fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57
  58 #include "nfs4_fs.h"
  59 #include "delegation.h"
  60 #include "internal.h"
  61 #include "iostat.h"
  62 #include "callback.h"
  63 #include "pnfs.h"
  64 #include "netns.h"
  65 #include "nfs4idmap.h"
  66 #include "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #include "nfs4trace.h"
  70
  71 #define NFSDBG_FACILITY         NFSDBG_PROC
  72
  73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  75
  76 /* file attributes which can be mapped to nfs attributes */
  77 #define NFS4_VALID_ATTRS (ATTR_MODE \
  78         | ATTR_UID \
  79         | ATTR_GID \
  80         | ATTR_SIZE \
  81         | ATTR_ATIME \
  82         | ATTR_MTIME \
  83         | ATTR_CTIME \
  84         | ATTR_ATIME_SET \
  85         | ATTR_MTIME_SET)
  86
  87 struct nfs4_opendata;
  88 static int _nfs4_proc_open(struct nfs4_opendata *data);
  89 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  90 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  91 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  92 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  93 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  94 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  95                             struct nfs_fattr *fattr, struct iattr *sattr,
  96                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  97                             struct nfs4_label *olabel);
  98 #ifdef CONFIG_NFS_V4_1
  99 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 100                 struct rpc_cred *);
 101 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 102                 struct rpc_cred *, bool);
 103 #endif
 104
 105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 106 static inline struct nfs4_label *
 107 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 108         struct iattr *sattr, struct nfs4_label *label)
 109 {
 110         int err;
 111
 112         if (label == NULL)
 113                 return NULL;
 114
 115         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 116                 return NULL;
 117
 118         err = security_dentry_init_security(dentry, sattr->ia_mode,
 119                                 &dentry->d_name, (void **)&label->label, &label->len);
 120         if (err == 0)
 121                 return label;
 122
 123         return NULL;
 124 }
 125 static inline void
 126 nfs4_label_release_security(struct nfs4_label *label)
 127 {
 128         if (label)
 129                 security_release_secctx(label->label, label->len);
 130 }
 131 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 132 {
 133         if (label)
 134                 return server->attr_bitmask;
 135
 136         return server->attr_bitmask_nl;
 137 }
 138 #else
 139 static inline struct nfs4_label *
 140 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 141         struct iattr *sattr, struct nfs4_label *l)
 142 { return NULL; }
 143 static inline void
 144 nfs4_label_release_security(struct nfs4_label *label)
 145 { return; }
 146 static inline u32 *
 147 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 148 { return server->attr_bitmask; }
 149 #endif
 150
 151 /* Prevent leaks of NFSv4 errors into userland */
 152 static int nfs4_map_errors(int err)
 153 {
 154         if (err >= -1000)
 155                 return err;
 156         switch (err) {
 157         case -NFS4ERR_RESOURCE:
 158         case -NFS4ERR_LAYOUTTRYLATER:
 159         case -NFS4ERR_RECALLCONFLICT:
 160                 return -EREMOTEIO;
 161         case -NFS4ERR_WRONGSEC:
 162         case -NFS4ERR_WRONG_CRED:
 163                 return -EPERM;
 164         case -NFS4ERR_BADOWNER:
 165         case -NFS4ERR_BADNAME:
 166                 return -EINVAL;
 167         case -NFS4ERR_SHARE_DENIED:
 168                 return -EACCES;
 169         case -NFS4ERR_MINOR_VERS_MISMATCH:
 170                 return -EPROTONOSUPPORT;
 171         case -NFS4ERR_FILE_OPEN:
 172                 return -EBUSY;
 173         default:
 174                 dprintk("%s could not handle NFSv4 error %d\n",
 175                                 __func__, -err);
 176                 break;
 177         }
 178         return -EIO;
 179 }
 180
 181 /*
 182  * This is our standard bitmap for GETATTR requests.
 183  */
 184 const u32 nfs4_fattr_bitmap[3] = {
 185         FATTR4_WORD0_TYPE
 186         | FATTR4_WORD0_CHANGE
 187         | FATTR4_WORD0_SIZE
 188         | FATTR4_WORD0_FSID
 189         | FATTR4_WORD0_FILEID,
 190         FATTR4_WORD1_MODE
 191         | FATTR4_WORD1_NUMLINKS
 192         | FATTR4_WORD1_OWNER
 193         | FATTR4_WORD1_OWNER_GROUP
 194         | FATTR4_WORD1_RAWDEV
 195         | FATTR4_WORD1_SPACE_USED
 196         | FATTR4_WORD1_TIME_ACCESS
 197         | FATTR4_WORD1_TIME_METADATA
 198         | FATTR4_WORD1_TIME_MODIFY
 199         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 200 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 201         FATTR4_WORD2_SECURITY_LABEL
 202 #endif
 203 };
 204
 205 static const u32 nfs4_pnfs_open_bitmap[3] = {
 206         FATTR4_WORD0_TYPE
 207         | FATTR4_WORD0_CHANGE
 208         | FATTR4_WORD0_SIZE
 209         | FATTR4_WORD0_FSID
 210         | FATTR4_WORD0_FILEID,
 211         FATTR4_WORD1_MODE
 212         | FATTR4_WORD1_NUMLINKS
 213         | FATTR4_WORD1_OWNER
 214         | FATTR4_WORD1_OWNER_GROUP
 215         | FATTR4_WORD1_RAWDEV
 216         | FATTR4_WORD1_SPACE_USED
 217         | FATTR4_WORD1_TIME_ACCESS
 218         | FATTR4_WORD1_TIME_METADATA
 219         | FATTR4_WORD1_TIME_MODIFY,
 220         FATTR4_WORD2_MDSTHRESHOLD
 221 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 222         | FATTR4_WORD2_SECURITY_LABEL
 223 #endif
 224 };
 225
 226 static const u32 nfs4_open_noattr_bitmap[3] = {
 227         FATTR4_WORD0_TYPE
 228         | FATTR4_WORD0_FILEID,
 229 };
 230
 231 const u32 nfs4_statfs_bitmap[3] = {
 232         FATTR4_WORD0_FILES_AVAIL
 233         | FATTR4_WORD0_FILES_FREE
 234         | FATTR4_WORD0_FILES_TOTAL,
 235         FATTR4_WORD1_SPACE_AVAIL
 236         | FATTR4_WORD1_SPACE_FREE
 237         | FATTR4_WORD1_SPACE_TOTAL
 238 };
 239
 240 const u32 nfs4_pathconf_bitmap[3] = {
 241         FATTR4_WORD0_MAXLINK
 242         | FATTR4_WORD0_MAXNAME,
 243         0
 244 };
 245
 246 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 247                         | FATTR4_WORD0_MAXREAD
 248                         | FATTR4_WORD0_MAXWRITE
 249                         | FATTR4_WORD0_LEASE_TIME,
 250                         FATTR4_WORD1_TIME_DELTA
 251                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 252                         FATTR4_WORD2_LAYOUT_BLKSIZE
 253                         | FATTR4_WORD2_CLONE_BLKSIZE
 254 };
 255
 256 const u32 nfs4_fs_locations_bitmap[3] = {
 257         FATTR4_WORD0_TYPE
 258         | FATTR4_WORD0_CHANGE
 259         | FATTR4_WORD0_SIZE
 260         | FATTR4_WORD0_FSID
 261         | FATTR4_WORD0_FILEID
 262         | FATTR4_WORD0_FS_LOCATIONS,
 263         FATTR4_WORD1_MODE
 264         | FATTR4_WORD1_NUMLINKS
 265         | FATTR4_WORD1_OWNER
 266         | FATTR4_WORD1_OWNER_GROUP
 267         | FATTR4_WORD1_RAWDEV
 268         | FATTR4_WORD1_SPACE_USED
 269         | FATTR4_WORD1_TIME_ACCESS
 270         | FATTR4_WORD1_TIME_METADATA
 271         | FATTR4_WORD1_TIME_MODIFY
 272         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 273 };
 274
 275 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 276                 struct nfs4_readdir_arg *readdir)
 277 {
 278         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 279         __be32 *start, *p;
 280
 281         if (cookie > 2) {
 282                 readdir->cookie = cookie;
 283                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 284                 return;
 285         }
 286
 287         readdir->cookie = 0;
 288         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 289         if (cookie == 2)
 290                 return;
 291
 292         /*
 293          * NFSv4 servers do not return entries for '.' and '..'
 294          * Therefore, we fake these entries here.  We let '.'
 295          * have cookie 0 and '..' have cookie 1.  Note that
 296          * when talking to the server, we always send cookie 0
 297          * instead of 1 or 2.
 298          */
 299         start = p = kmap_atomic(*readdir->pages);
 300
 301         if (cookie == 0) {
 302                 *p++ = xdr_one;                                  /* next */
 303                 *p++ = xdr_zero;                   /* cookie, first word */
 304                 *p++ = xdr_one;                   /* cookie, second word */
 305                 *p++ = xdr_one;                             /* entry len */
 306                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 307                 p++;
 308                 *p++ = xdr_one;                         /* bitmap length */
 309                 *p++ = htonl(attrs);                           /* bitmap */
 310                 *p++ = htonl(12);             /* attribute buffer length */
 311                 *p++ = htonl(NF4DIR);
 312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 313         }
 314
 315         *p++ = xdr_one;                                  /* next */
 316         *p++ = xdr_zero;                   /* cookie, first word */
 317         *p++ = xdr_two;                   /* cookie, second word */
 318         *p++ = xdr_two;                             /* entry len */
 319         memcpy(p, "..\0\0", 4);                         /* entry */
 320         p++;
 321         *p++ = xdr_one;                         /* bitmap length */
 322         *p++ = htonl(attrs);                           /* bitmap */
 323         *p++ = htonl(12);             /* attribute buffer length */
 324         *p++ = htonl(NF4DIR);
 325         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 326
 327         readdir->pgbase = (char *)p - (char *)start;
 328         readdir->count -= readdir->pgbase;
 329         kunmap_atomic(start);
 330 }
 331
 332 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 333                 nfs4_stateid *stateid,
 334                 struct rpc_cred *cred)
 335 {
 336         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 337
 338         ops->test_and_free_expired(server, stateid, cred);
 339 }
 340
 341 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 342                 nfs4_stateid *stateid,
 343                 struct rpc_cred *cred)
 344 {
 345         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 346         nfs4_test_and_free_stateid(server, stateid, cred);
 347 }
 348
 349 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 350                 const nfs4_stateid *stateid,
 351                 struct rpc_cred *cred)
 352 {
 353         nfs4_stateid tmp;
 354
 355         nfs4_stateid_copy(&tmp, stateid);
 356         __nfs4_free_revoked_stateid(server, &tmp, cred);
 357 }
 358
 359 static long nfs4_update_delay(long *timeout)
 360 {
 361         long ret;
 362         if (!timeout)
 363                 return NFS4_POLL_RETRY_MAX;
 364         if (*timeout <= 0)
 365                 *timeout = NFS4_POLL_RETRY_MIN;
 366         if (*timeout > NFS4_POLL_RETRY_MAX)
 367                 *timeout = NFS4_POLL_RETRY_MAX;
 368         ret = *timeout;
 369         *timeout <<= 1;
 370         return ret;
 371 }
 372
 373 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 374 {
 375         int res = 0;
 376
 377         might_sleep();
 378
 379         freezable_schedule_timeout_killable_unsafe(
 380                 nfs4_update_delay(timeout));
 381         if (fatal_signal_pending(current))
 382                 res = -ERESTARTSYS;
 383         return res;
 384 }
 385
 386 /* This is the error handling routine for processes that are allowed
 387  * to sleep.
 388  */
 389 static int nfs4_do_handle_exception(struct nfs_server *server,
 390                 int errorcode, struct nfs4_exception *exception)
 391 {
 392         struct nfs_client *clp = server->nfs_client;
 393         struct nfs4_state *state = exception->state;
 394         const nfs4_stateid *stateid = exception->stateid;
 395         struct inode *inode = exception->inode;
 396         int ret = errorcode;
 397
 398         exception->delay = 0;
 399         exception->recovering = 0;
 400         exception->retry = 0;
 401
 402         if (stateid == NULL && state != NULL)
 403                 stateid = &state->stateid;
 404
 405         switch(errorcode) {
 406                 case 0:
 407                         return 0;
 408                 case -NFS4ERR_DELEG_REVOKED:
 409                 case -NFS4ERR_ADMIN_REVOKED:
 410                 case -NFS4ERR_EXPIRED:
 411                 case -NFS4ERR_BAD_STATEID:
 412                         if (inode != NULL && stateid != NULL) {
 413                                 nfs_inode_find_state_and_recover(inode,
 414                                                 stateid);
 415                                 goto wait_on_recovery;
 416                         }
 417                 case -NFS4ERR_OPENMODE:
 418                         if (inode) {
 419                                 int err;
 420
 421                                 err = nfs_async_inode_return_delegation(inode,
 422                                                 stateid);
 423                                 if (err == 0)
 424                                         goto wait_on_recovery;
 425                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 426                                         exception->retry = 1;
 427                                         break;
 428                                 }
 429                         }
 430                         if (state == NULL)
 431                                 break;
 432                         ret = nfs4_schedule_stateid_recovery(server, state);
 433                         if (ret < 0)
 434                                 break;
 435                         goto wait_on_recovery;
 436                 case -NFS4ERR_STALE_STATEID:
 437                 case -NFS4ERR_STALE_CLIENTID:
 438                         nfs4_schedule_lease_recovery(clp);
 439                         goto wait_on_recovery;
 440                 case -NFS4ERR_MOVED:
 441                         ret = nfs4_schedule_migration_recovery(server);
 442                         if (ret < 0)
 443                                 break;
 444                         goto wait_on_recovery;
 445                 case -NFS4ERR_LEASE_MOVED:
 446                         nfs4_schedule_lease_moved_recovery(clp);
 447                         goto wait_on_recovery;
 448 #if defined(CONFIG_NFS_V4_1)
 449                 case -NFS4ERR_BADSESSION:
 450                 case -NFS4ERR_BADSLOT:
 451                 case -NFS4ERR_BAD_HIGH_SLOT:
 452                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 453                 case -NFS4ERR_DEADSESSION:
 454                 case -NFS4ERR_SEQ_FALSE_RETRY:
 455                 case -NFS4ERR_SEQ_MISORDERED:
 456                         dprintk("%s ERROR: %d Reset session\n", __func__,
 457                                 errorcode);
 458                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 459                         goto wait_on_recovery;
 460 #endif /* defined(CONFIG_NFS_V4_1) */
 461                 case -NFS4ERR_FILE_OPEN:
 462                         if (exception->timeout > HZ) {
 463                                 /* We have retried a decent amount, time to
 464                                  * fail
 465                                  */
 466                                 ret = -EBUSY;
 467                                 break;
 468                         }
 469                 case -NFS4ERR_DELAY:
 470                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 471                 case -NFS4ERR_GRACE:
 472                 case -NFS4ERR_LAYOUTTRYLATER:
 473                 case -NFS4ERR_RECALLCONFLICT:
 474                         exception->delay = 1;
 475                         return 0;
 476
 477                 case -NFS4ERR_RETRY_UNCACHED_REP:
 478                 case -NFS4ERR_OLD_STATEID:
 479                         exception->retry = 1;
 480                         break;
 481                 case -NFS4ERR_BADOWNER:
 482                         /* The following works around a Linux server bug! */
 483                 case -NFS4ERR_BADNAME:
 484                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 485                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 486                                 exception->retry = 1;
 487                                 printk(KERN_WARNING "NFS: v4 server %s "
 488                                                 "does not accept raw "
 489                                                 "uid/gids. "
 490                                                 "Reenabling the idmapper.\n",
 491                                                 server->nfs_client->cl_hostname);
 492                         }
 493         }
 494         /* We failed to handle the error */
 495         return nfs4_map_errors(ret);
 496 wait_on_recovery:
 497         exception->recovering = 1;
 498         return 0;
 499 }
 500
 501 /* This is the error handling routine for processes that are allowed
 502  * to sleep.
 503  */
 504 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 505 {
 506         struct nfs_client *clp = server->nfs_client;
 507         int ret;
 508
 509         ret = nfs4_do_handle_exception(server, errorcode, exception);
 510         if (exception->delay) {
 511                 ret = nfs4_delay(server->client, &exception->timeout);
 512                 goto out_retry;
 513         }
 514         if (exception->recovering) {
 515                 ret = nfs4_wait_clnt_recover(clp);
 516                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 517                         return -EIO;
 518                 goto out_retry;
 519         }
 520         return ret;
 521 out_retry:
 522         if (ret == 0)
 523                 exception->retry = 1;
 524         return ret;
 525 }
 526
 527 static int
 528 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 529                 int errorcode, struct nfs4_exception *exception)
 530 {
 531         struct nfs_client *clp = server->nfs_client;
 532         int ret;
 533
 534         ret = nfs4_do_handle_exception(server, errorcode, exception);
 535         if (exception->delay) {
 536                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 537                 goto out_retry;
 538         }
 539         if (exception->recovering) {
 540                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 541                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 542                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 543                 goto out_retry;
 544         }
 545         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 546                 ret = -EIO;
 547         return ret;
 548 out_retry:
 549         if (ret == 0)
 550                 exception->retry = 1;
 551         return ret;
 552 }
 553
 554 static int
 555 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 556                         struct nfs4_state *state, long *timeout)
 557 {
 558         struct nfs4_exception exception = {
 559                 .state = state,
 560         };
 561
 562         if (task->tk_status >= 0)
 563                 return 0;
 564         if (timeout)
 565                 exception.timeout = *timeout;
 566         task->tk_status = nfs4_async_handle_exception(task, server,
 567                         task->tk_status,
 568                         &exception);
 569         if (exception.delay && timeout)
 570                 *timeout = exception.timeout;
 571         if (exception.retry)
 572                 return -EAGAIN;
 573         return 0;
 574 }
 575
 576 /*
 577  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 578  * or 'false' otherwise.
 579  */
 580 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 581 {
 582         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 583         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 584 }
 585
 586 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 587 {
 588         spin_lock(&clp->cl_lock);
 589         if (time_before(clp->cl_last_renewal,timestamp))
 590                 clp->cl_last_renewal = timestamp;
 591         spin_unlock(&clp->cl_lock);
 592 }
 593
 594 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 595 {
 596         struct nfs_client *clp = server->nfs_client;
 597
 598         if (!nfs4_has_session(clp))
 599                 do_renew_lease(clp, timestamp);
 600 }
 601
 602 struct nfs4_call_sync_data {
 603         const struct nfs_server *seq_server;
 604         struct nfs4_sequence_args *seq_args;
 605         struct nfs4_sequence_res *seq_res;
 606 };
 607
 608 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 609                         struct nfs4_sequence_res *res, int cache_reply)
 610 {
 611         args->sa_slot = NULL;
 612         args->sa_cache_this = cache_reply;
 613         args->sa_privileged = 0;
 614
 615         res->sr_slot = NULL;
 616 }
 617
 618 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 619 {
 620         args->sa_privileged = 1;
 621 }
 622
 623 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 624 {
 625         struct nfs4_slot *slot = res->sr_slot;
 626         struct nfs4_slot_table *tbl;
 627
 628         tbl = slot->table;
 629         spin_lock(&tbl->slot_tbl_lock);
 630         if (!nfs41_wake_and_assign_slot(tbl, slot))
 631                 nfs4_free_slot(tbl, slot);
 632         spin_unlock(&tbl->slot_tbl_lock);
 633
 634         res->sr_slot = NULL;
 635 }
 636
 637 static int nfs40_sequence_done(struct rpc_task *task,
 638                                struct nfs4_sequence_res *res)
 639 {
 640         if (res->sr_slot != NULL)
 641                 nfs40_sequence_free_slot(res);
 642         return 1;
 643 }
 644
 645 #if defined(CONFIG_NFS_V4_1)
 646
 647 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 648 {
 649         struct nfs4_session *session;
 650         struct nfs4_slot_table *tbl;
 651         struct nfs4_slot *slot = res->sr_slot;
 652         bool send_new_highest_used_slotid = false;
 653
 654         tbl = slot->table;
 655         session = tbl->session;
 656
 657         /* Bump the slot sequence number */
 658         if (slot->seq_done)
 659                 slot->seq_nr++;
 660         slot->seq_done = 0;
 661
 662         spin_lock(&tbl->slot_tbl_lock);
 663         /* Be nice to the server: try to ensure that the last transmitted
 664          * value for highest_user_slotid <= target_highest_slotid
 665          */
 666         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 667                 send_new_highest_used_slotid = true;
 668
 669         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 670                 send_new_highest_used_slotid = false;
 671                 goto out_unlock;
 672         }
 673         nfs4_free_slot(tbl, slot);
 674
 675         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 676                 send_new_highest_used_slotid = false;
 677 out_unlock:
 678         spin_unlock(&tbl->slot_tbl_lock);
 679         res->sr_slot = NULL;
 680         if (send_new_highest_used_slotid)
 681                 nfs41_notify_server(session->clp);
 682         if (waitqueue_active(&tbl->slot_waitq))
 683                 wake_up_all(&tbl->slot_waitq);
 684 }
 685
 686 static int nfs41_sequence_process(struct rpc_task *task,
 687                 struct nfs4_sequence_res *res)
 688 {
 689         struct nfs4_session *session;
 690         struct nfs4_slot *slot = res->sr_slot;
 691         struct nfs_client *clp;
 692         bool interrupted = false;
 693         int ret = 1;
 694
 695         if (slot == NULL)
 696                 goto out_noaction;
 697         /* don't increment the sequence number if the task wasn't sent */
 698         if (!RPC_WAS_SENT(task))
 699                 goto out;
 700
 701         session = slot->table->session;
 702
 703         if (slot->interrupted) {
 704                 if (res->sr_status != -NFS4ERR_DELAY)
 705                         slot->interrupted = 0;
 706                 interrupted = true;
 707         }
 708
 709         trace_nfs4_sequence_done(session, res);
 710         /* Check the SEQUENCE operation status */
 711         switch (res->sr_status) {
 712         case 0:
 713                 /* If previous op on slot was interrupted and we reused
 714                  * the seq# and got a reply from the cache, then retry
 715                  */
 716                 if (task->tk_status == -EREMOTEIO && interrupted) {
 717                         ++slot->seq_nr;
 718                         goto retry_nowait;
 719                 }
 720                 /* Update the slot's sequence and clientid lease timer */
 721                 slot->seq_done = 1;
 722                 clp = session->clp;
 723                 do_renew_lease(clp, res->sr_timestamp);
 724                 /* Check sequence flags */
 725                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 726                                 !!slot->privileged);
 727                 nfs41_update_target_slotid(slot->table, slot, res);
 728                 break;
 729         case 1:
 730                 /*
 731                  * sr_status remains 1 if an RPC level error occurred.
 732                  * The server may or may not have processed the sequence
 733                  * operation..
 734                  * Mark the slot as having hosted an interrupted RPC call.
 735                  */
 736                 slot->interrupted = 1;
 737                 goto out;
 738         case -NFS4ERR_DELAY:
 739                 /* The server detected a resend of the RPC call and
 740                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 741                  * of RFC5661.
 742                  */
 743                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 744                         __func__,
 745                         slot->slot_nr,
 746                         slot->seq_nr);
 747                 goto out_retry;
 748         case -NFS4ERR_BADSLOT:
 749                 /*
 750                  * The slot id we used was probably retired. Try again
 751                  * using a different slot id.
 752                  */
 753                 goto retry_nowait;
 754         case -NFS4ERR_SEQ_MISORDERED:
 755                 /*
 756                  * Was the last operation on this sequence interrupted?
 757                  * If so, retry after bumping the sequence number.
 758                  */
 759                 if (interrupted) {
 760                         ++slot->seq_nr;
 761                         goto retry_nowait;
 762                 }
 763                 /*
 764                  * Could this slot have been previously retired?
 765                  * If so, then the server may be expecting seq_nr = 1!
