_ctx->lockowner;
4209         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4210 }
4211 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4212
4213 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4214                 const struct nfs_open_context *ctx,
4215                 const struct nfs_lock_context *l_ctx,
4216                 fmode_t fmode)
4217 {
4218         nfs4_stateid current_stateid;
4219
4220         /* If the current stateid represents a lost lock, then exit */
4221         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4222                 return true;
4223         return nfs4_stateid_match(stateid, &current_stateid);
4224 }
4225
4226 static bool nfs4_error_stateid_expired(int err)
4227 {
4228         switch (err) {
4229         case -NFS4ERR_DELEG_REVOKED:
4230         case -NFS4ERR_ADMIN_REVOKED:
4231         case -NFS4ERR_BAD_STATEID:
4232         case -NFS4ERR_STALE_STATEID:
4233         case -NFS4ERR_OLD_STATEID:
4234         case -NFS4ERR_OPENMODE:
4235         case -NFS4ERR_EXPIRED:
4236                 return true;
4237         }
4238         return false;
4239 }
4240
4241 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4242 {
4243         nfs_invalidate_atime(hdr->inode);
4244 }
4245
4246 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4247 {
4248         struct nfs_server *server = NFS_SERVER(hdr->inode);
4249
4250         trace_nfs4_read(hdr, task->tk_status);
4251         if (nfs4_async_handle_error(task, server,
4252                                     hdr->args.context->state,
4253                                     NULL) == -EAGAIN) {
4254                 rpc_restart_call_prepare(task);
4255                 return -EAGAIN;
4256         }
4257
4258         __nfs4_read_done_cb(hdr);
4259         if (task->tk_status > 0)
4260                 renew_lease(server, hdr->timestamp);
4261         return 0;
4262 }
4263
4264 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4265                 struct nfs_pgio_args *args)
4266 {
4267
4268         if (!nfs4_error_stateid_expired(task->tk_status) ||
4269                 nfs4_stateid_is_current(&args->stateid,
4270                                 args->context,
4271                                 args->lock_context,
4272                                 FMODE_READ))
4273                 return false;
4274         rpc_restart_call_prepare(task);
4275         return true;
4276 }
4277
4278 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4279 {
4280
4281         dprintk("--> %s\n", __func__);
4282
4283         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4284                 return -EAGAIN;
4285         if (nfs4_read_stateid_changed(task, &hdr->args))
4286                 return -EAGAIN;
4287         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4288                                     nfs4_read_done_cb(task, hdr);
4289 }
4290
4291 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4292                                  struct rpc_message *msg)
4293 {
4294         hdr->timestamp   = jiffies;
4295         hdr->pgio_done_cb = nfs4_read_done_cb;
4296         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4297         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4298 }
4299
4300 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4301                                       struct nfs_pgio_header *hdr)
4302 {
4303         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4304                         &hdr->args.seq_args,
4305                         &hdr->res.seq_res,
4306                         task))
4307                 return 0;
4308         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4309                                 hdr->args.lock_context,
4310                                 hdr->rw_ops->rw_mode) == -EIO)
4311                 return -EIO;
4312         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4313                 return -EIO;
4314         return 0;
4315 }
4316
4317 static int nfs4_write_done_cb(struct rpc_task *task,
4318                               struct nfs_pgio_header *hdr)
4319 {
4320         struct inode *inode = hdr->inode;
4321
4322         trace_nfs4_write(hdr, task->tk_status);
4323         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4324                                     hdr->args.context->state,
4325                                     NULL) == -EAGAIN) {
4326                 rpc_restart_call_prepare(task);
4327                 return -EAGAIN;
4328         }
4329         if (task->tk_status >= 0) {
4330                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4331                 nfs_writeback_update_inode(hdr);
4332         }
4333         return 0;
4334 }
4335
4336 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4337                 struct nfs_pgio_args *args)
4338 {
4339
4340         if (!nfs4_error_stateid_expired(task->tk_status) ||
4341                 nfs4_stateid_is_current(&args->stateid,
4342                                 args->context,
4343                                 args->lock_context,
4344                                 FMODE_WRITE))
4345                 return false;
4346         rpc_restart_call_prepare(task);
4347         return true;
4348 }
4349
4350 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4351 {
4352         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4353                 return -EAGAIN;
4354         if (nfs4_write_stateid_changed(task, &hdr->args))
4355                 return -EAGAIN;
4356         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4357                 nfs4_write_done_cb(task, hdr);
4358 }
4359
4360 static
4361 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4362 {
4363         /* Don't request attributes for pNFS or O_DIRECT writes */
4364         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4365                 return false;
4366         /* Otherwise, request attributes if and only if we don't hold
4367          * a delegation
4368          */
4369         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4370 }
4371
4372 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4373                                   struct rpc_message *msg)
4374 {
4375         struct nfs_server *server = NFS_SERVER(hdr->inode);
4376
4377         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4378                 hdr->args.bitmask = NULL;
4379                 hdr->res.fattr = NULL;
4380         } else
4381                 hdr->args.bitmask = server->cache_consistency_bitmask;
4382
4383         if (!hdr->pgio_done_cb)
4384                 hdr->pgio_done_cb = nfs4_write_done_cb;
4385         hdr->res.server = server;
4386         hdr->timestamp   = jiffies;
4387
4388         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4389         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4390 }
4391
4392 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4393 {
4394         nfs4_setup_sequence(NFS_SERVER(data->inode),
4395                         &data->args.seq_args,
4396                         &data->res.seq_res,
4397                         task);
4398 }
4399
4400 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4401 {
4402         struct inode *inode = data->inode;
4403
4404         trace_nfs4_commit(data, task->tk_status);
4405         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4406                                     NULL, NULL) == -EAGAIN) {
4407                 rpc_restart_call_prepare(task);
4408                 return -EAGAIN;
4409         }
4410         return 0;
4411 }
4412
4413 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4414 {
4415         if (!nfs4_sequence_done(task, &data->res.seq_res))
4416                 return -EAGAIN;
4417         return data->commit_done_cb(task, data);
4418 }
4419
4420 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4421 {
4422         struct nfs_server *server = NFS_SERVER(data->inode);
4423
4424         if (data->commit_done_cb == NULL)
4425                 data->commit_done_cb = nfs4_commit_done_cb;
4426         data->res.server = server;
4427         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4428         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4429 }
4430
4431 struct nfs4_renewdata {
4432         struct nfs_client       *client;
4433         unsigned long           timestamp;
4434 };
4435
4436 /*
4437  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4438  * standalone procedure for queueing an asynchronous RENEW.
4439  */
4440 static void nfs4_renew_release(void *calldata)
4441 {
4442         struct nfs4_renewdata *data = calldata;
4443         struct nfs_client *clp = data->client;
4444
4445         if (atomic_read(&clp->cl_count) > 1)
4446                 nfs4_schedule_state_renewal(clp);
4447         nfs_put_client(clp);
4448         kfree(data);
4449 }
4450
4451 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4452 {
4453         struct nfs4_renewdata *data = calldata;
4454         struct nfs_client *clp = data->client;
4455         unsigned long timestamp = data->timestamp;
4456
4457         trace_nfs4_renew_async(clp, task->tk_status);
4458         switch (task->tk_status) {
4459         case 0:
4460                 break;
4461         case -NFS4ERR_LEASE_MOVED:
4462                 nfs4_schedule_lease_moved_recovery(clp);
4463                 break;
4464         default:
4465                 /* Unless we're shutting down, schedule state recovery! */
4466                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4467                         return;
4468                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4469                         nfs4_schedule_lease_recovery(clp);
4470                         return;
4471                 }
4472                 nfs4_schedule_path_down_recovery(clp);
4473         }
4474         do_renew_lease(clp, timestamp);
4475 }
4476
4477 static const struct rpc_call_ops nfs4_renew_ops = {
4478         .rpc_call_done = nfs4_renew_done,
4479         .rpc_release = nfs4_renew_release,
4480 };
4481
4482 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4483 {
4484         struct rpc_message msg = {
4485                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4486                 .rpc_argp       = clp,
4487                 .rpc_cred       = cred,
4488         };
4489         struct nfs4_renewdata *data;
4490
4491         if (renew_flags == 0)
4492                 return 0;
4493         if (!atomic_inc_not_zero(&clp->cl_count))
4494                 return -EIO;
4495         data = kmalloc(sizeof(*data), GFP_NOFS);
4496         if (data == NULL)
4497                 return -ENOMEM;
4498         data->client = clp;
4499         data->timestamp = jiffies;
4500         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4501                         &nfs4_renew_ops, data);
4502 }
4503
4504 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4505 {
4506         struct rpc_message msg = {
4507                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4508                 .rpc_argp       = clp,
4509                 .rpc_cred       = cred,
4510         };
4511         unsigned long now = jiffies;
4512         int status;
4513
4514         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4515         if (status < 0)
4516                 return status;
4517         do_renew_lease(clp, now);
4518         return 0;
4519 }
4520
4521 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4522 {
4523         return server->caps & NFS_CAP_ACLS;
4524 }
4525
4526 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4527  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4528  * the stack.
4529  */
4530 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4531
4532 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4533                 struct page **pages, unsigned int *pgbase)
4534 {
4535         struct page *newpage, **spages;
4536         int rc = 0;
4537         size_t len;
4538         spages = pages;
4539
4540         do {
4541                 len = min_t(size_t, PAGE_SIZE, buflen);
4542                 newpage = alloc_page(GFP_KERNEL);
4543
4544                 if (newpage == NULL)
4545                         goto unwind;
4546                 memcpy(page_address(newpage), buf, len);
4547                 buf += len;
4548                 buflen -= len;
4549                 *pages++ = newpage;
4550                 rc++;
4551         } while (buflen != 0);
4552
4553         return rc;
4554
4555 unwind:
4556         for(; rc > 0; rc--)
4557                 __free_page(spages[rc-1]);
4558         return -ENOMEM;
4559 }
4560
4561 struct nfs4_cached_acl {
4562         int cached;
4563         size_t len;
4564         char data[0];
4565 };
4566
4567 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4568 {
4569         struct nfs_inode *nfsi = NFS_I(inode);
4570
4571         spin_lock(&inode->i_lock);
4572         kfree(nfsi->nfs4_acl);
4573         nfsi->nfs4_acl = acl;
4574         spin_unlock(&inode->i_lock);
4575 }
4576
4577 static void nfs4_zap_acl_attr(struct inode *inode)
4578 {
4579         nfs4_set_cached_acl(inode, NULL);
4580 }
4581
4582 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4583 {
4584         struct nfs_inode *nfsi = NFS_I(inode);
4585         struct nfs4_cached_acl *acl;
4586         int ret = -ENOENT;
4587
4588         spin_lock(&inode->i_lock);
4589         acl = nfsi->nfs4_acl;
4590         if (acl == NULL)
4591                 goto out;
4592         if (buf == NULL) /* user is just asking for length */
4593                 goto out_len;
4594         if (acl->cached == 0)
4595                 goto out;
4596         ret = -ERANGE; /* see getxattr(2) man page */
4597         if (acl->len > buflen)
4598                 goto out;
4599         memcpy(buf, acl->data, acl->len);
4600 out_len:
4601         ret = acl->len;
4602 out:
4603         spin_unlock(&inode->i_lock);
4604         return ret;
4605 }
4606
4607 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4608 {
4609         struct nfs4_cached_acl *acl;
4610         size_t buflen = sizeof(*acl) + acl_len;
4611
4612         if (buflen <= PAGE_SIZE) {
4613                 acl = kmalloc(buflen, GFP_KERNEL);
4614                 if (acl == NULL)
4615                         goto out;
4616                 acl->cached = 1;
4617                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4618         } else {
4619                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4620                 if (acl == NULL)
4621                         goto out;
4622                 acl->cached = 0;
4623         }
4624         acl->len = acl_len;
4625 out:
4626         nfs4_set_cached_acl(inode, acl);
4627 }
4628
4629 /*
4630  * The getxattr API returns the required buffer length when called with a
4631  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4632  * the required buf.  On a NULL buf, we send a page of data to the server
4633  * guessing that the ACL request can be serviced by a page. If so, we cache
4634  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4635  * the cache. If not so, we throw away the page, and cache the required
4636  * length. The next getxattr call will then produce another round trip to
4637  * the server, this time with the input buf of the required size.
4638  */
4639 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4640 {
4641         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4642         struct nfs_getaclargs args = {
4643                 .fh = NFS_FH(inode),
4644                 .acl_pages = pages,
4645                 .acl_len = buflen,
4646         };
4647         struct nfs_getaclres res = {
4648                 .acl_len = buflen,
4649         };
4650         struct rpc_message msg = {
4651                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4652                 .rpc_argp = &args,
4653                 .rpc_resp = &res,
4654         };
4655         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4656         int ret = -ENOMEM, i;
4657
4658         /* As long as we're doing a round trip to the server anyway,
4659          * let's be prepared for a page of acl data. */
4660         if (npages == 0)
4661                 npages = 1;
4662         if (npages > ARRAY_SIZE(pages))
4663                 return -ERANGE;
4664
4665         for (i = 0; i < npages; i++) {
4666                 pages[i] = alloc_page(GFP_KERNEL);
4667                 if (!pages[i])
4668                         goto out_free;
4669         }
4670
4671         /* for decoding across pages */
4672         res.acl_scratch = alloc_page(GFP_KERNEL);
4673         if (!res.acl_scratch)
4674                 goto out_free;
4675
4676         args.acl_len = npages * PAGE_SIZE;
4677         args.acl_pgbase = 0;
4678
4679         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4680                 __func__, buf, buflen, npages, args.acl_len);
4681         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4682                              &msg, &args.seq_args, &res.seq_res, 0);
4683         if (ret)
4684                 goto out_free;
4685
4686         /* Handle the case where the passed-in buffer is too short */
4687         if (res.acl_flags & NFS4_ACL_TRUNC) {
4688                 /* Did the user only issue a request for the acl length? */
4689                 if (buf == NULL)
4690                         goto out_ok;
4691                 ret = -ERANGE;
4692                 goto out_free;
4693         }
4694         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4695         if (buf) {
4696                 if (res.acl_len > buflen) {
4697                         ret = -ERANGE;
4698                         goto out_free;
4699                 }
4700                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4701         }
4702 out_ok:
4703         ret = res.acl_len;
4704 out_free:
4705         for (i = 0; i < npages; i++)
4706                 if (pages[i])
4707                         __free_page(pages[i]);
4708         if (res.acl_scratch)
4709                 __free_page(res.acl_scratch);
4710         return ret;
4711 }
4712
4713 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4714 {
4715         struct nfs4_exception exception = { };
4716         ssize_t ret;
4717         do {
4718                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4719                 trace_nfs4_get_acl(inode, ret);
4720                 if (ret >= 0)
4721                         break;
4722                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4723         } while (exception.retry);
4724         return ret;
4725 }
4726
4727 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4728 {
4729         struct nfs_server *server = NFS_SERVER(inode);
4730         int ret;
4731
4732         if (!nfs4_server_supports_acls(server))
4733                 return -EOPNOTSUPP;
4734         ret = nfs_revalidate_inode(server, inode);
4735         if (ret < 0)
4736                 return ret;
4737         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4738                 nfs_zap_acl_cache(inode);
4739         ret = nfs4_read_cached_acl(inode, buf, buflen);
4740         if (ret != -ENOENT)
4741                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4742                  * but no cached acl data, just the acl length */
4743                 return ret;
4744         return nfs4_get_acl_uncached(inode, buf, buflen);
4745 }
4746
4747 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4748 {
4749         struct nfs_server *server = NFS_SERVER(inode);
4750         struct page *pages[NFS4ACL_MAXPAGES];
4751         struct nfs_setaclargs arg = {
4752                 .fh             = NFS_FH(inode),
4753                 .acl_pages      = pages,
4754                 .acl_len        = buflen,
4755         };
4756         struct nfs_setaclres res;
4757         struct rpc_message msg = {
4758                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4759                 .rpc_argp       = &arg,
4760                 .rpc_resp       = &res,
4761         };
4762         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4763         int ret, i;
4764
4765         if (!nfs4_server_supports_acls(server))
4766                 return -EOPNOTSUPP;
4767         if (npages > ARRAY_SIZE(pages))
4768                 return -ERANGE;
4769         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages, &arg.acl_pgbase);
4770         if (i < 0)
4771                 return i;
4772         nfs4_inode_return_delegation(inode);
4773         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4774
4775         /*
4776          * Free each page after tx, so the only ref left is
4777          * held by the network stack
4778          */
4779         for (; i > 0; i--)
4780                 put_page(pages[i-1]);
4781
4782         /*
4783          * Acl update can result in inode attribute update.
4784          * so mark the attribute cache invalid.
