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