 766                  */
 767                 if (slot->seq_nr != 1) {
 768                         slot->seq_nr = 1;
 769                         goto retry_nowait;
 770                 }
 771                 break;
 772         case -NFS4ERR_SEQ_FALSE_RETRY:
 773                 ++slot->seq_nr;
 774                 goto retry_nowait;
 775         default:
 776                 /* Just update the slot sequence no. */
 777                 slot->seq_done = 1;
 778         }
 779 out:
 780         /* The session may be reset by one of the error handlers. */
 781         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 782 out_noaction:
 783         return ret;
 784 retry_nowait:
 785         if (rpc_restart_call_prepare(task)) {
 786                 nfs41_sequence_free_slot(res);
 787                 task->tk_status = 0;
 788                 ret = 0;
 789         }
 790         goto out;
 791 out_retry:
 792         if (!rpc_restart_call(task))
 793                 goto out;
 794         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 795         return 0;
 796 }
 797
 798 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 799 {
 800         if (!nfs41_sequence_process(task, res))
 801                 return 0;
 802         if (res->sr_slot != NULL)
 803                 nfs41_sequence_free_slot(res);
 804         return 1;
 805
 806 }
 807 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 808
 809 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 810 {
 811         if (res->sr_slot == NULL)
 812                 return 1;
 813         if (res->sr_slot->table->session != NULL)
 814                 return nfs41_sequence_process(task, res);
 815         return nfs40_sequence_done(task, res);
 816 }
 817
 818 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 819 {
 820         if (res->sr_slot != NULL) {
 821                 if (res->sr_slot->table->session != NULL)
 822                         nfs41_sequence_free_slot(res);
 823                 else
 824                         nfs40_sequence_free_slot(res);
 825         }
 826 }
 827
 828 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 829 {
 830         if (res->sr_slot == NULL)
 831                 return 1;
 832         if (!res->sr_slot->table->session)
 833                 return nfs40_sequence_done(task, res);
 834         return nfs41_sequence_done(task, res);
 835 }
 836 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 837
 838 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 839 {
 840         struct nfs4_call_sync_data *data = calldata;
 841
 842         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 843
 844         nfs4_setup_sequence(data->seq_server->nfs_client,
 845                             data->seq_args, data->seq_res, task);
 846 }
 847
 848 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 849 {
 850         struct nfs4_call_sync_data *data = calldata;
 851
 852         nfs41_sequence_done(task, data->seq_res);
 853 }
 854
 855 static const struct rpc_call_ops nfs41_call_sync_ops = {
 856         .rpc_call_prepare = nfs41_call_sync_prepare,
 857         .rpc_call_done = nfs41_call_sync_done,
 858 };
 859
 860 #else   /* !CONFIG_NFS_V4_1 */
 861
 862 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 863 {
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL)
 870                 nfs40_sequence_free_slot(res);
 871 }
 872
 873 int nfs4_sequence_done(struct rpc_task *task,
 874                        struct nfs4_sequence_res *res)
 875 {
 876         return nfs40_sequence_done(task, res);
 877 }
 878 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 879
 880 #endif  /* !CONFIG_NFS_V4_1 */
 881
 882 int nfs4_setup_sequence(const struct nfs_client *client,
 883                         struct nfs4_sequence_args *args,
 884                         struct nfs4_sequence_res *res,
 885                         struct rpc_task *task)
 886 {
 887         struct nfs4_session *session = nfs4_get_session(client);
 888         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 889         struct nfs4_slot *slot;
 890
 891         /* slot already allocated? */
 892         if (res->sr_slot != NULL)
 893                 goto out_start;
 894
 895         if (session) {
 896                 tbl = &session->fc_slot_table;
 897                 task->tk_timeout = 0;
 898         }
 899
 900         spin_lock(&tbl->slot_tbl_lock);
 901         /* The state manager will wait until the slot table is empty */
 902         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 903                 goto out_sleep;
 904
 905         slot = nfs4_alloc_slot(tbl);
 906         if (IS_ERR(slot)) {
 907                 /* Try again in 1/4 second */
 908                 if (slot == ERR_PTR(-ENOMEM))
 909                         task->tk_timeout = HZ >> 2;
 910                 goto out_sleep;
 911         }
 912         spin_unlock(&tbl->slot_tbl_lock);
 913
 914         slot->privileged = args->sa_privileged ? 1 : 0;
 915         args->sa_slot = slot;
 916
 917         res->sr_slot = slot;
 918         if (session) {
 919                 res->sr_timestamp = jiffies;
 920                 res->sr_status_flags = 0;
 921                 res->sr_status = 1;
 922         }
 923
 924         trace_nfs4_setup_sequence(session, args);
 925 out_start:
 926         rpc_call_start(task);
 927         return 0;
 928
 929 out_sleep:
 930         if (args->sa_privileged)
 931                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 932                                 NULL, RPC_PRIORITY_PRIVILEGED);
 933         else
 934                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 935         spin_unlock(&tbl->slot_tbl_lock);
 936         return -EAGAIN;
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
 939
 940 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 941 {
 942         struct nfs4_call_sync_data *data = calldata;
 943         nfs4_setup_sequence(data->seq_server->nfs_client,
 944                                 data->seq_args, data->seq_res, task);
 945 }
 946
 947 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950         nfs4_sequence_done(task, data->seq_res);
 951 }
 952
 953 static const struct rpc_call_ops nfs40_call_sync_ops = {
 954         .rpc_call_prepare = nfs40_call_sync_prepare,
 955         .rpc_call_done = nfs40_call_sync_done,
 956 };
 957
 958 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 959                                    struct nfs_server *server,
 960                                    struct rpc_message *msg,
 961                                    struct nfs4_sequence_args *args,
 962                                    struct nfs4_sequence_res *res)
 963 {
 964         int ret;
 965         struct rpc_task *task;
 966         struct nfs_client *clp = server->nfs_client;
 967         struct nfs4_call_sync_data data = {
 968                 .seq_server = server,
 969                 .seq_args = args,
 970                 .seq_res = res,
 971         };
 972         struct rpc_task_setup task_setup = {
 973                 .rpc_client = clnt,
 974                 .rpc_message = msg,
 975                 .callback_ops = clp->cl_mvops->call_sync_ops,
 976                 .callback_data = &data
 977         };
 978
 979         task = rpc_run_task(&task_setup);
 980         if (IS_ERR(task))
 981                 ret = PTR_ERR(task);
 982         else {
 983                 ret = task->tk_status;
 984                 rpc_put_task(task);
 985         }
 986         return ret;
 987 }
 988
 989 int nfs4_call_sync(struct rpc_clnt *clnt,
 990                    struct nfs_server *server,
 991                    struct rpc_message *msg,
 992                    struct nfs4_sequence_args *args,
 993                    struct nfs4_sequence_res *res,
 994                    int cache_reply)
 995 {
 996         nfs4_init_sequence(args, res, cache_reply);
 997         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 998 }
 999
1000 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1001                 unsigned long timestamp)
1002 {
1003         struct nfs_inode *nfsi = NFS_I(dir);
1004
1005         spin_lock(&dir->i_lock);
1006         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1007         if (cinfo->atomic && cinfo->before == dir->i_version) {
1008                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1009                 nfsi->attrtimeo_timestamp = jiffies;
1010         } else {
1011                 nfs_force_lookup_revalidate(dir);
1012                 if (cinfo->before != dir->i_version)
1013                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1014                                 NFS_INO_INVALID_ACL;
1015         }
1016         dir->i_version = cinfo->after;
1017         nfsi->read_cache_jiffies = timestamp;
1018         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1019         nfs_fscache_invalidate(dir);
1020         spin_unlock(&dir->i_lock);
1021 }
1022
1023 struct nfs4_opendata {
1024         struct kref kref;
1025         struct nfs_openargs o_arg;
1026         struct nfs_openres o_res;
1027         struct nfs_open_confirmargs c_arg;
1028         struct nfs_open_confirmres c_res;
1029         struct nfs4_string owner_name;
1030         struct nfs4_string group_name;
1031         struct nfs4_label *a_label;
1032         struct nfs_fattr f_attr;
1033         struct nfs4_label *f_label;
1034         struct dentry *dir;
1035         struct dentry *dentry;
1036         struct nfs4_state_owner *owner;
1037         struct nfs4_state *state;
1038         struct iattr attrs;
1039         unsigned long timestamp;
1040         bool rpc_done;
1041         bool file_created;
1042         bool is_recover;
1043         bool cancelled;
1044         int rpc_status;
1045 };
1046
1047 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1048                 int err, struct nfs4_exception *exception)
1049 {
1050         if (err != -EINVAL)
1051                 return false;
1052         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1053                 return false;
1054         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1055         exception->retry = 1;
1056         return true;
1057 }
1058
1059 static u32
1060 nfs4_map_atomic_open_share(struct nfs_server *server,
1061                 fmode_t fmode, int openflags)
1062 {
1063         u32 res = 0;
1064
1065         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1066         case FMODE_READ:
1067                 res = NFS4_SHARE_ACCESS_READ;
1068                 break;
1069         case FMODE_WRITE:
1070                 res = NFS4_SHARE_ACCESS_WRITE;
1071                 break;
1072         case FMODE_READ|FMODE_WRITE:
1073                 res = NFS4_SHARE_ACCESS_BOTH;
1074         }
1075         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1076                 goto out;
1077         /* Want no delegation if we're using O_DIRECT */
1078         if (openflags & O_DIRECT)
1079                 res |= NFS4_SHARE_WANT_NO_DELEG;
1080 out:
1081         return res;
1082 }
1083
1084 static enum open_claim_type4
1085 nfs4_map_atomic_open_claim(struct nfs_server *server,
1086                 enum open_claim_type4 claim)
1087 {
1088         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1089                 return claim;
1090         switch (claim) {
1091         default:
1092                 return claim;
1093         case NFS4_OPEN_CLAIM_FH:
1094                 return NFS4_OPEN_CLAIM_NULL;
1095         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1096                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1097         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1098                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1099         }
1100 }
1101
1102 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1103 {
1104         p->o_res.f_attr = &p->f_attr;
1105         p->o_res.f_label = p->f_label;
1106         p->o_res.seqid = p->o_arg.seqid;
1107         p->c_res.seqid = p->c_arg.seqid;
1108         p->o_res.server = p->o_arg.server;
1109         p->o_res.access_request = p->o_arg.access;
1110         nfs_fattr_init(&p->f_attr);
1111         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1112 }
1113
1114 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1115                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1116                 const struct iattr *attrs,
1117                 struct nfs4_label *label,
1118                 enum open_claim_type4 claim,
1119                 gfp_t gfp_mask)
1120 {
1121         struct dentry *parent = dget_parent(dentry);
1122         struct inode *dir = d_inode(parent);
1123         struct nfs_server *server = NFS_SERVER(dir);
1124         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1125         struct nfs4_opendata *p;
1126
1127         p = kzalloc(sizeof(*p), gfp_mask);
1128         if (p == NULL)
1129                 goto err;
1130
1131         p->f_label = nfs4_label_alloc(server, gfp_mask);
1132         if (IS_ERR(p->f_label))
1133                 goto err_free_p;
1134
1135         p->a_label = nfs4_label_alloc(server, gfp_mask);
1136         if (IS_ERR(p->a_label))
1137                 goto err_free_f;
1138
1139         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1140         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1141         if (IS_ERR(p->o_arg.seqid))
1142                 goto err_free_label;
1143         nfs_sb_active(dentry->d_sb);
1144         p->dentry = dget(dentry);
1145         p->dir = parent;
1146         p->owner = sp;
1147         atomic_inc(&sp->so_count);
1148         p->o_arg.open_flags = flags;
1149         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1150         p->o_arg.umask = current_umask();
1151         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1152         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1153                         fmode, flags);
1154         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1155          * will return permission denied for all bits until close */
1156         if (!(flags & O_EXCL)) {
1157                 /* ask server to check for all possible rights as results
1158                  * are cached */
1159                 switch (p->o_arg.claim) {
1160                 default:
1161                         break;
1162                 case NFS4_OPEN_CLAIM_NULL:
1163                 case NFS4_OPEN_CLAIM_FH:
1164                         p->o_arg.access = NFS4_ACCESS_READ |
1165                                 NFS4_ACCESS_MODIFY |
1166                                 NFS4_ACCESS_EXTEND |
1167                                 NFS4_ACCESS_EXECUTE;
1168                 }
1169         }
1170         p->o_arg.clientid = server->nfs_client->cl_clientid;
1171         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1172         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1173         p->o_arg.name = &dentry->d_name;
1174         p->o_arg.server = server;
1175         p->o_arg.bitmask = nfs4_bitmask(server, label);
1176         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1177         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1178         switch (p->o_arg.claim) {
1179         case NFS4_OPEN_CLAIM_NULL:
1180         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1181         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1182                 p->o_arg.fh = NFS_FH(dir);
1183                 break;
1184         case NFS4_OPEN_CLAIM_PREVIOUS:
1185         case NFS4_OPEN_CLAIM_FH:
1186         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1187         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1188                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1189         }
1190         if (attrs != NULL && attrs->ia_valid != 0) {
1191                 __u32 verf[2];
1192
1193                 p->o_arg.u.attrs = &p->attrs;
1194                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1195
1196                 verf[0] = jiffies;
1197                 verf[1] = current->pid;
1198                 memcpy(p->o_arg.u.verifier.data, verf,
1199                                 sizeof(p->o_arg.u.verifier.data));
1200         }
1201         p->c_arg.fh = &p->o_res.fh;
1202         p->c_arg.stateid = &p->o_res.stateid;
1203         p->c_arg.seqid = p->o_arg.seqid;
1204         nfs4_init_opendata_res(p);
1205         kref_init(&p->kref);
1206         return p;
1207
1208 err_free_label:
1209         nfs4_label_free(p->a_label);
1210 err_free_f:
1211         nfs4_label_free(p->f_label);
1212 err_free_p:
1213         kfree(p);
1214 err:
1215         dput(parent);
1216         return NULL;
1217 }
1218
1219 static void nfs4_opendata_free(struct kref *kref)
1220 {
1221         struct nfs4_opendata *p = container_of(kref,
1222                         struct nfs4_opendata, kref);
1223         struct super_block *sb = p->dentry->d_sb;
1224
1225         nfs_free_seqid(p->o_arg.seqid);
1226         nfs4_sequence_free_slot(&p->o_res.seq_res);
1227         if (p->state != NULL)
1228                 nfs4_put_open_state(p->state);
1229         nfs4_put_state_owner(p->owner);
1230
1231         nfs4_label_free(p->a_label);
1232         nfs4_label_free(p->f_label);
1233
1234         dput(p->dir);
1235         dput(p->dentry);
1236         nfs_sb_deactive(sb);
1237         nfs_fattr_free_names(&p->f_attr);
1238         kfree(p->f_attr.mdsthreshold);
1239         kfree(p);
1240 }
1241
1242 static void nfs4_opendata_put(struct nfs4_opendata *p)
1243 {
1244         if (p != NULL)
1245                 kref_put(&p->kref, nfs4_opendata_free);
1246 }
1247
1248 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1249                 fmode_t fmode)
1250 {
1251         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1252         case FMODE_READ|FMODE_WRITE:
1253                 return state->n_rdwr != 0;
1254         case FMODE_WRITE:
1255                 return state->n_wronly != 0;
1256         case FMODE_READ:
1257                 return state->n_rdonly != 0;
1258         }
1259         WARN_ON_ONCE(1);
1260         return false;
1261 }
1262
1263 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1264 {
1265         int ret = 0;
1266
1267         if (open_mode & (O_EXCL|O_TRUNC))
1268                 goto out;
1269         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1270                 case FMODE_READ:
1271                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1272                                 && state->n_rdonly != 0;
1273                         break;
1274                 case FMODE_WRITE:
1275                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1276                                 && state->n_wronly != 0;
1277                         break;
1278                 case FMODE_READ|FMODE_WRITE:
1279                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1280                                 && state->n_rdwr != 0;
1281         }
1282 out:
1283         return ret;
1284 }
1285
1286 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1287                 enum open_claim_type4 claim)
1288 {
1289         if (delegation == NULL)
1290                 return 0;
1291         if ((delegation->type & fmode) != fmode)
1292                 return 0;
1293         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1294                 return 0;
1295         switch (claim) {
1296         case NFS4_OPEN_CLAIM_NULL:
1297         case NFS4_OPEN_CLAIM_FH:
1298                 break;
1299         case NFS4_OPEN_CLAIM_PREVIOUS:
1300                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1301                         break;
1302         default:
1303                 return 0;
1304         }
1305         nfs_mark_delegation_referenced(delegation);
1306         return 1;
1307 }
1308
1309 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1310 {
1311         switch (fmode) {
1312                 case FMODE_WRITE:
1313                         state->n_wronly++;
1314                         break;
1315                 case FMODE_READ:
1316                         state->n_rdonly++;
1317                         break;
1318                 case FMODE_READ|FMODE_WRITE:
1319                         state->n_rdwr++;
1320         }
1321         nfs4_state_set_mode_locked(state, state->state | fmode);
1322 }
1323
1324 #ifdef CONFIG_NFS_V4_1
1325 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1326 {
1327         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1328                 return true;
1329         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1330                 return true;
1331         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1332                 return true;
1333         return false;
1334 }
1335 #endif /* CONFIG_NFS_V4_1 */
1336
1337 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1338 {
1339         struct nfs_client *clp = state->owner->so_server->nfs_client;
1340         bool need_recover = false;
1341
1342         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1343                 need_recover = true;
1344         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1345                 need_recover = true;
1346         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1347                 need_recover = true;
1348         if (need_recover)
1349                 nfs4_state_mark_reclaim_nograce(clp, state);
1350 }
1351
1352 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1353                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1354 {
1355         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1356                 return true;
1357         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1358                 nfs4_stateid_copy(freeme, &state->open_stateid);
1359                 nfs_test_and_clear_all_open_stateid(state);
1360                 return true;
1361         }
1362         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1363                 return true;
1364         return false;
1365 }
1366
1367 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1368 {
1369         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1370                 return;
1371         if (state->n_wronly)
1372                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1373         if (state->n_rdonly)
1374                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1375         if (state->n_rdwr)
1376                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1377         set_bit(NFS_OPEN_STATE, &state->flags);
1378 }
1379
1380 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1381                 nfs4_stateid *stateid, fmode_t fmode)
1382 {
1383         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1384         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1385         case FMODE_WRITE:
1386                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1387                 break;
1388         case FMODE_READ:
1389                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1390                 break;
1391         case 0:
1392                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1393                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1394                 clear_bit(NFS_OPEN_STATE, &state->flags);
1395         }
1396         if (stateid == NULL)
1397                 return;
1398         /* Handle OPEN+OPEN_DOWNGRADE races */
1399         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1400             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1401                 nfs_resync_open_stateid_locked(state);
1402                 return;
1403         }
1404         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1405                 nfs4_stateid_copy(&state->stateid, stateid);
1406         nfs4_stateid_copy(&state->open_stateid, stateid);
1407 }
1408
1409 static void nfs_clear_open_stateid(struct nfs4_state *state,
1410         nfs4_stateid *arg_stateid,
1411         nfs4_stateid *stateid, fmode_t fmode)
1412 {
1413         write_seqlock(&state->seqlock);
1414         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1415         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1416                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1417         write_sequnlock(&state->seqlock);
1418         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1419                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1420 }
1421
1422 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1423                 const nfs4_stateid *stateid, fmode_t fmode,
1424                 nfs4_stateid *freeme)
1425 {
1426         switch (fmode) {
1427                 case FMODE_READ:
1428                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1429                         break;
1430                 case FMODE_WRITE:
1431                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1432                         break;
1433                 case FMODE_READ|FMODE_WRITE:
1434                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1435         }
1436         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1437                 return;
1438         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1439                 nfs4_stateid_copy(&state->stateid, stateid);
1440         nfs4_stateid_copy(&state->open_stateid, stateid);
1441 }
1442
1443 static void __update_open_stateid(struct nfs4_state *state,
1444                 const nfs4_stateid *open_stateid,
1445                 const nfs4_stateid *deleg_stateid,
1446                 fmode_t fmode,
1447                 nfs4_stateid *freeme)
1448 {
1449         /*
1450          * Protect the call to nfs4_state_set_mode_locked and
1451          * serialise the stateid update
1452          */
1453         spin_lock(&state->owner->so_lock);
1454         write_seqlock(&state->seqlock);
1455         if (deleg_stateid != NULL) {
1456                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1457                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1458         }
1459         if (open_stateid != NULL)
1460                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1461         write_sequnlock(&state->seqlock);
1462         update_open_stateflags(state, fmode);
1463         spin_unlock(&state->owner->so_lock);
1464 }
1465
1466 static int update_open_stateid(struct nfs4_state *state,
1467                 const nfs4_stateid *open_stateid,
1468                 const nfs4_stateid *delegation,
1469                 fmode_t fmode)
1470 {
1471         struct nfs_server *server = NFS_SERVER(state->inode);
1472         struct nfs_client *clp = server->nfs_client;
1473         struct nfs_inode *nfsi = NFS_I(state->inode);
1474         struct nfs_delegation *deleg_cur;
1475         nfs4_stateid freeme = { };
1476         int ret = 0;
1477
1478         fmode &= (FMODE_READ|FMODE_WRITE);
1479
1480         rcu_read_lock();
1481         deleg_cur = rcu_dereference(nfsi->delegation);
1482         if (deleg_cur == NULL)
1483                 goto no_delegation;
1484
1485         spin_lock(&deleg_cur->lock);
1486         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1487            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1488             (deleg_cur->type & fmode) != fmode)
1489                 goto no_delegation_unlock;
1490
1491         if (delegation == NULL)
1492                 delegation = &deleg_cur->stateid;
1493         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1494                 goto no_delegation_unlock;
1495
1496         nfs_mark_delegation_referenced(deleg_cur);
1497         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1498                         fmode, &freeme);
1499         ret = 1;
1500 no_delegation_unlock:
1501         spin_unlock(&deleg_cur->lock);
1502 no_delegation:
1503         rcu_read_unlock();
1504
1505         if (!ret && open_stateid != NULL) {
1506                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1507                 ret = 1;
1508         }
1509         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1510                 nfs4_schedule_state_manager(clp);
1511         if (freeme.type != 0)
1512                 nfs4_test_and_free_stateid(server, &freeme,
1513                                 state->owner->so_cred);
1514
1515         return ret;
1516 }
1517
1518 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1519                 const nfs4_stateid *stateid)
1520 {
1521         struct nfs4_state *state = lsp->ls_state;
1522         bool ret = false;
1523
1524         spin_lock(&state->state_lock);
1525         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1526                 goto out_noupdate;
1527         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1528                 goto out_noupdate;
1529         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1530         ret = true;
1531 out_noupdate:
1532         spin_unlock(&state->state_lock);
1533         return ret;
1534 }
1535
1536 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1537 {
1538         struct nfs_delegation *delegation;
1539
1540         rcu_read_lock();
1541         delegation = rcu_dereference(NFS_I(inode)->delegation);
1542         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1543                 rcu_read_unlock();
1544                 return;
1545         }
1546         rcu_read_unlock();
1547         nfs4_inode_return_delegation(inode);
1548 }
1549
1550 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1551 {
1552         struct nfs4_state *state = opendata->state;
1553         struct nfs_inode *nfsi = NFS_I(state->inode);
1554         struct nfs_delegation *delegation;
1555         int open_mode = opendata->o_arg.open_flags;
1556         fmode_t fmode = opendata->o_arg.fmode;
1557         enum open_claim_type4 claim = opendata->o_arg.claim;
1558         nfs4_stateid stateid;
1559         int ret = -EAGAIN;
1560
1561         for (;;) {
1562                 spin_lock(&state->owner->so_lock);
1563                 if (can_open_cached(state, fmode, open_mode)) {
1564                         update_open_stateflags(state, fmode);
1565                         spin_unlock(&state->owner->so_lock);
1566                         goto out_return_state;
1567                 }
1568                 spin_unlock(&state->owner->so_lock);
1569                 rcu_read_lock();
1570                 delegation = rcu_dereference(nfsi->delegation);
1571                 if (!can_open_delegated(delegation, fmode, claim)) {
1572                         rcu_read_unlock();
1573                         break;
1574                 }
1575                 /* Save the delegation */
1576                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1577                 rcu_read_unlock();
1578                 nfs_release_seqid(opendata->o_arg.seqid);
1579                 if (!opendata->is_recover) {
1580                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1581                         if (ret != 0)
1582                                 goto out;
1583                 }
1584                 ret = -EAGAIN;
1585
1586                 /* Try to update the stateid using the delegation */
1587                 if (update_open_stateid(state, NULL, &stateid, fmode))
1588                         goto out_return_state;
1589         }
1590 out:
1591         return ERR_PTR(ret);
1592 out_return_state:
1593         atomic_inc(&state->count);
1594         return state;
1595 }
1596
1597 static void
1598 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1599 {
1600         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1601         struct nfs_delegation *delegation;
1602         int delegation_flags = 0;
1603
1604         rcu_read_lock();
1605         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1606         if (delegation)
1607                 delegation_flags = delegation->flags;
1608         rcu_read_unlock();
1609         switch (data->o_arg.claim) {
1610         default:
1611                 break;
1612         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1613         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1614                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1615                                    "returning a delegation for "
1616                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1617                                    clp->cl_hostname);
1618                 return;
1619         }
1620         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1621                 nfs_inode_set_delegation(state->inode,
1622                                          data->owner->so_cred,
1623                                          &data->o_res);
1624         else
1625                 nfs_inode_reclaim_delegation(state->inode,
1626                                              data->owner->so_cred,
1627                                              &data->o_res);
1628 }
1629
1630 /*
1631  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1632  * and update the nfs4_state.
1633  */
1634 static struct nfs4_state *
1635 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1636 {
1637         struct inode *inode = data->state->inode;
1638         struct nfs4_state *state = data->state;
1639         int ret;
1640
1641         if (!data->rpc_done) {
1642                 if (data->rpc_status)
1643                         return ERR_PTR(data->rpc_status);
1644                 /* cached opens have already been processed */
1645                 goto update;
1646         }
1647
1648         ret = nfs_refresh_inode(inode, &data->f_attr);
1649         if (ret)
1650                 return ERR_PTR(ret);
1651
1652         if (data->o_res.delegation_type != 0)
1653                 nfs4_opendata_check_deleg(data, state);
1654 update:
1655         update_open_stateid(state, &data->o_res.stateid, NULL,
1656                             data->o_arg.fmode);
1657         atomic_inc(&state->count);
1658
1659         return state;
1660 }
1661
1662 static struct nfs4_state *
1663 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1664 {
1665         struct inode *inode;
1666         struct nfs4_state *state = NULL;
1667         int ret;
1668
1669         if (!data->rpc_done) {
1670                 state = nfs4_try_open_cached(data);
1671                 trace_nfs4_cached_open(data->state);
1672                 goto out;
1673         }
1674
1675         ret = -EAGAIN;
1676         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1677                 goto err;
1678         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1679         ret = PTR_ERR(inode);
1680         if (IS_ERR(inode))
1681                 goto err;
1682         ret = -ENOMEM;
1683         state = nfs4_get_open_state(inode, data->owner);
1684         if (state == NULL)
1685                 goto err_put_inode;
1686         if (data->o_res.delegation_type != 0)
1687                 nfs4_opendata_check_deleg(data, state);
1688         update_open_stateid(state, &data->o_res.stateid, NULL,
1689                         data->o_arg.fmode);
1690         iput(inode);
1691 out:
1692         nfs_release_seqid(data->o_arg.seqid);
1693         return state;
1694 err_put_inode:
1695         iput(inode);
1696 err:
1697         return ERR_PTR(ret);
1698 }
1699
1700 static struct nfs4_state *
1701 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1702 {
1703         struct nfs4_state *ret;
1704
1705         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1706                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1707         else
1708                 ret = _nfs4_opendata_to_nfs4_state(data);
1709         nfs4_sequence_free_slot(&data->o_res.seq_res);
1710         return ret;
1711 }
1712
1713 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1714 {
1715         struct nfs_inode *nfsi = NFS_I(state->inode);
1716         struct nfs_open_context *ctx;
1717
1718         spin_lock(&state->inode->i_lock);
1719         list_for_each_entry(ctx, &nfsi->open_files, list) {
1720                 if (ctx->state != state)
1721                         continue;
1722                 get_nfs_open_context(ctx);
1723                 spin_unlock(&state->inode->i_lock);
1724                 return ctx;
1725         }
1726         spin_unlock(&state->inode->i_lock);
1727         return ERR_PTR(-ENOENT);
1728 }
1729
1730 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1731                 struct nfs4_state *state, enum open_claim_type4 claim)
1732 {
1733         struct nfs4_opendata *opendata;
1734
1735         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1736                         NULL, NULL, claim, GFP_NOFS);
1737         if (opendata == NULL)
1738                 return ERR_PTR(-ENOMEM);
1739         opendata->state = state;
1740         atomic_inc(&state->count);
1741         return opendata;
1742 }
1743
1744 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1745                 fmode_t fmode)
1746 {
1747         struct nfs4_state *newstate;
1748         int ret;
1749
1750         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1751                 return 0;
1752         opendata->o_arg.open_flags = 0;
1753         opendata->o_arg.fmode = fmode;
1754         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1755                         NFS_SB(opendata->dentry->d_sb),
1756                         fmode, 0);
1757         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1758         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1759         nfs4_init_opendata_res(opendata);
1760         ret = _nfs4_recover_proc_open(opendata);
1761         if (ret != 0)
1762                 return ret;
1763         newstate = nfs4_opendata_to_nfs4_state(opendata);
1764         if (IS_ERR(newstate))
1765                 return PTR_ERR(newstate);
1766         if (newstate != opendata->state)
1767                 ret = -ESTALE;
1768         nfs4_close_state(newstate, fmode);
1769         return ret;
1770 }
1771
1772 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1773 {
1774         int ret;
1775
1776         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1777         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1778         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1779         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1780         /* memory barrier prior to reading state->n_* */
1781         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1782         clear_bit(NFS_OPEN_STATE, &state->flags);
1783         smp_rmb();
1784         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1785         if (ret != 0)
1786                 return ret;
1787         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1788         if (ret != 0)
1789                 return ret;
1790         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1791         if (ret != 0)
1792                 return ret;
1793         /*
1794          * We may have performed cached opens for all three recoveries.
1795          * Check if we need to update the current stateid.
1796          */
1797         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1798             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1799                 write_seqlock(&state->seqlock);
1800                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1801                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1802                 write_sequnlock(&state->seqlock);
1803         }
1804         return 0;
1805 }
1806
1807 /*
1808  * OPEN_RECLAIM:
1809  *      reclaim state on the server after a reboot.
1810  */
1811 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1812 {
1813         struct nfs_delegation *delegation;
1814         struct nfs4_opendata *opendata;
1815         fmode_t delegation_type = 0;
1816         int status;
1817
1818         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1819                         NFS4_OPEN_CLAIM_PREVIOUS);
1820         if (IS_ERR(opendata))
1821                 return PTR_ERR(opendata);
1822         rcu_read_lock();
1823         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1824         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1825                 delegation_type = delegation->type;
1826         rcu_read_unlock();
1827         opendata->o_arg.u.delegation_type = delegation_type;
1828         status = nfs4_open_recover(opendata, state);
1829         nfs4_opendata_put(opendata);
1830         return status;
1831 }
1832
1833 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1834 {
1835         struct nfs_server *server = NFS_SERVER(state->inode);
1836         struct nfs4_exception exception = { };
1837         int err;
1838         do {
1839                 err = _nfs4_do_open_reclaim(ctx, state);
1840                 trace_nfs4_open_reclaim(ctx, 0, err);
1841                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1842                         continue;
1843                 if (err != -NFS4ERR_DELAY)
1844                         break;
1845                 nfs4_handle_exception(server, err, &exception);
1846         } while (exception.retry);
1847         return err;
1848 }
1849
1850 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1851 {
1852         struct nfs_open_context *ctx;
1853         int ret;
1854
1855         ctx = nfs4_state_find_open_context(state);
1856         if (IS_ERR(ctx))
1857                 return -EAGAIN;
1858         ret = nfs4_do_open_reclaim(ctx, state);
1859         put_nfs_open_context(ctx);
1860         return ret;
1861 }
1862
1863 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1864 {
1865         switch (err) {
1866                 default:
1867                         printk(KERN_ERR "NFS: %s: unhandled error "
1868                                         "%d.\n", __func__, err);
1869                 case 0:
1870                 case -ENOENT:
1871                 case -EAGAIN:
1872                 case -ESTALE:
1873                         break;
1874                 case -NFS4ERR_BADSESSION:
1875                 case -NFS4ERR_BADSLOT:
1876                 case -NFS4ERR_BAD_HIGH_SLOT:
1877                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1878                 case -NFS4ERR_DEADSESSION:
1879                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1880                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1881                         return -EAGAIN;
1882                 case -NFS4ERR_STALE_CLIENTID:
1883                 case -NFS4ERR_STALE_STATEID:
1884                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1885                         /* Don't recall a delegation if it was lost */
1886                         nfs4_schedule_lease_recovery(server->nfs_client);
1887                         return -EAGAIN;
1888                 case -NFS4ERR_MOVED:
1889                         nfs4_schedule_migration_recovery(server);
1890                         return -EAGAIN;
1891                 case -NFS4ERR_LEASE_MOVED:
1892                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1893                         return -EAGAIN;
1894                 case -NFS4ERR_DELEG_REVOKED:
1895                 case -NFS4ERR_ADMIN_REVOKED:
1896                 case -NFS4ERR_EXPIRED:
1897                 case -NFS4ERR_BAD_STATEID:
1898                 case -NFS4ERR_OPENMODE:
1899                         nfs_inode_find_state_and_recover(state->inode,
1900                                         stateid);
1901                         nfs4_schedule_stateid_recovery(server, state);
1902                         return -EAGAIN;
1903                 case -NFS4ERR_DELAY:
1904                 case -NFS4ERR_GRACE:
1905                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1906                         ssleep(1);
1907                         return -EAGAIN;
1908                 case -ENOMEM:
1909                 case -NFS4ERR_DENIED:
1910                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1911                         return 0;
1912         }
1913         return err;
1914 }
1915
1916 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1917                 struct nfs4_state *state, const nfs4_stateid *stateid,
1918                 fmode_t type)
1919 {
1920         struct nfs_server *server = NFS_SERVER(state->inode);
1921         struct nfs4_opendata *opendata;
1922         int err = 0;
1923
1924         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1925                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1926         if (IS_ERR(opendata))
1927                 return PTR_ERR(opendata);
1928         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1929         write_seqlock(&state->seqlock);
1930         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1931         write_sequnlock(&state->seqlock);
1932         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1933         switch (type & (FMODE_READ|FMODE_WRITE)) {
1934         case FMODE_READ|FMODE_WRITE:
1935         case FMODE_WRITE:
1936                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1937                 if (err)
1938                         break;
1939                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1940                 if (err)
1941                         break;
1942         case FMODE_READ:
1943                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1944         }
1945         nfs4_opendata_put(opendata);
1946         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1947 }
1948
1949 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1950 {
1951         struct nfs4_opendata *data = calldata;
1952
1953         nfs4_setup_sequence(data->o_arg.server->nfs_client,
1954                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
1955 }
1956
1957 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1958 {
1959         struct nfs4_opendata *data = calldata;
1960
1961         nfs40_sequence_done(task, &data->c_res.seq_res);
1962
1963         data->rpc_status = task->tk_status;
1964         if (data->rpc_status == 0) {
1965                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1966                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1967                 renew_lease(data->o_res.server, data->timestamp);
1968                 data->rpc_done = true;
1969         }
1970 }
1971
1972 static void nfs4_open_confirm_release(void *calldata)
1973 {
1974         struct nfs4_opendata *data = calldata;
1975         struct nfs4_state *state = NULL;
1976
1977         /* If this request hasn't been cancelled, do nothing */
1978         if (!data->cancelled)
1979                 goto out_free;
1980         /* In case of error, no cleanup! */
1981         if (!data->rpc_done)
1982                 goto out_free;
1983         state = nfs4_opendata_to_nfs4_state(data);
1984         if (!IS_ERR(state))
1985                 nfs4_close_state(state, data->o_arg.fmode);
1986 out_free:
1987         nfs4_opendata_put(data);
1988 }
1989
1990 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1991         .rpc_call_prepare = nfs4_open_confirm_prepare,
1992         .rpc_call_done = nfs4_open_confirm_done,
1993         .rpc_release = nfs4_open_confirm_release,
1994 };
1995
1996 /*
1997  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1998  */
1999 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2000 {
2001         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2002         struct rpc_task *task;
2003         struct  rpc_message msg = {
2004                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2005                 .rpc_argp = &data->c_arg,
2006                 .rpc_resp = &data->c_res,
2007                 .rpc_cred = data->owner->so_cred,
2008         };
2009         struct rpc_task_setup task_setup_data = {
2010                 .rpc_client = server->client,
2011                 .rpc_message = &msg,
2012                 .callback_ops = &nfs4_open_confirm_ops,
2013                 .callback_data = data,
2014                 .workqueue = nfsiod_workqueue,
2015                 .flags = RPC_TASK_ASYNC,
2016         };
2017         int status;
2018
2019         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2020         kref_get(&data->kref);
2021         data->rpc_done = false;
2022         data->rpc_status = 0;
2023         data->timestamp = jiffies;
2024         if (data->is_recover)
2025                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2026         task = rpc_run_task(&task_setup_data);
2027         if (IS_ERR(task))
2028                 return PTR_ERR(task);
2029         status = rpc_wait_for_completion_task(task);
2030         if (status != 0) {
2031                 data->cancelled = true;
2032                 smp_wmb();
2033         } else
2034                 status = data->rpc_status;
2035         rpc_put_task(task);
2036         return status;
2037 }
2038
2039 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2040 {
2041         struct nfs4_opendata *data = calldata;
2042         struct nfs4_state_owner *sp = data->owner;
2043         struct nfs_client *clp = sp->so_server->nfs_client;
2044         enum open_claim_type4 claim = data->o_arg.claim;
2045
2046         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2047                 goto out_wait;
2048         /*
2049          * Check if we still need to send an OPEN call, or if we can use
2050          * a delegation instead.
2051          */
2052         if (data->state != NULL) {
2053                 struct nfs_delegation *delegation;
2054
2055                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2056                         goto out_no_action;
2057                 rcu_read_lock();
2058                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2059                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2060                         goto unlock_no_action;
2061                 rcu_read_unlock();
2062         }
2063         /* Update client id. */
2064         data->o_arg.clientid = clp->cl_clientid;
2065         switch (claim) {
2066         default:
2067                 break;
2068         case NFS4_OPEN_CLAIM_PREVIOUS:
2069         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2070         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2071                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2072         case NFS4_OPEN_CLAIM_FH:
2073                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2074                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2075         }
2076         data->timestamp = jiffies;
2077         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2078                                 &data->o_arg.seq_args,
2079                                 &data->o_res.seq_res,
2080                                 task) != 0)
2081                 nfs_release_seqid(data->o_arg.seqid);
2082
2083         /* Set the create mode (note dependency on the session type) */
2084         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2085         if (data->o_arg.open_flags & O_EXCL) {
2086                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2087                 if (nfs4_has_persistent_session(clp))
2088                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2089                 else if (clp->cl_mvops->minor_version > 0)
2090                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2091         }
2092         return;
2093 unlock_no_action:
2094         trace_nfs4_cached_open(data->state);
2095         rcu_read_unlock();
2096 out_no_action:
2097         task->tk_action = NULL;
2098 out_wait:
2099         nfs4_sequence_done(task, &data->o_res.seq_res);
2100 }
2101
2102 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2103 {
2104         struct nfs4_opendata *data = calldata;
2105
2106         data->rpc_status = task->tk_status;
2107
2108         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2109                 return;
2110
2111         if (task->tk_status == 0) {
2112                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2113                         switch (data->o_res.f_attr->mode & S_IFMT) {
2114                         case S_IFREG:
2115                                 break;
2116                         case S_IFLNK:
2117                                 data->rpc_status = -ELOOP;
2118                                 break;
2119                         case S_IFDIR:
2120                                 data->rpc_status = -EISDIR;
2121                                 break;
2122                         default:
2123                                 data->rpc_status = -ENOTDIR;
2124                         }
2125                 }
2126                 renew_lease(data->o_res.server, data->timestamp);
2127                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2128                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2129         }
2130         data->rpc_done = true;
2131 }
2132
2133 static void nfs4_open_release(void *calldata)
2134 {
2135         struct nfs4_opendata *data = calldata;
2136         struct nfs4_state *state = NULL;
2137
2138         /* If this request hasn't been cancelled, do nothing */
2139         if (!data->cancelled)
2140                 goto out_free;
2141         /* In case of error, no cleanup! */
2142         if (data->rpc_status != 0 || !data->rpc_done)
2143                 goto out_free;
2144         /* In case we need an open_confirm, no cleanup! */
2145         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2146                 goto out_free;
2147         state = nfs4_opendata_to_nfs4_state(data);
2148         if (!IS_ERR(state))
2149                 nfs4_close_state(state, data->o_arg.fmode);
2150 out_free:
2151         nfs4_opendata_put(data);
2152 }
2153
2154 static const struct rpc_call_ops nfs4_open_ops = {
2155         .rpc_call_prepare = nfs4_open_prepare,
2156         .rpc_call_done = nfs4_open_done,
2157         .rpc_release = nfs4_open_release,
2158 };
2159
2160 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2161 {
2162         struct inode *dir = d_inode(data->dir);
2163         struct nfs_server *server = NFS_SERVER(dir);
2164         struct nfs_openargs *o_arg = &data->o_arg;
2165         struct nfs_openres *o_res = &data->o_res;
2166         struct rpc_task *task;
2167         struct rpc_message msg = {
2168                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2169                 .rpc_argp = o_arg,
2170                 .rpc_resp = o_res,
2171                 .rpc_cred = data->owner->so_cred,
2172         };
2173         struct rpc_task_setup task_setup_data = {
2174                 .rpc_client = server->client,
2175                 .rpc_message = &msg,
2176                 .callback_ops = &nfs4_open_ops,
2177                 .callback_data = data,
2178                 .workqueue = nfsiod_workqueue,
2179                 .flags = RPC_TASK_ASYNC,
2180         };
2181         int status;
2182
2183         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2184         kref_get(&data->kref);
2185         data->rpc_done = false;
2186         data->rpc_status = 0;
2187         data->cancelled = false;
2188         data->is_recover = false;
2189         if (isrecover) {
2190                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2191                 data->is_recover = true;
2192         }
2193         task = rpc_run_task(&task_setup_data);
2194         if (IS_ERR(task))
2195                 return PTR_ERR(task);
2196         status = rpc_wait_for_completion_task(task);
2197         if (status != 0) {
2198                 data->cancelled = true;
2199                 smp_wmb();
2200         } else
2201                 status = data->rpc_status;
2202         rpc_put_task(task);
2203
2204         return status;
2205 }
2206
2207 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2208 {
2209         struct inode *dir = d_inode(data->dir);
2210         struct nfs_openres *o_res = &data->o_res;
2211         int status;
2212
2213         status = nfs4_run_open_task(data, 1);
2214         if (status != 0 || !data->rpc_done)
2215                 return status;
2216
2217         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2218
2219         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2220                 status = _nfs4_proc_open_confirm(data);
2221
2222         return status;
2223 }
2224
2225 /*
2226  * Additional permission checks in order to distinguish between an
2227  * open for read, and an open for execute. This works around the
2228  * fact that NFSv4 OPEN treats read and execute permissions as being
2229  * the same.