4785          */
4786         spin_lock(&inode->i_lock);
4787         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4788         spin_unlock(&inode->i_lock);
4789         nfs_access_zap_cache(inode);
4790         nfs_zap_acl_cache(inode);
4791         return ret;
4792 }
4793
4794 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4795 {
4796         struct nfs4_exception exception = { };
4797         int err;
4798         do {
4799                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4800                 trace_nfs4_set_acl(inode, err);
4801                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4802                                 &exception);
4803         } while (exception.retry);
4804         return err;
4805 }
4806
4807 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4808 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4809                                         size_t buflen)
4810 {
4811         struct nfs_server *server = NFS_SERVER(inode);
4812         struct nfs_fattr fattr;
4813         struct nfs4_label label = {0, 0, buflen, buf};
4814
4815         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4816         struct nfs4_getattr_arg arg = {
4817                 .fh             = NFS_FH(inode),
4818                 .bitmask        = bitmask,
4819         };
4820         struct nfs4_getattr_res res = {
4821                 .fattr          = &fattr,
4822                 .label          = &label,
4823                 .server         = server,
4824         };
4825         struct rpc_message msg = {
4826                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4827                 .rpc_argp       = &arg,
4828                 .rpc_resp       = &res,
4829         };
4830         int ret;
4831
4832         nfs_fattr_init(&fattr);
4833
4834         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4835         if (ret)
4836                 return ret;
4837         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4838                 return -ENOENT;
4839         if (buflen < label.len)
4840                 return -ERANGE;
4841         return 0;
4842 }
4843
4844 static int nfs4_get_security_label(struct inode *inode, void *buf,
4845                                         size_t buflen)
4846 {
4847         struct nfs4_exception exception = { };
4848         int err;
4849
4850         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4851                 return -EOPNOTSUPP;
4852
4853         do {
4854                 err = _nfs4_get_security_label(inode, buf, buflen);
4855                 trace_nfs4_get_security_label(inode, err);
4856                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4857                                 &exception);
4858         } while (exception.retry);
4859         return err;
4860 }
4861
4862 static int _nfs4_do_set_security_label(struct inode *inode,
4863                 struct nfs4_label *ilabel,
4864                 struct nfs_fattr *fattr,
4865                 struct nfs4_label *olabel)
4866 {
4867
4868         struct iattr sattr = {0};
4869         struct nfs_server *server = NFS_SERVER(inode);
4870         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4871         struct nfs_setattrargs arg = {
4872                 .fh             = NFS_FH(inode),
4873                 .iap            = &sattr,
4874                 .server         = server,
4875                 .bitmask        = bitmask,
4876                 .label          = ilabel,
4877         };
4878         struct nfs_setattrres res = {
4879                 .fattr          = fattr,
4880                 .label          = olabel,
4881                 .server         = server,
4882         };
4883         struct rpc_message msg = {
4884                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4885                 .rpc_argp       = &arg,
4886                 .rpc_resp       = &res,
4887         };
4888         int status;
4889
4890         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4891
4892         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4893         if (status)
4894                 dprintk("%s failed: %d\n", __func__, status);
4895
4896         return status;
4897 }
4898
4899 static int nfs4_do_set_security_label(struct inode *inode,
4900                 struct nfs4_label *ilabel,
4901                 struct nfs_fattr *fattr,
4902                 struct nfs4_label *olabel)
4903 {
4904         struct nfs4_exception exception = { };
4905         int err;
4906
4907         do {
4908                 err = _nfs4_do_set_security_label(inode, ilabel,
4909                                 fattr, olabel);
4910                 trace_nfs4_set_security_label(inode, err);
4911                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4912                                 &exception);
4913         } while (exception.retry);
4914         return err;
4915 }
4916
4917 static int
4918 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4919 {
4920         struct nfs4_label ilabel, *olabel = NULL;
4921         struct nfs_fattr fattr;
4922         struct rpc_cred *cred;
4923         struct inode *inode = d_inode(dentry);
4924         int status;
4925
4926         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4927                 return -EOPNOTSUPP;
4928
4929         nfs_fattr_init(&fattr);
4930
4931         ilabel.pi = 0;
4932         ilabel.lfs = 0;
4933         ilabel.label = (char *)buf;
4934         ilabel.len = buflen;
4935
4936         cred = rpc_lookup_cred();
4937         if (IS_ERR(cred))
4938                 return PTR_ERR(cred);
4939
4940         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4941         if (IS_ERR(olabel)) {
4942                 status = -PTR_ERR(olabel);
4943                 goto out;
4944         }
4945
4946         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
4947         if (status == 0)
4948                 nfs_setsecurity(inode, &fattr, olabel);
4949
4950         nfs4_label_free(olabel);
4951 out:
4952         put_rpccred(cred);
4953         return status;
4954 }
4955 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
4956
4957
4958 static int
4959 nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server,
4960                         struct nfs4_state *state, long *timeout)
4961 {
4962         struct nfs_client *clp = server->nfs_client;
4963
4964         if (task->tk_status >= 0)
4965                 return 0;
4966         switch(task->tk_status) {
4967                 case -NFS4ERR_DELEG_REVOKED:
4968                 case -NFS4ERR_ADMIN_REVOKED:
4969                 case -NFS4ERR_BAD_STATEID:
4970                 case -NFS4ERR_OPENMODE:
4971                         if (state == NULL)
4972                                 break;
4973                         if (nfs4_schedule_stateid_recovery(server, state) < 0)
4974                                 goto recovery_failed;
4975                         goto wait_on_recovery;
4976                 case -NFS4ERR_EXPIRED:
4977                         if (state != NULL) {
4978                                 if (nfs4_schedule_stateid_recovery(server, state) < 0)
4979                                         goto recovery_failed;
4980                         }
4981                 case -NFS4ERR_STALE_STATEID:
4982                 case -NFS4ERR_STALE_CLIENTID:
4983                         nfs4_schedule_lease_recovery(clp);
4984                         goto wait_on_recovery;
4985                 case -NFS4ERR_MOVED:
4986                         if (nfs4_schedule_migration_recovery(server) < 0)
4987                                 goto recovery_failed;
4988                         goto wait_on_recovery;
4989                 case -NFS4ERR_LEASE_MOVED:
4990                         nfs4_schedule_lease_moved_recovery(clp);
4991                         goto wait_on_recovery;
4992 #if defined(CONFIG_NFS_V4_1)
4993                 case -NFS4ERR_BADSESSION:
4994                 case -NFS4ERR_BADSLOT:
4995                 case -NFS4ERR_BAD_HIGH_SLOT:
4996                 case -NFS4ERR_DEADSESSION:
4997                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
4998                 case -NFS4ERR_SEQ_FALSE_RETRY:
4999                 case -NFS4ERR_SEQ_MISORDERED:
5000                         dprintk("%s ERROR %d, Reset session\n", __func__,
5001                                 task->tk_status);
5002                         nfs4_schedule_session_recovery(clp->cl_session, task->tk_status);
5003                         goto wait_on_recovery;
5004 #endif /* CONFIG_NFS_V4_1 */
5005                 case -NFS4ERR_DELAY:
5006                         nfs_inc_server_stats(server, NFSIOS_DELAY);
5007                         rpc_delay(task, nfs4_update_delay(timeout));
5008                         goto restart_call;
5009                 case -NFS4ERR_GRACE:
5010                         rpc_delay(task, NFS4_POLL_RETRY_MAX);
5011                 case -NFS4ERR_RETRY_UNCACHED_REP:
5012                 case -NFS4ERR_OLD_STATEID:
5013                         goto restart_call;
5014         }
5015         task->tk_status = nfs4_map_errors(task->tk_status);
5016         return 0;
5017 recovery_failed:
5018         task->tk_status = -EIO;
5019         return 0;
5020 wait_on_recovery:
5021         rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
5022         if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
5023                 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
5024         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
5025                 goto recovery_failed;
5026 restart_call:
5027         task->tk_status = 0;
5028         return -EAGAIN;
5029 }
5030
5031 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5032                                     nfs4_verifier *bootverf)
5033 {
5034         __be32 verf[2];
5035
5036         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5037                 /* An impossible timestamp guarantees this value
5038                  * will never match a generated boot time. */
5039                 verf[0] = 0;
5040                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5041         } else {
5042                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5043                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5044                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5045         }
5046         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5047 }
5048
5049 static int
5050 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5051 {
5052         int result;
5053         size_t len;
5054         char *str;
5055
5056         if (clp->cl_owner_id != NULL)
5057                 return 0;
5058
5059         rcu_read_lock();
5060         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5061                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5062                 1 +
5063                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5064                 1;
5065         rcu_read_unlock();
5066
5067         if (len > NFS4_OPAQUE_LIMIT + 1)
5068                 return -EINVAL;
5069
5070         /*
5071          * Since this string is allocated at mount time, and held until the
5072          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5073          * about a memory-reclaim deadlock.
5074          */
5075         str = kmalloc(len, GFP_KERNEL);
5076         if (!str)
5077                 return -ENOMEM;
5078
5079         rcu_read_lock();
5080         result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5081                         clp->cl_ipaddr,
5082                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5083                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5084         rcu_read_unlock();
5085
5086         clp->cl_owner_id = str;
5087         return 0;
5088 }
5089
5090 static int
5091 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5092 {
5093         int result;
5094         size_t len;
5095         char *str;
5096
5097         len = 10 + 10 + 1 + 10 + 1 +
5098                 strlen(nfs4_client_id_uniquifier) + 1 +
5099                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5100
5101         if (len > NFS4_OPAQUE_LIMIT + 1)
5102                 return -EINVAL;
5103
5104         /*
5105          * Since this string is allocated at mount time, and held until the
5106          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5107          * about a memory-reclaim deadlock.
5108          */
5109         str = kmalloc(len, GFP_KERNEL);
5110         if (!str)
5111                 return -ENOMEM;
5112
5113         result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5114                         clp->rpc_ops->version, clp->cl_minorversion,
5115                         nfs4_client_id_uniquifier,
5116                         clp->cl_rpcclient->cl_nodename);
5117         clp->cl_owner_id = str;
5118         return 0;
5119 }
5120
5121 static int
5122 nfs4_init_uniform_client_string(struct nfs_client *clp)
5123 {
5124         int result;
5125         size_t len;
5126         char *str;
5127
5128         if (clp->cl_owner_id != NULL)
5129                 return 0;
5130
5131         if (nfs4_client_id_uniquifier[0] != '\0')
5132                 return nfs4_init_uniquifier_client_string(clp);
5133
5134         len = 10 + 10 + 1 + 10 + 1 +
5135                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5136
5137         if (len > NFS4_OPAQUE_LIMIT + 1)
5138                 return -EINVAL;
5139
5140         /*
5141          * Since this string is allocated at mount time, and held until the
5142          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5143          * about a memory-reclaim deadlock.
5144          */
5145         str = kmalloc(len, GFP_KERNEL);
5146         if (!str)
5147                 return -ENOMEM;
5148
5149         result = scnprintf(str, len, "Linux NFSv%u.%u %s",
5150                         clp->rpc_ops->version, clp->cl_minorversion,
5151                         clp->cl_rpcclient->cl_nodename);
5152         clp->cl_owner_id = str;
5153         return 0;
5154 }
5155
5156 /*
5157  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5158  * services.  Advertise one based on the address family of the
5159  * clientaddr.
5160  */
5161 static unsigned int
5162 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5163 {
5164         if (strchr(clp->cl_ipaddr, ':') != NULL)
5165                 return scnprintf(buf, len, "tcp6");
5166         else
5167                 return scnprintf(buf, len, "tcp");
5168 }
5169
5170 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5171 {
5172         struct nfs4_setclientid *sc = calldata;
5173
5174         if (task->tk_status == 0)
5175                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5176 }
5177
5178 static const struct rpc_call_ops nfs4_setclientid_ops = {
5179         .rpc_call_done = nfs4_setclientid_done,
5180 };
5181
5182 /**
5183  * nfs4_proc_setclientid - Negotiate client ID
5184  * @clp: state data structure
5185  * @program: RPC program for NFSv4 callback service
5186  * @port: IP port number for NFS4 callback service
5187  * @cred: RPC credential to use for this call
5188  * @res: where to place the result
5189  *
5190  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5191  */
5192 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5193                 unsigned short port, struct rpc_cred *cred,
5194                 struct nfs4_setclientid_res *res)
5195 {
5196         nfs4_verifier sc_verifier;
5197         struct nfs4_setclientid setclientid = {
5198                 .sc_verifier = &sc_verifier,
5199                 .sc_prog = program,
5200                 .sc_clnt = clp,
5201         };
5202         struct rpc_message msg = {
5203                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5204                 .rpc_argp = &setclientid,
5205                 .rpc_resp = res,
5206                 .rpc_cred = cred,
5207         };
5208         struct rpc_task *task;
5209         struct rpc_task_setup task_setup_data = {
5210                 .rpc_client = clp->cl_rpcclient,
5211                 .rpc_message = &msg,
5212                 .callback_ops = &nfs4_setclientid_ops,
5213                 .callback_data = &setclientid,
5214                 .flags = RPC_TASK_TIMEOUT,
5215         };
5216         int status;
5217
5218         /* nfs_client_id4 */
5219         nfs4_init_boot_verifier(clp, &sc_verifier);
5220
5221         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5222                 status = nfs4_init_uniform_client_string(clp);
5223         else
5224                 status = nfs4_init_nonuniform_client_string(clp);
5225
5226         if (status)
5227                 goto out;
5228
5229         /* cb_client4 */
5230         setclientid.sc_netid_len =
5231                                 nfs4_init_callback_netid(clp,
5232                                                 setclientid.sc_netid,
5233                                                 sizeof(setclientid.sc_netid));
5234         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5235                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5236                                 clp->cl_ipaddr, port >> 8, port & 255);
5237
5238         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5239                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5240                 clp->cl_owner_id);
5241         task = rpc_run_task(&task_setup_data);
5242         if (IS_ERR(task)) {
5243                 status = PTR_ERR(task);
5244                 goto out;
5245         }
5246         status = task->tk_status;
5247         if (setclientid.sc_cred) {
5248                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5249                 put_rpccred(setclientid.sc_cred);
5250         }
5251         rpc_put_task(task);
5252 out:
5253         trace_nfs4_setclientid(clp, status);
5254         dprintk("NFS reply setclientid: %d\n", status);
5255         return status;
5256 }
5257
5258 /**
5259  * nfs4_proc_setclientid_confirm - Confirm client ID
5260  * @clp: state data structure
5261  * @res: result of a previous SETCLIENTID
5262  * @cred: RPC credential to use for this call
5263  *
5264  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5265  */
5266 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5267                 struct nfs4_setclientid_res *arg,
5268                 struct rpc_cred *cred)
5269 {
5270         struct rpc_message msg = {
5271                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5272                 .rpc_argp = arg,
5273                 .rpc_cred = cred,
5274         };
5275         int status;
5276
5277         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5278                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5279                 clp->cl_clientid);
5280         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5281         trace_nfs4_setclientid_confirm(clp, status);
5282         dprintk("NFS reply setclientid_confirm: %d\n", status);
5283         return status;
5284 }
5285
5286 struct nfs4_delegreturndata {
5287         struct nfs4_delegreturnargs args;
5288         struct nfs4_delegreturnres res;
5289         struct nfs_fh fh;
5290         nfs4_stateid stateid;
5291         unsigned long timestamp;
5292         struct nfs_fattr fattr;
5293         int rpc_status;
5294         struct inode *inode;
5295         bool roc;
5296         u32 roc_barrier;
5297 };
5298
5299 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5300 {
5301         struct nfs4_delegreturndata *data = calldata;
5302
5303         if (!nfs4_sequence_done(task, &data->res.seq_res))
5304                 return;
5305
5306         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5307         switch (task->tk_status) {
5308         case 0:
5309                 renew_lease(data->res.server, data->timestamp);
5310         case -NFS4ERR_ADMIN_REVOKED:
5311         case -NFS4ERR_DELEG_REVOKED:
5312         case -NFS4ERR_BAD_STATEID:
5313         case -NFS4ERR_OLD_STATEID:
5314         case -NFS4ERR_STALE_STATEID:
5315         case -NFS4ERR_EXPIRED:
5316                 task->tk_status = 0;
5317                 if (data->roc)
5318                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5319                 break;
5320         default:
5321                 if (nfs4_async_handle_error(task, data->res.server,
5322                                             NULL, NULL) == -EAGAIN) {
5323                         rpc_restart_call_prepare(task);
5324                         return;
5325                 }
5326         }
5327         data->rpc_status = task->tk_status;
5328 }
5329
5330 static void nfs4_delegreturn_release(void *calldata)
5331 {
5332         struct nfs4_delegreturndata *data = calldata;
5333         struct inode *inode = data->inode;
5334
5335         if (inode) {
5336                 if (data->roc)
5337                         pnfs_roc_release(inode);
5338                 nfs_iput_and_deactive(inode);
5339         }
5340         kfree(calldata);
5341 }
5342
5343 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5344 {
5345         struct nfs4_delegreturndata *d_data;
5346
5347         d_data = (struct nfs4_delegreturndata *)data;
5348
5349         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5350                 return;
5351
5352         if (d_data->roc)
5353                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5354
5355         nfs4_setup_sequence(d_data->res.server,
5356                         &d_data->args.seq_args,
5357                         &d_data->res.seq_res,
5358                         task);
5359 }
5360
5361 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5362         .rpc_call_prepare = nfs4_delegreturn_prepare,
5363         .rpc_call_done = nfs4_delegreturn_done,
5364         .rpc_release = nfs4_delegreturn_release,
5365 };
5366
5367 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5368 {
5369         struct nfs4_delegreturndata *data;
5370         struct nfs_server *server = NFS_SERVER(inode);
5371         struct rpc_task *task;
5372         struct rpc_message msg = {
5373                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5374                 .rpc_cred = cred,
5375         };
5376         struct rpc_task_setup task_setup_data = {
5377                 .rpc_client = server->client,
5378                 .rpc_message = &msg,
5379                 .callback_ops = &nfs4_delegreturn_ops,
5380                 .flags = RPC_TASK_ASYNC,
5381         };
5382         int status = 0;
5383
5384         data = kzalloc(sizeof(*data), GFP_NOFS);
5385         if (data == NULL)
5386                 return -ENOMEM;
5387         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5388         data->args.fhandle = &data->fh;
5389         data->args.stateid = &data->stateid;
5390         data->args.bitmask = server->cache_consistency_bitmask;
5391         nfs_copy_fh(&data->fh, NFS_FH(inode));
5392         nfs4_stateid_copy(&data->stateid, stateid);
5393         data->res.fattr = &data->fattr;
5394         data->res.server = server;
5395         nfs_fattr_init(data->res.fattr);
5396         data->timestamp = jiffies;
5397         data->rpc_status = 0;
5398         data->inode = nfs_igrab_and_active(inode);
5399         if (data->inode)
5400                 data->roc = nfs4_roc(inode);
5401
5402         task_setup_data.callback_data = data;
5403         msg.rpc_argp = &data->args;
5404         msg.rpc_resp = &data->res;
5405         task = rpc_run_task(&task_setup_data);
5406         if (IS_ERR(task))
5407                 return PTR_ERR(task);
5408         if (!issync)
5409                 goto out;
5410         status = nfs4_wait_for_completion_rpc_task(task);
5411         if (status != 0)
5412                 goto out;
5413         status = data->rpc_status;
5414         if (status == 0)
5415                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5416         else
5417                 nfs_refresh_inode(inode, &data->fattr);
5418 out:
5419         rpc_put_task(task);
5420         return status;
5421 }
5422
5423 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5424 {
5425         struct nfs_server *server = NFS_SERVER(inode);
5426         struct nfs4_exception exception = { };
5427         int err;
5428         do {
5429                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5430                 trace_nfs4_delegreturn(inode, err);
5431                 switch (err) {
5432                         case -NFS4ERR_STALE_STATEID:
5433                         case -NFS4ERR_EXPIRED:
5434                         case 0:
5435                                 return 0;
5436                 }
5437                 err = nfs4_handle_exception(server, err, &exception);
5438         } while (exception.retry);
5439         return err;
5440 }
5441
5442 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5443 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5444
5445 /*
5446  * sleep, with exponential backoff, and retry the LOCK operation.