2230  * Note that in the non-execute case, we want to turn off permission
2231  * checking if we just created a new file (POSIX open() semantics).
2232  */
2233 static int nfs4_opendata_access(struct rpc_cred *cred,
2234                                 struct nfs4_opendata *opendata,
2235                                 struct nfs4_state *state, fmode_t fmode,
2236                                 int openflags)
2237 {
2238         struct nfs_access_entry cache;
2239         u32 mask;
2240
2241         /* access call failed or for some reason the server doesn't
2242          * support any access modes -- defer access call until later */
2243         if (opendata->o_res.access_supported == 0)
2244                 return 0;
2245
2246         mask = 0;
2247         /*
2248          * Use openflags to check for exec, because fmode won't
2249          * always have FMODE_EXEC set when file open for exec.
2250          */
2251         if (openflags & __FMODE_EXEC) {
2252                 /* ONLY check for exec rights */
2253                 mask = MAY_EXEC;
2254         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2255                 mask = MAY_READ;
2256
2257         cache.cred = cred;
2258         cache.jiffies = jiffies;
2259         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2260         nfs_access_add_cache(state->inode, &cache);
2261
2262         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2263                 return 0;
2264
2265         return -EACCES;
2266 }
2267
2268 /*
2269  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2270  */
2271 static int _nfs4_proc_open(struct nfs4_opendata *data)
2272 {
2273         struct inode *dir = d_inode(data->dir);
2274         struct nfs_server *server = NFS_SERVER(dir);
2275         struct nfs_openargs *o_arg = &data->o_arg;
2276         struct nfs_openres *o_res = &data->o_res;
2277         int status;
2278
2279         status = nfs4_run_open_task(data, 0);
2280         if (!data->rpc_done)
2281                 return status;
2282         if (status != 0) {
2283                 if (status == -NFS4ERR_BADNAME &&
2284                                 !(o_arg->open_flags & O_CREAT))
2285                         return -ENOENT;
2286                 return status;
2287         }
2288
2289         nfs_fattr_map_and_free_names(server, &data->f_attr);
2290
2291         if (o_arg->open_flags & O_CREAT) {
2292                 if (o_arg->open_flags & O_EXCL)
2293                         data->file_created = true;
2294                 else if (o_res->cinfo.before != o_res->cinfo.after)
2295                         data->file_created = true;
2296                 if (data->file_created || dir->i_version != o_res->cinfo.after)
2297                         update_changeattr(dir, &o_res->cinfo,
2298                                         o_res->f_attr->time_start);
2299         }
2300         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2301                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2302         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2303                 status = _nfs4_proc_open_confirm(data);
2304                 if (status != 0)
2305                         return status;
2306         }
2307         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2308                 nfs4_sequence_free_slot(&o_res->seq_res);
2309                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2310         }
2311         return 0;
2312 }
2313
2314 /*
2315  * OPEN_EXPIRED:
2316  *      reclaim state on the server after a network partition.
2317  *      Assumes caller holds the appropriate lock
2318  */
2319 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2320 {
2321         struct nfs4_opendata *opendata;
2322         int ret;
2323
2324         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2325                         NFS4_OPEN_CLAIM_FH);
2326         if (IS_ERR(opendata))
2327                 return PTR_ERR(opendata);
2328         ret = nfs4_open_recover(opendata, state);
2329         if (ret == -ESTALE)
2330                 d_drop(ctx->dentry);
2331         nfs4_opendata_put(opendata);
2332         return ret;
2333 }
2334
2335 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2336 {
2337         struct nfs_server *server = NFS_SERVER(state->inode);
2338         struct nfs4_exception exception = { };
2339         int err;
2340
2341         do {
2342                 err = _nfs4_open_expired(ctx, state);
2343                 trace_nfs4_open_expired(ctx, 0, err);
2344                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2345                         continue;
2346                 switch (err) {
2347                 default:
2348                         goto out;
2349                 case -NFS4ERR_GRACE:
2350                 case -NFS4ERR_DELAY:
2351                         nfs4_handle_exception(server, err, &exception);
2352                         err = 0;
2353                 }
2354         } while (exception.retry);
2355 out:
2356         return err;
2357 }
2358
2359 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2360 {
2361         struct nfs_open_context *ctx;
2362         int ret;
2363
2364         ctx = nfs4_state_find_open_context(state);
2365         if (IS_ERR(ctx))
2366                 return -EAGAIN;
2367         ret = nfs4_do_open_expired(ctx, state);
2368         put_nfs_open_context(ctx);
2369         return ret;
2370 }
2371
2372 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2373                 const nfs4_stateid *stateid)
2374 {
2375         nfs_remove_bad_delegation(state->inode, stateid);
2376         write_seqlock(&state->seqlock);
2377         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2378         write_sequnlock(&state->seqlock);
2379         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2380 }
2381
2382 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2383 {
2384         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2385                 nfs_finish_clear_delegation_stateid(state, NULL);
2386 }
2387
2388 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2389 {
2390         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2391         nfs40_clear_delegation_stateid(state);
2392         return nfs4_open_expired(sp, state);
2393 }
2394
2395 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2396                 nfs4_stateid *stateid,
2397                 struct rpc_cred *cred)
2398 {
2399         return -NFS4ERR_BAD_STATEID;
2400 }
2401
2402 #if defined(CONFIG_NFS_V4_1)
2403 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2404                 nfs4_stateid *stateid,
2405                 struct rpc_cred *cred)
2406 {
2407         int status;
2408
2409         switch (stateid->type) {
2410         default:
2411                 break;
2412         case NFS4_INVALID_STATEID_TYPE:
2413         case NFS4_SPECIAL_STATEID_TYPE:
2414                 return -NFS4ERR_BAD_STATEID;
2415         case NFS4_REVOKED_STATEID_TYPE:
2416                 goto out_free;
2417         }
2418
2419         status = nfs41_test_stateid(server, stateid, cred);
2420         switch (status) {
2421         case -NFS4ERR_EXPIRED:
2422         case -NFS4ERR_ADMIN_REVOKED:
2423         case -NFS4ERR_DELEG_REVOKED:
2424                 break;
2425         default:
2426                 return status;
2427         }
2428 out_free:
2429         /* Ack the revoked state to the server */
2430         nfs41_free_stateid(server, stateid, cred, true);
2431         return -NFS4ERR_EXPIRED;
2432 }
2433
2434 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2435 {
2436         struct nfs_server *server = NFS_SERVER(state->inode);
2437         nfs4_stateid stateid;
2438         struct nfs_delegation *delegation;
2439         struct rpc_cred *cred;
2440         int status;
2441
2442         /* Get the delegation credential for use by test/free_stateid */
2443         rcu_read_lock();
2444         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2445         if (delegation == NULL) {
2446                 rcu_read_unlock();
2447                 return;
2448         }
2449
2450         nfs4_stateid_copy(&stateid, &delegation->stateid);
2451         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2452                 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2453                         &delegation->flags)) {
2454                 rcu_read_unlock();
2455                 nfs_finish_clear_delegation_stateid(state, &stateid);
2456                 return;
2457         }
2458
2459         cred = get_rpccred(delegation->cred);
2460         rcu_read_unlock();
2461         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2462         trace_nfs4_test_delegation_stateid(state, NULL, status);
2463         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2464                 nfs_finish_clear_delegation_stateid(state, &stateid);
2465
2466         put_rpccred(cred);
2467 }
2468
2469 /**
2470  * nfs41_check_expired_locks - possibly free a lock stateid
2471  *
2472  * @state: NFSv4 state for an inode
2473  *
2474  * Returns NFS_OK if recovery for this stateid is now finished.
2475  * Otherwise a negative NFS4ERR value is returned.
2476  */
2477 static int nfs41_check_expired_locks(struct nfs4_state *state)
2478 {
2479         int status, ret = NFS_OK;
2480         struct nfs4_lock_state *lsp, *prev = NULL;
2481         struct nfs_server *server = NFS_SERVER(state->inode);
2482
2483         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2484                 goto out;
2485
2486         spin_lock(&state->state_lock);
2487         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2488                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2489                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2490
2491                         atomic_inc(&lsp->ls_count);
2492                         spin_unlock(&state->state_lock);
2493
2494                         nfs4_put_lock_state(prev);
2495                         prev = lsp;
2496
2497                         status = nfs41_test_and_free_expired_stateid(server,
2498                                         &lsp->ls_stateid,
2499                                         cred);
2500                         trace_nfs4_test_lock_stateid(state, lsp, status);
2501                         if (status == -NFS4ERR_EXPIRED ||
2502                             status == -NFS4ERR_BAD_STATEID) {
2503                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2504                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2505                                 if (!recover_lost_locks)
2506                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2507                         } else if (status != NFS_OK) {
2508                                 ret = status;
2509                                 nfs4_put_lock_state(prev);
2510                                 goto out;
2511                         }
2512                         spin_lock(&state->state_lock);
2513                 }
2514         }
2515         spin_unlock(&state->state_lock);
2516         nfs4_put_lock_state(prev);
2517 out:
2518         return ret;
2519 }
2520
2521 /**
2522  * nfs41_check_open_stateid - possibly free an open stateid
2523  *
2524  * @state: NFSv4 state for an inode
2525  *
2526  * Returns NFS_OK if recovery for this stateid is now finished.
2527  * Otherwise a negative NFS4ERR value is returned.
2528  */
2529 static int nfs41_check_open_stateid(struct nfs4_state *state)
2530 {
2531         struct nfs_server *server = NFS_SERVER(state->inode);
2532         nfs4_stateid *stateid = &state->open_stateid;
2533         struct rpc_cred *cred = state->owner->so_cred;
2534         int status;
2535
2536         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2537                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2538                         if (nfs4_have_delegation(state->inode, state->state))
2539                                 return NFS_OK;
2540                         return -NFS4ERR_OPENMODE;
2541                 }
2542                 return -NFS4ERR_BAD_STATEID;
2543         }
2544         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2545         trace_nfs4_test_open_stateid(state, NULL, status);
2546         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2547                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2548                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2549                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2550                 clear_bit(NFS_OPEN_STATE, &state->flags);
2551                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2552         }
2553         if (status != NFS_OK)
2554                 return status;
2555         if (nfs_open_stateid_recover_openmode(state))
2556                 return -NFS4ERR_OPENMODE;
2557         return NFS_OK;
2558 }
2559
2560 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2561 {
2562         int status;
2563
2564         nfs41_check_delegation_stateid(state);
2565         status = nfs41_check_expired_locks(state);
2566         if (status != NFS_OK)
2567                 return status;
2568         status = nfs41_check_open_stateid(state);
2569         if (status != NFS_OK)
2570                 status = nfs4_open_expired(sp, state);
2571         return status;
2572 }
2573 #endif
2574
2575 /*
2576  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2577  * fields corresponding to attributes that were used to store the verifier.
2578  * Make sure we clobber those fields in the later setattr call
2579  */
2580 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2581                                 struct iattr *sattr, struct nfs4_label **label)
2582 {
2583         const u32 *attrset = opendata->o_res.attrset;
2584
2585         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2586             !(sattr->ia_valid & ATTR_ATIME_SET))
2587                 sattr->ia_valid |= ATTR_ATIME;
2588
2589         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2590             !(sattr->ia_valid & ATTR_MTIME_SET))
2591                 sattr->ia_valid |= ATTR_MTIME;
2592
2593         /* Except MODE, it seems harmless of setting twice. */
2594         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2595                 (attrset[1] & FATTR4_WORD1_MODE ||
2596                  attrset[2] & FATTR4_WORD2_MODE_UMASK))
2597                 sattr->ia_valid &= ~ATTR_MODE;
2598
2599         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2600                 *label = NULL;
2601 }
2602
2603 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2604                 fmode_t fmode,
2605                 int flags,
2606                 struct nfs_open_context *ctx)
2607 {
2608         struct nfs4_state_owner *sp = opendata->owner;
2609         struct nfs_server *server = sp->so_server;
2610         struct dentry *dentry;
2611         struct nfs4_state *state;
2612         unsigned int seq;
2613         int ret;
2614
2615         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2616
2617         ret = _nfs4_proc_open(opendata);
2618         if (ret != 0)
2619                 goto out;
2620
2621         state = nfs4_opendata_to_nfs4_state(opendata);
2622         ret = PTR_ERR(state);
2623         if (IS_ERR(state))
2624                 goto out;
2625         ctx->state = state;
2626         if (server->caps & NFS_CAP_POSIX_LOCK)
2627                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2628         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2629                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2630
2631         dentry = opendata->dentry;
2632         if (d_really_is_negative(dentry)) {
2633                 struct dentry *alias;
2634                 d_drop(dentry);
2635                 alias = d_exact_alias(dentry, state->inode);
2636                 if (!alias)
2637                         alias = d_splice_alias(igrab(state->inode), dentry);
2638                 /* d_splice_alias() can't fail here - it's a non-directory */
2639                 if (alias) {
2640                         dput(ctx->dentry);
2641                         ctx->dentry = dentry = alias;
2642                 }
2643                 nfs_set_verifier(dentry,
2644                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2645         }
2646
2647         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2648         if (ret != 0)
2649                 goto out;
2650
2651         if (d_inode(dentry) == state->inode) {
2652                 nfs_inode_attach_open_context(ctx);
2653                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2654                         nfs4_schedule_stateid_recovery(server, state);
2655         }
2656 out:
2657         return ret;
2658 }
2659
2660 /*
2661  * Returns a referenced nfs4_state
2662  */
2663 static int _nfs4_do_open(struct inode *dir,
2664                         struct nfs_open_context *ctx,
2665                         int flags,
2666                         struct iattr *sattr,
2667                         struct nfs4_label *label,
2668                         int *opened)
2669 {
2670         struct nfs4_state_owner  *sp;
2671         struct nfs4_state     *state = NULL;
2672         struct nfs_server       *server = NFS_SERVER(dir);
2673         struct nfs4_opendata *opendata;
2674         struct dentry *dentry = ctx->dentry;
2675         struct rpc_cred *cred = ctx->cred;
2676         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2677         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2678         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2679         struct nfs4_label *olabel = NULL;
2680         int status;
2681
2682         /* Protect against reboot recovery conflicts */
2683         status = -ENOMEM;
2684         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2685         if (sp == NULL) {
2686                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2687                 goto out_err;
2688         }
2689         status = nfs4_client_recover_expired_lease(server->nfs_client);
2690         if (status != 0)
2691                 goto err_put_state_owner;
2692         if (d_really_is_positive(dentry))
2693                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2694         status = -ENOMEM;
2695         if (d_really_is_positive(dentry))
2696                 claim = NFS4_OPEN_CLAIM_FH;
2697         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2698                         label, claim, GFP_KERNEL);
2699         if (opendata == NULL)
2700                 goto err_put_state_owner;
2701
2702         if (label) {
2703                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2704                 if (IS_ERR(olabel)) {
2705                         status = PTR_ERR(olabel);
2706                         goto err_opendata_put;
2707                 }
2708         }
2709
2710         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2711                 if (!opendata->f_attr.mdsthreshold) {
2712                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2713                         if (!opendata->f_attr.mdsthreshold)
2714                                 goto err_free_label;
2715                 }
2716                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2717         }
2718         if (d_really_is_positive(dentry))
2719                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2720
2721         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2722         if (status != 0)
2723                 goto err_free_label;
2724         state = ctx->state;
2725
2726         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2727             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2728                 nfs4_exclusive_attrset(opendata, sattr, &label);
2729                 /*
2730                  * send create attributes which was not set by open
2731                  * with an extra setattr.
2732                  */
2733                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2734                         nfs_fattr_init(opendata->o_res.f_attr);
2735                         status = nfs4_do_setattr(state->inode, cred,
2736                                         opendata->o_res.f_attr, sattr,
2737                                         ctx, label, olabel);
2738                         if (status == 0) {
2739                                 nfs_setattr_update_inode(state->inode, sattr,
2740                                                 opendata->o_res.f_attr);
2741                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2742                         }
2743                 }
2744         }
2745         if (opened && opendata->file_created)
2746                 *opened |= FILE_CREATED;
2747
2748         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2749                 *ctx_th = opendata->f_attr.mdsthreshold;
2750                 opendata->f_attr.mdsthreshold = NULL;
2751         }
2752
2753         nfs4_label_free(olabel);
2754
2755         nfs4_opendata_put(opendata);
2756         nfs4_put_state_owner(sp);
2757         return 0;
2758 err_free_label:
2759         nfs4_label_free(olabel);
2760 err_opendata_put:
2761         nfs4_opendata_put(opendata);
2762 err_put_state_owner:
2763         nfs4_put_state_owner(sp);
2764 out_err:
2765         return status;
2766 }
2767
2768
2769 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2770                                         struct nfs_open_context *ctx,
2771                                         int flags,
2772                                         struct iattr *sattr,
2773                                         struct nfs4_label *label,
2774                                         int *opened)
2775 {
2776         struct nfs_server *server = NFS_SERVER(dir);
2777         struct nfs4_exception exception = { };
2778         struct nfs4_state *res;
2779         int status;
2780
2781         do {
2782                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2783                 res = ctx->state;
2784                 trace_nfs4_open_file(ctx, flags, status);
2785                 if (status == 0)
2786                         break;
2787                 /* NOTE: BAD_SEQID means the server and client disagree about the
2788                  * book-keeping w.r.t. state-changing operations
2789                  * (OPEN/CLOSE/LOCK/LOCKU...)
2790                  * It is actually a sign of a bug on the client or on the server.
2791                  *
2792                  * If we receive a BAD_SEQID error in the particular case of
2793                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2794                  * have unhashed the old state_owner for us, and that we can
2795                  * therefore safely retry using a new one. We should still warn
2796                  * the user though...
2797                  */
2798                 if (status == -NFS4ERR_BAD_SEQID) {
2799                         pr_warn_ratelimited("NFS: v4 server %s "
2800                                         " returned a bad sequence-id error!\n",
2801                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2802                         exception.retry = 1;
2803                         continue;
2804                 }
2805                 /*
2806                  * BAD_STATEID on OPEN means that the server cancelled our
2807                  * state before it received the OPEN_CONFIRM.
2808                  * Recover by retrying the request as per the discussion
2809                  * on Page 181 of RFC3530.
2810                  */
2811                 if (status == -NFS4ERR_BAD_STATEID) {
2812                         exception.retry = 1;
2813                         continue;
2814                 }
2815                 if (status == -EAGAIN) {
2816                         /* We must have found a delegation */
2817                         exception.retry = 1;
2818                         continue;
2819                 }
2820                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2821                         continue;
2822                 res = ERR_PTR(nfs4_handle_exception(server,
2823                                         status, &exception));
2824         } while (exception.retry);
2825         return res;
2826 }
2827
2828 static int _nfs4_do_setattr(struct inode *inode,
2829                             struct nfs_setattrargs *arg,
2830                             struct nfs_setattrres *res,
2831                             struct rpc_cred *cred,
2832                             struct nfs_open_context *ctx)
2833 {
2834         struct nfs_server *server = NFS_SERVER(inode);
2835         struct rpc_message msg = {
2836                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2837                 .rpc_argp       = arg,
2838                 .rpc_resp       = res,
2839                 .rpc_cred       = cred,
2840         };
2841         struct rpc_cred *delegation_cred = NULL;
2842         unsigned long timestamp = jiffies;
2843         fmode_t fmode;
2844         bool truncate;
2845         int status;
2846
2847         nfs_fattr_init(res->fattr);
2848
2849         /* Servers should only apply open mode checks for file size changes */
2850         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2851         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2852
2853         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2854                 /* Use that stateid */
2855         } else if (truncate && ctx != NULL) {
2856                 struct nfs_lock_context *l_ctx;
2857                 if (!nfs4_valid_open_stateid(ctx->state))
2858                         return -EBADF;
2859                 l_ctx = nfs_get_lock_context(ctx);
2860                 if (IS_ERR(l_ctx))
2861                         return PTR_ERR(l_ctx);
2862                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2863                                                 &arg->stateid, &delegation_cred);
2864                 nfs_put_lock_context(l_ctx);
2865                 if (status == -EIO)
2866                         return -EBADF;
2867         } else
2868                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2869         if (delegation_cred)
2870                 msg.rpc_cred = delegation_cred;
2871
2872         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2873
2874         put_rpccred(delegation_cred);
2875         if (status == 0 && ctx != NULL)
2876                 renew_lease(server, timestamp);
2877         trace_nfs4_setattr(inode, &arg->stateid, status);
2878         return status;
2879 }
2880
2881 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2882                            struct nfs_fattr *fattr, struct iattr *sattr,
2883                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2884                            struct nfs4_label *olabel)
2885 {
2886         struct nfs_server *server = NFS_SERVER(inode);
2887         struct nfs4_state *state = ctx ? ctx->state : NULL;
2888         struct nfs_setattrargs  arg = {
2889                 .fh             = NFS_FH(inode),
2890                 .iap            = sattr,
2891                 .server         = server,
2892                 .bitmask = server->attr_bitmask,
2893                 .label          = ilabel,
2894         };
2895         struct nfs_setattrres  res = {
2896                 .fattr          = fattr,
2897                 .label          = olabel,
2898                 .server         = server,
2899         };
2900         struct nfs4_exception exception = {
2901                 .state = state,
2902                 .inode = inode,
2903                 .stateid = &arg.stateid,
2904         };
2905         int err;
2906
2907         arg.bitmask = nfs4_bitmask(server, ilabel);
2908         if (ilabel)
2909                 arg.bitmask = nfs4_bitmask(server, olabel);
2910
2911         do {
2912                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
2913                 switch (err) {
2914                 case -NFS4ERR_OPENMODE:
2915                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2916                                 pr_warn_once("NFSv4: server %s is incorrectly "
2917                                                 "applying open mode checks to "
2918                                                 "a SETATTR that is not "
2919                                                 "changing file size.\n",
2920                                                 server->nfs_client->cl_hostname);
2921                         }
2922                         if (state && !(state->state & FMODE_WRITE)) {
2923                                 err = -EBADF;
2924                                 if (sattr->ia_valid & ATTR_OPEN)
2925                                         err = -EACCES;
2926                                 goto out;
2927                         }
2928                 }
2929                 err = nfs4_handle_exception(server, err, &exception);
2930         } while (exception.retry);
2931 out:
2932         return err;
2933 }
2934
2935 static bool
2936 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2937 {
2938         if (inode == NULL || !nfs_have_layout(inode))
2939                 return false;
2940
2941         return pnfs_wait_on_layoutreturn(inode, task);
2942 }
2943
2944 struct nfs4_closedata {
2945         struct inode *inode;
2946         struct nfs4_state *state;
2947         struct nfs_closeargs arg;
2948         struct nfs_closeres res;
2949         struct {
2950                 struct nfs4_layoutreturn_args arg;
2951                 struct nfs4_layoutreturn_res res;
2952                 struct nfs4_xdr_opaque_data ld_private;
2953                 u32 roc_barrier;
2954                 bool roc;
2955         } lr;
2956         struct nfs_fattr fattr;
2957         unsigned long timestamp;
2958 };
2959
2960 static void nfs4_free_closedata(void *data)
2961 {
2962         struct nfs4_closedata *calldata = data;
2963         struct nfs4_state_owner *sp = calldata->state->owner;
2964         struct super_block *sb = calldata->state->inode->i_sb;
2965
2966         if (calldata->lr.roc)
2967                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
2968                                 calldata->res.lr_ret);
2969         nfs4_put_open_state(calldata->state);
2970         nfs_free_seqid(calldata->arg.seqid);
2971         nfs4_put_state_owner(sp);
2972         nfs_sb_deactive(sb);
2973         kfree(calldata);
2974 }
2975
2976 static void nfs4_close_done(struct rpc_task *task, void *data)
2977 {
2978         struct nfs4_closedata *calldata = data;
2979         struct nfs4_state *state = calldata->state;
2980         struct nfs_server *server = NFS_SERVER(calldata->inode);
2981         nfs4_stateid *res_stateid = NULL;
2982
2983         dprintk("%s: begin!\n", __func__);
2984         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2985                 return;
2986         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2987
2988         /* Handle Layoutreturn errors */
2989         if (calldata->arg.lr_args && task->tk_status != 0) {
2990                 switch (calldata->res.lr_ret) {
2991                 default:
2992                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
2993                         break;
2994                 case 0:
2995                         calldata->arg.lr_args = NULL;
2996                         calldata->res.lr_res = NULL;
2997                         break;
2998                 case -NFS4ERR_ADMIN_REVOKED:
2999                 case -NFS4ERR_DELEG_REVOKED:
3000                 case -NFS4ERR_EXPIRED:
3001                 case -NFS4ERR_BAD_STATEID:
3002                 case -NFS4ERR_OLD_STATEID:
3003                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3004                 case -NFS4ERR_WRONG_CRED:
3005                         calldata->arg.lr_args = NULL;
3006                         calldata->res.lr_res = NULL;
3007                         calldata->res.lr_ret = 0;
3008                         rpc_restart_call_prepare(task);
3009                         return;
3010                 }
3011         }
3012
3013         /* hmm. we are done with the inode, and in the process of freeing
3014          * the state_owner. we keep this around to process errors
3015          */
3016         switch (task->tk_status) {
3017                 case 0:
3018                         res_stateid = &calldata->res.stateid;
3019                         renew_lease(server, calldata->timestamp);
3020                         break;
3021                 case -NFS4ERR_ACCESS:
3022                         if (calldata->arg.bitmask != NULL) {
3023                                 calldata->arg.bitmask = NULL;
3024                                 calldata->res.fattr = NULL;
3025                                 task->tk_status = 0;
3026                                 rpc_restart_call_prepare(task);
3027                                 goto out_release;
3028
3029                         }
3030                         break;
3031                 case -NFS4ERR_ADMIN_REVOKED:
3032                 case -NFS4ERR_STALE_STATEID:
3033                 case -NFS4ERR_EXPIRED:
3034                         nfs4_free_revoked_stateid(server,
3035                                         &calldata->arg.stateid,
3036                                         task->tk_msg.rpc_cred);
3037                 case -NFS4ERR_OLD_STATEID:
3038                 case -NFS4ERR_BAD_STATEID:
3039                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3040                                                 &state->open_stateid)) {
3041                                 rpc_restart_call_prepare(task);
3042                                 goto out_release;
3043                         }
3044                         if (calldata->arg.fmode == 0)
3045                                 break;
3046                 default:
3047                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3048                                 rpc_restart_call_prepare(task);
3049                                 goto out_release;
3050                         }
3051         }
3052         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3053                         res_stateid, calldata->arg.fmode);
3054 out_release:
3055         nfs_release_seqid(calldata->arg.seqid);
3056         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3057         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3058 }
3059
3060 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3061 {
3062         struct nfs4_closedata *calldata = data;
3063         struct nfs4_state *state = calldata->state;
3064         struct inode *inode = calldata->inode;
3065         bool is_rdonly, is_wronly, is_rdwr;
3066         int call_close = 0;
3067
3068         dprintk("%s: begin!\n", __func__);
3069         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3070                 goto out_wait;
3071
3072         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3073         spin_lock(&state->owner->so_lock);
3074         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3075         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3076         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3077         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3078         /* Calculate the change in open mode */
3079         calldata->arg.fmode = 0;
3080         if (state->n_rdwr == 0) {
3081                 if (state->n_rdonly == 0)
3082                         call_close |= is_rdonly;
3083                 else if (is_rdonly)
3084                         calldata->arg.fmode |= FMODE_READ;
3085                 if (state->n_wronly == 0)
3086                         call_close |= is_wronly;
3087                 else if (is_wronly)
3088                         calldata->arg.fmode |= FMODE_WRITE;
3089                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3090                         call_close |= is_rdwr;
3091         } else if (is_rdwr)
3092                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3093
3094         if (!nfs4_valid_open_stateid(state) ||
3095             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3096                 call_close = 0;
3097         spin_unlock(&state->owner->so_lock);
3098
3099         if (!call_close) {
3100                 /* Note: exit _without_ calling nfs4_close_done */
3101                 goto out_no_action;
3102         }
3103
3104         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3105                 nfs_release_seqid(calldata->arg.seqid);
3106                 goto out_wait;
3107         }
3108
3109         if (calldata->arg.fmode == 0)
3110                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3111
3112         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3113                 /* Close-to-open cache consistency revalidation */
3114                 if (!nfs4_have_delegation(inode, FMODE_READ))
3115                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3116                 else
3117                         calldata->arg.bitmask = NULL;
3118         }
3119
3120         calldata->arg.share_access =
3121                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3122                                 calldata->arg.fmode, 0);
3123
3124         if (calldata->res.fattr == NULL)
3125                 calldata->arg.bitmask = NULL;
3126         else if (calldata->arg.bitmask == NULL)
3127                 calldata->res.fattr = NULL;
3128         calldata->timestamp = jiffies;
3129         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3130                                 &calldata->arg.seq_args,
3131                                 &calldata->res.seq_res,
3132                                 task) != 0)
3133                 nfs_release_seqid(calldata->arg.seqid);
3134         dprintk("%s: done!\n", __func__);
3135         return;
3136 out_no_action:
3137         task->tk_action = NULL;
3138 out_wait:
3139         nfs4_sequence_done(task, &calldata->res.seq_res);
3140 }
3141
3142 static const struct rpc_call_ops nfs4_close_ops = {
3143         .rpc_call_prepare = nfs4_close_prepare,
3144         .rpc_call_done = nfs4_close_done,
3145         .rpc_release = nfs4_free_closedata,
3146 };
3147
3148 /*
3149  * It is possible for data to be read/written from a mem-mapped file
3150  * after the sys_close call (which hits the vfs layer as a flush).
3151  * This means that we can't safely call nfsv4 close on a file until
3152  * the inode is cleared. This in turn means that we are not good
3153  * NFSv4 citizens - we do not indicate to the server to update the file's
3154  * share state even when we are done with one of the three share
3155  * stateid's in the inode.
3156  *
3157  * NOTE: Caller must be holding the sp->so_owner semaphore!