5447  */
5448 static unsigned long
5449 nfs4_set_lock_task_retry(unsigned long timeout)
5450 {
5451         freezable_schedule_timeout_killable_unsafe(timeout);
5452         timeout <<= 1;
5453         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5454                 return NFS4_LOCK_MAXTIMEOUT;
5455         return timeout;
5456 }
5457
5458 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5459 {
5460         struct inode *inode = state->inode;
5461         struct nfs_server *server = NFS_SERVER(inode);
5462         struct nfs_client *clp = server->nfs_client;
5463         struct nfs_lockt_args arg = {
5464                 .fh = NFS_FH(inode),
5465                 .fl = request,
5466         };
5467         struct nfs_lockt_res res = {
5468                 .denied = request,
5469         };
5470         struct rpc_message msg = {
5471                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5472                 .rpc_argp       = &arg,
5473                 .rpc_resp       = &res,
5474                 .rpc_cred       = state->owner->so_cred,
5475         };
5476         struct nfs4_lock_state *lsp;
5477         int status;
5478
5479         arg.lock_owner.clientid = clp->cl_clientid;
5480         status = nfs4_set_lock_state(state, request);
5481         if (status != 0)
5482                 goto out;
5483         lsp = request->fl_u.nfs4_fl.owner;
5484         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5485         arg.lock_owner.s_dev = server->s_dev;
5486         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5487         switch (status) {
5488                 case 0:
5489                         request->fl_type = F_UNLCK;
5490                         break;
5491                 case -NFS4ERR_DENIED:
5492                         status = 0;
5493         }
5494         request->fl_ops->fl_release_private(request);
5495         request->fl_ops = NULL;
5496 out:
5497         return status;
5498 }
5499
5500 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5501 {
5502         struct nfs4_exception exception = { };
5503         int err;
5504
5505         do {
5506                 err = _nfs4_proc_getlk(state, cmd, request);
5507                 trace_nfs4_get_lock(request, state, cmd, err);
5508                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5509                                 &exception);
5510         } while (exception.retry);
5511         return err;
5512 }
5513
5514 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5515 {
5516         int res = 0;
5517         switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
5518                 case FL_POSIX:
5519                         res = posix_lock_inode_wait(inode, fl);
5520                         break;
5521                 case FL_FLOCK:
5522                         res = flock_lock_inode_wait(inode, fl);
5523                         break;
5524                 default:
5525                         BUG();
5526         }
5527         return res;
5528 }
5529
5530 struct nfs4_unlockdata {
5531         struct nfs_locku_args arg;
5532         struct nfs_locku_res res;
5533         struct nfs4_lock_state *lsp;
5534         struct nfs_open_context *ctx;
5535         struct file_lock fl;
5536         const struct nfs_server *server;
5537         unsigned long timestamp;
5538 };
5539
5540 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5541                 struct nfs_open_context *ctx,
5542                 struct nfs4_lock_state *lsp,
5543                 struct nfs_seqid *seqid)
5544 {
5545         struct nfs4_unlockdata *p;
5546         struct inode *inode = lsp->ls_state->inode;
5547
5548         p = kzalloc(sizeof(*p), GFP_NOFS);
5549         if (p == NULL)
5550                 return NULL;
5551         p->arg.fh = NFS_FH(inode);
5552         p->arg.fl = &p->fl;
5553         p->arg.seqid = seqid;
5554         p->res.seqid = seqid;
5555         p->lsp = lsp;
5556         atomic_inc(&lsp->ls_count);
5557         /* Ensure we don't close file until we're done freeing locks! */
5558         p->ctx = get_nfs_open_context(ctx);
5559         memcpy(&p->fl, fl, sizeof(p->fl));
5560         p->server = NFS_SERVER(inode);
5561         return p;
5562 }
5563
5564 static void nfs4_locku_release_calldata(void *data)
5565 {
5566         struct nfs4_unlockdata *calldata = data;
5567         nfs_free_seqid(calldata->arg.seqid);
5568         nfs4_put_lock_state(calldata->lsp);
5569         put_nfs_open_context(calldata->ctx);
5570         kfree(calldata);
5571 }
5572
5573 static void nfs4_locku_done(struct rpc_task *task, void *data)
5574 {
5575         struct nfs4_unlockdata *calldata = data;
5576
5577         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5578                 return;
5579         switch (task->tk_status) {
5580                 case 0:
5581                         renew_lease(calldata->server, calldata->timestamp);
5582                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5583                         if (nfs4_update_lock_stateid(calldata->lsp,
5584                                         &calldata->res.stateid))
5585                                 break;
5586                 case -NFS4ERR_BAD_STATEID:
5587                 case -NFS4ERR_OLD_STATEID:
5588                 case -NFS4ERR_STALE_STATEID:
5589                 case -NFS4ERR_EXPIRED:
5590                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5591                                                 &calldata->lsp->ls_stateid))
5592                                 rpc_restart_call_prepare(task);
5593                         break;
5594                 default:
5595                         if (nfs4_async_handle_error(task, calldata->server,
5596                                                     NULL, NULL) == -EAGAIN)
5597                                 rpc_restart_call_prepare(task);
5598         }
5599         nfs_release_seqid(calldata->arg.seqid);
5600 }
5601
5602 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5603 {
5604         struct nfs4_unlockdata *calldata = data;
5605
5606         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5607                 goto out_wait;
5608         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5609         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5610                 /* Note: exit _without_ running nfs4_locku_done */
5611                 goto out_no_action;
5612         }
5613         calldata->timestamp = jiffies;
5614         if (nfs4_setup_sequence(calldata->server,
5615                                 &calldata->arg.seq_args,
5616                                 &calldata->res.seq_res,
5617                                 task) != 0)
5618                 nfs_release_seqid(calldata->arg.seqid);
5619         return;
5620 out_no_action:
5621         task->tk_action = NULL;
5622 out_wait:
5623         nfs4_sequence_done(task, &calldata->res.seq_res);
5624 }
5625
5626 static const struct rpc_call_ops nfs4_locku_ops = {
5627         .rpc_call_prepare = nfs4_locku_prepare,
5628         .rpc_call_done = nfs4_locku_done,
5629         .rpc_release = nfs4_locku_release_calldata,
5630 };
5631
5632 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5633                 struct nfs_open_context *ctx,
5634                 struct nfs4_lock_state *lsp,
5635                 struct nfs_seqid *seqid)
5636 {
5637         struct nfs4_unlockdata *data;
5638         struct rpc_message msg = {
5639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5640                 .rpc_cred = ctx->cred,
5641         };
5642         struct rpc_task_setup task_setup_data = {
5643                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5644                 .rpc_message = &msg,
5645                 .callback_ops = &nfs4_locku_ops,
5646                 .workqueue = nfsiod_workqueue,
5647                 .flags = RPC_TASK_ASYNC,
5648         };
5649
5650         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5651                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5652
5653         /* Ensure this is an unlock - when canceling a lock, the
5654          * canceled lock is passed in, and it won't be an unlock.
5655          */
5656         fl->fl_type = F_UNLCK;
5657
5658         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5659         if (data == NULL) {
5660                 nfs_free_seqid(seqid);
5661                 return ERR_PTR(-ENOMEM);
5662         }
5663
5664         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5665         msg.rpc_argp = &data->arg;
5666         msg.rpc_resp = &data->res;
5667         task_setup_data.callback_data = data;
5668         return rpc_run_task(&task_setup_data);
5669 }
5670
5671 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5672 {
5673         struct inode *inode = state->inode;
5674         struct nfs4_state_owner *sp = state->owner;
5675         struct nfs_inode *nfsi = NFS_I(inode);
5676         struct nfs_seqid *seqid;
5677         struct nfs4_lock_state *lsp;
5678         struct rpc_task *task;
5679         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5680         int status = 0;
5681         unsigned char fl_flags = request->fl_flags;
5682
5683         status = nfs4_set_lock_state(state, request);
5684         /* Unlock _before_ we do the RPC call */
5685         request->fl_flags |= FL_EXISTS;
5686         /* Exclude nfs_delegation_claim_locks() */
5687         mutex_lock(&sp->so_delegreturn_mutex);
5688         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5689         down_read(&nfsi->rwsem);
5690         if (do_vfs_lock(inode, request) == -ENOENT) {
5691                 up_read(&nfsi->rwsem);
5692                 mutex_unlock(&sp->so_delegreturn_mutex);
5693                 goto out;
5694         }
5695         up_read(&nfsi->rwsem);
5696         mutex_unlock(&sp->so_delegreturn_mutex);
5697         if (status != 0)
5698                 goto out;
5699         /* Is this a delegated lock? */
5700         lsp = request->fl_u.nfs4_fl.owner;
5701         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5702                 goto out;
5703         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5704         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5705         status = -ENOMEM;
5706         if (IS_ERR(seqid))
5707                 goto out;
5708         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5709         status = PTR_ERR(task);
5710         if (IS_ERR(task))
5711                 goto out;
5712         status = nfs4_wait_for_completion_rpc_task(task);
5713         rpc_put_task(task);
5714 out:
5715         request->fl_flags = fl_flags;
5716         trace_nfs4_unlock(request, state, F_SETLK, status);
5717         return status;
5718 }
5719
5720 struct nfs4_lockdata {
5721         struct nfs_lock_args arg;
5722         struct nfs_lock_res res;
5723         struct nfs4_lock_state *lsp;
5724         struct nfs_open_context *ctx;
5725         struct file_lock fl;
5726         unsigned long timestamp;
5727         int rpc_status;
5728         int cancelled;
5729         struct nfs_server *server;
5730 };
5731
5732 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5733                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5734                 gfp_t gfp_mask)
5735 {
5736         struct nfs4_lockdata *p;
5737         struct inode *inode = lsp->ls_state->inode;
5738         struct nfs_server *server = NFS_SERVER(inode);
5739         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5740
5741         p = kzalloc(sizeof(*p), gfp_mask);
5742         if (p == NULL)
5743                 return NULL;
5744
5745         p->arg.fh = NFS_FH(inode);
5746         p->arg.fl = &p->fl;
5747         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
5748         if (IS_ERR(p->arg.open_seqid))
5749                 goto out_free;
5750         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
5751         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
5752         if (IS_ERR(p->arg.lock_seqid))
5753                 goto out_free_seqid;
5754         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
5755         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
5756         p->arg.lock_owner.s_dev = server->s_dev;
5757         p->res.lock_seqid = p->arg.lock_seqid;
5758         p->lsp = lsp;
5759         p->server = server;
5760         atomic_inc(&lsp->ls_count);
5761         p->ctx = get_nfs_open_context(ctx);
5762         get_file(fl->fl_file);
5763         memcpy(&p->fl, fl, sizeof(p->fl));
5764         return p;
5765 out_free_seqid:
5766         nfs_free_seqid(p->arg.open_seqid);
5767 out_free:
5768         kfree(p);
5769         return NULL;
5770 }
5771
5772 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
5773 {
5774         struct nfs4_lockdata *data = calldata;
5775         struct nfs4_state *state = data->lsp->ls_state;
5776
5777         dprintk("%s: begin!\n", __func__);
5778         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
5779                 goto out_wait;
5780         /* Do we need to do an open_to_lock_owner? */
5781         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
5782                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
5783                         goto out_release_lock_seqid;
5784                 }
5785                 nfs4_stateid_copy(&data->arg.open_stateid,
5786                                 &state->open_stateid);
5787                 data->arg.new_lock_owner = 1;
5788                 data->res.open_seqid = data->arg.open_seqid;
5789         } else {
5790                 data->arg.new_lock_owner = 0;
5791                 nfs4_stateid_copy(&data->arg.lock_stateid,
5792                                 &data->lsp->ls_stateid);
5793         }
5794         if (!nfs4_valid_open_stateid(state)) {
5795                 data->rpc_status = -EBADF;
5796                 task->tk_action = NULL;
5797                 goto out_release_open_seqid;
5798         }
5799         data->timestamp = jiffies;
5800         if (nfs4_setup_sequence(data->server,
5801                                 &data->arg.seq_args,
5802                                 &data->res.seq_res,
5803                                 task) == 0)
5804                 return;
5805 out_release_open_seqid:
5806         nfs_release_seqid(data->arg.open_seqid);
5807 out_release_lock_seqid:
5808         nfs_release_seqid(data->arg.lock_seqid);
5809 out_wait:
5810         nfs4_sequence_done(task, &data->res.seq_res);
5811         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
5812 }
5813
5814 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
5815 {
5816         struct nfs4_lockdata *data = calldata;
5817         struct nfs4_lock_state *lsp = data->lsp;
5818
5819         dprintk("%s: begin!\n", __func__);
5820
5821         if (!nfs4_sequence_done(task, &data->res.seq_res))
5822                 return;
5823
5824         data->rpc_status = task->tk_status;
5825         switch (task->tk_status) {
5826         case 0:
5827                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
5828                                 data->timestamp);
5829                 if (data->arg.new_lock) {
5830                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
5831                         if (do_vfs_lock(lsp->ls_state->inode, &data->fl) < 0) {
5832                                 rpc_restart_call_prepare(task);
5833                                 break;
5834                         }
5835                 }
5836                 if (data->arg.new_lock_owner != 0) {
5837                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
5838                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
5839                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
5840                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
5841                         rpc_restart_call_prepare(task);
5842                 break;
5843         case -NFS4ERR_BAD_STATEID:
5844         case -NFS4ERR_OLD_STATEID:
5845         case -NFS4ERR_STALE_STATEID:
5846         case -NFS4ERR_EXPIRED:
5847                 if (data->arg.new_lock_owner != 0) {
5848                         if (!nfs4_stateid_match(&data->arg.open_stateid,
5849                                                 &lsp->ls_state->open_stateid))
5850                                 rpc_restart_call_prepare(task);
5851                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
5852                                                 &lsp->ls_stateid))
5853                                 rpc_restart_call_prepare(task);
5854         }
5855         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
5856 }
5857
5858 static void nfs4_lock_release(void *calldata)
5859 {
5860         struct nfs4_lockdata *data = calldata;
5861
5862         dprintk("%s: begin!\n", __func__);
5863         nfs_free_seqid(data->arg.open_seqid);
5864         if (data->cancelled != 0) {
5865                 struct rpc_task *task;
5866                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
5867                                 data->arg.lock_seqid);
5868                 if (!IS_ERR(task))
5869                         rpc_put_task_async(task);
5870                 dprintk("%s: cancelling lock!\n", __func__);
5871         } else
5872                 nfs_free_seqid(data->arg.lock_seqid);
5873         nfs4_put_lock_state(data->lsp);
5874         put_nfs_open_context(data->ctx);
5875         fput(data->fl.fl_file);
5876         kfree(data);
5877         dprintk("%s: done!\n", __func__);
5878 }
5879
5880 static const struct rpc_call_ops nfs4_lock_ops = {
5881         .rpc_call_prepare = nfs4_lock_prepare,
5882         .rpc_call_done = nfs4_lock_done,
5883         .rpc_release = nfs4_lock_release,
5884 };
5885
5886 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
5887 {
5888         switch (error) {
5889         case -NFS4ERR_ADMIN_REVOKED:
5890         case -NFS4ERR_BAD_STATEID:
5891                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5892                 if (new_lock_owner != 0 ||
5893                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
5894                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
5895                 break;
5896         case -NFS4ERR_STALE_STATEID:
5897                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
5898         case -NFS4ERR_EXPIRED:
5899                 nfs4_schedule_lease_recovery(server->nfs_client);
5900         };
5901 }
5902
5903 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
5904 {
5905         struct nfs4_lockdata *data;
5906         struct rpc_task *task;
5907         struct rpc_message msg = {
5908                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
5909                 .rpc_cred = state->owner->so_cred,
5910         };
5911         struct rpc_task_setup task_setup_data = {
5912                 .rpc_client = NFS_CLIENT(state->inode),
5913                 .rpc_message = &msg,
5914                 .callback_ops = &nfs4_lock_ops,
5915                 .workqueue = nfsiod_workqueue,
5916                 .flags = RPC_TASK_ASYNC,
5917         };
5918         int ret;
5919
5920         dprintk("%s: begin!\n", __func__);
5921         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
5922                         fl->fl_u.nfs4_fl.owner,
5923                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
5924         if (data == NULL)
5925                 return -ENOMEM;
5926         if (IS_SETLKW(cmd))
5927                 data->arg.block = 1;
5928         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5929         msg.rpc_argp = &data->arg;
5930         msg.rpc_resp = &data->res;
5931         task_setup_data.callback_data = data;
5932         if (recovery_type > NFS_LOCK_NEW) {
5933                 if (recovery_type == NFS_LOCK_RECLAIM)
5934                         data->arg.reclaim = NFS_LOCK_RECLAIM;
5935                 nfs4_set_sequence_privileged(&data->arg.seq_args);
5936         } else
5937                 data->arg.new_lock = 1;
5938         task = rpc_run_task(&task_setup_data);
5939         if (IS_ERR(task))
5940                 return PTR_ERR(task);
5941         ret = nfs4_wait_for_completion_rpc_task(task);
5942         if (ret == 0) {
5943                 ret = data->rpc_status;
5944                 if (ret)
5945                         nfs4_handle_setlk_error(data->server, data->lsp,
5946                                         data->arg.new_lock_owner, ret);
5947         } else
5948                 data->cancelled = 1;
5949         rpc_put_task(task);
5950         dprintk("%s: done, ret = %d!\n", __func__, ret);
5951         return ret;
5952 }
5953
5954 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
5955 {
5956         struct nfs_server *server = NFS_SERVER(state->inode);
5957         struct nfs4_exception exception = {
5958                 .inode = state->inode,
5959         };
5960         int err;
5961
5962         do {
5963                 /* Cache the lock if possible... */
5964                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5965                         return 0;
5966                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
5967                 trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
5968                 if (err != -NFS4ERR_DELAY)
5969                         break;
5970                 nfs4_handle_exception(server, err, &exception);
5971         } while (exception.retry);
5972         return err;
5973 }
5974
5975 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
5976 {
5977         struct nfs_server *server = NFS_SERVER(state->inode);
5978         struct nfs4_exception exception = {
5979                 .inode = state->inode,
5980         };
5981         int err;
5982
5983         err = nfs4_set_lock_state(state, request);
5984         if (err != 0)
5985                 return err;
5986         if (!recover_lost_locks) {
5987                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
5988                 return 0;
5989         }
5990         do {
5991                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
5992                         return 0;
5993                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
5994                 trace_nfs4_lock_expired(request, state, F_SETLK, err);
5995                 switch (err) {
5996                 default:
5997                         goto out;
5998                 case -NFS4ERR_GRACE:
5999                 case -NFS4ERR_DELAY:
6000                         nfs4_handle_exception(server, err, &exception);
6001                         err = 0;
6002                 }
6003         } while (exception.retry);
6004 out:
6005         return err;
6006 }
6007
6008 #if defined(CONFIG_NFS_V4_1)
6009 /**
6010  * nfs41_check_expired_locks - possibly free a lock stateid
6011  *
6012  * @state: NFSv4 state for an inode
6013  *
6014  * Returns NFS_OK if recovery for this stateid is now finished.
6015  * Otherwise a negative NFS4ERR value is returned.
6016  */
6017 static int nfs41_check_expired_locks(struct nfs4_state *state)
6018 {
6019         int status, ret = -NFS4ERR_BAD_STATEID;
6020         struct nfs4_lock_state *lsp;
6021         struct nfs_server *server = NFS_SERVER(state->inode);
6022
6023         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
6024                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
6025                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
6026
6027                         status = nfs41_test_stateid(server,
6028                                         &lsp->ls_stateid,
6029                                         cred);
6030                         trace_nfs4_test_lock_stateid(state, lsp, status);
6031                         if (status != NFS_OK) {
6032                                 /* Free the stateid unless the server
6033                                  * informs us the stateid is unrecognized. */
6034                                 if (status != -NFS4ERR_BAD_STATEID)
6035                                         nfs41_free_stateid(server,
6036                                                         &lsp->ls_stateid,
6037                                                         cred);
6038                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6039                                 ret = status;
6040                         }
6041                 }
6042         };
6043
6044         return ret;
6045 }
6046
6047 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6048 {
6049         int status = NFS_OK;
6050
6051         if (test_bit(LK_STATE_IN_USE, &state->flags))
6052                 status = nfs41_check_expired_locks(state);
6053         if (status != NFS_OK)
6054                 status = nfs4_lock_expired(state, request);
6055         return status;
6056 }
6057 #endif
6058
6059 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6060 {
6061         struct nfs_inode *nfsi = NFS_I(state->inode);
6062         unsigned char fl_flags = request->fl_flags;
6063         int status = -ENOLCK;
6064
6065         if ((fl_flags & FL_POSIX) &&
6066                         !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6067                 goto out;
6068         /* Is this a delegated open? */
6069         status = nfs4_set_lock_state(state, request);
6070         if (status != 0)
6071                 goto out;
6072         request->fl_flags |= FL_ACCESS;
6073         status = do_vfs_lock(state->inode, request);
6074         if (status < 0)
6075                 goto out;
6076         down_read(&nfsi->rwsem);
6077         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6078                 /* Yes: cache locks! */
6079                 /* ...but avoid races with delegation recall... */
6080                 request->fl_flags = fl_flags & ~FL_SLEEP;
6081                 status = do_vfs_lock(state->inode, request);
6082                 up_read(&nfsi->rwsem);
6083                 goto out;
6084         }
6085         up_read(&nfsi->rwsem);
6086         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6087 out:
6088         request->fl_flags = fl_flags;
6089         return status;
6090 }
6091
6092 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6093 {
6094         struct nfs4_exception exception = {
6095                 .state = state,
6096                 .inode = state->inode,
6097         };
6098         int err;
6099
6100         do {
6101                 err = _nfs4_proc_setlk(state, cmd, request);
6102                 trace_nfs4_set_lock(request, state, cmd, err);
6103                 if (err == -NFS4ERR_DENIED)
6104                         err = -EAGAIN;
6105                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6106                                 err, &exception);
6107         } while (exception.retry);
6108         return err;
6109 }
6110
6111 static int
6112 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6113 {
6114         struct nfs_open_context *ctx;
6115         struct nfs4_state *state;
6116         unsigned long timeout = NFS4_LOCK_MINTIMEOUT;
6117         int status;
6118
6119         /* verify open state */
6120         ctx = nfs_file_open_context(filp);
6121         state = ctx->state;
6122
6123         if (request->fl_start < 0 || request->fl_end < 0)
6124                 return -EINVAL;
6125
6126         if (IS_GETLK(cmd)) {
6127                 if (state != NULL)
6128                         return nfs4_proc_getlk(state, F_GETLK, request);
6129                 return 0;
6130         }
6131
6132         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6133                 return -EINVAL;
6134
6135         if (request->fl_type == F_UNLCK) {
6136                 if (state != NULL)
6137                         return nfs4_proc_unlck(state, cmd, request);
6138                 return 0;
6139         }
6140
6141         if (state == NULL)
6142                 return -ENOLCK;
6143         /*
6144          * Don't rely on the VFS having checked the file open mode,
6145          * since it won't do this for flock() locks.
6146          */
6147         switch (request->fl_type) {
6148         case F_RDLCK:
6149                 if (!(filp->f_mode & FMODE_READ))
6150                         return -EBADF;
6151                 break;
6152         case F_WRLCK:
6153                 if (!(filp->f_mode & FMODE_WRITE))
6154                         return -EBADF;
6155         }
6156
6157         do {
6158                 status = nfs4_proc_setlk(state, cmd, request);
6159                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6160                         break;
6161                 timeout = nfs4_set_lock_task_retry(timeout);
6162                 status = -ERESTARTSYS;
6163                 if (signalled())
6164                         break;
6165         } while(status < 0);
6166         return status;
6167 }
6168
6169 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6170 {
6171         struct nfs_server *server = NFS_SERVER(state->inode);
6172         int err;
6173
6174         err = nfs4_set_lock_state(state, fl);
6175         if (err != 0)
6176                 return err;
6177         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6178         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6179 }
6180
6181 struct nfs_release_lockowner_data {
6182         struct nfs4_lock_state *lsp;
6183         struct nfs_server *server;
6184         struct nfs_release_lockowner_args args;
6185         struct nfs_release_lockowner_res res;
6186         unsigned long timestamp;
6187 };
6188
6189 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6190 {
6191         struct nfs_release_lockowner_data *data = calldata;
6192         struct nfs_server *server = data->server;
6193         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6194                              &data->args.seq_args, &data->res.seq_res, task);
6195         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6196         data->timestamp = jiffies;
6197 }
6198
6199 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6200 {
6201         struct nfs_release_lockowner_data *data = calldata;
6202         struct nfs_server *server = data->server;
6203
6204         nfs40_sequence_done(task, &data->res.seq_res);
6205
6206         switch (task->tk_status) {
6207         case 0:
6208                 renew_lease(server, data->timestamp);
6209                 break;
6210         case -NFS4ERR_STALE_CLIENTID:
6211         case -NFS4ERR_EXPIRED:
6212                 nfs4_schedule_lease_recovery(server->nfs_client);
6213                 break;
6214         case -NFS4ERR_LEASE_MOVED:
6215         case -NFS4ERR_DELAY:
6216                 if (nfs4_async_handle_error(task, server,
6217                                             NULL, NULL) == -EAGAIN)
6218                         rpc_restart_call_prepare(task);
6219         }
6220 }
6221
6222 static void nfs4_release_lockowner_release(void *calldata)
6223 {
6224         struct nfs_release_lockowner_data *data = calldata;
6225         nfs4_free_lock_state(data->server, data->lsp);
6226         kfree(calldata);
6227 }
6228
6229 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6230         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6231         .rpc_call_done = nfs4_release_lockowner_done,
6232         .rpc_release = nfs4_release_lockowner_release,
6233 };
6234
6235 static void
6236 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6237 {
6238         struct nfs_release_lockowner_data *data;
6239         struct rpc_message msg = {
6240                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6241         };
6242
6243         if (server->nfs_client->cl_mvops->minor_version != 0)
6244                 return;
6245
6246         data = kmalloc(sizeof(*data), GFP_NOFS);
6247         if (!data)
6248                 return;
6249         data->lsp = lsp;
6250         data->server = server;
6251         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6252         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6253         data->args.lock_owner.s_dev = server->s_dev;
6254
6255         msg.rpc_argp = &data->args;
6256         msg.rpc_resp = &data->res;
6257         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6258         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6259 }
6260
6261 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6262
6263 static int nfs4_xattr_set_nfs4_acl(struct dentry *dentry, const char *key,
6264                                    const void *buf, size_t buflen,
6265                                    int flags, int type)
6266 {
6267         if (strcmp(key, "") != 0)
6268                 return -EINVAL;
6269
6270         return nfs4_proc_set_acl(d_inode(dentry), buf, buflen);
6271 }
6272
6273 static int nfs4_xattr_get_nfs4_acl(struct dentry *dentry, const char *key,
6274                                    void *buf, size_t buflen, int type)
6275 {
6276         if (strcmp(key, "") != 0)
6277                 return -EINVAL;
6278
6279         return nfs4_proc_get_acl(d_inode(dentry), buf, buflen);
6280 }
6281
6282 static size_t nfs4_xattr_list_nfs4_acl(struct dentry *dentry, char *list,
6283                                        size_t list_len, const char *name,
6284                                        size_t name_len, int type)
6285 {
6286         size_t len = sizeof(XATTR_NAME_NFSV4_ACL);
6287
6288         if (!nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry))))
6289                 return 0;
6290
6291         if (list && len <= list_len)
6292                 memcpy(list, XATTR_NAME_NFSV4_ACL, len);
6293         return len;
6294 }
6295
6296 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6297 static inline int nfs4_server_supports_labels(struct nfs_server *server)
6298 {
6299         return server->caps & NFS_CAP_SECURITY_LABEL;
6300 }
6301
6302 static int nfs4_xattr_set_nfs4_label(struct dentry *dentry, const char *key,
6303                                    const void *buf, size_t buflen,
6304                                    int flags, int type)
6305 {
6306         if (security_ismaclabel(key))
6307                 return nfs4_set_security_label(dentry, buf, buflen);
6308
6309         return -EOPNOTSUPP;
6310 }
6311
6312 static int nfs4_xattr_get_nfs4_label(struct dentry *dentry, const char *key,
6313                                    void *buf, size_t buflen, int type)
6314 {
6315         if (security_ismaclabel(key))
6316                 return nfs4_get_security_label(d_inode(dentry), buf, buflen);
6317         return -EOPNOTSUPP;
6318 }
6319
6320 static size_t nfs4_xattr_list_nfs4_label(struct dentry *dentry, char *list,
6321                                        size_t list_len, const char *name,
6322                                        size_t name_len, int type)
6323 {
6324         size_t len = 0;
6325
6326         if (nfs_server_capable(d_inode(dentry), NFS_CAP_SECURITY_LABEL)) {
6327                 len = security_inode_listsecurity(d_inode(dentry), NULL, 0);
6328                 if (list && len <= list_len)
6329                         security_inode_listsecurity(d_inode(dentry), list, len);
6330         }
6331         return len;
6332 }
6333
6334 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6335         .prefix = XATTR_SECURITY_PREFIX,
6336         .list   = nfs4_xattr_list_nfs4_label,
6337         .get    = nfs4_xattr_get_nfs4_label,
6338         .set    = nfs4_xattr_set_nfs4_label,
6339 };
6340 #endif
6341
6342
6343 /*
6344  * nfs_fhget will use either the mounted_on_fileid or the fileid
6345  */
6346 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6347 {
6348         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6349                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6350               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6351               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6352                 return;
6353
6354         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6355                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6356         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6357         fattr->nlink = 2;
6358 }
6359
6360 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6361                                    const struct qstr *name,
6362                                    struct nfs4_fs_locations *fs_locations,
6363                                    struct page *page)
6364 {
6365         struct nfs_server *server = NFS_SERVER(dir);
6366         u32 bitmask[3] = {
6367                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6368         };
6369         struct nfs4_fs_locations_arg args = {
6370                 .dir_fh = NFS_FH(dir),
6371                 .name = name,
6372                 .page = page,
6373                 .bitmask = bitmask,
6374         };
6375         struct nfs4_fs_locations_res res = {
6376                 .fs_locations = fs_locations,
6377         };
6378         struct rpc_message msg = {
6379                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6380                 .rpc_argp = &args,
6381                 .rpc_resp = &res,
6382         };
6383         int status;
6384
6385         dprintk("%s: start\n", __func__);
6386
6387         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6388          * is not supported */
6389         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6390                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6391         else
6392                 bitmask[0] |= FATTR4_WORD0_FILEID;
6393
6394         nfs_fattr_init(&fs_locations->fattr);
6395         fs_locations->server = server;
6396         fs_locations->nlocations = 0;
6397         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6398         dprintk("%s: returned status = %d\n", __func__, status);
6399         return status;
6400 }
6401
6402 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6403                            const struct qstr *name,
6404                            struct nfs4_fs_locations *fs_locations,
6405                            struct page *page)
6406 {
6407         struct nfs4_exception exception = { };
6408         int err;
6409         do {
6410                 err = _nfs4_proc_fs_locations(client, dir, name,
6411                                 fs_locations, page);
6412                 trace_nfs4_get_fs_locations(dir, name, err);
6413                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6414                                 &exception);
6415         } while (exception.retry);
6416         return err;
6417 }
6418
6419 /*
6420  * This operation also signals the server that this client is
6421  * performing migration recovery.  The server can stop returning
6422  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6423  * appended to this compound to identify the client ID which is
6424  * performing recovery.
6425  */
6426 static int _nfs40_proc_get_locations(struct inode *inode,
6427                                      struct nfs4_fs_locations *locations,
6428                                      struct page *page, struct rpc_cred *cred)
6429 {
6430         struct nfs_server *server = NFS_SERVER(inode);
6431         struct rpc_clnt *clnt = server->client;
6432         u32 bitmask[2] = {
6433                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6434         };
6435         struct nfs4_fs_locations_arg args = {
6436                 .clientid       = server->nfs_client->cl_clientid,
6437                 .fh             = NFS_FH(inode),
6438                 .page           = page,
6439                 .bitmask        = bitmask,
6440                 .migration      = 1,            /* skip LOOKUP */
6441                 .renew          = 1,            /* append RENEW */
6442         };
6443         struct nfs4_fs_locations_res res = {
6444                 .fs_locations   = locations,
6445                 .migration      = 1,
6446                 .renew          = 1,
6447         };
6448         struct rpc_message msg = {
6449                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6450                 .rpc_argp       = &args,
6451                 .rpc_resp       = &res,
6452                 .rpc_cred       = cred,
6453         };
6454         unsigned long now = jiffies;
6455         int status;
6456
6457         nfs_fattr_init(&locations->fattr);
6458         locations->server = server;
6459         locations->nlocations = 0;
6460
6461         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6462         nfs4_set_sequence_privileged(&args.seq_args);
6463         status = nfs4_call_sync_sequence(clnt, server, &msg,
6464                                         &args.seq_args, &res.seq_res);
6465         if (status)
6466                 return status;
6467
6468         renew_lease(server, now);
6469         return 0;
6470 }
6471
6472 #ifdef CONFIG_NFS_V4_1
6473
6474 /*
6475  * This operation also signals the server that this client is
6476  * performing migration recovery.  The server can stop asserting
6477  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6478  * performing this operation is identified in the SEQUENCE
6479  * operation in this compound.
6480  *
6481  * When the client supports GETATTR(fs_locations_info), it can
6482  * be plumbed in here.
6483  */
6484 static int _nfs41_proc_get_locations(struct inode *inode,
6485                                      struct nfs4_fs_locations *locations,
6486                                      struct page *page, struct rpc_cred *cred)
6487 {
6488         struct nfs_server *server = NFS_SERVER(inode);
6489         struct rpc_clnt *clnt = server->client;
6490         u32 bitmask[2] = {
6491                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6492         };
6493         struct nfs4_fs_locations_arg args = {
6494                 .fh             = NFS_FH(inode),
6495                 .page           = page,
6496                 .bitmask        = bitmask,
6497                 .migration      = 1,            /* skip LOOKUP */
6498         };
6499         struct nfs4_fs_locations_res res = {
6500                 .fs_locations   = locations,
6501                 .migration      = 1,
6502         };
6503         struct rpc_message msg = {
6504                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6505                 .rpc_argp       = &args,
6506                 .rpc_resp       = &res,
6507                 .rpc_cred       = cred,
6508         };
6509         int status;
6510
6511         nfs_fattr_init(&locations->fattr);
6512         locations->server = server;
6513         locations->nlocations = 0;
6514
6515         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6516         nfs4_set_sequence_privileged(&args.seq_args);
6517         status = nfs4_call_sync_sequence(clnt, server, &msg,
6518                                         &args.seq_args, &res.seq_res);
6519         if (status == NFS4_OK &&
6520             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6521                 status = -NFS4ERR_LEASE_MOVED;
6522         return status;
6523 }
6524
6525 #endif  /* CONFIG_NFS_V4_1 */
6526
6527 /**
6528  * nfs4_proc_get_locations - discover locations for a migrated FSID
6529  * @inode: inode on FSID that is migrating
6530  * @locations: result of query
6531  * @page: buffer
6532  * @cred: credential to use for this operation
6533  *
6534  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6535  * operation failed, or a negative errno if a local error occurred.
6536  *
6537  * On success, "locations" is filled in, but if the server has
6538  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6539  * asserted.
6540  *
6541  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6542  * from this client that require migration recovery.
6543  */
6544 int nfs4_proc_get_locations(struct inode *inode,
6545                             struct nfs4_fs_locations *locations,
6546                             struct page *page, struct rpc_cred *cred)
6547 {
6548         struct nfs_server *server = NFS_SERVER(inode);
6549         struct nfs_client *clp = server->nfs_client;
6550         const struct nfs4_mig_recovery_ops *ops =
6551                                         clp->cl_mvops->mig_recovery_ops;
6552         struct nfs4_exception exception = { };
6553         int status;
6554
6555         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6556                 (unsigned long long)server->fsid.major,
6557                 (unsigned long long)server->fsid.minor,
6558                 clp->cl_hostname);
6559         nfs_display_fhandle(NFS_FH(inode), __func__);
6560
6561         do {
6562                 status = ops->get_locations(inode, locations, page, cred);
6563                 if (status != -NFS4ERR_DELAY)
6564                         break;
6565                 nfs4_handle_exception(server, status, &exception);
6566         } while (exception.retry);
6567         return status;
6568 }
6569
6570 /*
6571  * This operation also signals the server that this client is
6572  * performing "lease moved" recovery.  The server can stop
6573  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6574  * is appended to this compound to identify the client ID which is
6575  * performing recovery.