3158  */
3159 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3160 {
3161         struct nfs_server *server = NFS_SERVER(state->inode);
3162         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3163         struct nfs4_closedata *calldata;
3164         struct nfs4_state_owner *sp = state->owner;
3165         struct rpc_task *task;
3166         struct rpc_message msg = {
3167                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3168                 .rpc_cred = state->owner->so_cred,
3169         };
3170         struct rpc_task_setup task_setup_data = {
3171                 .rpc_client = server->client,
3172                 .rpc_message = &msg,
3173                 .callback_ops = &nfs4_close_ops,
3174                 .workqueue = nfsiod_workqueue,
3175                 .flags = RPC_TASK_ASYNC,
3176         };
3177         int status = -ENOMEM;
3178
3179         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3180                 &task_setup_data.rpc_client, &msg);
3181
3182         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3183         if (calldata == NULL)
3184                 goto out;
3185         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3186         calldata->inode = state->inode;
3187         calldata->state = state;
3188         calldata->arg.fh = NFS_FH(state->inode);
3189         /* Serialization for the sequence id */
3190         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3191         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3192         if (IS_ERR(calldata->arg.seqid))
3193                 goto out_free_calldata;
3194         nfs_fattr_init(&calldata->fattr);
3195         calldata->arg.fmode = 0;
3196         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3197         calldata->res.fattr = &calldata->fattr;
3198         calldata->res.seqid = calldata->arg.seqid;
3199         calldata->res.server = server;
3200         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3201         calldata->lr.roc = pnfs_roc(state->inode,
3202                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3203         if (calldata->lr.roc) {
3204                 calldata->arg.lr_args = &calldata->lr.arg;
3205                 calldata->res.lr_res = &calldata->lr.res;
3206         }
3207         nfs_sb_active(calldata->inode->i_sb);
3208
3209         msg.rpc_argp = &calldata->arg;
3210         msg.rpc_resp = &calldata->res;
3211         task_setup_data.callback_data = calldata;
3212         task = rpc_run_task(&task_setup_data);
3213         if (IS_ERR(task))
3214                 return PTR_ERR(task);
3215         status = 0;
3216         if (wait)
3217                 status = rpc_wait_for_completion_task(task);
3218         rpc_put_task(task);
3219         return status;
3220 out_free_calldata:
3221         kfree(calldata);
3222 out:
3223         nfs4_put_open_state(state);
3224         nfs4_put_state_owner(sp);
3225         return status;
3226 }
3227
3228 static struct inode *
3229 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3230                 int open_flags, struct iattr *attr, int *opened)
3231 {
3232         struct nfs4_state *state;
3233         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3234
3235         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3236
3237         /* Protect against concurrent sillydeletes */
3238         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3239
3240         nfs4_label_release_security(label);
3241
3242         if (IS_ERR(state))
3243                 return ERR_CAST(state);
3244         return state->inode;
3245 }
3246
3247 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3248 {
3249         if (ctx->state == NULL)
3250                 return;
3251         if (is_sync)
3252                 nfs4_close_sync(ctx->state, ctx->mode);
3253         else
3254                 nfs4_close_state(ctx->state, ctx->mode);
3255 }
3256
3257 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3258 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3259 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3260
3261 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3262 {
3263         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3264         struct nfs4_server_caps_arg args = {
3265                 .fhandle = fhandle,
3266                 .bitmask = bitmask,
3267         };
3268         struct nfs4_server_caps_res res = {};
3269         struct rpc_message msg = {
3270                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3271                 .rpc_argp = &args,
3272                 .rpc_resp = &res,
3273         };
3274         int status;
3275         int i;
3276
3277         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3278                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3279                      FATTR4_WORD0_LINK_SUPPORT |
3280                      FATTR4_WORD0_SYMLINK_SUPPORT |
3281                      FATTR4_WORD0_ACLSUPPORT;
3282         if (minorversion)
3283                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3284
3285         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3286         if (status == 0) {
3287                 /* Sanity check the server answers */
3288                 switch (minorversion) {
3289                 case 0:
3290                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3291                         res.attr_bitmask[2] = 0;
3292                         break;
3293                 case 1:
3294                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3295                         break;
3296                 case 2:
3297                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3298                 }
3299                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3300                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3301                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3302                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3303                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3304                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3305                                 NFS_CAP_SECURITY_LABEL);
3306                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3307                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3308                         server->caps |= NFS_CAP_ACLS;
3309                 if (res.has_links != 0)
3310                         server->caps |= NFS_CAP_HARDLINKS;
3311                 if (res.has_symlinks != 0)
3312                         server->caps |= NFS_CAP_SYMLINKS;
3313                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3314                         server->caps |= NFS_CAP_FILEID;
3315                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3316                         server->caps |= NFS_CAP_MODE;
3317                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3318                         server->caps |= NFS_CAP_NLINK;
3319                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3320                         server->caps |= NFS_CAP_OWNER;
3321                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3322                         server->caps |= NFS_CAP_OWNER_GROUP;
3323                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3324                         server->caps |= NFS_CAP_ATIME;
3325                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3326                         server->caps |= NFS_CAP_CTIME;
3327                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3328                         server->caps |= NFS_CAP_MTIME;
3329 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3330                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3331                         server->caps |= NFS_CAP_SECURITY_LABEL;
3332 #endif
3333                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3334                                 sizeof(server->attr_bitmask));
3335                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3336
3337                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3338                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3339                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3340                 server->cache_consistency_bitmask[2] = 0;
3341
3342                 /* Avoid a regression due to buggy server */
3343                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3344                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3345                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3346                         sizeof(server->exclcreat_bitmask));
3347
3348                 server->acl_bitmask = res.acl_bitmask;
3349                 server->fh_expire_type = res.fh_expire_type;
3350         }
3351
3352         return status;
3353 }
3354
3355 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3356 {
3357         struct nfs4_exception exception = { };
3358         int err;
3359         do {
3360                 err = nfs4_handle_exception(server,
3361                                 _nfs4_server_capabilities(server, fhandle),
3362                                 &exception);
3363         } while (exception.retry);
3364         return err;
3365 }
3366
3367 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3368                 struct nfs_fsinfo *info)
3369 {
3370         u32 bitmask[3];
3371         struct nfs4_lookup_root_arg args = {
3372                 .bitmask = bitmask,
3373         };
3374         struct nfs4_lookup_res res = {
3375                 .server = server,
3376                 .fattr = info->fattr,
3377                 .fh = fhandle,
3378         };
3379         struct rpc_message msg = {
3380                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3381                 .rpc_argp = &args,
3382                 .rpc_resp = &res,
3383         };
3384
3385         bitmask[0] = nfs4_fattr_bitmap[0];
3386         bitmask[1] = nfs4_fattr_bitmap[1];
3387         /*
3388          * Process the label in the upcoming getfattr
3389          */
3390         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3391
3392         nfs_fattr_init(info->fattr);
3393         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3394 }
3395
3396 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3397                 struct nfs_fsinfo *info)
3398 {
3399         struct nfs4_exception exception = { };
3400         int err;
3401         do {
3402                 err = _nfs4_lookup_root(server, fhandle, info);
3403                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3404                 switch (err) {
3405                 case 0:
3406                 case -NFS4ERR_WRONGSEC:
3407                         goto out;
3408                 default:
3409                         err = nfs4_handle_exception(server, err, &exception);
3410                 }
3411         } while (exception.retry);
3412 out:
3413         return err;
3414 }
3415
3416 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3417                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3418 {
3419         struct rpc_auth_create_args auth_args = {
3420                 .pseudoflavor = flavor,
3421         };
3422         struct rpc_auth *auth;
3423
3424         auth = rpcauth_create(&auth_args, server->client);
3425         if (IS_ERR(auth))
3426                 return -EACCES;
3427         return nfs4_lookup_root(server, fhandle, info);
3428 }
3429
3430 /*
3431  * Retry pseudoroot lookup with various security flavors.  We do this when:
3432  *
3433  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3434  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3435  *
3436  * Returns zero on success, or a negative NFS4ERR value, or a
3437  * negative errno value.
3438  */
3439 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3440                               struct nfs_fsinfo *info)
3441 {
3442         /* Per 3530bis 15.33.5 */
3443         static const rpc_authflavor_t flav_array[] = {
3444                 RPC_AUTH_GSS_KRB5P,
3445                 RPC_AUTH_GSS_KRB5I,
3446                 RPC_AUTH_GSS_KRB5,
3447                 RPC_AUTH_UNIX,                  /* courtesy */
3448                 RPC_AUTH_NULL,
3449         };
3450         int status = -EPERM;
3451         size_t i;
3452
3453         if (server->auth_info.flavor_len > 0) {
3454                 /* try each flavor specified by user */
3455                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3456                         status = nfs4_lookup_root_sec(server, fhandle, info,
3457                                                 server->auth_info.flavors[i]);
3458                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3459                                 continue;
3460                         break;
3461                 }
3462         } else {
3463                 /* no flavors specified by user, try default list */
3464                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3465                         status = nfs4_lookup_root_sec(server, fhandle, info,
3466                                                       flav_array[i]);
3467                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3468                                 continue;
3469                         break;
3470                 }
3471         }
3472
3473         /*
3474          * -EACCESS could mean that the user doesn't have correct permissions
3475          * to access the mount.  It could also mean that we tried to mount
3476          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3477          * existing mount programs don't handle -EACCES very well so it should
3478          * be mapped to -EPERM instead.
3479          */
3480         if (status == -EACCES)
3481                 status = -EPERM;
3482         return status;
3483 }
3484
3485 /**
3486  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3487  * @server: initialized nfs_server handle
3488  * @fhandle: we fill in the pseudo-fs root file handle
3489  * @info: we fill in an FSINFO struct
3490  * @auth_probe: probe the auth flavours
3491  *
3492  * Returns zero on success, or a negative errno.
3493  */
3494 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3495                          struct nfs_fsinfo *info,
3496                          bool auth_probe)
3497 {
3498         int status = 0;
3499
3500         if (!auth_probe)
3501                 status = nfs4_lookup_root(server, fhandle, info);
3502
3503         if (auth_probe || status == NFS4ERR_WRONGSEC)
3504                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3505                                 fhandle, info);
3506
3507         if (status == 0)
3508                 status = nfs4_server_capabilities(server, fhandle);
3509         if (status == 0)
3510                 status = nfs4_do_fsinfo(server, fhandle, info);
3511
3512         return nfs4_map_errors(status);
3513 }
3514
3515 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3516                               struct nfs_fsinfo *info)
3517 {
3518         int error;
3519         struct nfs_fattr *fattr = info->fattr;
3520         struct nfs4_label *label = NULL;
3521
3522         error = nfs4_server_capabilities(server, mntfh);
3523         if (error < 0) {
3524                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3525                 return error;
3526         }
3527
3528         label = nfs4_label_alloc(server, GFP_KERNEL);
3529         if (IS_ERR(label))
3530                 return PTR_ERR(label);
3531
3532         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3533         if (error < 0) {
3534                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3535                 goto err_free_label;
3536         }
3537
3538         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3539             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3540                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3541
3542 err_free_label:
3543         nfs4_label_free(label);
3544
3545         return error;
3546 }
3547
3548 /*
3549  * Get locations and (maybe) other attributes of a referral.
3550  * Note that we'll actually follow the referral later when
3551  * we detect fsid mismatch in inode revalidation
3552  */
3553 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3554                              const struct qstr *name, struct nfs_fattr *fattr,
3555                              struct nfs_fh *fhandle)
3556 {
3557         int status = -ENOMEM;
3558         struct page *page = NULL;
3559         struct nfs4_fs_locations *locations = NULL;
3560
3561         page = alloc_page(GFP_KERNEL);
3562         if (page == NULL)
3563                 goto out;
3564         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3565         if (locations == NULL)
3566                 goto out;
3567
3568         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3569         if (status != 0)
3570                 goto out;
3571
3572         /*
3573          * If the fsid didn't change, this is a migration event, not a
3574          * referral.  Cause us to drop into the exception handler, which
3575          * will kick off migration recovery.
3576          */
3577         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3578                 dprintk("%s: server did not return a different fsid for"
3579                         " a referral at %s\n", __func__, name->name);
3580                 status = -NFS4ERR_MOVED;
3581                 goto out;
3582         }
3583         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3584         nfs_fixup_referral_attributes(&locations->fattr);
3585
3586         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3587         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3588         memset(fhandle, 0, sizeof(struct nfs_fh));
3589 out:
3590         if (page)
3591                 __free_page(page);
3592         kfree(locations);
3593         return status;
3594 }
3595
3596 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3597                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3598 {
3599         struct nfs4_getattr_arg args = {
3600                 .fh = fhandle,
3601                 .bitmask = server->attr_bitmask,
3602         };
3603         struct nfs4_getattr_res res = {
3604                 .fattr = fattr,
3605                 .label = label,
3606                 .server = server,
3607         };
3608         struct rpc_message msg = {
3609                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3610                 .rpc_argp = &args,
3611                 .rpc_resp = &res,
3612         };
3613
3614         args.bitmask = nfs4_bitmask(server, label);
3615
3616         nfs_fattr_init(fattr);
3617         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3618 }
3619
3620 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3621                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3622 {
3623         struct nfs4_exception exception = { };
3624         int err;
3625         do {
3626                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3627                 trace_nfs4_getattr(server, fhandle, fattr, err);
3628                 err = nfs4_handle_exception(server, err,
3629                                 &exception);
3630         } while (exception.retry);
3631         return err;
3632 }
3633
3634 /*
3635  * The file is not closed if it is opened due to the a request to change
3636  * the size of the file. The open call will not be needed once the
3637  * VFS layer lookup-intents are implemented.
3638  *
3639  * Close is called when the inode is destroyed.
3640  * If we haven't opened the file for O_WRONLY, we
3641  * need to in the size_change case to obtain a stateid.
3642  *
3643  * Got race?
3644  * Because OPEN is always done by name in nfsv4, it is
3645  * possible that we opened a different file by the same
3646  * name.  We can recognize this race condition, but we
3647  * can't do anything about it besides returning an error.
3648  *
3649  * This will be fixed with VFS changes (lookup-intent).
3650  */
3651 static int
3652 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3653                   struct iattr *sattr)
3654 {
3655         struct inode *inode = d_inode(dentry);
3656         struct rpc_cred *cred = NULL;
3657         struct nfs_open_context *ctx = NULL;
3658         struct nfs4_label *label = NULL;
3659         int status;
3660
3661         if (pnfs_ld_layoutret_on_setattr(inode) &&
3662             sattr->ia_valid & ATTR_SIZE &&
3663             sattr->ia_size < i_size_read(inode))
3664                 pnfs_commit_and_return_layout(inode);
3665
3666         nfs_fattr_init(fattr);
3667
3668         /* Deal with open(O_TRUNC) */
3669         if (sattr->ia_valid & ATTR_OPEN)
3670                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3671
3672         /* Optimization: if the end result is no change, don't RPC */
3673         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3674                 return 0;
3675
3676         /* Search for an existing open(O_WRITE) file */
3677         if (sattr->ia_valid & ATTR_FILE) {
3678
3679                 ctx = nfs_file_open_context(sattr->ia_file);
3680                 if (ctx)
3681                         cred = ctx->cred;
3682         }
3683
3684         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3685         if (IS_ERR(label))
3686                 return PTR_ERR(label);
3687
3688         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3689         if (status == 0) {
3690                 nfs_setattr_update_inode(inode, sattr, fattr);
3691                 nfs_setsecurity(inode, fattr, label);
3692         }
3693         nfs4_label_free(label);
3694         return status;
3695 }
3696
3697 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3698                 const struct qstr *name, struct nfs_fh *fhandle,
3699                 struct nfs_fattr *fattr, struct nfs4_label *label)
3700 {
3701         struct nfs_server *server = NFS_SERVER(dir);
3702         int                    status;
3703         struct nfs4_lookup_arg args = {
3704                 .bitmask = server->attr_bitmask,
3705                 .dir_fh = NFS_FH(dir),
3706                 .name = name,
3707         };
3708         struct nfs4_lookup_res res = {
3709                 .server = server,
3710                 .fattr = fattr,
3711                 .label = label,
3712                 .fh = fhandle,
3713         };
3714         struct rpc_message msg = {
3715                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3716                 .rpc_argp = &args,
3717                 .rpc_resp = &res,
3718         };
3719
3720         args.bitmask = nfs4_bitmask(server, label);
3721
3722         nfs_fattr_init(fattr);
3723
3724         dprintk("NFS call  lookup %s\n", name->name);
3725         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3726         dprintk("NFS reply lookup: %d\n", status);
3727         return status;
3728 }
3729
3730 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3731 {
3732         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3733                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3734         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3735         fattr->nlink = 2;
3736 }
3737
3738 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3739                                    const struct qstr *name, struct nfs_fh *fhandle,
3740                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3741 {
3742         struct nfs4_exception exception = { };
3743         struct rpc_clnt *client = *clnt;
3744         int err;
3745         do {
3746                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3747                 trace_nfs4_lookup(dir, name, err);
3748                 switch (err) {
3749                 case -NFS4ERR_BADNAME:
3750                         err = -ENOENT;
3751                         goto out;
3752                 case -NFS4ERR_MOVED:
3753                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3754                         if (err == -NFS4ERR_MOVED)
3755                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3756                         goto out;
3757                 case -NFS4ERR_WRONGSEC:
3758                         err = -EPERM;
3759                         if (client != *clnt)
3760                                 goto out;
3761                         client = nfs4_negotiate_security(client, dir, name);
3762                         if (IS_ERR(client))
3763                                 return PTR_ERR(client);
3764
3765                         exception.retry = 1;
3766                         break;
3767                 default:
3768                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3769                 }
3770         } while (exception.retry);
3771
3772 out:
3773         if (err == 0)
3774                 *clnt = client;
3775         else if (client != *clnt)
3776                 rpc_shutdown_client(client);
3777
3778         return err;
3779 }
3780
3781 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3782                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3783                             struct nfs4_label *label)
3784 {
3785         int status;
3786         struct rpc_clnt *client = NFS_CLIENT(dir);
3787
3788         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3789         if (client != NFS_CLIENT(dir)) {
3790                 rpc_shutdown_client(client);
3791                 nfs_fixup_secinfo_attributes(fattr);
3792         }
3793         return status;
3794 }
3795
3796 struct rpc_clnt *
3797 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3798                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3799 {
3800         struct rpc_clnt *client = NFS_CLIENT(dir);
3801         int status;
3802
3803         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3804         if (status < 0)
3805                 return ERR_PTR(status);
3806         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3807 }
3808
3809 static int _nfs4_proc_lookupp(struct inode *inode,
3810                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3811                 struct nfs4_label *label)
3812 {
3813         struct rpc_clnt *clnt = NFS_CLIENT(inode);
3814         struct nfs_server *server = NFS_SERVER(inode);
3815         int                    status;
3816         struct nfs4_lookupp_arg args = {
3817                 .bitmask = server->attr_bitmask,
3818                 .fh = NFS_FH(inode),
3819         };
3820         struct nfs4_lookupp_res res = {
3821                 .server = server,
3822                 .fattr = fattr,
3823                 .label = label,
3824                 .fh = fhandle,
3825         };
3826         struct rpc_message msg = {
3827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
3828                 .rpc_argp = &args,
3829                 .rpc_resp = &res,
3830         };
3831
3832         args.bitmask = nfs4_bitmask(server, label);
3833
3834         nfs_fattr_init(fattr);
3835
3836         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
3837         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
3838                                 &res.seq_res, 0);
3839         dprintk("NFS reply lookupp: %d\n", status);
3840         return status;
3841 }
3842
3843 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
3844                              struct nfs_fattr *fattr, struct nfs4_label *label)
3845 {
3846         struct nfs4_exception exception = { };
3847         int err;
3848         do {
3849                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
3850                 trace_nfs4_lookupp(inode, err);
3851                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3852                                 &exception);
3853         } while (exception.retry);
3854         return err;
3855 }
3856
3857 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3858 {
3859         struct nfs_server *server = NFS_SERVER(inode);
3860         struct nfs4_accessargs args = {
3861                 .fh = NFS_FH(inode),
3862                 .bitmask = server->cache_consistency_bitmask,
3863         };
3864         struct nfs4_accessres res = {
3865                 .server = server,
3866         };
3867         struct rpc_message msg = {
3868                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3869                 .rpc_argp = &args,
3870                 .rpc_resp = &res,
3871                 .rpc_cred = entry->cred,
3872         };
3873         int mode = entry->mask;
3874         int status = 0;
3875
3876         /*
3877          * Determine which access bits we want to ask for...
3878          */
3879         if (mode & MAY_READ)
3880                 args.access |= NFS4_ACCESS_READ;
3881         if (S_ISDIR(inode->i_mode)) {
3882                 if (mode & MAY_WRITE)
3883                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3884                 if (mode & MAY_EXEC)
3885                         args.access |= NFS4_ACCESS_LOOKUP;
3886         } else {
3887                 if (mode & MAY_WRITE)
3888                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3889                 if (mode & MAY_EXEC)
3890                         args.access |= NFS4_ACCESS_EXECUTE;
3891         }
3892
3893         res.fattr = nfs_alloc_fattr();
3894         if (res.fattr == NULL)
3895                 return -ENOMEM;
3896
3897         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3898         if (!status) {
3899                 nfs_access_set_mask(entry, res.access);
3900                 nfs_refresh_inode(inode, res.fattr);
3901         }
3902         nfs_free_fattr(res.fattr);
3903         return status;
3904 }
3905
3906 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3907 {
3908         struct nfs4_exception exception = { };
3909         int err;
3910         do {
3911                 err = _nfs4_proc_access(inode, entry);
3912                 trace_nfs4_access(inode, err);
3913                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3914                                 &exception);
3915         } while (exception.retry);
3916         return err;
3917 }
3918
3919 /*
3920  * TODO: For the time being, we don't try to get any attributes
3921  * along with any of the zero-copy operations READ, READDIR,
3922  * READLINK, WRITE.
3923  *
3924  * In the case of the first three, we want to put the GETATTR
3925  * after the read-type operation -- this is because it is hard
3926  * to predict the length of a GETATTR response in v4, and thus
3927  * align the READ data correctly.  This means that the GETATTR
3928  * may end up partially falling into the page cache, and we should
3929  * shift it into the 'tail' of the xdr_buf before processing.
3930  * To do this efficiently, we need to know the total length
3931  * of data received, which doesn't seem to be available outside
3932  * of the RPC layer.
3933  *
3934  * In the case of WRITE, we also want to put the GETATTR after
3935  * the operation -- in this case because we want to make sure
3936  * we get the post-operation mtime and size.
3937  *
3938  * Both of these changes to the XDR layer would in fact be quite
3939  * minor, but I decided to leave them for a subsequent patch.
3940  */
3941 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3942                 unsigned int pgbase, unsigned int pglen)
3943 {
3944         struct nfs4_readlink args = {
3945                 .fh       = NFS_FH(inode),
3946                 .pgbase   = pgbase,
3947                 .pglen    = pglen,
3948                 .pages    = &page,
3949         };
3950         struct nfs4_readlink_res res;
3951         struct rpc_message msg = {
3952                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3953                 .rpc_argp = &args,
3954                 .rpc_resp = &res,
3955         };
3956
3957         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3958 }
3959
3960 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3961                 unsigned int pgbase, unsigned int pglen)
3962 {
3963         struct nfs4_exception exception = { };
3964         int err;
3965         do {
3966                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3967                 trace_nfs4_readlink(inode, err);
3968                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3969                                 &exception);
3970         } while (exception.retry);
3971         return err;
3972 }
3973
3974 /*
3975  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3976  */
3977 static int
3978 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3979                  int flags)
3980 {
3981         struct nfs_server *server = NFS_SERVER(dir);
3982         struct nfs4_label l, *ilabel = NULL;
3983         struct nfs_open_context *ctx;
3984         struct nfs4_state *state;
3985         int status = 0;
3986
3987         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
3988         if (IS_ERR(ctx))
3989                 return PTR_ERR(ctx);
3990
3991         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3992
3993         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
3994                 sattr->ia_mode &= ~current_umask();
3995         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3996         if (IS_ERR(state)) {
3997                 status = PTR_ERR(state);
3998                 goto out;
3999         }
4000 out:
4001         nfs4_label_release_security(ilabel);
4002         put_nfs_open_context(ctx);
4003         return status;
4004 }
4005
4006 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4007 {
4008         struct nfs_server *server = NFS_SERVER(dir);
4009         struct nfs_removeargs args = {
4010                 .fh = NFS_FH(dir),
4011                 .name = *name,
4012         };
4013         struct nfs_removeres res = {
4014                 .server = server,
4015         };
4016         struct rpc_message msg = {
4017                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4018                 .rpc_argp = &args,
4019                 .rpc_resp = &res,
4020         };
4021         unsigned long timestamp = jiffies;
4022         int status;
4023
4024         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4025         if (status == 0)
4026                 update_changeattr(dir, &res.cinfo, timestamp);
4027         return status;
4028 }
4029
4030 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4031 {
4032         struct nfs4_exception exception = { };
4033         int err;
4034         do {
4035                 err = _nfs4_proc_remove(dir, name);
4036                 trace_nfs4_remove(dir, name, err);
4037                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4038                                 &exception);
4039         } while (exception.retry);
4040         return err;
4041 }
4042
4043 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4044 {
4045         struct nfs_server *server = NFS_SERVER(dir);
4046         struct nfs_removeargs *args = msg->rpc_argp;
4047         struct nfs_removeres *res = msg->rpc_resp;
4048
4049         res->server = server;
4050         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4051         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4052
4053         nfs_fattr_init(res->dir_attr);
4054 }
4055
4056 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4057 {
4058         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4059                         &data->args.seq_args,
4060                         &data->res.seq_res,
4061                         task);
4062 }
4063
4064 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4065 {
4066         struct nfs_unlinkdata *data = task->tk_calldata;
4067         struct nfs_removeres *res = &data->res;
4068
4069         if (!nfs4_sequence_done(task, &res->seq_res))
4070                 return 0;
4071         if (nfs4_async_handle_error(task, res->server, NULL,
4072                                     &data->timeout) == -EAGAIN)
4073                 return 0;
4074         if (task->tk_status == 0)
4075                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4076         return 1;
4077 }
4078
4079 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4080 {
4081         struct nfs_server *server = NFS_SERVER(dir);
4082         struct nfs_renameargs *arg = msg->rpc_argp;
4083         struct nfs_renameres *res = msg->rpc_resp;
4084
4085         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4086         res->server = server;
4087         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4088 }
4089
4090 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4091 {
4092         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4093                         &data->args.seq_args,
4094                         &data->res.seq_res,
4095                         task);
4096 }
4097
4098 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4099                                  struct inode *new_dir)
4100 {
4101         struct nfs_renamedata *data = task->tk_calldata;
4102         struct nfs_renameres *res = &data->res;
4103
4104         if (!nfs4_sequence_done(task, &res->seq_res))
4105                 return 0;
4106         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4107                 return 0;
4108
4109         if (task->tk_status == 0) {
4110                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4111                 if (new_dir != old_dir)
4112                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4113         }
4114         return 1;
4115 }
4116
4117 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4118 {
4119         struct nfs_server *server = NFS_SERVER(inode);
4120         struct nfs4_link_arg arg = {
4121                 .fh     = NFS_FH(inode),
4122                 .dir_fh = NFS_FH(dir),
4123                 .name   = name,
4124                 .bitmask = server->attr_bitmask,
4125         };
4126         struct nfs4_link_res res = {
4127                 .server = server,
4128                 .label = NULL,
4129         };
4130         struct rpc_message msg = {
4131                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4132                 .rpc_argp = &arg,
4133                 .rpc_resp = &res,
4134         };
4135         int status = -ENOMEM;
4136
4137         res.fattr = nfs_alloc_fattr();
4138         if (res.fattr == NULL)
4139                 goto out;
4140
4141         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4142         if (IS_ERR(res.label)) {
4143                 status = PTR_ERR(res.label);
4144                 goto out;
4145         }
4146         arg.bitmask = nfs4_bitmask(server, res.label);
4147
4148         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4149         if (!status) {
4150                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4151                 status = nfs_post_op_update_inode(inode, res.fattr);
4152                 if (!status)
4153                         nfs_setsecurity(inode, res.fattr, res.label);
4154         }
4155
4156
4157         nfs4_label_free(res.label);
4158
4159 out:
4160         nfs_free_fattr(res.fattr);
4161         return status;
4162 }
4163
4164 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4165 {
4166         struct nfs4_exception exception = { };
4167         int err;
4168         do {
4169                 err = nfs4_handle_exception(NFS_SERVER(inode),
4170                                 _nfs4_proc_link(inode, dir, name),
4171                                 &exception);
4172         } while (exception.retry);
4173         return err;
4174 }
4175
4176 struct nfs4_createdata {
4177         struct rpc_message msg;
4178         struct nfs4_create_arg arg;
4179         struct nfs4_create_res res;
4180         struct nfs_fh fh;
4181         struct nfs_fattr fattr;
4182         struct nfs4_label *label;
4183 };
4184
4185 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4186                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4187 {
4188         struct nfs4_createdata *data;
4189
4190         data = kzalloc(sizeof(*data), GFP_KERNEL);
4191         if (data != NULL) {
4192                 struct nfs_server *server = NFS_SERVER(dir);
4193
4194                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4195                 if (IS_ERR(data->label))
4196                         goto out_free;
4197
4198                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4199                 data->msg.rpc_argp = &data->arg;
4200                 data->msg.rpc_resp = &data->res;
4201                 data->arg.dir_fh = NFS_FH(dir);
4202                 data->arg.server = server;
4203                 data->arg.name = name;
4204                 data->arg.attrs = sattr;
4205                 data->arg.ftype = ftype;
4206                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4207                 data->arg.umask = current_umask();
4208                 data->res.server = server;
4209                 data->res.fh = &data->fh;
4210                 data->res.fattr = &data->fattr;
4211                 data->res.label = data->label;
4212                 nfs_fattr_init(data->res.fattr);
4213         }
4214         return data;
4215 out_free:
4216         kfree(data);
4217         return NULL;
4218 }
4219
4220 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4221 {
4222         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4223                                     &data->arg.seq_args, &data->res.seq_res, 1);
4224         if (status == 0) {
4225                 update_changeattr(dir, &data->res.dir_cinfo,
4226                                 data->res.fattr->time_start);
4227                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4228         }
4229         return status;
4230 }
4231
4232 static void nfs4_free_createdata(struct nfs4_createdata *data)
4233 {
4234         nfs4_label_free(data->label);
4235         kfree(data);
4236 }
4237
4238 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4239                 struct page *page, unsigned int len, struct iattr *sattr,
4240                 struct nfs4_label *label)
4241 {
4242         struct nfs4_createdata *data;
4243         int status = -ENAMETOOLONG;
4244
4245         if (len > NFS4_MAXPATHLEN)
4246                 goto out;
4247
4248         status = -ENOMEM;
4249         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4250         if (data == NULL)
4251                 goto out;
4252
4253         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4254         data->arg.u.symlink.pages = &page;
4255         data->arg.u.symlink.len = len;
4256         data->arg.label = label;
4257
4258         status = nfs4_do_create(dir, dentry, data);
4259
4260         nfs4_free_createdata(data);
4261 out:
4262         return status;
4263 }
4264
4265 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4266                 struct page *page, unsigned int len, struct iattr *sattr)
4267 {
4268         struct nfs4_exception exception = { };
4269         struct nfs4_label l, *label = NULL;
4270         int err;
4271
4272         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4273
4274         do {
4275                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4276                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4277                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4278                                 &exception);
4279         } while (exception.retry);
4280
4281         nfs4_label_release_security(label);
4282         return err;
4283 }
4284
4285 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4286                 struct iattr *sattr, struct nfs4_label *label)
4287 {
4288         struct nfs4_createdata *data;
4289         int status = -ENOMEM;
4290
4291         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4292         if (data == NULL)
4293                 goto out;
4294
4295         data->arg.label = label;
4296         status = nfs4_do_create(dir, dentry, data);
4297
4298         nfs4_free_createdata(data);
4299 out:
4300         return status;
4301 }
4302
4303 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4304                 struct iattr *sattr)
4305 {
4306         struct nfs_server *server = NFS_SERVER(dir);
4307         struct nfs4_exception exception = { };
4308         struct nfs4_label l, *label = NULL;
4309         int err;
4310
4311         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4312
4313         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4314                 sattr->ia_mode &= ~current_umask();
4315         do {
4316                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4317                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4318                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4319                                 &exception);
4320         } while (exception.retry);
4321         nfs4_label_release_security(label);
4322
4323         return err;
4324 }
4325
4326 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4327                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4328 {
4329         struct inode            *dir = d_inode(dentry);
4330         struct nfs4_readdir_arg args = {
4331                 .fh = NFS_FH(dir),
4332                 .pages = pages,
4333                 .pgbase = 0,
4334                 .count = count,
4335                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4336                 .plus = plus,
4337         };
4338         struct nfs4_readdir_res res;
4339         struct rpc_message msg = {
4340                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4341                 .rpc_argp = &args,
4342                 .rpc_resp = &res,
4343                 .rpc_cred = cred,
4344         };
4345         int                     status;
4346
4347         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4348                         dentry,
4349                         (unsigned long long)cookie);
4350         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4351         res.pgbase = args.pgbase;
4352         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4353         if (status >= 0) {
4354                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4355                 status += args.pgbase;
4356         }
4357
4358         nfs_invalidate_atime(dir);
4359
4360         dprintk("%s: returns %d\n", __func__, status);
4361         return status;
4362 }
4363
4364 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4365                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4366 {
4367         struct nfs4_exception exception = { };
4368         int err;
4369         do {
4370                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4371                                 pages, count, plus);
4372                 trace_nfs4_readdir(d_inode(dentry), err);
4373                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4374                                 &exception);
4375         } while (exception.retry);
4376         return err;
4377 }
4378
4379 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4380                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4381 {
4382         struct nfs4_createdata *data;
4383         int mode = sattr->ia_mode;
4384         int status = -ENOMEM;
4385
4386         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4387         if (data == NULL)
4388                 goto out;
4389
4390         if (S_ISFIFO(mode))
4391                 data->arg.ftype = NF4FIFO;
4392         else if (S_ISBLK(mode)) {
4393                 data->arg.ftype = NF4BLK;
4394                 data->arg.u.device.specdata1 = MAJOR(rdev);
4395                 data->arg.u.device.specdata2 = MINOR(rdev);
4396         }
4397         else if (S_ISCHR(mode)) {
4398                 data->arg.ftype = NF4CHR;
4399                 data->arg.u.device.specdata1 = MAJOR(rdev);
4400                 data->arg.u.device.specdata2 = MINOR(rdev);
4401         } else if (!S_ISSOCK(mode)) {
4402                 status = -EINVAL;
4403                 goto out_free;
4404         }
4405
4406         data->arg.label = label;
4407         status = nfs4_do_create(dir, dentry, data);
4408 out_free:
4409         nfs4_free_createdata(data);
4410 out:
4411         return status;
4412 }
4413
4414 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4415                 struct iattr *sattr, dev_t rdev)
4416 {
4417         struct nfs_server *server = NFS_SERVER(dir);
4418         struct nfs4_exception exception = { };
4419         struct nfs4_label l, *label = NULL;
4420         int err;
4421
4422         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4423
4424         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4425                 sattr->ia_mode &= ~current_umask();
4426         do {
4427                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4428                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4429                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4430                                 &exception);
4431         } while (exception.retry);
4432
4433         nfs4_label_release_security(label);
4434
4435         return err;
4436 }
4437
4438 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4439                  struct nfs_fsstat *fsstat)
4440 {
4441         struct nfs4_statfs_arg args = {
4442                 .fh = fhandle,
4443                 .bitmask = server->attr_bitmask,
4444         };
4445         struct nfs4_statfs_res res = {
4446                 .fsstat = fsstat,
4447         };
4448         struct rpc_message msg = {
4449                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4450                 .rpc_argp = &args,
4451                 .rpc_resp = &res,
4452         };
4453
4454         nfs_fattr_init(fsstat->fattr);
4455         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4456 }
4457
4458 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4459 {
4460         struct nfs4_exception exception = { };
4461         int err;
4462         do {
4463                 err = nfs4_handle_exception(server,
4464                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4465                                 &exception);
4466         } while (exception.retry);
4467         return err;
4468 }
4469
4470 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4471                 struct nfs_fsinfo *fsinfo)
4472 {
4473         struct nfs4_fsinfo_arg args = {
4474                 .fh = fhandle,
4475                 .bitmask = server->attr_bitmask,
4476         };
4477         struct nfs4_fsinfo_res res = {
4478                 .fsinfo = fsinfo,
4479         };
4480         struct rpc_message msg = {
4481                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4482                 .rpc_argp = &args,
4483                 .rpc_resp = &res,
4484         };
4485
4486         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4487 }
4488
4489 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4490 {
4491         struct nfs4_exception exception = { };
4492         unsigned long now = jiffies;
4493         int err;
4494
4495         do {
4496                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4497                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4498                 if (err == 0) {
4499                         nfs4_set_lease_period(server->nfs_client,
4500                                         fsinfo->lease_time * HZ,
4501                                         now);
4502                         break;
4503                 }
4504                 err = nfs4_handle_exception(server, err, &exception);
4505         } while (exception.retry);
4506         return err;
4507 }
4508
4509 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4510 {
4511         int error;
4512
4513         nfs_fattr_init(fsinfo->fattr);
4514         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4515         if (error == 0) {
4516                 /* block layout checks this! */
4517                 server->pnfs_blksize = fsinfo->blksize;
4518                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4519         }
4520
4521         return error;
4522 }
4523
4524 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4525                 struct nfs_pathconf *pathconf)
4526 {
4527         struct nfs4_pathconf_arg args = {
4528                 .fh = fhandle,
4529                 .bitmask = server->attr_bitmask,
4530         };
4531         struct nfs4_pathconf_res res = {
4532                 .pathconf = pathconf,
4533         };
4534         struct rpc_message msg = {
4535                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4536                 .rpc_argp = &args,
4537                 .rpc_resp = &res,
4538         };
4539
4540         /* None of the pathconf attributes are mandatory to implement */
4541         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4542                 memset(pathconf, 0, sizeof(*pathconf));
4543                 return 0;
4544         }
4545
4546         nfs_fattr_init(pathconf->fattr);
4547         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4548 }
4549
4550 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4551                 struct nfs_pathconf *pathconf)
4552 {
4553         struct nfs4_exception exception = { };
4554         int err;
4555
4556         do {
4557                 err = nfs4_handle_exception(server,
4558                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4559                                 &exception);
4560         } while (exception.retry);
4561         return err;
4562 }
4563
4564 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4565                 const struct nfs_open_context *ctx,
4566                 const struct nfs_lock_context *l_ctx,
4567                 fmode_t fmode)
4568 {
4569         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4570 }
4571 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4572
4573 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4574                 const struct nfs_open_context *ctx,
4575                 const struct nfs_lock_context *l_ctx,
4576                 fmode_t fmode)
4577 {
4578         nfs4_stateid current_stateid;
4579
4580         /* If the current stateid represents a lost lock, then exit */
4581         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4582                 return true;
4583         return nfs4_stateid_match(stateid, &current_stateid);
4584 }
4585
4586 static bool nfs4_error_stateid_expired(int err)
4587 {
4588         switch (err) {
4589         case -NFS4ERR_DELEG_REVOKED:
4590         case -NFS4ERR_ADMIN_REVOKED:
4591         case -NFS4ERR_BAD_STATEID:
4592         case -NFS4ERR_STALE_STATEID:
4593         case -NFS4ERR_OLD_STATEID:
4594         case -NFS4ERR_OPENMODE:
4595         case -NFS4ERR_EXPIRED:
4596                 return true;
4597         }
4598         return false;
4599 }
4600
4601 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4602 {
4603         struct nfs_server *server = NFS_SERVER(hdr->inode);
4604
4605         trace_nfs4_read(hdr, task->tk_status);
4606         if (task->tk_status < 0) {
4607                 struct nfs4_exception exception = {
4608                         .inode = hdr->inode,
4609                         .state = hdr->args.context->state,
4610                         .stateid = &hdr->args.stateid,
4611                 };
4612                 task->tk_status = nfs4_async_handle_exception(task,
4613                                 server, task->tk_status, &exception);
4614                 if (exception.retry) {
4615                         rpc_restart_call_prepare(task);
4616                         return -EAGAIN;
4617                 }
4618         }
4619
4620         if (task->tk_status > 0)
4621                 renew_lease(server, hdr->timestamp);
4622         return 0;
4623 }
4624
4625 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4626                 struct nfs_pgio_args *args)
4627 {
4628
4629         if (!nfs4_error_stateid_expired(task->tk_status) ||
4630                 nfs4_stateid_is_current(&args->stateid,
4631                                 args->context,
4632                                 args->lock_context,
4633                                 FMODE_READ))
4634                 return false;
4635         rpc_restart_call_prepare(task);
4636         return true;
4637 }
4638
4639 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4640 {
4641
4642         dprintk("--> %s\n", __func__);
4643
4644         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4645                 return -EAGAIN;
4646         if (nfs4_read_stateid_changed(task, &hdr->args))
4647                 return -EAGAIN;
4648         if (task->tk_status > 0)
4649                 nfs_invalidate_atime(hdr->inode);
4650         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4651                                     nfs4_read_done_cb(task, hdr);
4652 }
4653
4654 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4655                                  struct rpc_message *msg)
4656 {
4657         hdr->timestamp   = jiffies;
4658         if (!hdr->pgio_done_cb)
4659                 hdr->pgio_done_cb = nfs4_read_done_cb;
4660         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4661         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4662 }
4663
4664 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4665                                       struct nfs_pgio_header *hdr)
4666 {
4667         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4668                         &hdr->args.seq_args,
4669                         &hdr->res.seq_res,
4670                         task))
4671                 return 0;
4672         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4673                                 hdr->args.lock_context,
4674                                 hdr->rw_mode) == -EIO)
4675                 return -EIO;
4676         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4677                 return -EIO;
4678         return 0;
4679 }
4680
4681 static int nfs4_write_done_cb(struct rpc_task *task,
4682                               struct nfs_pgio_header *hdr)
4683 {
4684         struct inode *inode = hdr->inode;
4685
4686         trace_nfs4_write(hdr, task->tk_status);
4687         if (task->tk_status < 0) {
4688                 struct nfs4_exception exception = {
4689                         .inode = hdr->inode,
4690                         .state = hdr->args.context->state,
4691                         .stateid = &hdr->args.stateid,
4692                 };
4693                 task->tk_status = nfs4_async_handle_exception(task,
4694                                 NFS_SERVER(inode), task->tk_status,
4695                                 &exception);
4696                 if (exception.retry) {
4697                         rpc_restart_call_prepare(task);
4698                         return -EAGAIN;
4699                 }
4700         }
4701         if (task->tk_status >= 0) {
4702                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4703                 nfs_writeback_update_inode(hdr);
4704         }
4705         return 0;
4706 }
4707
4708 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4709                 struct nfs_pgio_args *args)
4710 {
4711
4712         if (!nfs4_error_stateid_expired(task->tk_status) ||
4713                 nfs4_stateid_is_current(&args->stateid,
4714                                 args->context,
4715                                 args->lock_context,
4716                                 FMODE_WRITE))
4717                 return false;
4718         rpc_restart_call_prepare(task);
4719         return true;
4720 }
4721
4722 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4723 {
4724         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4725                 return -EAGAIN;
4726         if (nfs4_write_stateid_changed(task, &hdr->args))
4727                 return -EAGAIN;
4728         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4729                 nfs4_write_done_cb(task, hdr);
4730 }
4731
4732 static
4733 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4734 {
4735         /* Don't request attributes for pNFS or O_DIRECT writes */
4736         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4737                 return false;
4738         /* Otherwise, request attributes if and only if we don't hold
4739          * a delegation
4740          */
4741         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4742 }
4743
4744 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4745                                   struct rpc_message *msg)
4746 {
4747         struct nfs_server *server = NFS_SERVER(hdr->inode);
4748
4749         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4750                 hdr->args.bitmask = NULL;
4751                 hdr->res.fattr = NULL;
4752         } else
4753                 hdr->args.bitmask = server->cache_consistency_bitmask;
4754
4755         if (!hdr->pgio_done_cb)
4756                 hdr->pgio_done_cb = nfs4_write_done_cb;
4757         hdr->res.server = server;
4758         hdr->timestamp   = jiffies;
4759
4760         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4761         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4762 }
4763
4764 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4765 {
4766         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
4767                         &data->args.seq_args,
4768                         &data->res.seq_res,
4769                         task);
4770 }
4771
4772 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4773 {
4774         struct inode *inode = data->inode;
4775
4776         trace_nfs4_commit(data, task->tk_status);
4777         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4778                                     NULL, NULL) == -EAGAIN) {
4779                 rpc_restart_call_prepare(task);
4780                 return -EAGAIN;
4781         }
4782         return 0;
4783 }
4784
4785 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4786 {
4787         if (!nfs4_sequence_done(task, &data->res.seq_res))
4788                 return -EAGAIN;
4789         return data->commit_done_cb(task, data);
4790 }
4791
4792 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4793 {
4794         struct nfs_server *server = NFS_SERVER(data->inode);
4795
4796         if (data->commit_done_cb == NULL)
4797                 data->commit_done_cb = nfs4_commit_done_cb;
4798         data->res.server = server;
4799         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4800         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4801 }
4802
4803 struct nfs4_renewdata {
4804         struct nfs_client       *client;
4805         unsigned long           timestamp;
4806 };
4807
4808 /*
4809  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4810  * standalone procedure for queueing an asynchronous RENEW.