6576  */
6577 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6578 {
6579         struct nfs_server *server = NFS_SERVER(inode);
6580         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6581         struct rpc_clnt *clnt = server->client;
6582         struct nfs4_fsid_present_arg args = {
6583                 .fh             = NFS_FH(inode),
6584                 .clientid       = clp->cl_clientid,
6585                 .renew          = 1,            /* append RENEW */
6586         };
6587         struct nfs4_fsid_present_res res = {
6588                 .renew          = 1,
6589         };
6590         struct rpc_message msg = {
6591                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6592                 .rpc_argp       = &args,
6593                 .rpc_resp       = &res,
6594                 .rpc_cred       = cred,
6595         };
6596         unsigned long now = jiffies;
6597         int status;
6598
6599         res.fh = nfs_alloc_fhandle();
6600         if (res.fh == NULL)
6601                 return -ENOMEM;
6602
6603         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6604         nfs4_set_sequence_privileged(&args.seq_args);
6605         status = nfs4_call_sync_sequence(clnt, server, &msg,
6606                                                 &args.seq_args, &res.seq_res);
6607         nfs_free_fhandle(res.fh);
6608         if (status)
6609                 return status;
6610
6611         do_renew_lease(clp, now);
6612         return 0;
6613 }
6614
6615 #ifdef CONFIG_NFS_V4_1
6616
6617 /*
6618  * This operation also signals the server that this client is
6619  * performing "lease moved" recovery.  The server can stop asserting
6620  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6621  * this operation is identified in the SEQUENCE operation in this
6622  * compound.
6623  */
6624 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6625 {
6626         struct nfs_server *server = NFS_SERVER(inode);
6627         struct rpc_clnt *clnt = server->client;
6628         struct nfs4_fsid_present_arg args = {
6629                 .fh             = NFS_FH(inode),
6630         };
6631         struct nfs4_fsid_present_res res = {
6632         };
6633         struct rpc_message msg = {
6634                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6635                 .rpc_argp       = &args,
6636                 .rpc_resp       = &res,
6637                 .rpc_cred       = cred,
6638         };
6639         int status;
6640
6641         res.fh = nfs_alloc_fhandle();
6642         if (res.fh == NULL)
6643                 return -ENOMEM;
6644
6645         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6646         nfs4_set_sequence_privileged(&args.seq_args);
6647         status = nfs4_call_sync_sequence(clnt, server, &msg,
6648                                                 &args.seq_args, &res.seq_res);
6649         nfs_free_fhandle(res.fh);
6650         if (status == NFS4_OK &&
6651             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6652                 status = -NFS4ERR_LEASE_MOVED;
6653         return status;
6654 }
6655
6656 #endif  /* CONFIG_NFS_V4_1 */
6657
6658 /**
6659  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
6660  * @inode: inode on FSID to check
6661  * @cred: credential to use for this operation
6662  *
6663  * Server indicates whether the FSID is present, moved, or not
6664  * recognized.  This operation is necessary to clear a LEASE_MOVED
6665  * condition for this client ID.
6666  *
6667  * Returns NFS4_OK if the FSID is present on this server,
6668  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
6669  *  NFS4ERR code if some error occurred on the server, or a
6670  *  negative errno if a local failure occurred.
6671  */
6672 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6673 {
6674         struct nfs_server *server = NFS_SERVER(inode);
6675         struct nfs_client *clp = server->nfs_client;
6676         const struct nfs4_mig_recovery_ops *ops =
6677                                         clp->cl_mvops->mig_recovery_ops;
6678         struct nfs4_exception exception = { };
6679         int status;
6680
6681         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6682                 (unsigned long long)server->fsid.major,
6683                 (unsigned long long)server->fsid.minor,
6684                 clp->cl_hostname);
6685         nfs_display_fhandle(NFS_FH(inode), __func__);
6686
6687         do {
6688                 status = ops->fsid_present(inode, cred);
6689                 if (status != -NFS4ERR_DELAY)
6690                         break;
6691                 nfs4_handle_exception(server, status, &exception);
6692         } while (exception.retry);
6693         return status;
6694 }
6695
6696 /**
6697  * If 'use_integrity' is true and the state managment nfs_client
6698  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
6699  * and the machine credential as per RFC3530bis and RFC5661 Security
6700  * Considerations sections. Otherwise, just use the user cred with the
6701  * filesystem's rpc_client.
6702  */
6703 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
6704 {
6705         int status;
6706         struct nfs4_secinfo_arg args = {
6707                 .dir_fh = NFS_FH(dir),
6708                 .name   = name,
6709         };
6710         struct nfs4_secinfo_res res = {
6711                 .flavors     = flavors,
6712         };
6713         struct rpc_message msg = {
6714                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
6715                 .rpc_argp = &args,
6716                 .rpc_resp = &res,
6717         };
6718         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
6719         struct rpc_cred *cred = NULL;
6720
6721         if (use_integrity) {
6722                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
6723                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
6724                 msg.rpc_cred = cred;
6725         }
6726
6727         dprintk("NFS call  secinfo %s\n", name->name);
6728
6729         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
6730                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
6731
6732         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
6733                                 &res.seq_res, 0);
6734         dprintk("NFS reply  secinfo: %d\n", status);
6735
6736         if (cred)
6737                 put_rpccred(cred);
6738
6739         return status;
6740 }
6741
6742 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
6743                       struct nfs4_secinfo_flavors *flavors)
6744 {
6745         struct nfs4_exception exception = { };
6746         int err;
6747         do {
6748                 err = -NFS4ERR_WRONGSEC;
6749
6750                 /* try to use integrity protection with machine cred */
6751                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
6752                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
6753
6754                 /*
6755                  * if unable to use integrity protection, or SECINFO with
6756                  * integrity protection returns NFS4ERR_WRONGSEC (which is
6757                  * disallowed by spec, but exists in deployed servers) use
6758                  * the current filesystem's rpc_client and the user cred.
6759                  */
6760                 if (err == -NFS4ERR_WRONGSEC)
6761                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
6762
6763                 trace_nfs4_secinfo(dir, name, err);
6764                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6765                                 &exception);
6766         } while (exception.retry);
6767         return err;
6768 }
6769
6770 #ifdef CONFIG_NFS_V4_1
6771 /*
6772  * Check the exchange flags returned by the server for invalid flags, having
6773  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
6774  * DS flags set.
6775  */
6776 static int nfs4_check_cl_exchange_flags(u32 flags)
6777 {
6778         if (flags & ~EXCHGID4_FLAG_MASK_R)
6779                 goto out_inval;
6780         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
6781             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
6782                 goto out_inval;
6783         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
6784                 goto out_inval;
6785         return NFS_OK;
6786 out_inval:
6787         return -NFS4ERR_INVAL;
6788 }
6789
6790 static bool
6791 nfs41_same_server_scope(struct nfs41_server_scope *a,
6792                         struct nfs41_server_scope *b)
6793 {
6794         if (a->server_scope_sz == b->server_scope_sz &&
6795             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
6796                 return true;
6797
6798         return false;
6799 }
6800
6801 /*
6802  * nfs4_proc_bind_conn_to_session()
6803  *
6804  * The 4.1 client currently uses the same TCP connection for the
6805  * fore and backchannel.
6806  */
6807 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
6808 {
6809         int status;
6810         struct nfs41_bind_conn_to_session_args args = {
6811                 .client = clp,
6812                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
6813         };
6814         struct nfs41_bind_conn_to_session_res res;
6815         struct rpc_message msg = {
6816                 .rpc_proc =
6817                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
6818                 .rpc_argp = &args,
6819                 .rpc_resp = &res,
6820                 .rpc_cred = cred,
6821         };
6822
6823         dprintk("--> %s\n", __func__);
6824
6825         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
6826         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
6827                 args.dir = NFS4_CDFC4_FORE;
6828
6829         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
6830         trace_nfs4_bind_conn_to_session(clp, status);
6831         if (status == 0) {
6832                 if (memcmp(res.sessionid.data,
6833                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
6834                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
6835                         status = -EIO;
6836                         goto out;
6837                 }
6838                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
6839                         dprintk("NFS: %s: Unexpected direction from server\n",
6840                                 __func__);
6841                         status = -EIO;
6842                         goto out;
6843                 }
6844                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
6845                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
6846                                 __func__);
6847                         status = -EIO;
6848                         goto out;
6849                 }
6850         }
6851 out:
6852         dprintk("<-- %s status= %d\n", __func__, status);
6853         return status;
6854 }
6855
6856 /*
6857  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
6858  * and operations we'd like to see to enable certain features in the allow map
6859  */
6860 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
6861         .how = SP4_MACH_CRED,
6862         .enforce.u.words = {
6863                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6864                       1 << (OP_EXCHANGE_ID - 32) |
6865                       1 << (OP_CREATE_SESSION - 32) |
6866                       1 << (OP_DESTROY_SESSION - 32) |
6867                       1 << (OP_DESTROY_CLIENTID - 32)
6868         },
6869         .allow.u.words = {
6870                 [0] = 1 << (OP_CLOSE) |
6871                       1 << (OP_LOCKU) |
6872                       1 << (OP_COMMIT),
6873                 [1] = 1 << (OP_SECINFO - 32) |
6874                       1 << (OP_SECINFO_NO_NAME - 32) |
6875                       1 << (OP_TEST_STATEID - 32) |
6876                       1 << (OP_FREE_STATEID - 32) |
6877                       1 << (OP_WRITE - 32)
6878         }
6879 };
6880
6881 /*
6882  * Select the state protection mode for client `clp' given the server results
6883  * from exchange_id in `sp'.
6884  *
6885  * Returns 0 on success, negative errno otherwise.
6886  */
6887 static int nfs4_sp4_select_mode(struct nfs_client *clp,
6888                                  struct nfs41_state_protection *sp)
6889 {
6890         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
6891                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
6892                       1 << (OP_EXCHANGE_ID - 32) |
6893                       1 << (OP_CREATE_SESSION - 32) |
6894                       1 << (OP_DESTROY_SESSION - 32) |
6895                       1 << (OP_DESTROY_CLIENTID - 32)
6896         };
6897         unsigned int i;
6898
6899         if (sp->how == SP4_MACH_CRED) {
6900                 /* Print state protect result */
6901                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
6902                 for (i = 0; i <= LAST_NFS4_OP; i++) {
6903                         if (test_bit(i, sp->enforce.u.longs))
6904                                 dfprintk(MOUNT, "  enforce op %d\n", i);
6905                         if (test_bit(i, sp->allow.u.longs))
6906                                 dfprintk(MOUNT, "  allow op %d\n", i);
6907                 }
6908
6909                 /* make sure nothing is on enforce list that isn't supported */
6910                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
6911                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
6912                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6913                                 return -EINVAL;
6914                         }
6915                 }
6916
6917                 /*
6918                  * Minimal mode - state operations are allowed to use machine
6919                  * credential.  Note this already happens by default, so the
6920                  * client doesn't have to do anything more than the negotiation.
6921                  *
6922                  * NOTE: we don't care if EXCHANGE_ID is in the list -
6923                  *       we're already using the machine cred for exchange_id
6924                  *       and will never use a different cred.
6925                  */
6926                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
6927                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
6928                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
6929                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
6930                         dfprintk(MOUNT, "sp4_mach_cred:\n");
6931                         dfprintk(MOUNT, "  minimal mode enabled\n");
6932                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
6933                 } else {
6934                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
6935                         return -EINVAL;
6936                 }
6937
6938                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
6939                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
6940                         dfprintk(MOUNT, "  cleanup mode enabled\n");
6941                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
6942                 }
6943
6944                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
6945                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
6946                         dfprintk(MOUNT, "  secinfo mode enabled\n");
6947                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
6948                 }
6949
6950                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
6951                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
6952                         dfprintk(MOUNT, "  stateid mode enabled\n");
6953                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
6954                 }
6955
6956                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
6957                         dfprintk(MOUNT, "  write mode enabled\n");
6958                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
6959                 }
6960
6961                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
6962                         dfprintk(MOUNT, "  commit mode enabled\n");
6963                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
6964                 }
6965         }
6966
6967         return 0;
6968 }
6969
6970 /*
6971  * _nfs4_proc_exchange_id()
6972  *
6973  * Wrapper for EXCHANGE_ID operation.
6974  */
6975 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
6976         u32 sp4_how)
6977 {
6978         nfs4_verifier verifier;
6979         struct nfs41_exchange_id_args args = {
6980                 .verifier = &verifier,
6981                 .client = clp,
6982 #ifdef CONFIG_NFS_V4_1_MIGRATION
6983                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6984                          EXCHGID4_FLAG_BIND_PRINC_STATEID |
6985                          EXCHGID4_FLAG_SUPP_MOVED_MIGR,
6986 #else
6987                 .flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
6988                          EXCHGID4_FLAG_BIND_PRINC_STATEID,
6989 #endif
6990         };
6991         struct nfs41_exchange_id_res res = {
6992                 0
6993         };
6994         int status;
6995         struct rpc_message msg = {
6996                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
6997                 .rpc_argp = &args,
6998                 .rpc_resp = &res,
6999                 .rpc_cred = cred,
7000         };
7001
7002         nfs4_init_boot_verifier(clp, &verifier);
7003
7004         status = nfs4_init_uniform_client_string(clp);
7005         if (status)
7006                 goto out;
7007
7008         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7009                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7010                 clp->cl_owner_id);
7011
7012         res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7013                                         GFP_NOFS);
7014         if (unlikely(res.server_owner == NULL)) {
7015                 status = -ENOMEM;
7016                 goto out;
7017         }
7018
7019         res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7020                                         GFP_NOFS);
7021         if (unlikely(res.server_scope == NULL)) {
7022                 status = -ENOMEM;
7023                 goto out_server_owner;
7024         }
7025
7026         res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7027         if (unlikely(res.impl_id == NULL)) {
7028                 status = -ENOMEM;
7029                 goto out_server_scope;
7030         }
7031
7032         switch (sp4_how) {
7033         case SP4_NONE:
7034                 args.state_protect.how = SP4_NONE;
7035                 break;
7036
7037         case SP4_MACH_CRED:
7038                 args.state_protect = nfs4_sp4_mach_cred_request;
7039                 break;
7040
7041         default:
7042                 /* unsupported! */
7043                 WARN_ON_ONCE(1);
7044                 status = -EINVAL;
7045                 goto out_impl_id;
7046         }
7047
7048         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7049         trace_nfs4_exchange_id(clp, status);
7050         if (status == 0)
7051                 status = nfs4_check_cl_exchange_flags(res.flags);
7052
7053         if (status == 0)
7054                 status = nfs4_sp4_select_mode(clp, &res.state_protect);
7055
7056         if (status == 0) {
7057                 clp->cl_clientid = res.clientid;
7058                 clp->cl_exchange_flags = res.flags;
7059                 /* Client ID is not confirmed */
7060                 if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7061                         clear_bit(NFS4_SESSION_ESTABLISHED,
7062                                         &clp->cl_session->session_state);
7063                         clp->cl_seqid = res.seqid;
7064                 }
7065
7066                 kfree(clp->cl_serverowner);
7067                 clp->cl_serverowner = res.server_owner;
7068                 res.server_owner = NULL;
7069
7070                 /* use the most recent implementation id */
7071                 kfree(clp->cl_implid);
7072                 clp->cl_implid = res.impl_id;
7073                 res.impl_id = NULL;
7074
7075                 if (clp->cl_serverscope != NULL &&
7076                     !nfs41_same_server_scope(clp->cl_serverscope,
7077                                              res.server_scope)) {
7078                         dprintk("%s: server_scope mismatch detected\n",
7079                                 __func__);
7080                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7081                         kfree(clp->cl_serverscope);
7082                         clp->cl_serverscope = NULL;
7083                 }
7084
7085                 if (clp->cl_serverscope == NULL) {
7086                         clp->cl_serverscope = res.server_scope;
7087                         res.server_scope = NULL;
7088                 }
7089         }
7090
7091 out_impl_id:
7092         kfree(res.impl_id);
7093 out_server_scope:
7094         kfree(res.server_scope);
7095 out_server_owner:
7096         kfree(res.server_owner);
7097 out:
7098         if (clp->cl_implid != NULL)
7099                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7100                         "domain: %s, name: %s, date: %llu,%u\n",
7101                         clp->cl_implid->domain, clp->cl_implid->name,
7102                         clp->cl_implid->date.seconds,
7103                         clp->cl_implid->date.nseconds);
7104         dprintk("NFS reply exchange_id: %d\n", status);
7105         return status;
7106 }
7107
7108 /*
7109  * nfs4_proc_exchange_id()
7110  *
7111  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7112  *
7113  * Since the clientid has expired, all compounds using sessions
7114  * associated with the stale clientid will be returning
7115  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7116  * be in some phase of session reset.
7117  *
7118  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7119  */
7120 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7121 {
7122         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7123         int status;
7124
7125         /* try SP4_MACH_CRED if krb5i/p */
7126         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7127             authflavor == RPC_AUTH_GSS_KRB5P) {
7128                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
7129                 if (!status)
7130                         return 0;
7131         }
7132
7133         /* try SP4_NONE */
7134         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
7135 }
7136
7137 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7138                 struct rpc_cred *cred)
7139 {
7140         struct rpc_message msg = {
7141                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7142                 .rpc_argp = clp,
7143                 .rpc_cred = cred,
7144         };
7145         int status;
7146
7147         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7148         trace_nfs4_destroy_clientid(clp, status);
7149         if (status)
7150                 dprintk("NFS: Got error %d from the server %s on "
7151                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7152         return status;
7153 }
7154
7155 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7156                 struct rpc_cred *cred)
7157 {
7158         unsigned int loop;
7159         int ret;
7160
7161         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7162                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7163                 switch (ret) {
7164                 case -NFS4ERR_DELAY:
7165                 case -NFS4ERR_CLIENTID_BUSY:
7166                         ssleep(1);
7167                         break;
7168                 default:
7169                         return ret;
7170                 }
7171         }
7172         return 0;
7173 }
7174
7175 int nfs4_destroy_clientid(struct nfs_client *clp)
7176 {
7177         struct rpc_cred *cred;
7178         int ret = 0;
7179
7180         if (clp->cl_mvops->minor_version < 1)
7181                 goto out;
7182         if (clp->cl_exchange_flags == 0)
7183                 goto out;
7184         if (clp->cl_preserve_clid)
7185                 goto out;
7186         cred = nfs4_get_clid_cred(clp);
7187         ret = nfs4_proc_destroy_clientid(clp, cred);
7188         if (cred)
7189                 put_rpccred(cred);
7190         switch (ret) {
7191         case 0:
7192         case -NFS4ERR_STALE_CLIENTID:
7193                 clp->cl_exchange_flags = 0;
7194         }
7195 out:
7196         return ret;
7197 }
7198
7199 struct nfs4_get_lease_time_data {
7200         struct nfs4_get_lease_time_args *args;
7201         struct nfs4_get_lease_time_res *res;
7202         struct nfs_client *clp;
7203 };
7204
7205 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7206                                         void *calldata)
7207 {
7208         struct nfs4_get_lease_time_data *data =
7209                         (struct nfs4_get_lease_time_data *)calldata;
7210
7211         dprintk("--> %s\n", __func__);
7212         /* just setup sequence, do not trigger session recovery
7213            since we're invoked within one */
7214         nfs41_setup_sequence(data->clp->cl_session,
7215                         &data->args->la_seq_args,
7216                         &data->res->lr_seq_res,
7217                         task);
7218         dprintk("<-- %s\n", __func__);
7219 }
7220
7221 /*
7222  * Called from nfs4_state_manager thread for session setup, so don't recover
7223  * from sequence operation or clientid errors.
7224  */
7225 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7226 {
7227         struct nfs4_get_lease_time_data *data =
7228                         (struct nfs4_get_lease_time_data *)calldata;
7229
7230         dprintk("--> %s\n", __func__);
7231         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7232                 return;
7233         switch (task->tk_status) {
7234         case -NFS4ERR_DELAY:
7235         case -NFS4ERR_GRACE:
7236                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7237                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7238                 task->tk_status = 0;
7239                 /* fall through */
7240         case -NFS4ERR_RETRY_UNCACHED_REP:
7241                 rpc_restart_call_prepare(task);
7242                 return;
7243         }
7244         dprintk("<-- %s\n", __func__);
7245 }
7246
7247 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7248         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7249         .rpc_call_done = nfs4_get_lease_time_done,
7250 };
7251
7252 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7253 {
7254         struct rpc_task *task;
7255         struct nfs4_get_lease_time_args args;
7256         struct nfs4_get_lease_time_res res = {
7257                 .lr_fsinfo = fsinfo,
7258         };
7259         struct nfs4_get_lease_time_data data = {
7260                 .args = &args,
7261                 .res = &res,
7262                 .clp = clp,
7263         };
7264         struct rpc_message msg = {
7265                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7266                 .rpc_argp = &args,
7267                 .rpc_resp = &res,
7268         };
7269         struct rpc_task_setup task_setup = {
7270                 .rpc_client = clp->cl_rpcclient,
7271                 .rpc_message = &msg,
7272                 .callback_ops = &nfs4_get_lease_time_ops,
7273                 .callback_data = &data,
7274                 .flags = RPC_TASK_TIMEOUT,
7275         };
7276         int status;
7277
7278         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7279         nfs4_set_sequence_privileged(&args.la_seq_args);
7280         dprintk("--> %s\n", __func__);
7281         task = rpc_run_task(&task_setup);
7282
7283         if (IS_ERR(task))
7284                 status = PTR_ERR(task);
7285         else {
7286                 status = task->tk_status;
7287                 rpc_put_task(task);
7288         }
7289         dprintk("<-- %s return %d\n", __func__, status);
7290
7291         return status;
7292 }
7293
7294 /*
7295  * Initialize the values to be used by the client in CREATE_SESSION
7296  * If nfs4_init_session set the fore channel request and response sizes,
7297  * use them.
7298  *
7299  * Set the back channel max_resp_sz_cached to zero to force the client to
7300  * always set csa_cachethis to FALSE because the current implementation
7301  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7302  */
7303 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args)
7304 {
7305         unsigned int max_rqst_sz, max_resp_sz;
7306
7307         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7308         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7309
7310         /* Fore channel attributes */
7311         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7312         args->fc_attrs.max_resp_sz = max_resp_sz;
7313         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7314         args->fc_attrs.max_reqs = max_session_slots;
7315
7316         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7317                 "max_ops=%u max_reqs=%u\n",
7318                 __func__,
7319                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7320                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7321
7322         /* Back channel attributes */
7323         args->bc_attrs.max_rqst_sz = PAGE_SIZE;
7324         args->bc_attrs.max_resp_sz = PAGE_SIZE;
7325         args->bc_attrs.max_resp_sz_cached = 0;
7326         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7327         args->bc_attrs.max_reqs = 1;
7328
7329         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7330                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7331                 __func__,
7332                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7333                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7334                 args->bc_attrs.max_reqs);
7335 }
7336
7337 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7338                 struct nfs41_create_session_res *res)
7339 {
7340         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7341         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7342
7343         if (rcvd->max_resp_sz > sent->max_resp_sz)
7344                 return -EINVAL;
7345         /*
7346          * Our requested max_ops is the minimum we need; we're not
7347          * prepared to break up compounds into smaller pieces than that.
7348          * So, no point even trying to continue if the server won't
7349          * cooperate:
7350          */
7351         if (rcvd->max_ops < sent->max_ops)
7352                 return -EINVAL;
7353         if (rcvd->max_reqs == 0)
7354                 return -EINVAL;
7355         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7356                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7357         return 0;
7358 }
7359
7360 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7361                 struct nfs41_create_session_res *res)
7362 {
7363         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7364         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7365
7366         if (!(res->flags & SESSION4_BACK_CHAN))
7367                 goto out;
7368         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7369                 return -EINVAL;
7370         if (rcvd->max_resp_sz < sent->max_resp_sz)
7371                 return -EINVAL;
7372         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7373                 return -EINVAL;
7374         /* These would render the backchannel useless: */
7375         if (rcvd->max_ops != sent->max_ops)
7376                 return -EINVAL;
7377         if (rcvd->max_reqs != sent->max_reqs)
7378                 return -EINVAL;
7379 out:
7380         return 0;
7381 }
7382
7383 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7384                                      struct nfs41_create_session_res *res)
7385 {
7386         int ret;
7387
7388         ret = nfs4_verify_fore_channel_attrs(args, res);
7389         if (ret)
7390                 return ret;
7391         return nfs4_verify_back_channel_attrs(args, res);
7392 }
7393
7394 static void nfs4_update_session(struct nfs4_session *session,
7395                 struct nfs41_create_session_res *res)
7396 {
7397         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7398         /* Mark client id and session as being confirmed */
7399         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7400         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7401         session->flags = res->flags;
7402         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7403         if (res->flags & SESSION4_BACK_CHAN)
7404                 memcpy(&session->bc_attrs, &res->bc_attrs,
7405                                 sizeof(session->bc_attrs));
7406 }
7407
7408 static int _nfs4_proc_create_session(struct nfs_client *clp,
7409                 struct rpc_cred *cred)
7410 {
7411         struct nfs4_session *session = clp->cl_session;
7412         struct nfs41_create_session_args args = {
7413                 .client = clp,
7414                 .clientid = clp->cl_clientid,
7415                 .seqid = clp->cl_seqid,
7416                 .cb_program = NFS4_CALLBACK,
7417         };
7418         struct nfs41_create_session_res res;
7419
7420         struct rpc_message msg = {
7421                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7422                 .rpc_argp = &args,
7423                 .rpc_resp = &res,
7424                 .rpc_cred = cred,
7425         };
7426         int status;
7427
7428         nfs4_init_channel_attrs(&args);
7429         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7430
7431         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7432         trace_nfs4_create_session(clp, status);
7433
7434         if (!status) {
7435                 /* Verify the session's negotiated channel_attrs values */
7436                 status = nfs4_verify_channel_attrs(&args, &res);
7437                 /* Increment the clientid slot sequence id */
7438                 if (clp->cl_seqid == res.seqid)
7439                         clp->cl_seqid++;
7440                 if (status)
7441                         goto out;
7442                 nfs4_update_session(session, &res);
7443         }
7444 out:
7445         return status;
7446 }
7447
7448 /*
7449  * Issues a CREATE_SESSION operation to the server.
7450  * It is the responsibility of the caller to verify the session is
7451  * expired before calling this routine.
7452  */
7453 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7454 {
7455         int status;
7456         unsigned *ptr;
7457         struct nfs4_session *session = clp->cl_session;
7458
7459         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7460
7461         status = _nfs4_proc_create_session(clp, cred);
7462         if (status)
7463                 goto out;
7464
7465         /* Init or reset the session slot tables */
7466         status = nfs4_setup_session_slot_tables(session);
7467         dprintk("slot table setup returned %d\n", status);
7468         if (status)
7469                 goto out;
7470
7471         ptr = (unsigned *)&session->sess_id.data[0];
7472         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7473                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7474 out:
7475         dprintk("<-- %s\n", __func__);
7476         return status;
7477 }
7478
7479 /*
7480  * Issue the over-the-wire RPC DESTROY_SESSION.
7481  * The caller must serialize access to this routine.
7482  */
7483 int nfs4_proc_destroy_session(struct nfs4_session *session,
7484                 struct rpc_cred *cred)
7485 {
7486         struct rpc_message msg = {
7487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
7488                 .rpc_argp = session,
7489                 .rpc_cred = cred,
7490         };
7491         int status = 0;
7492
7493         dprintk("--> nfs4_proc_destroy_session\n");
7494
7495         /* session is still being setup */
7496         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
7497                 return 0;
7498
7499         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7500         trace_nfs4_destroy_session(session->clp, status);
7501
7502         if (status)
7503                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
7504                         "Session has been destroyed regardless...\n", status);
7505
7506         dprintk("<-- nfs4_proc_destroy_session\n");
7507         return status;
7508 }
7509
7510 /*
7511  * Renew the cl_session lease.
7512  */
7513 struct nfs4_sequence_data {
7514         struct nfs_client *clp;
7515         struct nfs4_sequence_args args;
7516         struct nfs4_sequence_res res;
7517 };
7518
7519 static void nfs41_sequence_release(void *data)
7520 {
7521         struct nfs4_sequence_data *calldata = data;
7522         struct nfs_client *clp = calldata->clp;
7523
7524         if (atomic_read(&clp->cl_count) > 1)
7525                 nfs4_schedule_state_renewal(clp);
7526         nfs_put_client(clp);
7527         kfree(calldata);
7528 }
7529
7530 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7531 {
7532         switch(task->tk_status) {
7533         case -NFS4ERR_DELAY:
7534                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7535                 return -EAGAIN;
7536         default:
7537                 nfs4_schedule_lease_recovery(clp);
7538         }
7539         return 0;
7540 }
7541
7542 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
7543 {
7544         struct nfs4_sequence_data *calldata = data;
7545         struct nfs_client *clp = calldata->clp;
7546
7547         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
7548                 return;
7549
7550         trace_nfs4_sequence(clp, task->tk_status);
7551         if (task->tk_status < 0) {
7552                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
7553                 if (atomic_read(&clp->cl_count) == 1)
7554                         goto out;
7555
7556                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
7557                         rpc_restart_call_prepare(task);
7558                         return;
7559                 }
7560         }
7561         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
7562 out:
7563         dprintk("<-- %s\n", __func__);
7564 }
7565
7566 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
7567 {
7568         struct nfs4_sequence_data *calldata = data;
7569         struct nfs_client *clp = calldata->clp;
7570         struct nfs4_sequence_args *args;
7571         struct nfs4_sequence_res *res;
7572
7573         args = task->tk_msg.rpc_argp;
7574         res = task->tk_msg.rpc_resp;
7575
7576         nfs41_setup_sequence(clp->cl_session, args, res, task);
7577 }
7578
7579 static const struct rpc_call_ops nfs41_sequence_ops = {
7580         .rpc_call_done = nfs41_sequence_call_done,
7581         .rpc_call_prepare = nfs41_sequence_prepare,
7582         .rpc_release = nfs41_sequence_release,
7583 };
7584
7585 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
7586                 struct rpc_cred *cred,
7587                 bool is_privileged)
7588 {
7589         struct nfs4_sequence_data *calldata;
7590         struct rpc_message msg = {
7591                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
7592                 .rpc_cred = cred,
7593         };
7594         struct rpc_task_setup task_setup_data = {
7595                 .rpc_client = clp->cl_rpcclient,
7596                 .rpc_message = &msg,
7597                 .callback_ops = &nfs41_sequence_ops,
7598                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7599         };
7600
7601         if (!atomic_inc_not_zero(&clp->cl_count))
7602                 return ERR_PTR(-EIO);
7603         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7604         if (calldata == NULL) {
7605                 nfs_put_client(clp);
7606                 return ERR_PTR(-ENOMEM);
7607         }
7608         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
7609         if (is_privileged)
7610                 nfs4_set_sequence_privileged(&calldata->args);
7611         msg.rpc_argp = &calldata->args;
7612         msg.rpc_resp = &calldata->res;
7613         calldata->clp = clp;
7614         task_setup_data.callback_data = calldata;
7615
7616         return rpc_run_task(&task_setup_data);
7617 }
7618
7619 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
7620 {
7621         struct rpc_task *task;
7622         int ret = 0;
7623
7624         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
7625                 return -EAGAIN;
7626         task = _nfs41_proc_sequence(clp, cred, false);
7627         if (IS_ERR(task))
7628                 ret = PTR_ERR(task);
7629         else
7630                 rpc_put_task_async(task);
7631         dprintk("<-- %s status=%d\n", __func__, ret);
7632         return ret;
7633 }
7634
7635 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
7636 {
7637         struct rpc_task *task;
7638         int ret;
7639
7640         task = _nfs41_proc_sequence(clp, cred, true);
7641         if (IS_ERR(task)) {
7642                 ret = PTR_ERR(task);
7643                 goto out;
7644         }
7645         ret = rpc_wait_for_completion_task(task);
7646         if (!ret)
7647                 ret = task->tk_status;
7648         rpc_put_task(task);
7649 out:
7650         dprintk("<-- %s status=%d\n", __func__, ret);
7651         return ret;
7652 }
7653
7654 struct nfs4_reclaim_complete_data {
7655         struct nfs_client *clp;
7656         struct nfs41_reclaim_complete_args arg;
7657         struct nfs41_reclaim_complete_res res;
7658 };
7659
7660 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
7661 {
7662         struct nfs4_reclaim_complete_data *calldata = data;
7663
7664         nfs41_setup_sequence(calldata->clp->cl_session,
7665                         &calldata->arg.seq_args,
7666                         &calldata->res.seq_res,
7667                         task);
7668 }
7669
7670 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
7671 {
7672         switch(task->tk_status) {
7673         case 0:
7674         case -NFS4ERR_COMPLETE_ALREADY:
7675         case -NFS4ERR_WRONG_CRED: /* What to do here? */
7676                 break;
7677         case -NFS4ERR_DELAY:
7678                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
7679                 /* fall through */
7680         case -NFS4ERR_RETRY_UNCACHED_REP:
7681                 return -EAGAIN;
7682         default:
7683                 nfs4_schedule_lease_recovery(clp);
7684         }
7685         return 0;
7686 }
7687
7688 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
7689 {
7690         struct nfs4_reclaim_complete_data *calldata = data;
7691         struct nfs_client *clp = calldata->clp;
7692         struct nfs4_sequence_res *res = &calldata->res.seq_res;
7693
7694         dprintk("--> %s\n", __func__);
7695         if (!nfs41_sequence_done(task, res))
7696                 return;
7697
7698         trace_nfs4_reclaim_complete(clp, task->tk_status);
7699         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
7700                 rpc_restart_call_prepare(task);
7701                 return;
7702         }
7703         dprintk("<-- %s\n", __func__);
7704 }
7705
7706 static void nfs4_free_reclaim_complete_data(void *data)
7707 {
7708         struct nfs4_reclaim_complete_data *calldata = data;
7709
7710         kfree(calldata);
7711 }
7712
7713 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
7714         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
7715         .rpc_call_done = nfs4_reclaim_complete_done,
7716         .rpc_release = nfs4_free_reclaim_complete_data,
7717 };
7718
7719 /*
7720  * Issue a global reclaim complete.
7721  */
7722 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
7723                 struct rpc_cred *cred)
7724 {
7725         struct nfs4_reclaim_complete_data *calldata;
7726         struct rpc_task *task;
7727         struct rpc_message msg = {
7728                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
7729                 .rpc_cred = cred,
7730         };
7731         struct rpc_task_setup task_setup_data = {
7732                 .rpc_client = clp->cl_rpcclient,
7733                 .rpc_message = &msg,
7734                 .callback_ops = &nfs4_reclaim_complete_call_ops,
7735                 .flags = RPC_TASK_ASYNC,
7736         };
7737         int status = -ENOMEM;
7738
7739         dprintk("--> %s\n", __func__);
7740         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7741         if (calldata == NULL)
7742                 goto out;
7743         calldata->clp = clp;
7744         calldata->arg.one_fs = 0;
7745
7746         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
7747         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
7748         msg.rpc_argp = &calldata->arg;
7749         msg.rpc_resp = &calldata->res;
7750         task_setup_data.callback_data = calldata;
7751         task = rpc_run_task(&task_setup_data);
7752         if (IS_ERR(task)) {
7753                 status = PTR_ERR(task);
7754                 goto out;
7755         }
7756         status = nfs4_wait_for_completion_rpc_task(task);
7757         if (status == 0)
7758                 status = task->tk_status;
7759         rpc_put_task(task);
7760         return 0;
7761 out:
7762         dprintk("<-- %s status=%d\n", __func__, status);
7763         return status;
7764 }
7765
7766 static void
7767 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
7768 {
7769         struct nfs4_layoutget *lgp = calldata;
7770         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
7771         struct nfs4_session *session = nfs4_get_session(server);
7772
7773         dprintk("--> %s\n", __func__);
7774         /* Note the is a race here, where a CB_LAYOUTRECALL can come in
7775          * right now covering the LAYOUTGET we are about to send.