4811  */
4812 static void nfs4_renew_release(void *calldata)
4813 {
4814         struct nfs4_renewdata *data = calldata;
4815         struct nfs_client *clp = data->client;
4816
4817         if (atomic_read(&clp->cl_count) > 1)
4818                 nfs4_schedule_state_renewal(clp);
4819         nfs_put_client(clp);
4820         kfree(data);
4821 }
4822
4823 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4824 {
4825         struct nfs4_renewdata *data = calldata;
4826         struct nfs_client *clp = data->client;
4827         unsigned long timestamp = data->timestamp;
4828
4829         trace_nfs4_renew_async(clp, task->tk_status);
4830         switch (task->tk_status) {
4831         case 0:
4832                 break;
4833         case -NFS4ERR_LEASE_MOVED:
4834                 nfs4_schedule_lease_moved_recovery(clp);
4835                 break;
4836         default:
4837                 /* Unless we're shutting down, schedule state recovery! */
4838                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4839                         return;
4840                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4841                         nfs4_schedule_lease_recovery(clp);
4842                         return;
4843                 }
4844                 nfs4_schedule_path_down_recovery(clp);
4845         }
4846         do_renew_lease(clp, timestamp);
4847 }
4848
4849 static const struct rpc_call_ops nfs4_renew_ops = {
4850         .rpc_call_done = nfs4_renew_done,
4851         .rpc_release = nfs4_renew_release,
4852 };
4853
4854 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4855 {
4856         struct rpc_message msg = {
4857                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4858                 .rpc_argp       = clp,
4859                 .rpc_cred       = cred,
4860         };
4861         struct nfs4_renewdata *data;
4862
4863         if (renew_flags == 0)
4864                 return 0;
4865         if (!atomic_inc_not_zero(&clp->cl_count))
4866                 return -EIO;
4867         data = kmalloc(sizeof(*data), GFP_NOFS);
4868         if (data == NULL) {
4869                 nfs_put_client(clp);
4870                 return -ENOMEM;
4871         }
4872         data->client = clp;
4873         data->timestamp = jiffies;
4874         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4875                         &nfs4_renew_ops, data);
4876 }
4877
4878 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4879 {
4880         struct rpc_message msg = {
4881                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4882                 .rpc_argp       = clp,
4883                 .rpc_cred       = cred,
4884         };
4885         unsigned long now = jiffies;
4886         int status;
4887
4888         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4889         if (status < 0)
4890                 return status;
4891         do_renew_lease(clp, now);
4892         return 0;
4893 }
4894
4895 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4896 {
4897         return server->caps & NFS_CAP_ACLS;
4898 }
4899
4900 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4901  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4902  * the stack.
4903  */
4904 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4905
4906 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4907                 struct page **pages)
4908 {
4909         struct page *newpage, **spages;
4910         int rc = 0;
4911         size_t len;
4912         spages = pages;
4913
4914         do {
4915                 len = min_t(size_t, PAGE_SIZE, buflen);
4916                 newpage = alloc_page(GFP_KERNEL);
4917
4918                 if (newpage == NULL)
4919                         goto unwind;
4920                 memcpy(page_address(newpage), buf, len);
4921                 buf += len;
4922                 buflen -= len;
4923                 *pages++ = newpage;
4924                 rc++;
4925         } while (buflen != 0);
4926
4927         return rc;
4928
4929 unwind:
4930         for(; rc > 0; rc--)
4931                 __free_page(spages[rc-1]);
4932         return -ENOMEM;
4933 }
4934
4935 struct nfs4_cached_acl {
4936         int cached;
4937         size_t len;
4938         char data[0];
4939 };
4940
4941 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4942 {
4943         struct nfs_inode *nfsi = NFS_I(inode);
4944
4945         spin_lock(&inode->i_lock);
4946         kfree(nfsi->nfs4_acl);
4947         nfsi->nfs4_acl = acl;
4948         spin_unlock(&inode->i_lock);
4949 }
4950
4951 static void nfs4_zap_acl_attr(struct inode *inode)
4952 {
4953         nfs4_set_cached_acl(inode, NULL);
4954 }
4955
4956 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4957 {
4958         struct nfs_inode *nfsi = NFS_I(inode);
4959         struct nfs4_cached_acl *acl;
4960         int ret = -ENOENT;
4961
4962         spin_lock(&inode->i_lock);
4963         acl = nfsi->nfs4_acl;
4964         if (acl == NULL)
4965                 goto out;
4966         if (buf == NULL) /* user is just asking for length */
4967                 goto out_len;
4968         if (acl->cached == 0)
4969                 goto out;
4970         ret = -ERANGE; /* see getxattr(2) man page */
4971         if (acl->len > buflen)
4972                 goto out;
4973         memcpy(buf, acl->data, acl->len);
4974 out_len:
4975         ret = acl->len;
4976 out:
4977         spin_unlock(&inode->i_lock);
4978         return ret;
4979 }
4980
4981 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4982 {
4983         struct nfs4_cached_acl *acl;
4984         size_t buflen = sizeof(*acl) + acl_len;
4985
4986         if (buflen <= PAGE_SIZE) {
4987                 acl = kmalloc(buflen, GFP_KERNEL);
4988                 if (acl == NULL)
4989                         goto out;
4990                 acl->cached = 1;
4991                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4992         } else {
4993                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4994                 if (acl == NULL)
4995                         goto out;
4996                 acl->cached = 0;
4997         }
4998         acl->len = acl_len;
4999 out:
5000         nfs4_set_cached_acl(inode, acl);
5001 }
5002
5003 /*
5004  * The getxattr API returns the required buffer length when called with a
5005  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5006  * the required buf.  On a NULL buf, we send a page of data to the server
5007  * guessing that the ACL request can be serviced by a page. If so, we cache
5008  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5009  * the cache. If not so, we throw away the page, and cache the required
5010  * length. The next getxattr call will then produce another round trip to
5011  * the server, this time with the input buf of the required size.
5012  */
5013 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5014 {
5015         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5016         struct nfs_getaclargs args = {
5017                 .fh = NFS_FH(inode),
5018                 .acl_pages = pages,
5019                 .acl_len = buflen,
5020         };
5021         struct nfs_getaclres res = {
5022                 .acl_len = buflen,
5023         };
5024         struct rpc_message msg = {
5025                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5026                 .rpc_argp = &args,
5027                 .rpc_resp = &res,
5028         };
5029         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5030         int ret = -ENOMEM, i;
5031
5032         if (npages > ARRAY_SIZE(pages))
5033                 return -ERANGE;
5034
5035         for (i = 0; i < npages; i++) {
5036                 pages[i] = alloc_page(GFP_KERNEL);
5037                 if (!pages[i])
5038                         goto out_free;
5039         }
5040
5041         /* for decoding across pages */
5042         res.acl_scratch = alloc_page(GFP_KERNEL);
5043         if (!res.acl_scratch)
5044                 goto out_free;
5045
5046         args.acl_len = npages * PAGE_SIZE;
5047
5048         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5049                 __func__, buf, buflen, npages, args.acl_len);
5050         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5051                              &msg, &args.seq_args, &res.seq_res, 0);
5052         if (ret)
5053                 goto out_free;
5054
5055         /* Handle the case where the passed-in buffer is too short */
5056         if (res.acl_flags & NFS4_ACL_TRUNC) {
5057                 /* Did the user only issue a request for the acl length? */
5058                 if (buf == NULL)
5059                         goto out_ok;
5060                 ret = -ERANGE;
5061                 goto out_free;
5062         }
5063         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5064         if (buf) {
5065                 if (res.acl_len > buflen) {
5066                         ret = -ERANGE;
5067                         goto out_free;
5068                 }
5069                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5070         }
5071 out_ok:
5072         ret = res.acl_len;
5073 out_free:
5074         for (i = 0; i < npages; i++)
5075                 if (pages[i])
5076                         __free_page(pages[i]);
5077         if (res.acl_scratch)
5078                 __free_page(res.acl_scratch);
5079         return ret;
5080 }
5081
5082 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5083 {
5084         struct nfs4_exception exception = { };
5085         ssize_t ret;
5086         do {
5087                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5088                 trace_nfs4_get_acl(inode, ret);
5089                 if (ret >= 0)
5090                         break;
5091                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5092         } while (exception.retry);
5093         return ret;
5094 }
5095
5096 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5097 {
5098         struct nfs_server *server = NFS_SERVER(inode);
5099         int ret;
5100
5101         if (!nfs4_server_supports_acls(server))
5102                 return -EOPNOTSUPP;
5103         ret = nfs_revalidate_inode(server, inode);
5104         if (ret < 0)
5105                 return ret;
5106         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5107                 nfs_zap_acl_cache(inode);
5108         ret = nfs4_read_cached_acl(inode, buf, buflen);
5109         if (ret != -ENOENT)
5110                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5111                  * but no cached acl data, just the acl length */
5112                 return ret;
5113         return nfs4_get_acl_uncached(inode, buf, buflen);
5114 }
5115
5116 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5117 {
5118         struct nfs_server *server = NFS_SERVER(inode);
5119         struct page *pages[NFS4ACL_MAXPAGES];
5120         struct nfs_setaclargs arg = {
5121                 .fh             = NFS_FH(inode),
5122                 .acl_pages      = pages,
5123                 .acl_len        = buflen,
5124         };
5125         struct nfs_setaclres res;
5126         struct rpc_message msg = {
5127                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5128                 .rpc_argp       = &arg,
5129                 .rpc_resp       = &res,
5130         };
5131         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5132         int ret, i;
5133
5134         if (!nfs4_server_supports_acls(server))
5135                 return -EOPNOTSUPP;
5136         if (npages > ARRAY_SIZE(pages))
5137                 return -ERANGE;
5138         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5139         if (i < 0)
5140                 return i;
5141         nfs4_inode_return_delegation(inode);
5142         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5143
5144         /*
5145          * Free each page after tx, so the only ref left is
5146          * held by the network stack
5147          */
5148         for (; i > 0; i--)
5149                 put_page(pages[i-1]);
5150
5151         /*
5152          * Acl update can result in inode attribute update.
5153          * so mark the attribute cache invalid.
5154          */
5155         spin_lock(&inode->i_lock);
5156         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5157         spin_unlock(&inode->i_lock);
5158         nfs_access_zap_cache(inode);
5159         nfs_zap_acl_cache(inode);
5160         return ret;
5161 }
5162
5163 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5164 {
5165         struct nfs4_exception exception = { };
5166         int err;
5167         do {
5168                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5169                 trace_nfs4_set_acl(inode, err);
5170                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5171                                 &exception);
5172         } while (exception.retry);
5173         return err;
5174 }
5175
5176 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5177 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5178                                         size_t buflen)
5179 {
5180         struct nfs_server *server = NFS_SERVER(inode);
5181         struct nfs_fattr fattr;
5182         struct nfs4_label label = {0, 0, buflen, buf};
5183
5184         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5185         struct nfs4_getattr_arg arg = {
5186                 .fh             = NFS_FH(inode),
5187                 .bitmask        = bitmask,
5188         };
5189         struct nfs4_getattr_res res = {
5190                 .fattr          = &fattr,
5191                 .label          = &label,
5192                 .server         = server,
5193         };
5194         struct rpc_message msg = {
5195                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5196                 .rpc_argp       = &arg,
5197                 .rpc_resp       = &res,
5198         };
5199         int ret;
5200
5201         nfs_fattr_init(&fattr);
5202
5203         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5204         if (ret)
5205                 return ret;
5206         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5207                 return -ENOENT;
5208         if (buflen < label.len)
5209                 return -ERANGE;
5210         return 0;
5211 }
5212
5213 static int nfs4_get_security_label(struct inode *inode, void *buf,
5214                                         size_t buflen)
5215 {
5216         struct nfs4_exception exception = { };
5217         int err;
5218
5219         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5220                 return -EOPNOTSUPP;
5221
5222         do {
5223                 err = _nfs4_get_security_label(inode, buf, buflen);
5224                 trace_nfs4_get_security_label(inode, err);
5225                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5226                                 &exception);
5227         } while (exception.retry);
5228         return err;
5229 }
5230
5231 static int _nfs4_do_set_security_label(struct inode *inode,
5232                 struct nfs4_label *ilabel,
5233                 struct nfs_fattr *fattr,
5234                 struct nfs4_label *olabel)
5235 {
5236
5237         struct iattr sattr = {0};
5238         struct nfs_server *server = NFS_SERVER(inode);
5239         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5240         struct nfs_setattrargs arg = {
5241                 .fh             = NFS_FH(inode),
5242                 .iap            = &sattr,
5243                 .server         = server,
5244                 .bitmask        = bitmask,
5245                 .label          = ilabel,
5246         };
5247         struct nfs_setattrres res = {
5248                 .fattr          = fattr,
5249                 .label          = olabel,
5250                 .server         = server,
5251         };
5252         struct rpc_message msg = {
5253                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5254                 .rpc_argp       = &arg,
5255                 .rpc_resp       = &res,
5256         };
5257         int status;
5258
5259         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5260
5261         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5262         if (status)
5263                 dprintk("%s failed: %d\n", __func__, status);
5264
5265         return status;
5266 }
5267
5268 static int nfs4_do_set_security_label(struct inode *inode,
5269                 struct nfs4_label *ilabel,
5270                 struct nfs_fattr *fattr,
5271                 struct nfs4_label *olabel)
5272 {
5273         struct nfs4_exception exception = { };
5274         int err;
5275
5276         do {
5277                 err = _nfs4_do_set_security_label(inode, ilabel,
5278                                 fattr, olabel);
5279                 trace_nfs4_set_security_label(inode, err);
5280                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5281                                 &exception);
5282         } while (exception.retry);
5283         return err;
5284 }
5285
5286 static int
5287 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5288 {
5289         struct nfs4_label ilabel, *olabel = NULL;
5290         struct nfs_fattr fattr;
5291         struct rpc_cred *cred;
5292         int status;
5293
5294         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5295                 return -EOPNOTSUPP;
5296
5297         nfs_fattr_init(&fattr);
5298
5299         ilabel.pi = 0;
5300         ilabel.lfs = 0;
5301         ilabel.label = (char *)buf;
5302         ilabel.len = buflen;
5303
5304         cred = rpc_lookup_cred();
5305         if (IS_ERR(cred))
5306                 return PTR_ERR(cred);
5307
5308         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5309         if (IS_ERR(olabel)) {
5310                 status = -PTR_ERR(olabel);
5311                 goto out;
5312         }
5313
5314         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5315         if (status == 0)
5316                 nfs_setsecurity(inode, &fattr, olabel);
5317
5318         nfs4_label_free(olabel);
5319 out:
5320         put_rpccred(cred);
5321         return status;
5322 }
5323 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5324
5325
5326 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5327                                     nfs4_verifier *bootverf)
5328 {
5329         __be32 verf[2];
5330
5331         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5332                 /* An impossible timestamp guarantees this value
5333                  * will never match a generated boot time. */
5334                 verf[0] = cpu_to_be32(U32_MAX);
5335                 verf[1] = cpu_to_be32(U32_MAX);
5336         } else {
5337                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5338                 u64 ns = ktime_to_ns(nn->boot_time);
5339
5340                 verf[0] = cpu_to_be32(ns >> 32);
5341                 verf[1] = cpu_to_be32(ns);
5342         }
5343         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5344 }
5345
5346 static int
5347 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5348 {
5349         size_t len;
5350         char *str;
5351
5352         if (clp->cl_owner_id != NULL)
5353                 return 0;
5354
5355         rcu_read_lock();
5356         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5357                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5358                 1 +
5359                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5360                 1;
5361         rcu_read_unlock();
5362
5363         if (len > NFS4_OPAQUE_LIMIT + 1)
5364                 return -EINVAL;
5365
5366         /*
5367          * Since this string is allocated at mount time, and held until the
5368          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5369          * about a memory-reclaim deadlock.
5370          */
5371         str = kmalloc(len, GFP_KERNEL);
5372         if (!str)
5373                 return -ENOMEM;
5374
5375         rcu_read_lock();
5376         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5377                         clp->cl_ipaddr,
5378                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5379                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5380         rcu_read_unlock();
5381
5382         clp->cl_owner_id = str;
5383         return 0;
5384 }
5385
5386 static int
5387 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5388 {
5389         size_t len;
5390         char *str;
5391
5392         len = 10 + 10 + 1 + 10 + 1 +
5393                 strlen(nfs4_client_id_uniquifier) + 1 +
5394                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5395
5396         if (len > NFS4_OPAQUE_LIMIT + 1)
5397                 return -EINVAL;
5398
5399         /*
5400          * Since this string is allocated at mount time, and held until the
5401          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5402          * about a memory-reclaim deadlock.
5403          */
5404         str = kmalloc(len, GFP_KERNEL);
5405         if (!str)
5406                 return -ENOMEM;
5407
5408         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5409                         clp->rpc_ops->version, clp->cl_minorversion,
5410                         nfs4_client_id_uniquifier,
5411                         clp->cl_rpcclient->cl_nodename);
5412         clp->cl_owner_id = str;
5413         return 0;
5414 }
5415
5416 static int
5417 nfs4_init_uniform_client_string(struct nfs_client *clp)
5418 {
5419         size_t len;
5420         char *str;
5421
5422         if (clp->cl_owner_id != NULL)
5423                 return 0;
5424
5425         if (nfs4_client_id_uniquifier[0] != '\0')
5426                 return nfs4_init_uniquifier_client_string(clp);
5427
5428         len = 10 + 10 + 1 + 10 + 1 +
5429                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5430
5431         if (len > NFS4_OPAQUE_LIMIT + 1)
5432                 return -EINVAL;
5433
5434         /*
5435          * Since this string is allocated at mount time, and held until the
5436          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5437          * about a memory-reclaim deadlock.
5438          */
5439         str = kmalloc(len, GFP_KERNEL);
5440         if (!str)
5441                 return -ENOMEM;
5442
5443         scnprintf(str, len, "Linux NFSv%u.%u %s",
5444                         clp->rpc_ops->version, clp->cl_minorversion,
5445                         clp->cl_rpcclient->cl_nodename);
5446         clp->cl_owner_id = str;
5447         return 0;
5448 }
5449
5450 /*
5451  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5452  * services.  Advertise one based on the address family of the
5453  * clientaddr.
5454  */
5455 static unsigned int
5456 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5457 {
5458         if (strchr(clp->cl_ipaddr, ':') != NULL)
5459                 return scnprintf(buf, len, "tcp6");
5460         else
5461                 return scnprintf(buf, len, "tcp");
5462 }
5463
5464 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5465 {
5466         struct nfs4_setclientid *sc = calldata;
5467
5468         if (task->tk_status == 0)
5469                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5470 }
5471
5472 static const struct rpc_call_ops nfs4_setclientid_ops = {
5473         .rpc_call_done = nfs4_setclientid_done,
5474 };
5475
5476 /**
5477  * nfs4_proc_setclientid - Negotiate client ID
5478  * @clp: state data structure
5479  * @program: RPC program for NFSv4 callback service
5480  * @port: IP port number for NFS4 callback service
5481  * @cred: RPC credential to use for this call
5482  * @res: where to place the result
5483  *
5484  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5485  */
5486 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5487                 unsigned short port, struct rpc_cred *cred,
5488                 struct nfs4_setclientid_res *res)
5489 {
5490         nfs4_verifier sc_verifier;
5491         struct nfs4_setclientid setclientid = {
5492                 .sc_verifier = &sc_verifier,
5493                 .sc_prog = program,
5494                 .sc_clnt = clp,
5495         };
5496         struct rpc_message msg = {
5497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5498                 .rpc_argp = &setclientid,
5499                 .rpc_resp = res,
5500                 .rpc_cred = cred,
5501         };
5502         struct rpc_task *task;
5503         struct rpc_task_setup task_setup_data = {
5504                 .rpc_client = clp->cl_rpcclient,
5505                 .rpc_message = &msg,
5506                 .callback_ops = &nfs4_setclientid_ops,
5507                 .callback_data = &setclientid,
5508                 .flags = RPC_TASK_TIMEOUT,
5509         };
5510         int status;
5511
5512         /* nfs_client_id4 */
5513         nfs4_init_boot_verifier(clp, &sc_verifier);
5514
5515         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5516                 status = nfs4_init_uniform_client_string(clp);
5517         else
5518                 status = nfs4_init_nonuniform_client_string(clp);
5519
5520         if (status)
5521                 goto out;
5522
5523         /* cb_client4 */
5524         setclientid.sc_netid_len =
5525                                 nfs4_init_callback_netid(clp,
5526                                                 setclientid.sc_netid,
5527                                                 sizeof(setclientid.sc_netid));
5528         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5529                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5530                                 clp->cl_ipaddr, port >> 8, port & 255);
5531
5532         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5533                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5534                 clp->cl_owner_id);
5535         task = rpc_run_task(&task_setup_data);
5536         if (IS_ERR(task)) {
5537                 status = PTR_ERR(task);
5538                 goto out;
5539         }
5540         status = task->tk_status;
5541         if (setclientid.sc_cred) {
5542                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5543                 put_rpccred(setclientid.sc_cred);
5544         }
5545         rpc_put_task(task);
5546 out:
5547         trace_nfs4_setclientid(clp, status);
5548         dprintk("NFS reply setclientid: %d\n", status);
5549         return status;
5550 }
5551
5552 /**
5553  * nfs4_proc_setclientid_confirm - Confirm client ID
5554  * @clp: state data structure
5555  * @res: result of a previous SETCLIENTID
5556  * @cred: RPC credential to use for this call
5557  *
5558  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5559  */
5560 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5561                 struct nfs4_setclientid_res *arg,
5562                 struct rpc_cred *cred)
5563 {
5564         struct rpc_message msg = {
5565                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5566                 .rpc_argp = arg,
5567                 .rpc_cred = cred,
5568         };
5569         int status;
5570
5571         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5572                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5573                 clp->cl_clientid);
5574         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5575         trace_nfs4_setclientid_confirm(clp, status);
5576         dprintk("NFS reply setclientid_confirm: %d\n", status);
5577         return status;
5578 }
5579
5580 struct nfs4_delegreturndata {
5581         struct nfs4_delegreturnargs args;
5582         struct nfs4_delegreturnres res;
5583         struct nfs_fh fh;
5584         nfs4_stateid stateid;
5585         unsigned long timestamp;
5586         struct {
5587                 struct nfs4_layoutreturn_args arg;
5588                 struct nfs4_layoutreturn_res res;
5589                 struct nfs4_xdr_opaque_data ld_private;
5590                 u32 roc_barrier;
5591                 bool roc;
5592         } lr;
5593         struct nfs_fattr fattr;
5594         int rpc_status;
5595         struct inode *inode;
5596 };
5597
5598 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5599 {
5600         struct nfs4_delegreturndata *data = calldata;
5601
5602         if (!nfs4_sequence_done(task, &data->res.seq_res))
5603                 return;
5604
5605         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5606
5607         /* Handle Layoutreturn errors */
5608         if (data->args.lr_args && task->tk_status != 0) {
5609                 switch(data->res.lr_ret) {
5610                 default:
5611                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5612                         break;
5613                 case 0:
5614                         data->args.lr_args = NULL;
5615                         data->res.lr_res = NULL;
5616                         break;
5617                 case -NFS4ERR_ADMIN_REVOKED:
5618                 case -NFS4ERR_DELEG_REVOKED:
5619                 case -NFS4ERR_EXPIRED:
5620                 case -NFS4ERR_BAD_STATEID:
5621                 case -NFS4ERR_OLD_STATEID:
5622                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5623                 case -NFS4ERR_WRONG_CRED:
5624                         data->args.lr_args = NULL;
5625                         data->res.lr_res = NULL;
5626                         data->res.lr_ret = 0;
5627                         rpc_restart_call_prepare(task);
5628                         return;
5629                 }
5630         }
5631
5632         switch (task->tk_status) {
5633         case 0:
5634                 renew_lease(data->res.server, data->timestamp);
5635                 break;
5636         case -NFS4ERR_ADMIN_REVOKED:
5637         case -NFS4ERR_DELEG_REVOKED:
5638         case -NFS4ERR_EXPIRED:
5639                 nfs4_free_revoked_stateid(data->res.server,
5640                                 data->args.stateid,
5641                                 task->tk_msg.rpc_cred);
5642         case -NFS4ERR_BAD_STATEID:
5643         case -NFS4ERR_OLD_STATEID:
5644         case -NFS4ERR_STALE_STATEID:
5645                 task->tk_status = 0;
5646                 break;
5647         case -NFS4ERR_ACCESS:
5648                 if (data->args.bitmask) {
5649                         data->args.bitmask = NULL;
5650                         data->res.fattr = NULL;
5651                         task->tk_status = 0;
5652                         rpc_restart_call_prepare(task);
5653                         return;
5654                 }
5655         default:
5656                 if (nfs4_async_handle_error(task, data->res.server,
5657                                             NULL, NULL) == -EAGAIN) {
5658                         rpc_restart_call_prepare(task);
5659                         return;
5660                 }
5661         }
5662         data->rpc_status = task->tk_status;
5663 }
5664
5665 static void nfs4_delegreturn_release(void *calldata)
5666 {
5667         struct nfs4_delegreturndata *data = calldata;
5668         struct inode *inode = data->inode;
5669
5670         if (inode) {
5671                 if (data->lr.roc)
5672                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5673                                         data->res.lr_ret);
5674                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5675                 nfs_iput_and_deactive(inode);
5676         }
5677         kfree(calldata);
5678 }
5679
5680 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5681 {
5682         struct nfs4_delegreturndata *d_data;
5683
5684         d_data = (struct nfs4_delegreturndata *)data;
5685
5686         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5687                 return;
5688
5689         nfs4_setup_sequence(d_data->res.server->nfs_client,
5690                         &d_data->args.seq_args,
5691                         &d_data->res.seq_res,
5692                         task);
5693 }
5694
5695 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5696         .rpc_call_prepare = nfs4_delegreturn_prepare,
5697         .rpc_call_done = nfs4_delegreturn_done,
5698         .rpc_release = nfs4_delegreturn_release,
5699 };
5700
5701 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5702 {
5703         struct nfs4_delegreturndata *data;
5704         struct nfs_server *server = NFS_SERVER(inode);
5705         struct rpc_task *task;
5706         struct rpc_message msg = {
5707                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5708                 .rpc_cred = cred,
5709         };
5710         struct rpc_task_setup task_setup_data = {
5711                 .rpc_client = server->client,
5712                 .rpc_message = &msg,
5713                 .callback_ops = &nfs4_delegreturn_ops,
5714                 .flags = RPC_TASK_ASYNC,
5715         };
5716         int status = 0;
5717
5718         data = kzalloc(sizeof(*data), GFP_NOFS);
5719         if (data == NULL)
5720                 return -ENOMEM;
5721         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5722
5723         nfs4_state_protect(server->nfs_client,
5724                         NFS_SP4_MACH_CRED_CLEANUP,
5725                         &task_setup_data.rpc_client, &msg);
5726
5727         data->args.fhandle = &data->fh;
5728         data->args.stateid = &data->stateid;
5729         data->args.bitmask = server->cache_consistency_bitmask;
5730         nfs_copy_fh(&data->fh, NFS_FH(inode));
5731         nfs4_stateid_copy(&data->stateid, stateid);
5732         data->res.fattr = &data->fattr;
5733         data->res.server = server;
5734         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5735         data->lr.arg.ld_private = &data->lr.ld_private;
5736         nfs_fattr_init(data->res.fattr);
5737         data->timestamp = jiffies;
5738         data->rpc_status = 0;
5739         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5740         data->inode = nfs_igrab_and_active(inode);
5741         if (data->inode) {
5742                 if (data->lr.roc) {
5743                         data->args.lr_args = &data->lr.arg;
5744                         data->res.lr_res = &data->lr.res;
5745                 }
5746         } else if (data->lr.roc) {
5747                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5748                 data->lr.roc = false;
5749         }
5750
5751         task_setup_data.callback_data = data;
5752         msg.rpc_argp = &data->args;
5753         msg.rpc_resp = &data->res;
5754         task = rpc_run_task(&task_setup_data);
5755         if (IS_ERR(task))
5756                 return PTR_ERR(task);
5757         if (!issync)
5758                 goto out;
5759         status = rpc_wait_for_completion_task(task);
5760         if (status != 0)
5761                 goto out;
5762         status = data->rpc_status;
5763 out:
5764         rpc_put_task(task);
5765         return status;
5766 }
5767
5768 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5769 {
5770         struct nfs_server *server = NFS_SERVER(inode);
5771         struct nfs4_exception exception = { };
5772         int err;
5773         do {
5774                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5775                 trace_nfs4_delegreturn(inode, stateid, err);
5776                 switch (err) {
5777                         case -NFS4ERR_STALE_STATEID:
5778                         case -NFS4ERR_EXPIRED:
5779                         case 0:
5780                                 return 0;
5781                 }
5782                 err = nfs4_handle_exception(server, err, &exception);
5783         } while (exception.retry);
5784         return err;
5785 }
5786
5787 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5788 {
5789         struct inode *inode = state->inode;
5790         struct nfs_server *server = NFS_SERVER(inode);
5791         struct nfs_client *clp = server->nfs_client;
5792         struct nfs_lockt_args arg = {
5793                 .fh = NFS_FH(inode),
5794                 .fl = request,
5795         };
5796         struct nfs_lockt_res res = {
5797                 .denied = request,
5798         };
5799         struct rpc_message msg = {
5800                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5801                 .rpc_argp       = &arg,
5802                 .rpc_resp       = &res,
5803                 .rpc_cred       = state->owner->so_cred,
5804         };
5805         struct nfs4_lock_state *lsp;
5806         int status;
5807
5808         arg.lock_owner.clientid = clp->cl_clientid;
5809         status = nfs4_set_lock_state(state, request);
5810         if (status != 0)
5811                 goto out;
5812         lsp = request->fl_u.nfs4_fl.owner;
5813         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5814         arg.lock_owner.s_dev = server->s_dev;
5815         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5816         switch (status) {
5817                 case 0:
5818                         request->fl_type = F_UNLCK;
5819                         break;
5820                 case -NFS4ERR_DENIED:
5821                         status = 0;
5822         }
5823         request->fl_ops->fl_release_private(request);
5824         request->fl_ops = NULL;
5825 out:
5826         return status;
5827 }
5828
5829 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5830 {
5831         struct nfs4_exception exception = { };
5832         int err;
5833
5834         do {
5835                 err = _nfs4_proc_getlk(state, cmd, request);
5836                 trace_nfs4_get_lock(request, state, cmd, err);
5837                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5838                                 &exception);
5839         } while (exception.retry);
5840         return err;
5841 }
5842
5843 struct nfs4_unlockdata {
5844         struct nfs_locku_args arg;
5845         struct nfs_locku_res res;
5846         struct nfs4_lock_state *lsp;
5847         struct nfs_open_context *ctx;
5848         struct nfs_lock_context *l_ctx;
5849         struct file_lock fl;
5850         struct nfs_server *server;
5851         unsigned long timestamp;
5852 };
5853
5854 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5855                 struct nfs_open_context *ctx,
5856                 struct nfs4_lock_state *lsp,
5857                 struct nfs_seqid *seqid)
5858 {
5859         struct nfs4_unlockdata *p;
5860         struct inode *inode = lsp->ls_state->inode;
5861
5862         p = kzalloc(sizeof(*p), GFP_NOFS);
5863         if (p == NULL)
5864                 return NULL;
5865         p->arg.fh = NFS_FH(inode);
5866         p->arg.fl = &p->fl;
5867         p->arg.seqid = seqid;
5868         p->res.seqid = seqid;
5869         p->lsp = lsp;
5870         atomic_inc(&lsp->ls_count);
5871         /* Ensure we don't close file until we're done freeing locks! */
5872         p->ctx = get_nfs_open_context(ctx);
5873         p->l_ctx = nfs_get_lock_context(ctx);
5874         memcpy(&p->fl, fl, sizeof(p->fl));
5875         p->server = NFS_SERVER(inode);
5876         return p;
5877 }
5878
5879 static void nfs4_locku_release_calldata(void *data)
5880 {
5881         struct nfs4_unlockdata *calldata = data;
5882         nfs_free_seqid(calldata->arg.seqid);
5883         nfs4_put_lock_state(calldata->lsp);
5884         nfs_put_lock_context(calldata->l_ctx);
5885         put_nfs_open_context(calldata->ctx);
5886         kfree(calldata);
5887 }
5888
5889 static void nfs4_locku_done(struct rpc_task *task, void *data)
5890 {
5891         struct nfs4_unlockdata *calldata = data;
5892
5893         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5894                 return;
5895         switch (task->tk_status) {
5896                 case 0:
5897                         renew_lease(calldata->server, calldata->timestamp);
5898                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5899                         if (nfs4_update_lock_stateid(calldata->lsp,
5900                                         &calldata->res.stateid))
5901                                 break;
5902                 case -NFS4ERR_ADMIN_REVOKED:
5903                 case -NFS4ERR_EXPIRED:
5904                         nfs4_free_revoked_stateid(calldata->server,
5905                                         &calldata->arg.stateid,
5906                                         task->tk_msg.rpc_cred);
5907                 case -NFS4ERR_BAD_STATEID:
5908                 case -NFS4ERR_OLD_STATEID:
5909                 case -NFS4ERR_STALE_STATEID:
5910                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5911                                                 &calldata->lsp->ls_stateid))
5912                                 rpc_restart_call_prepare(task);
5913                         break;
5914                 default:
5915                         if (nfs4_async_handle_error(task, calldata->server,
5916                                                     NULL, NULL) == -EAGAIN)
5917                                 rpc_restart_call_prepare(task);
5918         }
5919         nfs_release_seqid(calldata->arg.seqid);
5920 }
5921
5922 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5923 {
5924         struct nfs4_unlockdata *calldata = data;
5925
5926         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
5927                 nfs_async_iocounter_wait(task, calldata->l_ctx))
5928                 return;
5929
5930         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5931                 goto out_wait;
5932         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5933         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5934                 /* Note: exit _without_ running nfs4_locku_done */
5935                 goto out_no_action;
5936         }
5937         calldata->timestamp = jiffies;
5938         if (nfs4_setup_sequence(calldata->server->nfs_client,
5939                                 &calldata->arg.seq_args,
5940                                 &calldata->res.seq_res,
5941                                 task) != 0)
5942                 nfs_release_seqid(calldata->arg.seqid);
5943         return;
5944 out_no_action:
5945         task->tk_action = NULL;
5946 out_wait:
5947         nfs4_sequence_done(task, &calldata->res.seq_res);
5948 }
5949
5950 static const struct rpc_call_ops nfs4_locku_ops = {
5951         .rpc_call_prepare = nfs4_locku_prepare,
5952         .rpc_call_done = nfs4_locku_done,
5953         .rpc_release = nfs4_locku_release_calldata,
5954 };
5955
5956 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5957                 struct nfs_open_context *ctx,
5958                 struct nfs4_lock_state *lsp,
5959                 struct nfs_seqid *seqid)
5960 {
5961         struct nfs4_unlockdata *data;
5962         struct rpc_message msg = {
5963                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5964                 .rpc_cred = ctx->cred,
5965         };
5966         struct rpc_task_setup task_setup_data = {
5967                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5968                 .rpc_message = &msg,
5969                 .callback_ops = &nfs4_locku_ops,
5970                 .workqueue = nfsiod_workqueue,
5971                 .flags = RPC_TASK_ASYNC,
5972         };
5973
5974         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5975                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5976
5977         /* Ensure this is an unlock - when canceling a lock, the
5978          * canceled lock is passed in, and it won't be an unlock.
5979          */
5980         fl->fl_type = F_UNLCK;
5981         if (fl->fl_flags & FL_CLOSE)
5982                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
5983
5984         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5985         if (data == NULL) {
5986                 nfs_free_seqid(seqid);
5987                 return ERR_PTR(-ENOMEM);
5988         }
5989
5990         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5991         msg.rpc_argp = &data->arg;
5992         msg.rpc_resp = &data->res;
5993         task_setup_data.callback_data = data;
5994         return rpc_run_task(&task_setup_data);
5995 }
5996
5997 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5998 {
5999         struct inode *inode = state->inode;
6000         struct nfs4_state_owner *sp = state->owner;
6001         struct nfs_inode *nfsi = NFS_I(inode);
6002         struct nfs_seqid *seqid;
6003         struct nfs4_lock_state *lsp;
6004         struct rpc_task *task;
6005         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6006         int status = 0;
6007         unsigned char fl_flags = request->fl_flags;
6008
6009         status = nfs4_set_lock_state(state, request);
6010         /* Unlock _before_ we do the RPC call */
6011         request->fl_flags |= FL_EXISTS;
6012         /* Exclude nfs_delegation_claim_locks() */
6013         mutex_lock(&sp->so_delegreturn_mutex);
6014         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6015         down_read(&nfsi->rwsem);
6016         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6017                 up_read(&nfsi->rwsem);
6018                 mutex_unlock(&sp->so_delegreturn_mutex);
6019                 goto out;
6020         }
6021         up_read(&nfsi->rwsem);
6022         mutex_unlock(&sp->so_delegreturn_mutex);
6023         if (status != 0)
6024                 goto out;
6025         /* Is this a delegated lock? */
6026         lsp = request->fl_u.nfs4_fl.owner;
6027         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6028                 goto out;
6029         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6030         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6031         status = -ENOMEM;
6032         if (IS_ERR(seqid))
6033                 goto out;
6034         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6035         status = PTR_ERR(task);
6036         if (IS_ERR(task))
6037                 goto out;
6038         status = rpc_wait_for_completion_task(task);
6039         rpc_put_task(task);
6040 out:
6041         request->fl_flags = fl_flags;
6042         trace_nfs4_unlock(request, state, F_SETLK, status);
6043         return status;
6044 }
6045
6046 struct nfs4_lockdata {
6047         struct nfs_lock_args arg;
6048         struct nfs_lock_res res;
6049         struct nfs4_lock_state *lsp;
6050         struct nfs_open_context *ctx;
6051         struct file_lock fl;
6052         unsigned long timestamp;
6053         int rpc_status;
6054         int cancelled;
6055         struct nfs_server *server;
6056 };
6057
6058 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6059                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6060                 gfp_t gfp_mask)
6061 {
6062         struct nfs4_lockdata *p;
6063         struct inode *inode = lsp->ls_state->inode;
6064         struct nfs_server *server = NFS_SERVER(inode);
6065         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6066
6067         p = kzalloc(sizeof(*p), gfp_mask);
6068         if (p == NULL)
6069                 return NULL;
6070
6071         p->arg.fh = NFS_FH(inode);
6072         p->arg.fl = &p->fl;
6073         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6074         if (IS_ERR(p->arg.open_seqid))
6075                 goto out_free;
6076         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6077         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6078         if (IS_ERR(p->arg.lock_seqid))
6079                 goto out_free_seqid;
6080         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6081         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6082         p->arg.lock_owner.s_dev = server->s_dev;
6083         p->res.lock_seqid = p->arg.lock_seqid;
6084         p->lsp = lsp;
6085         p->server = server;
6086         atomic_inc(&lsp->ls_count);
6087         p->ctx = get_nfs_open_context(ctx);
6088         memcpy(&p->fl, fl, sizeof(p->fl));
6089         return p;
6090 out_free_seqid:
6091         nfs_free_seqid(p->arg.open_seqid);
6092 out_free:
6093         kfree(p);
6094         return NULL;
6095 }
6096
6097 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6098 {
6099         struct nfs4_lockdata *data = calldata;
6100         struct nfs4_state *state = data->lsp->ls_state;
6101
6102         dprintk("%s: begin!\n", __func__);
6103         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6104                 goto out_wait;
6105         /* Do we need to do an open_to_lock_owner? */
6106         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6107                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6108                         goto out_release_lock_seqid;
6109                 }
6110                 nfs4_stateid_copy(&data->arg.open_stateid,
6111                                 &state->open_stateid);
6112                 data->arg.new_lock_owner = 1;
6113                 data->res.open_seqid = data->arg.open_seqid;
6114         } else {
6115                 data->arg.new_lock_owner = 0;
6116                 nfs4_stateid_copy(&data->arg.lock_stateid,
6117                                 &data->lsp->ls_stateid);
6118         }
6119         if (!nfs4_valid_open_stateid(state)) {
6120                 data->rpc_status = -EBADF;
6121                 task->tk_action = NULL;
6122                 goto out_release_open_seqid;
6123         }
6124         data->timestamp = jiffies;
6125         if (nfs4_setup_sequence(data->server->nfs_client,
6126                                 &data->arg.seq_args,
6127                                 &data->res.seq_res,
6128                                 task) == 0)
6129                 return;
6130 out_release_open_seqid:
6131         nfs_release_seqid(data->arg.open_seqid);
6132 out_release_lock_seqid:
6133         nfs_release_seqid(data->arg.lock_seqid);
6134 out_wait:
6135         nfs4_sequence_done(task, &data->res.seq_res);
6136         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6137 }
6138
6139 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6140 {
6141         struct nfs4_lockdata *data = calldata;
6142         struct nfs4_lock_state *lsp = data->lsp;
6143
6144         dprintk("%s: begin!\n", __func__);
6145
6146         if (!nfs4_sequence_done(task, &data->res.seq_res))
6147                 return;
6148
6149         data->rpc_status = task->tk_status;
6150         switch (task->tk_status) {
6151         case 0:
6152                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6153                                 data->timestamp);
6154                 if (data->arg.new_lock) {
6155                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6156                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6157                                 rpc_restart_call_prepare(task);
6158                                 break;
6159                         }
6160                 }
6161                 if (data->arg.new_lock_owner != 0) {
6162                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6163                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6164                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6165                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6166                         rpc_restart_call_prepare(task);
6167                 break;
6168         case -NFS4ERR_BAD_STATEID:
6169         case -NFS4ERR_OLD_STATEID:
6170         case -NFS4ERR_STALE_STATEID:
6171         case -NFS4ERR_EXPIRED:
6172                 if (data->arg.new_lock_owner != 0) {
6173                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6174                                                 &lsp->ls_state->open_stateid))
6175                                 rpc_restart_call_prepare(task);
6176                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6177                                                 &lsp->ls_stateid))
6178                                 rpc_restart_call_prepare(task);
6179         }
6180         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6181 }
6182
6183 static void nfs4_lock_release(void *calldata)
6184 {
6185         struct nfs4_lockdata *data = calldata;
6186
6187         dprintk("%s: begin!\n", __func__);
6188         nfs_free_seqid(data->arg.open_seqid);
6189         if (data->cancelled) {
6190                 struct rpc_task *task;
6191                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6192                                 data->arg.lock_seqid);
6193                 if (!IS_ERR(task))
6194                         rpc_put_task_async(task);
6195                 dprintk("%s: cancelling lock!\n", __func__);
6196         } else
6197                 nfs_free_seqid(data->arg.lock_seqid);
6198         nfs4_put_lock_state(data->lsp);
6199         put_nfs_open_context(data->ctx);
6200         kfree(data);
6201         dprintk("%s: done!\n", __func__);
6202 }
6203
6204 static const struct rpc_call_ops nfs4_lock_ops = {
6205         .rpc_call_prepare = nfs4_lock_prepare,
6206         .rpc_call_done = nfs4_lock_done,
6207         .rpc_release = nfs4_lock_release,
6208 };
6209
6210 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6211 {
6212         switch (error) {
6213         case -NFS4ERR_ADMIN_REVOKED:
6214         case -NFS4ERR_EXPIRED:
6215         case -NFS4ERR_BAD_STATEID:
6216                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6217                 if (new_lock_owner != 0 ||
6218                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6219                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6220                 break;
6221         case -NFS4ERR_STALE_STATEID:
6222                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6223                 nfs4_schedule_lease_recovery(server->nfs_client);
6224         };
6225 }
6226
6227 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6228 {
6229         struct nfs4_lockdata *data;
6230         struct rpc_task *task;
6231         struct rpc_message msg = {
6232                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6233                 .rpc_cred = state->owner->so_cred,
6234         };
6235         struct rpc_task_setup task_setup_data = {
6236                 .rpc_client = NFS_CLIENT(state->inode),
6237                 .rpc_message = &msg,
6238                 .callback_ops = &nfs4_lock_ops,
6239                 .workqueue = nfsiod_workqueue,
6240                 .flags = RPC_TASK_ASYNC,
6241         };
6242         int ret;
6243
6244         dprintk("%s: begin!\n", __func__);
6245         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6246                         fl->fl_u.nfs4_fl.owner,
6247                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6248         if (data == NULL)
6249                 return -ENOMEM;
6250         if (IS_SETLKW(cmd))
6251                 data->arg.block = 1;
6252         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6253         msg.rpc_argp = &data->arg;
6254         msg.rpc_resp = &data->res;
6255         task_setup_data.callback_data = data;
6256         if (recovery_type > NFS_LOCK_NEW) {
6257                 if (recovery_type == NFS_LOCK_RECLAIM)
6258                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6259                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6260         } else
6261                 data->arg.new_lock = 1;
6262         task = rpc_run_task(&task_setup_data);
6263         if (IS_ERR(task))
6264                 return PTR_ERR(task);
6265         ret = rpc_wait_for_completion_task(task);
6266         if (ret == 0) {
6267                 ret = data->rpc_status;
6268                 if (ret)
6269                         nfs4_handle_setlk_error(data->server, data->lsp,
6270                                         data->arg.new_lock_owner, ret);
6271         } else
6272                 data->cancelled = true;
6273         rpc_put_task(task);
6274         dprintk("%s: done, ret = %d!\n", __func__, ret);
6275         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6276         return ret;
6277 }
6278
6279 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6280 {
6281         struct nfs_server *server = NFS_SERVER(state->inode);
6282         struct nfs4_exception exception = {
6283                 .inode = state->inode,
6284         };
6285         int err;
6286
6287         do {
6288                 /* Cache the lock if possible... */
6289                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6290                         return 0;
6291                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6292                 if (err != -NFS4ERR_DELAY)
6293                         break;
6294                 nfs4_handle_exception(server, err, &exception);
6295         } while (exception.retry);
6296         return err;
6297 }
6298
6299 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6300 {
6301         struct nfs_server *server = NFS_SERVER(state->inode);
6302         struct nfs4_exception exception = {
6303                 .inode = state->inode,
6304         };
6305         int err;
6306
6307         err = nfs4_set_lock_state(state, request);
6308         if (err != 0)
6309                 return err;
6310         if (!recover_lost_locks) {
6311                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6312                 return 0;
6313         }
6314         do {
6315                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6316                         return 0;
6317                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6318                 switch (err) {
6319                 default:
6320                         goto out;
6321                 case -NFS4ERR_GRACE:
6322                 case -NFS4ERR_DELAY:
6323                         nfs4_handle_exception(server, err, &exception);
6324                         err = 0;
6325                 }
6326         } while (exception.retry);
6327 out:
6328         return err;
6329 }
6330
6331 #if defined(CONFIG_NFS_V4_1)
6332 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6333 {
6334         struct nfs4_lock_state *lsp;
6335         int status;
6336
6337         status = nfs4_set_lock_state(state, request);
6338         if (status != 0)
6339                 return status;
6340         lsp = request->fl_u.nfs4_fl.owner;
6341         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6342             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6343                 return 0;
6344         return nfs4_lock_expired(state, request);
6345 }
6346 #endif
6347
6348 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6349 {
6350         struct nfs_inode *nfsi = NFS_I(state->inode);
6351         struct nfs4_state_owner *sp = state->owner;
6352         unsigned char fl_flags = request->fl_flags;
6353         int status;
6354
6355         request->fl_flags |= FL_ACCESS;
6356         status = locks_lock_inode_wait(state->inode, request);
6357         if (status < 0)
6358                 goto out;
6359         mutex_lock(&sp->so_delegreturn_mutex);
6360         down_read(&nfsi->rwsem);
6361         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6362                 /* Yes: cache locks! */
6363                 /* ...but avoid races with delegation recall... */
6364                 request->fl_flags = fl_flags & ~FL_SLEEP;
6365                 status = locks_lock_inode_wait(state->inode, request);
6366                 up_read(&nfsi->rwsem);
6367                 mutex_unlock(&sp->so_delegreturn_mutex);
6368                 goto out;
6369         }
6370         up_read(&nfsi->rwsem);
6371         mutex_unlock(&sp->so_delegreturn_mutex);
6372         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6373 out:
6374         request->fl_flags = fl_flags;
6375         return status;
6376 }
6377
6378 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6379 {
6380         struct nfs4_exception exception = {
6381                 .state = state,
6382                 .inode = state->inode,
6383         };
6384         int err;
6385
6386         do {
6387                 err = _nfs4_proc_setlk(state, cmd, request);
6388                 if (err == -NFS4ERR_DENIED)
6389                         err = -EAGAIN;
6390                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6391                                 err, &exception);
6392         } while (exception.retry);
6393         return err;
6394 }
6395
6396 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6397 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6398
6399 static int
6400 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6401                         struct file_lock *request)
6402 {
6403         int             status = -ERESTARTSYS;
6404         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6405
6406         while(!signalled()) {
6407                 status = nfs4_proc_setlk(state, cmd, request);
6408                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6409                         break;
6410                 freezable_schedule_timeout_interruptible(timeout);
6411                 timeout *= 2;
6412                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6413                 status = -ERESTARTSYS;
6414         }
6415         return status;
6416 }
6417
6418 #ifdef CONFIG_NFS_V4_1
6419 struct nfs4_lock_waiter {
6420         struct task_struct      *task;
6421         struct inode            *inode;
6422         struct nfs_lowner       *owner;
6423         bool                    notified;
6424 };
6425
6426 static int
6427 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6428 {
6429         int ret;
6430         struct cb_notify_lock_args *cbnl = key;
6431         struct nfs4_lock_waiter *waiter = wait->private;
6432         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6433                                 *wowner = waiter->owner;
6434
6435         /* Only wake if the callback was for the same owner */
6436         if (lowner->clientid != wowner->clientid ||
6437             lowner->id != wowner->id             ||
6438             lowner->s_dev != wowner->s_dev)
6439                 return 0;
6440
6441         /* Make sure it's for the right inode */
6442         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6443                 return 0;
6444
6445         waiter->notified = true;
6446
6447         /* override "private" so we can use default_wake_function */
6448         wait->private = waiter->task;
6449         ret = autoremove_wake_function(wait, mode, flags, key);
6450         wait->private = waiter;
6451         return ret;
6452 }
6453
6454 static int
6455 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6456 {
6457         int status = -ERESTARTSYS;
6458         unsigned long flags;
6459         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6460         struct nfs_server *server = NFS_SERVER(state->inode);
6461         struct nfs_client *clp = server->nfs_client;
6462         wait_queue_head_t *q = &clp->cl_lock_waitq;
6463         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6464                                     .id = lsp->ls_seqid.owner_id,
6465                                     .s_dev = server->s_dev };
6466         struct nfs4_lock_waiter waiter = { .task  = current,
6467                                            .inode = state->inode,
6468                                            .owner = &owner,
6469                                            .notified = false };
6470         wait_queue_entry_t wait;
6471
6472         /* Don't bother with waitqueue if we don't expect a callback */
6473         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6474                 return nfs4_retry_setlk_simple(state, cmd, request);
6475
6476         init_wait(&wait);
6477         wait.private = &waiter;
6478         wait.func = nfs4_wake_lock_waiter;
6479         add_wait_queue(q, &wait);
6480
6481         while(!signalled()) {
6482                 status = nfs4_proc_setlk(state, cmd, request);
6483                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6484                         break;
6485
6486                 status = -ERESTARTSYS;
6487                 spin_lock_irqsave(&q->lock, flags);
6488                 if (waiter.notified) {
6489                         spin_unlock_irqrestore(&q->lock, flags);
6490                         continue;
6491                 }
6492                 set_current_state(TASK_INTERRUPTIBLE);
6493                 spin_unlock_irqrestore(&q->lock, flags);
6494
6495                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6496         }
6497
6498         finish_wait(q, &wait);
6499         return status;
6500 }
6501 #else /* !CONFIG_NFS_V4_1 */
6502 static inline int
6503 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6504 {
6505         return nfs4_retry_setlk_simple(state, cmd, request);
6506 }
6507 #endif
6508
6509 static int
6510 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6511 {
6512         struct nfs_open_context *ctx;
6513         struct nfs4_state *state;
6514         int status;
6515
6516         /* verify open state */
6517         ctx = nfs_file_open_context(filp);
6518         state = ctx->state;
6519
6520         if (IS_GETLK(cmd)) {
6521                 if (state != NULL)
6522                         return nfs4_proc_getlk(state, F_GETLK, request);
6523                 return 0;
6524         }
6525
6526         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6527                 return -EINVAL;
6528
6529         if (request->fl_type == F_UNLCK) {
6530                 if (state != NULL)
6531                         return nfs4_proc_unlck(state, cmd, request);
6532                 return 0;
6533         }
6534
6535         if (state == NULL)
6536                 return -ENOLCK;
6537
6538         if ((request->fl_flags & FL_POSIX) &&
6539             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6540                 return -ENOLCK;
6541
6542         status = nfs4_set_lock_state(state, request);
6543         if (status != 0)
6544                 return status;
6545
6546         return nfs4_retry_setlk(state, cmd, request);
6547 }
6548
6549 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6550 {
6551         struct nfs_server *server = NFS_SERVER(state->inode);
6552         int err;
6553
6554         err = nfs4_set_lock_state(state, fl);
6555         if (err != 0)
6556                 return err;
6557         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6558         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6559 }
6560
6561 struct nfs_release_lockowner_data {
6562         struct nfs4_lock_state *lsp;
6563         struct nfs_server *server;
6564         struct nfs_release_lockowner_args args;
6565         struct nfs_release_lockowner_res res;
6566         unsigned long timestamp;
6567 };
6568
6569 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6570 {
6571         struct nfs_release_lockowner_data *data = calldata;
6572         struct nfs_server *server = data->server;
6573         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
6574                            &data->res.seq_res, task);
6575         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6576         data->timestamp = jiffies;
6577 }
6578
6579 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6580 {
6581         struct nfs_release_lockowner_data *data = calldata;
6582         struct nfs_server *server = data->server;
6583
6584         nfs40_sequence_done(task, &data->res.seq_res);
6585
6586         switch (task->tk_status) {
6587         case 0:
6588                 renew_lease(server, data->timestamp);
6589                 break;
6590         case -NFS4ERR_STALE_CLIENTID:
6591         case -NFS4ERR_EXPIRED:
6592                 nfs4_schedule_lease_recovery(server->nfs_client);
6593                 break;
6594         case -NFS4ERR_LEASE_MOVED:
6595         case -NFS4ERR_DELAY:
6596                 if (nfs4_async_handle_error(task, server,
6597                                             NULL, NULL) == -EAGAIN)
6598                         rpc_restart_call_prepare(task);
6599         }
6600 }
6601
6602 static void nfs4_release_lockowner_release(void *calldata)
6603 {
6604         struct nfs_release_lockowner_data *data = calldata;
6605         nfs4_free_lock_state(data->server, data->lsp);
6606         kfree(calldata);
6607 }
6608
6609 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6610         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6611         .rpc_call_done = nfs4_release_lockowner_done,
6612         .rpc_release = nfs4_release_lockowner_release,
6613 };
6614
6615 static void
6616 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6617 {
6618         struct nfs_release_lockowner_data *data;
6619         struct rpc_message msg = {
6620                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6621         };
6622
6623         if (server->nfs_client->cl_mvops->minor_version != 0)
6624                 return;
6625
6626         data = kmalloc(sizeof(*data), GFP_NOFS);
6627         if (!data)
6628                 return;
6629         data->lsp = lsp;
6630         data->server = server;
6631         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6632         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6633         data->args.lock_owner.s_dev = server->s_dev;
6634
6635         msg.rpc_argp = &data->args;
6636         msg.rpc_resp = &data->res;
6637         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6638         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6639 }
6640
6641 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6642
6643 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6644                                    struct dentry *unused, struct inode *inode,
6645                                    const char *key, const void *buf,
6646                                    size_t buflen, int flags)
6647 {
6648         return nfs4_proc_set_acl(inode, buf, buflen);
6649 }
6650
6651 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6652                                    struct dentry *unused, struct inode *inode,
6653                                    const char *key, void *buf, size_t buflen)
6654 {
6655         return nfs4_proc_get_acl(inode, buf, buflen);
6656 }
6657
6658 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6659 {
6660         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6661 }
6662
6663 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6664
6665 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6666                                      struct dentry *unused, struct inode *inode,
6667                                      const char *key, const void *buf,
6668                                      size_t buflen, int flags)
6669 {
6670         if (security_ismaclabel(key))
6671                 return nfs4_set_security_label(inode, buf, buflen);
6672
6673         return -EOPNOTSUPP;
6674 }
6675
6676 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6677                                      struct dentry *unused, struct inode *inode,
6678                                      const char *key, void *buf, size_t buflen)
6679 {
6680         if (security_ismaclabel(key))
6681                 return nfs4_get_security_label(inode, buf, buflen);
6682         return -EOPNOTSUPP;
6683 }
6684
6685 static ssize_t
6686 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6687 {
6688         int len = 0;
6689
6690         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6691                 len = security_inode_listsecurity(inode, list, list_len);
6692                 if (list_len && len > list_len)
6693                         return -ERANGE;
6694         }
6695         return len;
6696 }
6697
6698 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6699         .prefix = XATTR_SECURITY_PREFIX,
6700         .get    = nfs4_xattr_get_nfs4_label,
6701         .set    = nfs4_xattr_set_nfs4_label,
6702 };
6703
6704 #else
6705
6706 static ssize_t
6707 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6708 {
6709         return 0;
6710 }
6711
6712 #endif
6713
6714 /*
6715  * nfs_fhget will use either the mounted_on_fileid or the fileid
6716  */
6717 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6718 {
6719         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6720                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6721               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6722               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6723                 return;
6724
6725         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6726                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6727         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6728         fattr->nlink = 2;
6729 }
6730
6731 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6732                                    const struct qstr *name,
6733                                    struct nfs4_fs_locations *fs_locations,
6734                                    struct page *page)
6735 {
6736         struct nfs_server *server = NFS_SERVER(dir);
6737         u32 bitmask[3] = {
6738                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6739         };
6740         struct nfs4_fs_locations_arg args = {
6741                 .dir_fh = NFS_FH(dir),
6742                 .name = name,
6743                 .page = page,
6744                 .bitmask = bitmask,
6745         };
6746         struct nfs4_fs_locations_res res = {
6747                 .fs_locations = fs_locations,
6748         };
6749         struct rpc_message msg = {
6750                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6751                 .rpc_argp = &args,
6752                 .rpc_resp = &res,
6753         };
6754         int status;
6755
6756         dprintk("%s: start\n", __func__);
6757
6758         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6759          * is not supported */
6760         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6761                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6762         else
6763                 bitmask[0] |= FATTR4_WORD0_FILEID;
6764
6765         nfs_fattr_init(&fs_locations->fattr);
6766         fs_locations->server = server;
6767         fs_locations->nlocations = 0;
6768         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6769         dprintk("%s: returned status = %d\n", __func__, status);
6770         return status;
6771 }
6772
6773 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6774                            const struct qstr *name,
6775                            struct nfs4_fs_locations *fs_locations,
6776                            struct page *page)
6777 {
6778         struct nfs4_exception exception = { };
6779         int err;
6780         do {
6781                 err = _nfs4_proc_fs_locations(client, dir, name,
6782                                 fs_locations, page);
6783                 trace_nfs4_get_fs_locations(dir, name, err);
6784                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6785                                 &exception);
6786         } while (exception.retry);
6787         return err;
6788 }
6789
6790 /*
6791  * This operation also signals the server that this client is
6792  * performing migration recovery.  The server can stop returning
6793  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6794  * appended to this compound to identify the client ID which is
6795  * performing recovery.