7776          * However, that is not so catastrophic, and there seems
7777          * to be no way to prevent it completely.
7778          */
7779         if (nfs41_setup_sequence(session, &lgp->args.seq_args,
7780                                 &lgp->res.seq_res, task))
7781                 return;
7782         if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
7783                                           NFS_I(lgp->args.inode)->layout,
7784                                           &lgp->args.range,
7785                                           lgp->args.ctx->state)) {
7786                 rpc_exit(task, NFS4_OK);
7787         }
7788 }
7789
7790 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
7791 {
7792         struct nfs4_layoutget *lgp = calldata;
7793         struct inode *inode = lgp->args.inode;
7794         struct nfs_server *server = NFS_SERVER(inode);
7795         struct pnfs_layout_hdr *lo;
7796         struct nfs4_state *state = NULL;
7797         unsigned long timeo, now, giveup;
7798
7799         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
7800
7801         if (!nfs41_sequence_done(task, &lgp->res.seq_res))
7802                 goto out;
7803
7804         switch (task->tk_status) {
7805         case 0:
7806                 goto out;
7807         /*
7808          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
7809          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
7810          */
7811         case -NFS4ERR_BADLAYOUT:
7812                 goto out_overflow;
7813         /*
7814          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
7815          * (or clients) writing to the same RAID stripe except when
7816          * the minlength argument is 0 (see RFC5661 section 18.43.3).
7817          */
7818         case -NFS4ERR_LAYOUTTRYLATER:
7819                 if (lgp->args.minlength == 0)
7820                         goto out_overflow;
7821         /*
7822          * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall
7823          * existing layout before getting a new one).
7824          */
7825         case -NFS4ERR_RECALLCONFLICT:
7826                 timeo = rpc_get_timeout(task->tk_client);
7827                 giveup = lgp->args.timestamp + timeo;
7828                 now = jiffies;
7829                 if (time_after(giveup, now)) {
7830                         unsigned long delay;
7831
7832                         /* Delay for:
7833                          * - Not less then NFS4_POLL_RETRY_MIN.
7834                          * - One last time a jiffie before we give up
7835                          * - exponential backoff (time_now minus start_attempt)
7836                          */
7837                         delay = max_t(unsigned long, NFS4_POLL_RETRY_MIN,
7838                                     min((giveup - now - 1),
7839                                         now - lgp->args.timestamp));
7840
7841                         dprintk("%s: NFS4ERR_RECALLCONFLICT waiting %lu\n",
7842                                 __func__, delay);
7843                         rpc_delay(task, delay);
7844                         /* Do not call nfs4_async_handle_error() */
7845                         goto out_restart;
7846                 }
7847                 break;
7848         case -NFS4ERR_EXPIRED:
7849         case -NFS4ERR_BAD_STATEID:
7850                 spin_lock(&inode->i_lock);
7851                 if (nfs4_stateid_match(&lgp->args.stateid,
7852                                         &lgp->args.ctx->state->stateid)) {
7853                         spin_unlock(&inode->i_lock);
7854                         /* If the open stateid was bad, then recover it. */
7855                         state = lgp->args.ctx->state;
7856                         break;
7857                 }
7858                 lo = NFS_I(inode)->layout;
7859                 if (lo && nfs4_stateid_match(&lgp->args.stateid,
7860                                         &lo->plh_stateid)) {
7861                         LIST_HEAD(head);
7862
7863                         /*
7864                          * Mark the bad layout state as invalid, then retry
7865                          * with the current stateid.
7866                          */
7867                         set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
7868                         pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
7869                         spin_unlock(&inode->i_lock);
7870                         pnfs_free_lseg_list(&head);
7871                 } else
7872                         spin_unlock(&inode->i_lock);
7873                 goto out_restart;
7874         }
7875         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
7876                 goto out_restart;
7877 out:
7878         dprintk("<-- %s\n", __func__);
7879         return;
7880 out_restart:
7881         task->tk_status = 0;
7882         rpc_restart_call_prepare(task);
7883         return;
7884 out_overflow:
7885         task->tk_status = -EOVERFLOW;
7886         goto out;
7887 }
7888
7889 static size_t max_response_pages(struct nfs_server *server)
7890 {
7891         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
7892         return nfs_page_array_len(0, max_resp_sz);
7893 }
7894
7895 static void nfs4_free_pages(struct page **pages, size_t size)
7896 {
7897         int i;
7898
7899         if (!pages)
7900                 return;
7901
7902         for (i = 0; i < size; i++) {
7903                 if (!pages[i])
7904                         break;
7905                 __free_page(pages[i]);
7906         }
7907         kfree(pages);
7908 }
7909
7910 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
7911 {
7912         struct page **pages;
7913         int i;
7914
7915         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
7916         if (!pages) {
7917                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
7918                 return NULL;
7919         }
7920
7921         for (i = 0; i < size; i++) {
7922                 pages[i] = alloc_page(gfp_flags);
7923                 if (!pages[i]) {
7924                         dprintk("%s: failed to allocate page\n", __func__);
7925                         nfs4_free_pages(pages, size);
7926                         return NULL;
7927                 }
7928         }
7929
7930         return pages;
7931 }
7932
7933 static void nfs4_layoutget_release(void *calldata)
7934 {
7935         struct nfs4_layoutget *lgp = calldata;
7936         struct inode *inode = lgp->args.inode;
7937         struct nfs_server *server = NFS_SERVER(inode);
7938         size_t max_pages = max_response_pages(server);
7939
7940         dprintk("--> %s\n", __func__);
7941         nfs4_free_pages(lgp->args.layout.pages, max_pages);
7942         pnfs_put_layout_hdr(NFS_I(inode)->layout);
7943         put_nfs_open_context(lgp->args.ctx);
7944         kfree(calldata);
7945         dprintk("<-- %s\n", __func__);
7946 }
7947
7948 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
7949         .rpc_call_prepare = nfs4_layoutget_prepare,
7950         .rpc_call_done = nfs4_layoutget_done,
7951         .rpc_release = nfs4_layoutget_release,
7952 };
7953
7954 struct pnfs_layout_segment *
7955 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
7956 {
7957         struct inode *inode = lgp->args.inode;
7958         struct nfs_server *server = NFS_SERVER(inode);
7959         size_t max_pages = max_response_pages(server);
7960         struct rpc_task *task;
7961         struct rpc_message msg = {
7962                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
7963                 .rpc_argp = &lgp->args,
7964                 .rpc_resp = &lgp->res,
7965                 .rpc_cred = lgp->cred,
7966         };
7967         struct rpc_task_setup task_setup_data = {
7968                 .rpc_client = server->client,
7969                 .rpc_message = &msg,
7970                 .callback_ops = &nfs4_layoutget_call_ops,
7971                 .callback_data = lgp,
7972                 .flags = RPC_TASK_ASYNC,
7973         };
7974         struct pnfs_layout_segment *lseg = NULL;
7975         int status = 0;
7976
7977         dprintk("--> %s\n", __func__);
7978
7979         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
7980         pnfs_get_layout_hdr(NFS_I(inode)->layout);
7981
7982         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
7983         if (!lgp->args.layout.pages) {
7984                 nfs4_layoutget_release(lgp);
7985                 return ERR_PTR(-ENOMEM);
7986         }
7987         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
7988         lgp->args.timestamp = jiffies;
7989
7990         lgp->res.layoutp = &lgp->args.layout;
7991         lgp->res.seq_res.sr_slot = NULL;
7992         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
7993
7994         task = rpc_run_task(&task_setup_data);
7995         if (IS_ERR(task))
7996                 return ERR_CAST(task);
7997         status = nfs4_wait_for_completion_rpc_task(task);
7998         if (status == 0)
7999                 status = task->tk_status;
8000         trace_nfs4_layoutget(lgp->args.ctx,
8001                         &lgp->args.range,
8002                         &lgp->res.range,
8003                         status);
8004         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8005         if (status == 0 && lgp->res.layoutp->len)
8006                 lseg = pnfs_layout_process(lgp);
8007         rpc_put_task(task);
8008         dprintk("<-- %s status=%d\n", __func__, status);
8009         if (status)
8010                 return ERR_PTR(status);
8011         return lseg;
8012 }
8013
8014 static void
8015 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8016 {
8017         struct nfs4_layoutreturn *lrp = calldata;
8018
8019         dprintk("--> %s\n", __func__);
8020         nfs41_setup_sequence(lrp->clp->cl_session,
8021                         &lrp->args.seq_args,
8022                         &lrp->res.seq_res,
8023                         task);
8024 }
8025
8026 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8027 {
8028         struct nfs4_layoutreturn *lrp = calldata;
8029         struct nfs_server *server;
8030
8031         dprintk("--> %s\n", __func__);
8032
8033         if (!nfs41_sequence_done(task, &lrp->res.seq_res))
8034                 return;
8035
8036         server = NFS_SERVER(lrp->args.inode);
8037         switch (task->tk_status) {
8038         default:
8039                 task->tk_status = 0;
8040         case 0:
8041                 break;
8042         case -NFS4ERR_DELAY:
8043                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8044                         break;
8045                 rpc_restart_call_prepare(task);
8046                 return;
8047         }
8048         dprintk("<-- %s\n", __func__);
8049 }
8050
8051 static void nfs4_layoutreturn_release(void *calldata)
8052 {
8053         struct nfs4_layoutreturn *lrp = calldata;
8054         struct pnfs_layout_hdr *lo = lrp->args.layout;
8055         LIST_HEAD(freeme);
8056
8057         dprintk("--> %s\n", __func__);
8058         spin_lock(&lo->plh_inode->i_lock);
8059         if (lrp->res.lrs_present)
8060                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8061         pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
8062         pnfs_clear_layoutreturn_waitbit(lo);
8063         lo->plh_block_lgets--;
8064         spin_unlock(&lo->plh_inode->i_lock);
8065         pnfs_free_lseg_list(&freeme);
8066         pnfs_put_layout_hdr(lrp->args.layout);
8067         nfs_iput_and_deactive(lrp->inode);
8068         kfree(calldata);
8069         dprintk("<-- %s\n", __func__);
8070 }
8071
8072 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8073         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8074         .rpc_call_done = nfs4_layoutreturn_done,
8075         .rpc_release = nfs4_layoutreturn_release,
8076 };
8077
8078 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8079 {
8080         struct rpc_task *task;
8081         struct rpc_message msg = {
8082                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8083                 .rpc_argp = &lrp->args,
8084                 .rpc_resp = &lrp->res,
8085                 .rpc_cred = lrp->cred,
8086         };
8087         struct rpc_task_setup task_setup_data = {
8088                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8089                 .rpc_message = &msg,
8090                 .callback_ops = &nfs4_layoutreturn_call_ops,
8091                 .callback_data = lrp,
8092         };
8093         int status = 0;
8094
8095         dprintk("--> %s\n", __func__);
8096         if (!sync) {
8097                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8098                 if (!lrp->inode) {
8099                         nfs4_layoutreturn_release(lrp);
8100                         return -EAGAIN;
8101                 }
8102                 task_setup_data.flags |= RPC_TASK_ASYNC;
8103         }
8104         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8105         task = rpc_run_task(&task_setup_data);
8106         if (IS_ERR(task))
8107                 return PTR_ERR(task);
8108         if (sync)
8109                 status = task->tk_status;
8110         trace_nfs4_layoutreturn(lrp->args.inode, status);
8111         dprintk("<-- %s status=%d\n", __func__, status);
8112         rpc_put_task(task);
8113         return status;
8114 }
8115
8116 static int
8117 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8118                 struct pnfs_device *pdev,
8119                 struct rpc_cred *cred)
8120 {
8121         struct nfs4_getdeviceinfo_args args = {
8122                 .pdev = pdev,
8123                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8124                         NOTIFY_DEVICEID4_DELETE,
8125         };
8126         struct nfs4_getdeviceinfo_res res = {
8127                 .pdev = pdev,
8128         };
8129         struct rpc_message msg = {
8130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8131                 .rpc_argp = &args,
8132                 .rpc_resp = &res,
8133                 .rpc_cred = cred,
8134         };
8135         int status;
8136
8137         dprintk("--> %s\n", __func__);
8138         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8139         if (res.notification & ~args.notify_types)
8140                 dprintk("%s: unsupported notification\n", __func__);
8141         if (res.notification != args.notify_types)
8142                 pdev->nocache = 1;
8143
8144         dprintk("<-- %s status=%d\n", __func__, status);
8145
8146         return status;
8147 }
8148
8149 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8150                 struct pnfs_device *pdev,
8151                 struct rpc_cred *cred)
8152 {
8153         struct nfs4_exception exception = { };
8154         int err;
8155
8156         do {
8157                 err = nfs4_handle_exception(server,
8158                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8159                                         &exception);
8160         } while (exception.retry);
8161         return err;
8162 }
8163 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8164
8165 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8166 {
8167         struct nfs4_layoutcommit_data *data = calldata;
8168         struct nfs_server *server = NFS_SERVER(data->args.inode);
8169         struct nfs4_session *session = nfs4_get_session(server);
8170
8171         nfs41_setup_sequence(session,
8172                         &data->args.seq_args,
8173                         &data->res.seq_res,
8174                         task);
8175 }
8176
8177 static void
8178 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8179 {
8180         struct nfs4_layoutcommit_data *data = calldata;
8181         struct nfs_server *server = NFS_SERVER(data->args.inode);
8182
8183         if (!nfs41_sequence_done(task, &data->res.seq_res))
8184                 return;
8185
8186         switch (task->tk_status) { /* Just ignore these failures */
8187         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8188         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8189         case -NFS4ERR_BADLAYOUT:     /* no layout */
8190         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8191                 task->tk_status = 0;
8192         case 0:
8193                 break;
8194         default:
8195                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8196                         rpc_restart_call_prepare(task);
8197                         return;
8198                 }
8199         }
8200 }
8201
8202 static void nfs4_layoutcommit_release(void *calldata)
8203 {
8204         struct nfs4_layoutcommit_data *data = calldata;
8205
8206         pnfs_cleanup_layoutcommit(data);
8207         nfs_post_op_update_inode_force_wcc(data->args.inode,
8208                                            data->res.fattr);
8209         put_rpccred(data->cred);
8210         nfs_iput_and_deactive(data->inode);
8211         kfree(data);
8212 }
8213
8214 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8215         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8216         .rpc_call_done = nfs4_layoutcommit_done,
8217         .rpc_release = nfs4_layoutcommit_release,
8218 };
8219
8220 int
8221 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8222 {
8223         struct rpc_message msg = {
8224                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8225                 .rpc_argp = &data->args,
8226                 .rpc_resp = &data->res,
8227                 .rpc_cred = data->cred,
8228         };
8229         struct rpc_task_setup task_setup_data = {
8230                 .task = &data->task,
8231                 .rpc_client = NFS_CLIENT(data->args.inode),
8232                 .rpc_message = &msg,
8233                 .callback_ops = &nfs4_layoutcommit_ops,
8234                 .callback_data = data,
8235         };
8236         struct rpc_task *task;
8237         int status = 0;
8238
8239         dprintk("NFS: initiating layoutcommit call. sync %d "
8240                 "lbw: %llu inode %lu\n", sync,
8241                 data->args.lastbytewritten,
8242                 data->args.inode->i_ino);
8243
8244         if (!sync) {
8245                 data->inode = nfs_igrab_and_active(data->args.inode);
8246                 if (data->inode == NULL) {
8247                         nfs4_layoutcommit_release(data);
8248                         return -EAGAIN;
8249                 }
8250                 task_setup_data.flags = RPC_TASK_ASYNC;
8251         }
8252         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8253         task = rpc_run_task(&task_setup_data);
8254         if (IS_ERR(task))
8255                 return PTR_ERR(task);
8256         if (sync)
8257                 status = task->tk_status;
8258         trace_nfs4_layoutcommit(data->args.inode, status);
8259         dprintk("%s: status %d\n", __func__, status);
8260         rpc_put_task(task);
8261         return status;
8262 }
8263
8264 /**
8265  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8266  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8267  */
8268 static int
8269 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8270                     struct nfs_fsinfo *info,
8271                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8272 {
8273         struct nfs41_secinfo_no_name_args args = {
8274                 .style = SECINFO_STYLE_CURRENT_FH,
8275         };
8276         struct nfs4_secinfo_res res = {
8277                 .flavors = flavors,
8278         };
8279         struct rpc_message msg = {
8280                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8281                 .rpc_argp = &args,
8282                 .rpc_resp = &res,
8283         };
8284         struct rpc_clnt *clnt = server->client;
8285         struct rpc_cred *cred = NULL;
8286         int status;
8287
8288         if (use_integrity) {
8289                 clnt = server->nfs_client->cl_rpcclient;
8290                 cred = nfs4_get_clid_cred(server->nfs_client);
8291                 msg.rpc_cred = cred;
8292         }
8293
8294         dprintk("--> %s\n", __func__);
8295         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8296                                 &res.seq_res, 0);
8297         dprintk("<-- %s status=%d\n", __func__, status);
8298
8299         if (cred)
8300                 put_rpccred(cred);
8301
8302         return status;
8303 }
8304
8305 static int
8306 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8307                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8308 {
8309         struct nfs4_exception exception = { };
8310         int err;
8311         do {
8312                 /* first try using integrity protection */
8313                 err = -NFS4ERR_WRONGSEC;
8314
8315                 /* try to use integrity protection with machine cred */
8316                 if (_nfs4_is_integrity_protected(server->nfs_client))
8317                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8318                                                           flavors, true);
8319
8320                 /*
8321                  * if unable to use integrity protection, or SECINFO with
8322                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8323                  * disallowed by spec, but exists in deployed servers) use
8324                  * the current filesystem's rpc_client and the user cred.