6796  */
6797 static int _nfs40_proc_get_locations(struct inode *inode,
6798                                      struct nfs4_fs_locations *locations,
6799                                      struct page *page, struct rpc_cred *cred)
6800 {
6801         struct nfs_server *server = NFS_SERVER(inode);
6802         struct rpc_clnt *clnt = server->client;
6803         u32 bitmask[2] = {
6804                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6805         };
6806         struct nfs4_fs_locations_arg args = {
6807                 .clientid       = server->nfs_client->cl_clientid,
6808                 .fh             = NFS_FH(inode),
6809                 .page           = page,
6810                 .bitmask        = bitmask,
6811                 .migration      = 1,            /* skip LOOKUP */
6812                 .renew          = 1,            /* append RENEW */
6813         };
6814         struct nfs4_fs_locations_res res = {
6815                 .fs_locations   = locations,
6816                 .migration      = 1,
6817                 .renew          = 1,
6818         };
6819         struct rpc_message msg = {
6820                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6821                 .rpc_argp       = &args,
6822                 .rpc_resp       = &res,
6823                 .rpc_cred       = cred,
6824         };
6825         unsigned long now = jiffies;
6826         int status;
6827
6828         nfs_fattr_init(&locations->fattr);
6829         locations->server = server;
6830         locations->nlocations = 0;
6831
6832         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6833         nfs4_set_sequence_privileged(&args.seq_args);
6834         status = nfs4_call_sync_sequence(clnt, server, &msg,
6835                                         &args.seq_args, &res.seq_res);
6836         if (status)
6837                 return status;
6838
6839         renew_lease(server, now);
6840         return 0;
6841 }
6842
6843 #ifdef CONFIG_NFS_V4_1
6844
6845 /*
6846  * This operation also signals the server that this client is
6847  * performing migration recovery.  The server can stop asserting
6848  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6849  * performing this operation is identified in the SEQUENCE
6850  * operation in this compound.
6851  *
6852  * When the client supports GETATTR(fs_locations_info), it can
6853  * be plumbed in here.
6854  */
6855 static int _nfs41_proc_get_locations(struct inode *inode,
6856                                      struct nfs4_fs_locations *locations,
6857                                      struct page *page, struct rpc_cred *cred)
6858 {
6859         struct nfs_server *server = NFS_SERVER(inode);
6860         struct rpc_clnt *clnt = server->client;
6861         u32 bitmask[2] = {
6862                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6863         };
6864         struct nfs4_fs_locations_arg args = {
6865                 .fh             = NFS_FH(inode),
6866                 .page           = page,
6867                 .bitmask        = bitmask,
6868                 .migration      = 1,            /* skip LOOKUP */
6869         };
6870         struct nfs4_fs_locations_res res = {
6871                 .fs_locations   = locations,
6872                 .migration      = 1,
6873         };
6874         struct rpc_message msg = {
6875                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6876                 .rpc_argp       = &args,
6877                 .rpc_resp       = &res,
6878                 .rpc_cred       = cred,
6879         };
6880         int status;
6881
6882         nfs_fattr_init(&locations->fattr);
6883         locations->server = server;
6884         locations->nlocations = 0;
6885
6886         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6887         nfs4_set_sequence_privileged(&args.seq_args);
6888         status = nfs4_call_sync_sequence(clnt, server, &msg,
6889                                         &args.seq_args, &res.seq_res);
6890         if (status == NFS4_OK &&
6891             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6892                 status = -NFS4ERR_LEASE_MOVED;
6893         return status;
6894 }
6895
6896 #endif  /* CONFIG_NFS_V4_1 */
6897
6898 /**
6899  * nfs4_proc_get_locations - discover locations for a migrated FSID
6900  * @inode: inode on FSID that is migrating
6901  * @locations: result of query
6902  * @page: buffer
6903  * @cred: credential to use for this operation
6904  *
6905  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6906  * operation failed, or a negative errno if a local error occurred.
6907  *
6908  * On success, "locations" is filled in, but if the server has
6909  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6910  * asserted.
6911  *
6912  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6913  * from this client that require migration recovery.
6914  */
6915 int nfs4_proc_get_locations(struct inode *inode,
6916                             struct nfs4_fs_locations *locations,
6917                             struct page *page, struct rpc_cred *cred)
6918 {
6919         struct nfs_server *server = NFS_SERVER(inode);
6920         struct nfs_client *clp = server->nfs_client;
6921         const struct nfs4_mig_recovery_ops *ops =
6922                                         clp->cl_mvops->mig_recovery_ops;
6923         struct nfs4_exception exception = { };
6924         int status;
6925
6926         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6927                 (unsigned long long)server->fsid.major,
6928                 (unsigned long long)server->fsid.minor,
6929                 clp->cl_hostname);
6930         nfs_display_fhandle(NFS_FH(inode), __func__);
6931
6932         do {
6933                 status = ops->get_locations(inode, locations, page, cred);
6934                 if (status != -NFS4ERR_DELAY)
6935                         break;
6936                 nfs4_handle_exception(server, status, &exception);
6937         } while (exception.retry);
6938         return status;
6939 }
6940
6941 /*
6942  * This operation also signals the server that this client is
6943  * performing "lease moved" recovery.  The server can stop
6944  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6945  * is appended to this compound to identify the client ID which is
6946  * performing recovery.
6947  */
6948 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6949 {
6950         struct nfs_server *server = NFS_SERVER(inode);
6951         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6952         struct rpc_clnt *clnt = server->client;
6953         struct nfs4_fsid_present_arg args = {
6954                 .fh             = NFS_FH(inode),
6955                 .clientid       = clp->cl_clientid,
6956                 .renew          = 1,            /* append RENEW */
6957         };
6958         struct nfs4_fsid_present_res res = {
6959                 .renew          = 1,
6960         };
6961         struct rpc_message msg = {
6962                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6963                 .rpc_argp       = &args,
6964                 .rpc_resp       = &res,
6965                 .rpc_cred       = cred,
6966         };
6967         unsigned long now = jiffies;
6968         int status;
6969
6970         res.fh = nfs_alloc_fhandle();
6971         if (res.fh == NULL)
6972                 return -ENOMEM;
6973
6974         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6975         nfs4_set_sequence_privileged(&args.seq_args);
6976         status = nfs4_call_sync_sequence(clnt, server, &msg,
6977                                                 &args.seq_args, &res.seq_res);
6978         nfs_free_fhandle(res.fh);
6979         if (status)
6980                 return status;
6981
6982         do_renew_lease(clp, now);
6983         return 0;
6984 }
6985
6986 #ifdef CONFIG_NFS_V4_1
6987
6988 /*
6989  * This operation also signals the server that this client is
6990  * performing "lease moved" recovery.  The server can stop asserting
6991  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6992  * this operation is identified in the SEQUENCE operation in this
6993  * compound.
6994  */
6995 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6996 {
6997         struct nfs_server *server = NFS_SERVER(inode);
6998         struct rpc_clnt *clnt = server->client;
6999         struct nfs4_fsid_present_arg args = {
7000                 .fh             = NFS_FH(inode),
7001         };
7002         struct nfs4_fsid_present_res res = {
7003         };
7004         struct rpc_message msg = {
7005                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7006                 .rpc_argp       = &args,
7007                 .rpc_resp       = &res,
7008                 .rpc_cred       = cred,
7009         };
7010         int status;
7011
7012         res.fh = nfs_alloc_fhandle();
7013         if (res.fh == NULL)
7014                 return -ENOMEM;
7015
7016         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7017         nfs4_set_sequence_privileged(&args.seq_args);
7018         status = nfs4_call_sync_sequence(clnt, server, &msg,
7019                                                 &args.seq_args, &res.seq_res);
7020         nfs_free_fhandle(res.fh);
7021         if (status == NFS4_OK &&
7022             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7023                 status = -NFS4ERR_LEASE_MOVED;
7024         return status;
7025 }
7026
7027 #endif  /* CONFIG_NFS_V4_1 */
7028
7029 /**
7030  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7031  * @inode: inode on FSID to check
7032  * @cred: credential to use for this operation
7033  *
7034  * Server indicates whether the FSID is present, moved, or not
7035  * recognized.  This operation is necessary to clear a LEASE_MOVED
7036  * condition for this client ID.
7037  *
7038  * Returns NFS4_OK if the FSID is present on this server,
7039  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7040  *  NFS4ERR code if some error occurred on the server, or a
7041  *  negative errno if a local failure occurred.
7042  */
7043 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7044 {
7045         struct nfs_server *server = NFS_SERVER(inode);
7046         struct nfs_client *clp = server->nfs_client;
7047         const struct nfs4_mig_recovery_ops *ops =
7048                                         clp->cl_mvops->mig_recovery_ops;
7049         struct nfs4_exception exception = { };
7050         int status;
7051
7052         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7053                 (unsigned long long)server->fsid.major,
7054                 (unsigned long long)server->fsid.minor,
7055                 clp->cl_hostname);
7056         nfs_display_fhandle(NFS_FH(inode), __func__);
7057
7058         do {
7059                 status = ops->fsid_present(inode, cred);
7060                 if (status != -NFS4ERR_DELAY)
7061                         break;
7062                 nfs4_handle_exception(server, status, &exception);
7063         } while (exception.retry);
7064         return status;
7065 }
7066
7067 /**
7068  * If 'use_integrity' is true and the state managment nfs_client
7069  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7070  * and the machine credential as per RFC3530bis and RFC5661 Security
7071  * Considerations sections. Otherwise, just use the user cred with the
7072  * filesystem's rpc_client.
7073  */
7074 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7075 {
7076         int status;
7077         struct nfs4_secinfo_arg args = {
7078                 .dir_fh = NFS_FH(dir),
7079                 .name   = name,
7080         };
7081         struct nfs4_secinfo_res res = {
7082                 .flavors     = flavors,
7083         };
7084         struct rpc_message msg = {
7085                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7086                 .rpc_argp = &args,
7087                 .rpc_resp = &res,
7088         };
7089         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7090         struct rpc_cred *cred = NULL;
7091
7092         if (use_integrity) {
7093                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7094                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7095                 msg.rpc_cred = cred;
7096         }
7097
7098         dprintk("NFS call  secinfo %s\n", name->name);
7099
7100         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7101                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7102
7103         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7104                                 &res.seq_res, 0);
7105         dprintk("NFS reply  secinfo: %d\n", status);
7106
7107         if (cred)
7108                 put_rpccred(cred);
7109
7110         return status;
7111 }
7112
7113 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7114                       struct nfs4_secinfo_flavors *flavors)
7115 {
7116         struct nfs4_exception exception = { };
7117         int err;
7118         do {
7119                 err = -NFS4ERR_WRONGSEC;
7120
7121                 /* try to use integrity protection with machine cred */
7122                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7123                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7124
7125                 /*
7126                  * if unable to use integrity protection, or SECINFO with
7127                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7128                  * disallowed by spec, but exists in deployed servers) use
7129                  * the current filesystem's rpc_client and the user cred.
7130                  */
7131                 if (err == -NFS4ERR_WRONGSEC)
7132                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7133
7134                 trace_nfs4_secinfo(dir, name, err);
7135                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7136                                 &exception);
7137         } while (exception.retry);
7138         return err;
7139 }
7140
7141 #ifdef CONFIG_NFS_V4_1
7142 /*
7143  * Check the exchange flags returned by the server for invalid flags, having
7144  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7145  * DS flags set.
7146  */
7147 static int nfs4_check_cl_exchange_flags(u32 flags)
7148 {
7149         if (flags & ~EXCHGID4_FLAG_MASK_R)
7150                 goto out_inval;
7151         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7152             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7153                 goto out_inval;
7154         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7155                 goto out_inval;
7156         return NFS_OK;
7157 out_inval:
7158         return -NFS4ERR_INVAL;
7159 }
7160
7161 static bool
7162 nfs41_same_server_scope(struct nfs41_server_scope *a,
7163                         struct nfs41_server_scope *b)
7164 {
7165         if (a->server_scope_sz != b->server_scope_sz)
7166                 return false;
7167         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7168 }
7169
7170 static void
7171 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7172 {
7173 }
7174
7175 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7176         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7177 };
7178
7179 /*
7180  * nfs4_proc_bind_one_conn_to_session()
7181  *
7182  * The 4.1 client currently uses the same TCP connection for the
7183  * fore and backchannel.
7184  */
7185 static
7186 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7187                 struct rpc_xprt *xprt,
7188                 struct nfs_client *clp,
7189                 struct rpc_cred *cred)
7190 {
7191         int status;
7192         struct nfs41_bind_conn_to_session_args args = {
7193                 .client = clp,
7194                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7195         };
7196         struct nfs41_bind_conn_to_session_res res;
7197         struct rpc_message msg = {
7198                 .rpc_proc =
7199                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7200                 .rpc_argp = &args,
7201                 .rpc_resp = &res,
7202                 .rpc_cred = cred,
7203         };
7204         struct rpc_task_setup task_setup_data = {
7205                 .rpc_client = clnt,
7206                 .rpc_xprt = xprt,
7207                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7208                 .rpc_message = &msg,
7209                 .flags = RPC_TASK_TIMEOUT,
7210         };
7211         struct rpc_task *task;
7212
7213         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7214         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7215                 args.dir = NFS4_CDFC4_FORE;
7216
7217         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7218         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7219                 args.dir = NFS4_CDFC4_FORE;
7220
7221         task = rpc_run_task(&task_setup_data);
7222         if (!IS_ERR(task)) {
7223                 status = task->tk_status;
7224                 rpc_put_task(task);
7225         } else
7226                 status = PTR_ERR(task);
7227         trace_nfs4_bind_conn_to_session(clp, status);
7228         if (status == 0) {
7229                 if (memcmp(res.sessionid.data,
7230                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7231                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7232                         return -EIO;
7233                 }
7234                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7235                         dprintk("NFS: %s: Unexpected direction from server\n",
7236                                 __func__);
7237                         return -EIO;
7238                 }
7239                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7240                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7241                                 __func__);
7242                         return -EIO;
7243                 }
7244         }
7245
7246         return status;
7247 }
7248
7249 struct rpc_bind_conn_calldata {
7250         struct nfs_client *clp;
7251         struct rpc_cred *cred;
7252 };
7253
7254 static int
7255 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7256                 struct rpc_xprt *xprt,
7257                 void *calldata)
7258 {
7259         struct rpc_bind_conn_calldata *p = calldata;
7260
7261         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7262 }
7263
7264 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7265 {
7266         struct rpc_bind_conn_calldata data = {
7267                 .clp = clp,
7268                 .cred = cred,
7269         };
7270         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7271                         nfs4_proc_bind_conn_to_session_callback, &data);
7272 }
7273
7274 /*
7275  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7276  * and operations we'd like to see to enable certain features in the allow map
7277  */
7278 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7279         .how = SP4_MACH_CRED,
7280         .enforce.u.words = {
7281                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7282                       1 << (OP_EXCHANGE_ID - 32) |
7283                       1 << (OP_CREATE_SESSION - 32) |
7284                       1 << (OP_DESTROY_SESSION - 32) |
7285                       1 << (OP_DESTROY_CLIENTID - 32)
7286         },
7287         .allow.u.words = {
7288                 [0] = 1 << (OP_CLOSE) |
7289                       1 << (OP_OPEN_DOWNGRADE) |
7290                       1 << (OP_LOCKU) |
7291                       1 << (OP_DELEGRETURN) |
7292                       1 << (OP_COMMIT),
7293                 [1] = 1 << (OP_SECINFO - 32) |
7294                       1 << (OP_SECINFO_NO_NAME - 32) |
7295                       1 << (OP_LAYOUTRETURN - 32) |
7296                       1 << (OP_TEST_STATEID - 32) |
7297                       1 << (OP_FREE_STATEID - 32) |
7298                       1 << (OP_WRITE - 32)
7299         }
7300 };
7301
7302 /*
7303  * Select the state protection mode for client `clp' given the server results
7304  * from exchange_id in `sp'.
7305  *
7306  * Returns 0 on success, negative errno otherwise.
7307  */
7308 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7309                                  struct nfs41_state_protection *sp)
7310 {
7311         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7312                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7313                       1 << (OP_EXCHANGE_ID - 32) |
7314                       1 << (OP_CREATE_SESSION - 32) |
7315                       1 << (OP_DESTROY_SESSION - 32) |
7316                       1 << (OP_DESTROY_CLIENTID - 32)
7317         };
7318         unsigned int i;
7319
7320         if (sp->how == SP4_MACH_CRED) {
7321                 /* Print state protect result */
7322                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7323                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7324                         if (test_bit(i, sp->enforce.u.longs))
7325                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7326                         if (test_bit(i, sp->allow.u.longs))
7327                                 dfprintk(MOUNT, "  allow op %d\n", i);
7328                 }
7329
7330                 /* make sure nothing is on enforce list that isn't supported */
7331                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7332                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7333                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7334                                 return -EINVAL;
7335                         }
7336                 }
7337
7338                 /*
7339                  * Minimal mode - state operations are allowed to use machine
7340                  * credential.  Note this already happens by default, so the
7341                  * client doesn't have to do anything more than the negotiation.
7342                  *
7343                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7344                  *       we're already using the machine cred for exchange_id
7345                  *       and will never use a different cred.
7346                  */
7347                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7348                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7349                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7350                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7351                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7352                         dfprintk(MOUNT, "  minimal mode enabled\n");
7353                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7354                 } else {
7355                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7356                         return -EINVAL;
7357                 }
7358
7359                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7360                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7361                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7362                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7363                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7364                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7365                 }
7366
7367                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7368                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7369                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7370                                 &clp->cl_sp4_flags);
7371                 }
7372
7373                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7374                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7375                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7376                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7377                 }
7378
7379                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7380                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7381                         dfprintk(MOUNT, "  stateid mode enabled\n");
7382                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7383                 }
7384
7385                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7386                         dfprintk(MOUNT, "  write mode enabled\n");
7387                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7388                 }
7389
7390                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7391                         dfprintk(MOUNT, "  commit mode enabled\n");
7392                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7393                 }
7394         }
7395
7396         return 0;
7397 }
7398
7399 struct nfs41_exchange_id_data {
7400         struct nfs41_exchange_id_res res;
7401         struct nfs41_exchange_id_args args;
7402         struct rpc_xprt *xprt;
7403         int rpc_status;
7404 };
7405
7406 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7407 {
7408         struct nfs41_exchange_id_data *cdata =
7409                                         (struct nfs41_exchange_id_data *)data;
7410         struct nfs_client *clp = cdata->args.client;
7411         int status = task->tk_status;
7412
7413         trace_nfs4_exchange_id(clp, status);
7414
7415         if (status == 0)
7416                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7417
7418         if (cdata->xprt && status == 0) {
7419                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7420                                                       cdata->xprt);
7421                 goto out;
7422         }
7423
7424         if (status  == 0)
7425                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7426
7427         if (status == 0) {
7428                 clp->cl_clientid = cdata->res.clientid;
7429                 clp->cl_exchange_flags = cdata->res.flags;
7430                 clp->cl_seqid = cdata->res.seqid;
7431                 /* Client ID is not confirmed */
7432                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R))
7433                         clear_bit(NFS4_SESSION_ESTABLISHED,
7434                                   &clp->cl_session->session_state);
7435
7436                 kfree(clp->cl_serverowner);
7437                 clp->cl_serverowner = cdata->res.server_owner;
7438                 cdata->res.server_owner = NULL;
7439
7440                 /* use the most recent implementation id */
7441                 kfree(clp->cl_implid);
7442                 clp->cl_implid = cdata->res.impl_id;
7443                 cdata->res.impl_id = NULL;
7444
7445                 if (clp->cl_serverscope != NULL &&
7446                     !nfs41_same_server_scope(clp->cl_serverscope,
7447                                         cdata->res.server_scope)) {
7448                         dprintk("%s: server_scope mismatch detected\n",
7449                                 __func__);
7450                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7451                         kfree(clp->cl_serverscope);
7452                         clp->cl_serverscope = NULL;
7453                 }
7454
7455                 if (clp->cl_serverscope == NULL) {
7456                         clp->cl_serverscope = cdata->res.server_scope;
7457                         cdata->res.server_scope = NULL;
7458                 }
7459                 /* Save the EXCHANGE_ID verifier session trunk tests */
7460                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7461                        sizeof(clp->cl_confirm.data));
7462         }
7463 out:
7464         cdata->rpc_status = status;
7465         return;
7466 }
7467
7468 static void nfs4_exchange_id_release(void *data)
7469 {
7470         struct nfs41_exchange_id_data *cdata =
7471                                         (struct nfs41_exchange_id_data *)data;
7472
7473         if (cdata->xprt) {
7474                 xprt_put(cdata->xprt);
7475                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7476         }
7477         nfs_put_client(cdata->args.client);
7478         kfree(cdata->res.impl_id);
7479         kfree(cdata->res.server_scope);
7480         kfree(cdata->res.server_owner);
7481         kfree(cdata);
7482 }
7483
7484 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7485         .rpc_call_done = nfs4_exchange_id_done,
7486         .rpc_release = nfs4_exchange_id_release,
7487 };
7488
7489 /*
7490  * _nfs4_proc_exchange_id()
7491  *
7492  * Wrapper for EXCHANGE_ID operation.
7493  */
7494 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7495                         u32 sp4_how, struct rpc_xprt *xprt)
7496 {
7497         nfs4_verifier verifier;
7498         struct rpc_message msg = {
7499                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7500                 .rpc_cred = cred,
7501         };
7502         struct rpc_task_setup task_setup_data = {
7503                 .rpc_client = clp->cl_rpcclient,
7504                 .callback_ops = &nfs4_exchange_id_call_ops,
7505                 .rpc_message = &msg,
7506                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7507         };
7508         struct nfs41_exchange_id_data *calldata;
7509         struct rpc_task *task;
7510         int status;
7511
7512         if (!atomic_inc_not_zero(&clp->cl_count))
7513                 return -EIO;
7514
7515         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7516         if (!calldata) {
7517                 nfs_put_client(clp);
7518                 return -ENOMEM;
7519         }
7520
7521         if (!xprt)
7522                 nfs4_init_boot_verifier(clp, &verifier);
7523
7524         status = nfs4_init_uniform_client_string(clp);
7525         if (status)
7526                 goto out_calldata;
7527
7528         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7529                                                 GFP_NOFS);
7530         status = -ENOMEM;
7531         if (unlikely(calldata->res.server_owner == NULL))
7532                 goto out_calldata;
7533
7534         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7535                                         GFP_NOFS);
7536         if (unlikely(calldata->res.server_scope == NULL))
7537                 goto out_server_owner;
7538
7539         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7540         if (unlikely(calldata->res.impl_id == NULL))
7541                 goto out_server_scope;
7542
7543         switch (sp4_how) {
7544         case SP4_NONE:
7545                 calldata->args.state_protect.how = SP4_NONE;
7546                 break;
7547
7548         case SP4_MACH_CRED:
7549                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7550                 break;
7551
7552         default:
7553                 /* unsupported! */
7554                 WARN_ON_ONCE(1);
7555                 status = -EINVAL;
7556                 goto out_impl_id;
7557         }
7558         if (xprt) {
7559                 calldata->xprt = xprt;
7560                 task_setup_data.rpc_xprt = xprt;
7561                 task_setup_data.flags =
7562                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7563                 calldata->args.verifier = &clp->cl_confirm;
7564         } else {
7565                 calldata->args.verifier = &verifier;
7566         }
7567         calldata->args.client = clp;
7568 #ifdef CONFIG_NFS_V4_1_MIGRATION
7569         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7570         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7571         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7572 #else
7573         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7574         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7575 #endif
7576         msg.rpc_argp = &calldata->args;
7577         msg.rpc_resp = &calldata->res;
7578         task_setup_data.callback_data = calldata;
7579
7580         task = rpc_run_task(&task_setup_data);
7581         if (IS_ERR(task))
7582                 return PTR_ERR(task);
7583
7584         if (!xprt) {
7585                 status = rpc_wait_for_completion_task(task);
7586                 if (!status)
7587                         status = calldata->rpc_status;
7588         } else  /* session trunking test */
7589                 status = calldata->rpc_status;
7590
7591         rpc_put_task(task);
7592 out:
7593         return status;
7594
7595 out_impl_id:
7596         kfree(calldata->res.impl_id);
7597 out_server_scope:
7598         kfree(calldata->res.server_scope);
7599 out_server_owner:
7600         kfree(calldata->res.server_owner);
7601 out_calldata:
7602         kfree(calldata);
7603         nfs_put_client(clp);
7604         goto out;
7605 }
7606
7607 /*
7608  * nfs4_proc_exchange_id()
7609  *
7610  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7611  *
7612  * Since the clientid has expired, all compounds using sessions
7613  * associated with the stale clientid will be returning
7614  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7615  * be in some phase of session reset.
7616  *
7617  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7618  */
7619 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7620 {
7621         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7622         int status;
7623
7624         /* try SP4_MACH_CRED if krb5i/p */
7625         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7626             authflavor == RPC_AUTH_GSS_KRB5P) {
7627                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7628                 if (!status)
7629                         return 0;
7630         }
7631
7632         /* try SP4_NONE */
7633         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7634 }
7635
7636 /**
7637  * nfs4_test_session_trunk
7638  *
7639  * This is an add_xprt_test() test function called from
7640  * rpc_clnt_setup_test_and_add_xprt.
7641  *
7642  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7643  * and is dereferrenced in nfs4_exchange_id_release
7644  *
7645  * Upon success, add the new transport to the rpc_clnt
7646  *
7647  * @clnt: struct rpc_clnt to get new transport
7648  * @xprt: the rpc_xprt to test
7649  * @data: call data for _nfs4_proc_exchange_id.
7650  */
7651 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7652                             void *data)
7653 {
7654         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7655         u32 sp4_how;
7656
7657         dprintk("--> %s try %s\n", __func__,
7658                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7659
7660         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7661
7662         /* Test connection for session trunking. Async exchange_id call */
7663         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7664 }
7665 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7666
7667 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7668                 struct rpc_cred *cred)
7669 {
7670         struct rpc_message msg = {
7671                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7672                 .rpc_argp = clp,
7673                 .rpc_cred = cred,
7674         };
7675         int status;
7676
7677         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7678         trace_nfs4_destroy_clientid(clp, status);
7679         if (status)
7680                 dprintk("NFS: Got error %d from the server %s on "
7681                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7682         return status;
7683 }
7684
7685 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7686                 struct rpc_cred *cred)
7687 {
7688         unsigned int loop;
7689         int ret;
7690
7691         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7692                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7693                 switch (ret) {
7694                 case -NFS4ERR_DELAY:
7695                 case -NFS4ERR_CLIENTID_BUSY:
7696                         ssleep(1);
7697                         break;
7698                 default:
7699                         return ret;
7700                 }
7701         }
7702         return 0;
7703 }
7704
7705 int nfs4_destroy_clientid(struct nfs_client *clp)
7706 {
7707         struct rpc_cred *cred;
7708         int ret = 0;
7709
7710         if (clp->cl_mvops->minor_version < 1)
7711                 goto out;
7712         if (clp->cl_exchange_flags == 0)
7713                 goto out;
7714         if (clp->cl_preserve_clid)
7715                 goto out;
7716         cred = nfs4_get_clid_cred(clp);
7717         ret = nfs4_proc_destroy_clientid(clp, cred);
7718         if (cred)
7719                 put_rpccred(cred);
7720         switch (ret) {
7721         case 0:
7722         case -NFS4ERR_STALE_CLIENTID:
7723                 clp->cl_exchange_flags = 0;
7724         }
7725 out:
7726         return ret;
7727 }
7728
7729 struct nfs4_get_lease_time_data {
7730         struct nfs4_get_lease_time_args *args;
7731         struct nfs4_get_lease_time_res *res;
7732         struct nfs_client *clp;
7733 };
7734
7735 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7736                                         void *calldata)
7737 {
7738         struct nfs4_get_lease_time_data *data =
7739                         (struct nfs4_get_lease_time_data *)calldata;
7740
7741         dprintk("--> %s\n", __func__);
7742         /* just setup sequence, do not trigger session recovery
7743            since we're invoked within one */
7744         nfs4_setup_sequence(data->clp,
7745                         &data->args->la_seq_args,
7746                         &data->res->lr_seq_res,
7747                         task);
7748         dprintk("<-- %s\n", __func__);
7749 }
7750
7751 /*
7752  * Called from nfs4_state_manager thread for session setup, so don't recover
7753  * from sequence operation or clientid errors.
7754  */
7755 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7756 {
7757         struct nfs4_get_lease_time_data *data =
7758                         (struct nfs4_get_lease_time_data *)calldata;
7759
7760         dprintk("--> %s\n", __func__);
7761         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7762                 return;
7763         switch (task->tk_status) {
7764         case -NFS4ERR_DELAY:
7765         case -NFS4ERR_GRACE:
7766                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7767                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7768                 task->tk_status = 0;
7769                 /* fall through */
7770         case -NFS4ERR_RETRY_UNCACHED_REP:
7771                 rpc_restart_call_prepare(task);
7772                 return;
7773         }
7774         dprintk("<-- %s\n", __func__);
7775 }
7776
7777 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7778         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7779         .rpc_call_done = nfs4_get_lease_time_done,
7780 };
7781
7782 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7783 {
7784         struct rpc_task *task;
7785         struct nfs4_get_lease_time_args args;
7786         struct nfs4_get_lease_time_res res = {
7787                 .lr_fsinfo = fsinfo,
7788         };
7789         struct nfs4_get_lease_time_data data = {
7790                 .args = &args,
7791                 .res = &res,
7792                 .clp = clp,
7793         };
7794         struct rpc_message msg = {
7795                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7796                 .rpc_argp = &args,
7797                 .rpc_resp = &res,
7798         };
7799         struct rpc_task_setup task_setup = {
7800                 .rpc_client = clp->cl_rpcclient,
7801                 .rpc_message = &msg,
7802                 .callback_ops = &nfs4_get_lease_time_ops,
7803                 .callback_data = &data,
7804                 .flags = RPC_TASK_TIMEOUT,
7805         };
7806         int status;
7807
7808         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7809         nfs4_set_sequence_privileged(&args.la_seq_args);
7810         task = rpc_run_task(&task_setup);
7811
7812         if (IS_ERR(task))
7813                 return PTR_ERR(task);
7814
7815         status = task->tk_status;
7816         rpc_put_task(task);
7817         return status;
7818 }
7819
7820 /*
7821  * Initialize the values to be used by the client in CREATE_SESSION
7822  * If nfs4_init_session set the fore channel request and response sizes,
7823  * use them.
7824  *
7825  * Set the back channel max_resp_sz_cached to zero to force the client to
7826  * always set csa_cachethis to FALSE because the current implementation
7827  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7828  */
7829 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7830                                     struct rpc_clnt *clnt)
7831 {
7832         unsigned int max_rqst_sz, max_resp_sz;
7833         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7834
7835         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7836         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7837
7838         /* Fore channel attributes */
7839         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7840         args->fc_attrs.max_resp_sz = max_resp_sz;
7841         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7842         args->fc_attrs.max_reqs = max_session_slots;
7843
7844         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7845                 "max_ops=%u max_reqs=%u\n",
7846                 __func__,
7847                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7848                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7849
7850         /* Back channel attributes */
7851         args->bc_attrs.max_rqst_sz = max_bc_payload;
7852         args->bc_attrs.max_resp_sz = max_bc_payload;
7853         args->bc_attrs.max_resp_sz_cached = 0;
7854         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7855         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7856
7857         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7858                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7859                 __func__,
7860                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7861                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7862                 args->bc_attrs.max_reqs);
7863 }
7864
7865 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7866                 struct nfs41_create_session_res *res)
7867 {
7868         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7869         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7870
7871         if (rcvd->max_resp_sz > sent->max_resp_sz)
7872                 return -EINVAL;
7873         /*
7874          * Our requested max_ops is the minimum we need; we're not
7875          * prepared to break up compounds into smaller pieces than that.