8325                  */
8326                 if (err == -NFS4ERR_WRONGSEC)
8327                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8328                                                           flavors, false);
8329
8330                 switch (err) {
8331                 case 0:
8332                 case -NFS4ERR_WRONGSEC:
8333                 case -ENOTSUPP:
8334                         goto out;
8335                 default:
8336                         err = nfs4_handle_exception(server, err, &exception);
8337                 }
8338         } while (exception.retry);
8339 out:
8340         return err;
8341 }
8342
8343 static int
8344 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8345                     struct nfs_fsinfo *info)
8346 {
8347         int err;
8348         struct page *page;
8349         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8350         struct nfs4_secinfo_flavors *flavors;
8351         struct nfs4_secinfo4 *secinfo;
8352         int i;
8353
8354         page = alloc_page(GFP_KERNEL);
8355         if (!page) {
8356                 err = -ENOMEM;
8357                 goto out;
8358         }
8359
8360         flavors = page_address(page);
8361         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8362
8363         /*
8364          * Fall back on "guess and check" method if
8365          * the server doesn't support SECINFO_NO_NAME
8366          */
8367         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8368                 err = nfs4_find_root_sec(server, fhandle, info);
8369                 goto out_freepage;
8370         }
8371         if (err)
8372                 goto out_freepage;
8373
8374         for (i = 0; i < flavors->num_flavors; i++) {
8375                 secinfo = &flavors->flavors[i];
8376
8377                 switch (secinfo->flavor) {
8378                 case RPC_AUTH_NULL:
8379                 case RPC_AUTH_UNIX:
8380                 case RPC_AUTH_GSS:
8381                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8382                                         &secinfo->flavor_info);
8383                         break;
8384                 default:
8385                         flavor = RPC_AUTH_MAXFLAVOR;
8386                         break;
8387                 }
8388
8389                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8390                         flavor = RPC_AUTH_MAXFLAVOR;
8391
8392                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8393                         err = nfs4_lookup_root_sec(server, fhandle,
8394                                                    info, flavor);
8395                         if (!err)
8396                                 break;
8397                 }
8398         }
8399
8400         if (flavor == RPC_AUTH_MAXFLAVOR)
8401                 err = -EPERM;
8402
8403 out_freepage:
8404         put_page(page);
8405         if (err == -EACCES)
8406                 return -EPERM;
8407 out:
8408         return err;
8409 }
8410
8411 static int _nfs41_test_stateid(struct nfs_server *server,
8412                 nfs4_stateid *stateid,
8413                 struct rpc_cred *cred)
8414 {
8415         int status;
8416         struct nfs41_test_stateid_args args = {
8417                 .stateid = stateid,
8418         };
8419         struct nfs41_test_stateid_res res;
8420         struct rpc_message msg = {
8421                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8422                 .rpc_argp = &args,
8423                 .rpc_resp = &res,
8424                 .rpc_cred = cred,
8425         };
8426         struct rpc_clnt *rpc_client = server->client;
8427
8428         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8429                 &rpc_client, &msg);
8430
8431         dprintk("NFS call  test_stateid %p\n", stateid);
8432         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8433         nfs4_set_sequence_privileged(&args.seq_args);
8434         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8435                         &args.seq_args, &res.seq_res);
8436         if (status != NFS_OK) {
8437                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8438                 return status;
8439         }
8440         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8441         return -res.status;
8442 }
8443
8444 /**
8445  * nfs41_test_stateid - perform a TEST_STATEID operation
8446  *
8447  * @server: server / transport on which to perform the operation
8448  * @stateid: state ID to test
8449  * @cred: credential
8450  *
8451  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8452  * Otherwise a negative NFS4ERR value is returned if the operation
8453  * failed or the state ID is not currently valid.
8454  */
8455 static int nfs41_test_stateid(struct nfs_server *server,
8456                 nfs4_stateid *stateid,
8457                 struct rpc_cred *cred)
8458 {
8459         struct nfs4_exception exception = { };
8460         int err;
8461         do {
8462                 err = _nfs41_test_stateid(server, stateid, cred);
8463                 if (err != -NFS4ERR_DELAY)
8464                         break;
8465                 nfs4_handle_exception(server, err, &exception);
8466         } while (exception.retry);
8467         return err;
8468 }
8469
8470 struct nfs_free_stateid_data {
8471         struct nfs_server *server;
8472         struct nfs41_free_stateid_args args;
8473         struct nfs41_free_stateid_res res;
8474 };
8475
8476 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
8477 {
8478         struct nfs_free_stateid_data *data = calldata;
8479         nfs41_setup_sequence(nfs4_get_session(data->server),
8480                         &data->args.seq_args,
8481                         &data->res.seq_res,
8482                         task);
8483 }
8484
8485 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
8486 {
8487         struct nfs_free_stateid_data *data = calldata;
8488
8489         nfs41_sequence_done(task, &data->res.seq_res);
8490
8491         switch (task->tk_status) {
8492         case -NFS4ERR_DELAY:
8493                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
8494                         rpc_restart_call_prepare(task);
8495         }
8496 }
8497
8498 static void nfs41_free_stateid_release(void *calldata)
8499 {
8500         kfree(calldata);
8501 }
8502
8503 static const struct rpc_call_ops nfs41_free_stateid_ops = {
8504         .rpc_call_prepare = nfs41_free_stateid_prepare,
8505         .rpc_call_done = nfs41_free_stateid_done,
8506         .rpc_release = nfs41_free_stateid_release,
8507 };
8508
8509 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
8510                 nfs4_stateid *stateid,
8511                 struct rpc_cred *cred,
8512                 bool privileged)
8513 {
8514         struct rpc_message msg = {
8515                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
8516                 .rpc_cred = cred,
8517         };
8518         struct rpc_task_setup task_setup = {
8519                 .rpc_client = server->client,
8520                 .rpc_message = &msg,
8521                 .callback_ops = &nfs41_free_stateid_ops,
8522                 .flags = RPC_TASK_ASYNC,
8523         };
8524         struct nfs_free_stateid_data *data;
8525
8526         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8527                 &task_setup.rpc_client, &msg);
8528
8529         dprintk("NFS call  free_stateid %p\n", stateid);
8530         data = kmalloc(sizeof(*data), GFP_NOFS);
8531         if (!data)
8532                 return ERR_PTR(-ENOMEM);
8533         data->server = server;
8534         nfs4_stateid_copy(&data->args.stateid, stateid);
8535
8536         task_setup.callback_data = data;
8537
8538         msg.rpc_argp = &data->args;
8539         msg.rpc_resp = &data->res;
8540         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
8541         if (privileged)
8542                 nfs4_set_sequence_privileged(&data->args.seq_args);
8543
8544         return rpc_run_task(&task_setup);
8545 }
8546
8547 /**
8548  * nfs41_free_stateid - perform a FREE_STATEID operation
8549  *
8550  * @server: server / transport on which to perform the operation
8551  * @stateid: state ID to release
8552  * @cred: credential
8553  *
8554  * Returns NFS_OK if the server freed "stateid".  Otherwise a
8555  * negative NFS4ERR value is returned.
8556  */
8557 static int nfs41_free_stateid(struct nfs_server *server,
8558                 nfs4_stateid *stateid,
8559                 struct rpc_cred *cred)
8560 {
8561         struct rpc_task *task;
8562         int ret;
8563
8564         task = _nfs41_free_stateid(server, stateid, cred, true);
8565         if (IS_ERR(task))
8566                 return PTR_ERR(task);
8567         ret = rpc_wait_for_completion_task(task);
8568         if (!ret)
8569                 ret = task->tk_status;
8570         rpc_put_task(task);
8571         return ret;
8572 }
8573
8574 static void
8575 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
8576 {
8577         struct rpc_task *task;
8578         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
8579
8580         task = _nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
8581         nfs4_free_lock_state(server, lsp);
8582         if (IS_ERR(task))
8583                 return;
8584         rpc_put_task(task);
8585 }
8586
8587 static bool nfs41_match_stateid(const nfs4_stateid *s1,
8588                 const nfs4_stateid *s2)
8589 {
8590         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
8591                 return false;
8592
8593         if (s1->seqid == s2->seqid)
8594                 return true;
8595         if (s1->seqid == 0 || s2->seqid == 0)
8596                 return true;
8597
8598         return false;
8599 }
8600
8601 #endif /* CONFIG_NFS_V4_1 */
8602
8603 static bool nfs4_match_stateid(const nfs4_stateid *s1,
8604                 const nfs4_stateid *s2)
8605 {
8606         return nfs4_stateid_match(s1, s2);
8607 }
8608
8609
8610 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
8611         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8612         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8613         .recover_open   = nfs4_open_reclaim,
8614         .recover_lock   = nfs4_lock_reclaim,
8615         .establish_clid = nfs4_init_clientid,
8616         .detect_trunking = nfs40_discover_server_trunking,
8617 };
8618
8619 #if defined(CONFIG_NFS_V4_1)
8620 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
8621         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
8622         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
8623         .recover_open   = nfs4_open_reclaim,
8624         .recover_lock   = nfs4_lock_reclaim,
8625         .establish_clid = nfs41_init_clientid,
8626         .reclaim_complete = nfs41_proc_reclaim_complete,
8627         .detect_trunking = nfs41_discover_server_trunking,
8628 };
8629 #endif /* CONFIG_NFS_V4_1 */
8630
8631 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
8632         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8633         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8634         .recover_open   = nfs40_open_expired,
8635         .recover_lock   = nfs4_lock_expired,
8636         .establish_clid = nfs4_init_clientid,
8637 };
8638
8639 #if defined(CONFIG_NFS_V4_1)
8640 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
8641         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
8642         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
8643         .recover_open   = nfs41_open_expired,
8644         .recover_lock   = nfs41_lock_expired,
8645         .establish_clid = nfs41_init_clientid,
8646 };
8647 #endif /* CONFIG_NFS_V4_1 */
8648
8649 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
8650         .sched_state_renewal = nfs4_proc_async_renew,
8651         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
8652         .renew_lease = nfs4_proc_renew,
8653 };
8654
8655 #if defined(CONFIG_NFS_V4_1)
8656 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
8657         .sched_state_renewal = nfs41_proc_async_sequence,
8658         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
8659         .renew_lease = nfs4_proc_sequence,
8660 };
8661 #endif
8662
8663 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
8664         .get_locations = _nfs40_proc_get_locations,
8665         .fsid_present = _nfs40_proc_fsid_present,
8666 };
8667
8668 #if defined(CONFIG_NFS_V4_1)
8669 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
8670         .get_locations = _nfs41_proc_get_locations,
8671         .fsid_present = _nfs41_proc_fsid_present,
8672 };
8673 #endif  /* CONFIG_NFS_V4_1 */
8674
8675 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
8676         .minor_version = 0,
8677         .init_caps = NFS_CAP_READDIRPLUS
8678                 | NFS_CAP_ATOMIC_OPEN
8679                 | NFS_CAP_POSIX_LOCK,
8680         .init_client = nfs40_init_client,
8681         .shutdown_client = nfs40_shutdown_client,
8682         .match_stateid = nfs4_match_stateid,
8683         .find_root_sec = nfs4_find_root_sec,
8684         .free_lock_state = nfs4_release_lockowner,
8685         .alloc_seqid = nfs_alloc_seqid,
8686         .call_sync_ops = &nfs40_call_sync_ops,
8687         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
8688         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
8689         .state_renewal_ops = &nfs40_state_renewal_ops,
8690         .mig_recovery_ops = &nfs40_mig_recovery_ops,
8691 };
8692
8693 #if defined(CONFIG_NFS_V4_1)
8694 static struct nfs_seqid *
8695 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
8696 {
8697         return NULL;
8698 }
8699
8700 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
8701         .minor_version = 1,
8702         .init_caps = NFS_CAP_READDIRPLUS
8703                 | NFS_CAP_ATOMIC_OPEN
8704                 | NFS_CAP_POSIX_LOCK
8705                 | NFS_CAP_STATEID_NFSV41
8706                 | NFS_CAP_ATOMIC_OPEN_V1,
8707         .init_client = nfs41_init_client,
8708         .shutdown_client = nfs41_shutdown_client,
8709         .match_stateid = nfs41_match_stateid,
8710         .find_root_sec = nfs41_find_root_sec,
8711         .free_lock_state = nfs41_free_lock_state,
8712         .alloc_seqid = nfs_alloc_no_seqid,
8713         .call_sync_ops = &nfs41_call_sync_ops,
8714         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8715         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8716         .state_renewal_ops = &nfs41_state_renewal_ops,
8717         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8718 };
8719 #endif
8720
8721 #if defined(CONFIG_NFS_V4_2)
8722 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
8723         .minor_version = 2,
8724         .init_caps = NFS_CAP_READDIRPLUS
8725                 | NFS_CAP_ATOMIC_OPEN
8726                 | NFS_CAP_POSIX_LOCK
8727                 | NFS_CAP_STATEID_NFSV41
8728                 | NFS_CAP_ATOMIC_OPEN_V1
8729                 | NFS_CAP_ALLOCATE
8730                 | NFS_CAP_DEALLOCATE
8731                 | NFS_CAP_SEEK
8732                 | NFS_CAP_LAYOUTSTATS,
8733         .init_client = nfs41_init_client,
8734         .shutdown_client = nfs41_shutdown_client,
8735         .match_stateid = nfs41_match_stateid,
8736         .find_root_sec = nfs41_find_root_sec,
8737         .free_lock_state = nfs41_free_lock_state,
8738         .call_sync_ops = &nfs41_call_sync_ops,
8739         .alloc_seqid = nfs_alloc_no_seqid,
8740         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
8741         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
8742         .state_renewal_ops = &nfs41_state_renewal_ops,
8743         .mig_recovery_ops = &nfs41_mig_recovery_ops,
8744 };
8745 #endif
8746
8747 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
8748         [0] = &nfs_v4_0_minor_ops,
8749 #if defined(CONFIG_NFS_V4_1)
8750         [1] = &nfs_v4_1_minor_ops,
8751 #endif
8752 #if defined(CONFIG_NFS_V4_2)
8753         [2] = &nfs_v4_2_minor_ops,
8754 #endif
8755 };
8756
8757 static const struct inode_operations nfs4_dir_inode_operations = {
8758         .create         = nfs_create,
8759         .lookup         = nfs_lookup,
8760         .atomic_open    = nfs_atomic_open,
8761         .link           = nfs_link,
8762         .unlink         = nfs_unlink,
8763         .symlink        = nfs_symlink,
8764         .mkdir          = nfs_mkdir,
8765         .rmdir          = nfs_rmdir,
8766         .mknod          = nfs_mknod,
8767         .rename         = nfs_rename,
8768         .permission     = nfs_permission,
8769         .getattr        = nfs_getattr,
8770         .setattr        = nfs_setattr,
8771         .getxattr       = generic_getxattr,
8772         .setxattr       = generic_setxattr,
8773         .listxattr      = generic_listxattr,
8774         .removexattr    = generic_removexattr,
8775 };
8776
8777 static const struct inode_operations nfs4_file_inode_operations = {
8778         .permission     = nfs_permission,
8779         .getattr        = nfs_getattr,
8780         .setattr        = nfs_setattr,
8781         .getxattr       = generic_getxattr,
8782         .setxattr       = generic_setxattr,
8783         .listxattr      = generic_listxattr,
8784         .removexattr    = generic_removexattr,
8785 };
8786
8787 const struct nfs_rpc_ops nfs_v4_clientops = {
8788         .version        = 4,                    /* protocol version */
8789         .dentry_ops     = &nfs4_dentry_operations,
8790         .dir_inode_ops  = &nfs4_dir_inode_operations,
8791         .file_inode_ops = &nfs4_file_inode_operations,
8792         .file_ops       = &nfs4_file_operations,
8793         .getroot        = nfs4_proc_get_root,
8794         .submount       = nfs4_submount,
8795         .try_mount      = nfs4_try_mount,
8796         .getattr        = nfs4_proc_getattr,
8797         .setattr        = nfs4_proc_setattr,
8798         .lookup         = nfs4_proc_lookup,
8799         .access         = nfs4_proc_access,
8800         .readlink       = nfs4_proc_readlink,
8801         .create         = nfs4_proc_create,
8802         .remove         = nfs4_proc_remove,
8803         .unlink_setup   = nfs4_proc_unlink_setup,
8804         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
8805         .unlink_done    = nfs4_proc_unlink_done,
8806         .rename_setup   = nfs4_proc_rename_setup,
8807         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
8808         .rename_done    = nfs4_proc_rename_done,
8809         .link           = nfs4_proc_link,
8810         .symlink        = nfs4_proc_symlink,
8811         .mkdir          = nfs4_proc_mkdir,
8812         .rmdir          = nfs4_proc_remove,
8813         .readdir        = nfs4_proc_readdir,
8814         .mknod          = nfs4_proc_mknod,
8815         .statfs         = nfs4_proc_statfs,
8816         .fsinfo         = nfs4_proc_fsinfo,
8817         .pathconf       = nfs4_proc_pathconf,
8818         .set_capabilities = nfs4_server_capabilities,
8819         .decode_dirent  = nfs4_decode_dirent,
8820         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
8821         .read_setup     = nfs4_proc_read_setup,
8822         .read_done      = nfs4_read_done,
8823         .write_setup    = nfs4_proc_write_setup,
8824         .write_done     = nfs4_write_done,
8825         .commit_setup   = nfs4_proc_commit_setup,
8826         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
8827         .commit_done    = nfs4_commit_done,
8828         .lock           = nfs4_proc_lock,
8829         .clear_acl_cache = nfs4_zap_acl_attr,
8830         .close_context  = nfs4_close_context,
8831         .open_context   = nfs4_atomic_open,
8832         .have_delegation = nfs4_have_delegation,
8833         .return_delegation = nfs4_inode_return_delegation,
8834         .alloc_client   = nfs4_alloc_client,
8835         .init_client    = nfs4_init_client,
8836         .free_client    = nfs4_free_client,
8837         .create_server  = nfs4_create_server,
8838         .clone_server   = nfs_clone_server,
8839 };
8840
8841 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
8842         .prefix = XATTR_NAME_NFSV4_ACL,
8843         .list   = nfs4_xattr_list_nfs4_acl,
8844         .get    = nfs4_xattr_get_nfs4_acl,
8845         .set    = nfs4_xattr_set_nfs4_acl,
8846 };
8847
8848 const struct xattr_handler *nfs4_xattr_handlers[] = {
8849         &nfs4_xattr_nfs4_acl_handler,
8850 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
8851         &nfs4_xattr_nfs4_label_handler,
8852 #endif
8853         NULL
8854 };
8855
8856 /*
8857  * Local variables:
8858  *  c-basic-offset: 8
8859  * End:
8860  */