7876          * So, no point even trying to continue if the server won't
7877          * cooperate:
7878          */
7879         if (rcvd->max_ops < sent->max_ops)
7880                 return -EINVAL;
7881         if (rcvd->max_reqs == 0)
7882                 return -EINVAL;
7883         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7884                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7885         return 0;
7886 }
7887
7888 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7889                 struct nfs41_create_session_res *res)
7890 {
7891         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7892         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7893
7894         if (!(res->flags & SESSION4_BACK_CHAN))
7895                 goto out;
7896         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7897                 return -EINVAL;
7898         if (rcvd->max_resp_sz < sent->max_resp_sz)
7899                 return -EINVAL;
7900         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7901                 return -EINVAL;
7902         if (rcvd->max_ops > sent->max_ops)
7903                 return -EINVAL;
7904         if (rcvd->max_reqs > sent->max_reqs)
7905                 return -EINVAL;
7906 out:
7907         return 0;
7908 }
7909
7910 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7911                                      struct nfs41_create_session_res *res)
7912 {
7913         int ret;
7914
7915         ret = nfs4_verify_fore_channel_attrs(args, res);
7916         if (ret)
7917                 return ret;
7918         return nfs4_verify_back_channel_attrs(args, res);
7919 }
7920
7921 static void nfs4_update_session(struct nfs4_session *session,
7922                 struct nfs41_create_session_res *res)
7923 {
7924         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7925         /* Mark client id and session as being confirmed */
7926         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7927         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7928         session->flags = res->flags;
7929         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7930         if (res->flags & SESSION4_BACK_CHAN)
7931                 memcpy(&session->bc_attrs, &res->bc_attrs,
7932                                 sizeof(session->bc_attrs));
7933 }
7934
7935 static int _nfs4_proc_create_session(struct nfs_client *clp,
7936                 struct rpc_cred *cred)
7937 {
7938         struct nfs4_session *session = clp->cl_session;
7939         struct nfs41_create_session_args args = {
7940                 .client = clp,
7941                 .clientid = clp->cl_clientid,
7942                 .seqid = clp->cl_seqid,
7943                 .cb_program = NFS4_CALLBACK,
7944         };
7945         struct nfs41_create_session_res res;
7946
7947         struct rpc_message msg = {
7948                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7949                 .rpc_argp = &args,
7950                 .rpc_resp = &res,
7951                 .rpc_cred = cred,
7952         };
7953         int status;
7954
7955         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7956         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7957
7958         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7959         trace_nfs4_create_session(clp, status);
7960
7961         switch (status) {
7962         case -NFS4ERR_STALE_CLIENTID:
7963         case -NFS4ERR_DELAY:
7964         case -ETIMEDOUT:
7965         case -EACCES:
7966         case -EAGAIN:
7967                 goto out;
7968         };
7969
7970         clp->cl_seqid++;
7971         if (!status) {
7972                 /* Verify the session's negotiated channel_attrs values */
7973                 status = nfs4_verify_channel_attrs(&args, &res);
7974                 /* Increment the clientid slot sequence id */
7975                 if (status)
7976                         goto out;
7977                 nfs4_update_session(session, &res);
7978         }
7979 out:
7980         return status;
7981 }
7982
7983 /*
7984  * Issues a CREATE_SESSION operation to the server.
7985  * It is the responsibility of the caller to verify the session is
7986  * expired before calling this routine.
7987  */
7988 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7989 {
7990         int status;
7991         unsigned *ptr;
7992         struct nfs4_session *session = clp->cl_session;
7993
7994         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7995
7996         status = _nfs4_proc_create_session(clp, cred);
7997         if (status)
7998                 goto out;
7999
8000         /* Init or reset the session slot tables */
8001         status = nfs4_setup_session_slot_tables(session);
8002         dprintk("slot table setup returned %d\n", status);
8003         if (status)
8004                 goto out;
8005
8006         ptr = (unsigned *)&session->sess_id.data[0];
8007         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8008                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8009 out:
8010         dprintk("<-- %s\n", __func__);
8011         return status;
8012 }
8013
8014 /*
8015  * Issue the over-the-wire RPC DESTROY_SESSION.
8016  * The caller must serialize access to this routine.
8017  */
8018 int nfs4_proc_destroy_session(struct nfs4_session *session,
8019                 struct rpc_cred *cred)
8020 {
8021         struct rpc_message msg = {
8022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8023                 .rpc_argp = session,
8024                 .rpc_cred = cred,
8025         };
8026         int status = 0;
8027
8028         dprintk("--> nfs4_proc_destroy_session\n");
8029
8030         /* session is still being setup */
8031         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8032                 return 0;
8033
8034         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8035         trace_nfs4_destroy_session(session->clp, status);
8036
8037         if (status)
8038                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8039                         "Session has been destroyed regardless...\n", status);
8040
8041         dprintk("<-- nfs4_proc_destroy_session\n");
8042         return status;
8043 }
8044
8045 /*
8046  * Renew the cl_session lease.
8047  */
8048 struct nfs4_sequence_data {
8049         struct nfs_client *clp;
8050         struct nfs4_sequence_args args;
8051         struct nfs4_sequence_res res;
8052 };
8053
8054 static void nfs41_sequence_release(void *data)
8055 {
8056         struct nfs4_sequence_data *calldata = data;
8057         struct nfs_client *clp = calldata->clp;
8058
8059         if (atomic_read(&clp->cl_count) > 1)
8060                 nfs4_schedule_state_renewal(clp);
8061         nfs_put_client(clp);
8062         kfree(calldata);
8063 }
8064
8065 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8066 {
8067         switch(task->tk_status) {
8068         case -NFS4ERR_DELAY:
8069                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8070                 return -EAGAIN;
8071         default:
8072                 nfs4_schedule_lease_recovery(clp);
8073         }
8074         return 0;
8075 }
8076
8077 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8078 {
8079         struct nfs4_sequence_data *calldata = data;
8080         struct nfs_client *clp = calldata->clp;
8081
8082         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8083                 return;
8084
8085         trace_nfs4_sequence(clp, task->tk_status);
8086         if (task->tk_status < 0) {
8087                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8088                 if (atomic_read(&clp->cl_count) == 1)
8089                         goto out;
8090
8091                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8092                         rpc_restart_call_prepare(task);
8093                         return;
8094                 }
8095         }
8096         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8097 out:
8098         dprintk("<-- %s\n", __func__);
8099 }
8100
8101 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8102 {
8103         struct nfs4_sequence_data *calldata = data;
8104         struct nfs_client *clp = calldata->clp;
8105         struct nfs4_sequence_args *args;
8106         struct nfs4_sequence_res *res;
8107
8108         args = task->tk_msg.rpc_argp;
8109         res = task->tk_msg.rpc_resp;
8110
8111         nfs4_setup_sequence(clp, args, res, task);
8112 }
8113
8114 static const struct rpc_call_ops nfs41_sequence_ops = {
8115         .rpc_call_done = nfs41_sequence_call_done,
8116         .rpc_call_prepare = nfs41_sequence_prepare,
8117         .rpc_release = nfs41_sequence_release,
8118 };
8119
8120 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8121                 struct rpc_cred *cred,
8122                 bool is_privileged)
8123 {
8124         struct nfs4_sequence_data *calldata;
8125         struct rpc_message msg = {
8126                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8127                 .rpc_cred = cred,
8128         };
8129         struct rpc_task_setup task_setup_data = {
8130                 .rpc_client = clp->cl_rpcclient,
8131                 .rpc_message = &msg,
8132                 .callback_ops = &nfs41_sequence_ops,
8133                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8134         };
8135
8136         if (!atomic_inc_not_zero(&clp->cl_count))
8137                 return ERR_PTR(-EIO);
8138         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8139         if (calldata == NULL) {
8140                 nfs_put_client(clp);
8141                 return ERR_PTR(-ENOMEM);
8142         }
8143         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8144         if (is_privileged)
8145                 nfs4_set_sequence_privileged(&calldata->args);
8146         msg.rpc_argp = &calldata->args;
8147         msg.rpc_resp = &calldata->res;
8148         calldata->clp = clp;
8149         task_setup_data.callback_data = calldata;
8150
8151         return rpc_run_task(&task_setup_data);
8152 }
8153
8154 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8155 {
8156         struct rpc_task *task;
8157         int ret = 0;
8158
8159         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8160                 return -EAGAIN;
8161         task = _nfs41_proc_sequence(clp, cred, false);
8162         if (IS_ERR(task))
8163                 ret = PTR_ERR(task);
8164         else
8165                 rpc_put_task_async(task);
8166         dprintk("<-- %s status=%d\n", __func__, ret);
8167         return ret;
8168 }
8169
8170 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8171 {
8172         struct rpc_task *task;
8173         int ret;
8174
8175         task = _nfs41_proc_sequence(clp, cred, true);
8176         if (IS_ERR(task)) {
8177                 ret = PTR_ERR(task);
8178                 goto out;
8179         }
8180         ret = rpc_wait_for_completion_task(task);
8181         if (!ret)
8182                 ret = task->tk_status;
8183         rpc_put_task(task);
8184 out:
8185         dprintk("<-- %s status=%d\n", __func__, ret);
8186         return ret;
8187 }
8188
8189 struct nfs4_reclaim_complete_data {
8190         struct nfs_client *clp;
8191         struct nfs41_reclaim_complete_args arg;
8192         struct nfs41_reclaim_complete_res res;
8193 };
8194
8195 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8196 {
8197         struct nfs4_reclaim_complete_data *calldata = data;
8198
8199         nfs4_setup_sequence(calldata->clp,
8200                         &calldata->arg.seq_args,
8201                         &calldata->res.seq_res,
8202                         task);
8203 }
8204
8205 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8206 {
8207         switch(task->tk_status) {
8208         case 0:
8209         case -NFS4ERR_COMPLETE_ALREADY:
8210         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8211                 break;
8212         case -NFS4ERR_DELAY:
8213                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8214                 /* fall through */
8215         case -NFS4ERR_RETRY_UNCACHED_REP:
8216                 return -EAGAIN;
8217         case -NFS4ERR_BADSESSION:
8218         case -NFS4ERR_DEADSESSION:
8219         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8220                 nfs4_schedule_session_recovery(clp->cl_session,
8221                                 task->tk_status);
8222                 break;
8223         default:
8224                 nfs4_schedule_lease_recovery(clp);
8225         }
8226         return 0;
8227 }
8228
8229 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8230 {
8231         struct nfs4_reclaim_complete_data *calldata = data;
8232         struct nfs_client *clp = calldata->clp;
8233         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8234
8235         dprintk("--> %s\n", __func__);
8236         if (!nfs41_sequence_done(task, res))
8237                 return;
8238
8239         trace_nfs4_reclaim_complete(clp, task->tk_status);
8240         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8241                 rpc_restart_call_prepare(task);
8242                 return;
8243         }
8244         dprintk("<-- %s\n", __func__);
8245 }
8246
8247 static void nfs4_free_reclaim_complete_data(void *data)
8248 {
8249         struct nfs4_reclaim_complete_data *calldata = data;
8250
8251         kfree(calldata);
8252 }
8253
8254 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8255         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8256         .rpc_call_done = nfs4_reclaim_complete_done,
8257         .rpc_release = nfs4_free_reclaim_complete_data,
8258 };
8259
8260 /*
8261  * Issue a global reclaim complete.
8262  */
8263 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8264                 struct rpc_cred *cred)
8265 {
8266         struct nfs4_reclaim_complete_data *calldata;
8267         struct rpc_task *task;
8268         struct rpc_message msg = {
8269                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8270                 .rpc_cred = cred,
8271         };
8272         struct rpc_task_setup task_setup_data = {
8273                 .rpc_client = clp->cl_rpcclient,
8274                 .rpc_message = &msg,
8275                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8276                 .flags = RPC_TASK_ASYNC,
8277         };
8278         int status = -ENOMEM;
8279
8280         dprintk("--> %s\n", __func__);
8281         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8282         if (calldata == NULL)
8283                 goto out;
8284         calldata->clp = clp;
8285         calldata->arg.one_fs = 0;
8286
8287         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8288         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8289         msg.rpc_argp = &calldata->arg;
8290         msg.rpc_resp = &calldata->res;
8291         task_setup_data.callback_data = calldata;
8292         task = rpc_run_task(&task_setup_data);
8293         if (IS_ERR(task)) {
8294                 status = PTR_ERR(task);
8295                 goto out;
8296         }
8297         status = rpc_wait_for_completion_task(task);
8298         if (status == 0)
8299                 status = task->tk_status;
8300         rpc_put_task(task);
8301 out:
8302         dprintk("<-- %s status=%d\n", __func__, status);
8303         return status;
8304 }
8305
8306 static void
8307 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8308 {
8309         struct nfs4_layoutget *lgp = calldata;
8310         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8311
8312         dprintk("--> %s\n", __func__);
8313         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8314                                 &lgp->res.seq_res, task);
8315         dprintk("<-- %s\n", __func__);
8316 }
8317
8318 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8319 {
8320         struct nfs4_layoutget *lgp = calldata;
8321
8322         dprintk("--> %s\n", __func__);
8323         nfs41_sequence_process(task, &lgp->res.seq_res);
8324         dprintk("<-- %s\n", __func__);
8325 }
8326
8327 static int
8328 nfs4_layoutget_handle_exception(struct rpc_task *task,
8329                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8330 {
8331         struct inode *inode = lgp->args.inode;
8332         struct nfs_server *server = NFS_SERVER(inode);
8333         struct pnfs_layout_hdr *lo;
8334         int nfs4err = task->tk_status;
8335         int err, status = 0;
8336         LIST_HEAD(head);
8337
8338         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8339
8340         switch (nfs4err) {
8341         case 0:
8342                 goto out;
8343
8344         /*
8345          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8346          * on the file. set tk_status to -ENODATA to tell upper layer to
8347          * retry go inband.
8348          */
8349         case -NFS4ERR_LAYOUTUNAVAILABLE:
8350                 status = -ENODATA;
8351                 goto out;
8352         /*
8353          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8354          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8355          */
8356         case -NFS4ERR_BADLAYOUT:
8357                 status = -EOVERFLOW;
8358                 goto out;
8359         /*
8360          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8361          * (or clients) writing to the same RAID stripe except when
8362          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8363          *
8364          * Treat it like we would RECALLCONFLICT -- we retry for a little
8365          * while, and then eventually give up.
8366          */
8367         case -NFS4ERR_LAYOUTTRYLATER:
8368                 if (lgp->args.minlength == 0) {
8369                         status = -EOVERFLOW;
8370                         goto out;
8371                 }
8372                 status = -EBUSY;
8373                 break;
8374         case -NFS4ERR_RECALLCONFLICT:
8375                 status = -ERECALLCONFLICT;
8376                 break;
8377         case -NFS4ERR_DELEG_REVOKED:
8378         case -NFS4ERR_ADMIN_REVOKED:
8379         case -NFS4ERR_EXPIRED:
8380         case -NFS4ERR_BAD_STATEID:
8381                 exception->timeout = 0;
8382                 spin_lock(&inode->i_lock);
8383                 lo = NFS_I(inode)->layout;
8384                 /* If the open stateid was bad, then recover it. */
8385                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8386                     nfs4_stateid_match_other(&lgp->args.stateid,
8387                                         &lgp->args.ctx->state->stateid)) {
8388                         spin_unlock(&inode->i_lock);
8389                         exception->state = lgp->args.ctx->state;
8390                         exception->stateid = &lgp->args.stateid;
8391                         break;
8392                 }
8393
8394                 /*
8395                  * Mark the bad layout state as invalid, then retry
8396                  */
8397                 pnfs_mark_layout_stateid_invalid(lo, &head);
8398                 spin_unlock(&inode->i_lock);
8399                 nfs_commit_inode(inode, 0);
8400                 pnfs_free_lseg_list(&head);
8401                 status = -EAGAIN;
8402                 goto out;
8403         }
8404
8405         nfs4_sequence_free_slot(&lgp->res.seq_res);
8406         err = nfs4_handle_exception(server, nfs4err, exception);
8407         if (!status) {
8408                 if (exception->retry)
8409                         status = -EAGAIN;
8410                 else
8411                         status = err;
8412         }
8413 out:
8414         dprintk("<-- %s\n", __func__);
8415         return status;
8416 }
8417
8418 static size_t max_response_pages(struct nfs_server *server)
8419 {
8420         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8421         return nfs_page_array_len(0, max_resp_sz);
8422 }
8423
8424 static void nfs4_free_pages(struct page **pages, size_t size)
8425 {
8426         int i;
8427
8428         if (!pages)
8429                 return;
8430
8431         for (i = 0; i < size; i++) {
8432                 if (!pages[i])
8433                         break;
8434                 __free_page(pages[i]);
8435         }
8436         kfree(pages);
8437 }
8438
8439 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8440 {
8441         struct page **pages;
8442         int i;
8443
8444         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8445         if (!pages) {
8446                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8447                 return NULL;
8448         }
8449
8450         for (i = 0; i < size; i++) {
8451                 pages[i] = alloc_page(gfp_flags);
8452                 if (!pages[i]) {
8453                         dprintk("%s: failed to allocate page\n", __func__);
8454                         nfs4_free_pages(pages, size);
8455                         return NULL;
8456                 }
8457         }
8458
8459         return pages;
8460 }
8461
8462 static void nfs4_layoutget_release(void *calldata)
8463 {
8464         struct nfs4_layoutget *lgp = calldata;
8465         struct inode *inode = lgp->args.inode;
8466         struct nfs_server *server = NFS_SERVER(inode);
8467         size_t max_pages = max_response_pages(server);
8468
8469         dprintk("--> %s\n", __func__);
8470         nfs4_sequence_free_slot(&lgp->res.seq_res);
8471         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8472         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8473         put_nfs_open_context(lgp->args.ctx);
8474         kfree(calldata);
8475         dprintk("<-- %s\n", __func__);
8476 }
8477
8478 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8479         .rpc_call_prepare = nfs4_layoutget_prepare,
8480         .rpc_call_done = nfs4_layoutget_done,
8481         .rpc_release = nfs4_layoutget_release,
8482 };
8483
8484 struct pnfs_layout_segment *
8485 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8486 {
8487         struct inode *inode = lgp->args.inode;
8488         struct nfs_server *server = NFS_SERVER(inode);
8489         size_t max_pages = max_response_pages(server);
8490         struct rpc_task *task;
8491         struct rpc_message msg = {
8492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8493                 .rpc_argp = &lgp->args,
8494                 .rpc_resp = &lgp->res,
8495                 .rpc_cred = lgp->cred,
8496         };
8497         struct rpc_task_setup task_setup_data = {
8498                 .rpc_client = server->client,
8499                 .rpc_message = &msg,
8500                 .callback_ops = &nfs4_layoutget_call_ops,
8501                 .callback_data = lgp,
8502                 .flags = RPC_TASK_ASYNC,
8503         };
8504         struct pnfs_layout_segment *lseg = NULL;
8505         struct nfs4_exception exception = {
8506                 .inode = inode,
8507                 .timeout = *timeout,
8508         };
8509         int status = 0;
8510
8511         dprintk("--> %s\n", __func__);
8512
8513         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8514         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8515
8516         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8517         if (!lgp->args.layout.pages) {
8518                 nfs4_layoutget_release(lgp);
8519                 return ERR_PTR(-ENOMEM);
8520         }
8521         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8522
8523         lgp->res.layoutp = &lgp->args.layout;
8524         lgp->res.seq_res.sr_slot = NULL;
8525         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8526
8527         task = rpc_run_task(&task_setup_data);
8528         if (IS_ERR(task))
8529                 return ERR_CAST(task);
8530         status = rpc_wait_for_completion_task(task);
8531         if (status == 0) {
8532                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8533                 *timeout = exception.timeout;
8534         }
8535
8536         trace_nfs4_layoutget(lgp->args.ctx,
8537                         &lgp->args.range,
8538                         &lgp->res.range,
8539                         &lgp->res.stateid,
8540                         status);
8541
8542         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8543         if (status == 0 && lgp->res.layoutp->len)
8544                 lseg = pnfs_layout_process(lgp);
8545         rpc_put_task(task);
8546         dprintk("<-- %s status=%d\n", __func__, status);
8547         if (status)
8548                 return ERR_PTR(status);
8549         return lseg;
8550 }
8551
8552 static void
8553 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8554 {
8555         struct nfs4_layoutreturn *lrp = calldata;
8556
8557         dprintk("--> %s\n", __func__);
8558         nfs4_setup_sequence(lrp->clp,
8559                         &lrp->args.seq_args,
8560                         &lrp->res.seq_res,
8561                         task);
8562 }
8563
8564 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8565 {
8566         struct nfs4_layoutreturn *lrp = calldata;
8567         struct nfs_server *server;
8568
8569         dprintk("--> %s\n", __func__);
8570
8571         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8572                 return;
8573
8574         server = NFS_SERVER(lrp->args.inode);
8575         switch (task->tk_status) {
8576         default:
8577                 task->tk_status = 0;
8578         case 0:
8579                 break;
8580         case -NFS4ERR_DELAY:
8581                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8582                         break;
8583                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8584                 rpc_restart_call_prepare(task);
8585                 return;
8586         }
8587         dprintk("<-- %s\n", __func__);
8588 }
8589
8590 static void nfs4_layoutreturn_release(void *calldata)
8591 {
8592         struct nfs4_layoutreturn *lrp = calldata;
8593         struct pnfs_layout_hdr *lo = lrp->args.layout;
8594
8595         dprintk("--> %s\n", __func__);
8596         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8597                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8598         nfs4_sequence_free_slot(&lrp->res.seq_res);
8599         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8600                 lrp->ld_private.ops->free(&lrp->ld_private);
8601         pnfs_put_layout_hdr(lrp->args.layout);
8602         nfs_iput_and_deactive(lrp->inode);
8603         kfree(calldata);
8604         dprintk("<-- %s\n", __func__);
8605 }
8606
8607 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8608         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8609         .rpc_call_done = nfs4_layoutreturn_done,
8610         .rpc_release = nfs4_layoutreturn_release,
8611 };
8612
8613 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8614 {
8615         struct rpc_task *task;
8616         struct rpc_message msg = {
8617                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8618                 .rpc_argp = &lrp->args,
8619                 .rpc_resp = &lrp->res,
8620                 .rpc_cred = lrp->cred,
8621         };
8622         struct rpc_task_setup task_setup_data = {
8623                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8624                 .rpc_message = &msg,
8625                 .callback_ops = &nfs4_layoutreturn_call_ops,
8626                 .callback_data = lrp,
8627         };
8628         int status = 0;
8629
8630         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8631                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8632                         &task_setup_data.rpc_client, &msg);
8633
8634         dprintk("--> %s\n", __func__);
8635         if (!sync) {
8636                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8637                 if (!lrp->inode) {
8638                         nfs4_layoutreturn_release(lrp);
8639                         return -EAGAIN;
8640                 }
8641                 task_setup_data.flags |= RPC_TASK_ASYNC;
8642         }
8643         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8644         task = rpc_run_task(&task_setup_data);
8645         if (IS_ERR(task))
8646                 return PTR_ERR(task);
8647         if (sync)
8648                 status = task->tk_status;
8649         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8650         dprintk("<-- %s status=%d\n", __func__, status);
8651         rpc_put_task(task);
8652         return status;
8653 }
8654
8655 static int
8656 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8657                 struct pnfs_device *pdev,
8658                 struct rpc_cred *cred)
8659 {
8660         struct nfs4_getdeviceinfo_args args = {
8661                 .pdev = pdev,
8662                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8663                         NOTIFY_DEVICEID4_DELETE,
8664         };
8665         struct nfs4_getdeviceinfo_res res = {
8666                 .pdev = pdev,
8667         };
8668         struct rpc_message msg = {
8669                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8670                 .rpc_argp = &args,
8671                 .rpc_resp = &res,
8672                 .rpc_cred = cred,
8673         };
8674         int status;
8675
8676         dprintk("--> %s\n", __func__);
8677         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8678         if (res.notification & ~args.notify_types)
8679                 dprintk("%s: unsupported notification\n", __func__);
8680         if (res.notification != args.notify_types)
8681                 pdev->nocache = 1;
8682
8683         dprintk("<-- %s status=%d\n", __func__, status);
8684
8685         return status;
8686 }
8687
8688 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8689                 struct pnfs_device *pdev,
8690                 struct rpc_cred *cred)
8691 {
8692         struct nfs4_exception exception = { };
8693         int err;
8694
8695         do {
8696                 err = nfs4_handle_exception(server,
8697                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8698                                         &exception);
8699         } while (exception.retry);
8700         return err;
8701 }
8702 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8703
8704 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8705 {
8706         struct nfs4_layoutcommit_data *data = calldata;
8707         struct nfs_server *server = NFS_SERVER(data->args.inode);
8708
8709         nfs4_setup_sequence(server->nfs_client,
8710                         &data->args.seq_args,
8711                         &data->res.seq_res,
8712                         task);
8713 }
8714
8715 static void
8716 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8717 {
8718         struct nfs4_layoutcommit_data *data = calldata;
8719         struct nfs_server *server = NFS_SERVER(data->args.inode);
8720
8721         if (!nfs41_sequence_done(task, &data->res.seq_res))
8722                 return;
8723
8724         switch (task->tk_status) { /* Just ignore these failures */
8725         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8726         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8727         case -NFS4ERR_BADLAYOUT:     /* no layout */
8728         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8729                 task->tk_status = 0;
8730         case 0:
8731                 break;
8732         default:
8733                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8734                         rpc_restart_call_prepare(task);
8735                         return;
8736                 }
8737         }
8738 }
8739
8740 static void nfs4_layoutcommit_release(void *calldata)
8741 {
8742         struct nfs4_layoutcommit_data *data = calldata;
8743
8744         pnfs_cleanup_layoutcommit(data);
8745         nfs_post_op_update_inode_force_wcc(data->args.inode,
8746                                            data->res.fattr);
8747         put_rpccred(data->cred);
8748         nfs_iput_and_deactive(data->inode);
8749         kfree(data);
8750 }
8751
8752 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8753         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8754         .rpc_call_done = nfs4_layoutcommit_done,
8755         .rpc_release = nfs4_layoutcommit_release,
8756 };
8757
8758 int
8759 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8760 {
8761         struct rpc_message msg = {
8762                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8763                 .rpc_argp = &data->args,
8764                 .rpc_resp = &data->res,
8765                 .rpc_cred = data->cred,
8766         };
8767         struct rpc_task_setup task_setup_data = {
8768                 .task = &data->task,
8769                 .rpc_client = NFS_CLIENT(data->args.inode),
8770                 .rpc_message = &msg,
8771                 .callback_ops = &nfs4_layoutcommit_ops,
8772                 .callback_data = data,
8773         };
8774         struct rpc_task *task;
8775         int status = 0;
8776
8777         dprintk("NFS: initiating layoutcommit call. sync %d "
8778                 "lbw: %llu inode %lu\n", sync,
8779                 data->args.lastbytewritten,
8780                 data->args.inode->i_ino);
8781
8782         if (!sync) {
8783                 data->inode = nfs_igrab_and_active(data->args.inode);
8784                 if (data->inode == NULL) {
8785                         nfs4_layoutcommit_release(data);
8786                         return -EAGAIN;
8787                 }
8788                 task_setup_data.flags = RPC_TASK_ASYNC;
8789         }
8790         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8791         task = rpc_run_task(&task_setup_data);
8792         if (IS_ERR(task))
8793                 return PTR_ERR(task);
8794         if (sync)
8795                 status = task->tk_status;
8796         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8797         dprintk("%s: status %d\n", __func__, status);
8798         rpc_put_task(task);
8799         return status;
8800 }
8801
8802 /**
8803  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8804  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8805  */
8806 static int
8807 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8808                     struct nfs_fsinfo *info,
8809                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8810 {
8811         struct nfs41_secinfo_no_name_args args = {
8812                 .style = SECINFO_STYLE_CURRENT_FH,
8813         };
8814         struct nfs4_secinfo_res res = {
8815                 .flavors = flavors,
8816         };
8817         struct rpc_message msg = {
8818                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8819                 .rpc_argp = &args,
8820                 .rpc_resp = &res,
8821         };
8822         struct rpc_clnt *clnt = server->client;
8823         struct rpc_cred *cred = NULL;
8824         int status;
8825
8826         if (use_integrity) {
8827                 clnt = server->nfs_client->cl_rpcclient;
8828                 cred = nfs4_get_clid_cred(server->nfs_client);
8829                 msg.rpc_cred = cred;
8830         }
8831
8832         dprintk("--> %s\n", __func__);
8833         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8834                                 &res.seq_res, 0);
8835         dprintk("<-- %s status=%d\n", __func__, status);
8836
8837         if (cred)
8838                 put_rpccred(cred);
8839
8840         return status;
8841 }
8842
8843 static int
8844 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8845                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8846 {
8847         struct nfs4_exception exception = { };
8848         int err;
8849         do {
8850                 /* first try using integrity protection */
8851                 err = -NFS4ERR_WRONGSEC;
8852
8853                 /* try to use integrity protection with machine cred */
8854                 if (_nfs4_is_integrity_protected(server->nfs_client))
8855                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8856                                                           flavors, true);
8857
8858                 /*
8859                  * if unable to use integrity protection, or SECINFO with
8860                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8861                  * disallowed by spec, but exists in deployed servers) use
8862                  * the current filesystem's rpc_client and the user cred.
8863                  */
8864                 if (err == -NFS4ERR_WRONGSEC)
8865                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8866                                                           flavors, false);
8867
8868                 switch (err) {
8869                 case 0:
8870                 case -NFS4ERR_WRONGSEC:
8871                 case -ENOTSUPP:
8872                         goto out;
8873                 default:
8874                         err = nfs4_handle_exception(server, err, &exception);
8875                 }
8876         } while (exception.retry);
8877 out:
8878         return err;
8879 }
8880
8881 static int
8882 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8883                     struct nfs_fsinfo *info)
8884 {
8885         int err;
8886         struct page *page;
8887         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8888         struct nfs4_secinfo_flavors *flavors;
8889         struct nfs4_secinfo4 *secinfo;
8890         int i;
8891
8892         page = alloc_page(GFP_KERNEL);
8893         if (!page) {
8894                 err = -ENOMEM;
8895                 goto out;
8896         }
8897
8898         flavors = page_address(page);
8899         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8900
8901         /*
8902          * Fall back on "guess and check" method if
8903          * the server doesn't support SECINFO_NO_NAME
8904          */
8905         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8906                 err = nfs4_find_root_sec(server, fhandle, info);
8907                 goto out_freepage;
8908         }
8909         if (err)
8910                 goto out_freepage;
8911
8912         for (i = 0; i < flavors->num_flavors; i++) {
8913                 secinfo = &flavors->flavors[i];
8914
8915                 switch (secinfo->flavor) {
8916                 case RPC_AUTH_NULL:
8917                 case RPC_AUTH_UNIX:
8918                 case RPC_AUTH_GSS:
8919                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8920                                         &secinfo->flavor_info);
8921                         break;
8922                 default:
8923                         flavor = RPC_AUTH_MAXFLAVOR;
8924                         break;
8925                 }
8926
8927                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8928                         flavor = RPC_AUTH_MAXFLAVOR;
8929
8930                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8931                         err = nfs4_lookup_root_sec(server, fhandle,
8932                                                    info, flavor);
8933                         if (!err)
8934                                 break;
8935                 }
8936         }
8937
8938         if (flavor == RPC_AUTH_MAXFLAVOR)
8939                 err = -EPERM;
8940
8941 out_freepage:
8942         put_page(page);
8943         if (err == -EACCES)
8944                 return -EPERM;
8945 out:
8946         return err;
8947 }
8948
8949 static int _nfs41_test_stateid(struct nfs_server *server,
8950                 nfs4_stateid *stateid,
8951                 struct rpc_cred *cred)
8952 {
8953         int status;
8954         struct nfs41_test_stateid_args args = {
8955                 .stateid = stateid,
8956         };
8957         struct nfs41_test_stateid_res res;
8958         struct rpc_message msg = {
8959                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8960                 .rpc_argp = &args,
8961                 .rpc_resp = &res,
8962                 .rpc_cred = cred,
8963         };
8964         struct rpc_clnt *rpc_client = server->client;
8965
8966         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8967                 &rpc_client, &msg);
8968
8969         dprintk("NFS call  test_stateid %p\n", stateid);
8970         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8971         nfs4_set_sequence_privileged(&args.seq_args);
8972         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8973                         &args.seq_args, &res.seq_res);
8974         if (status != NFS_OK) {
8975                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8976                 return status;
8977         }
8978         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8979         return -res.status;
8980 }
8981
8982 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8983                 int err, struct nfs4_exception *exception)
8984 {
8985         exception->retry = 0;
8986         switch(err) {
8987         case -NFS4ERR_DELAY:
8988         case -NFS4ERR_RETRY_UNCACHED_REP:
8989                 nfs4_handle_exception(server, err, exception);
8990                 break;
8991         case -NFS4ERR_BADSESSION:
8992         case -NFS4ERR_BADSLOT:
8993         case -NFS4ERR_BAD_HIGH_SLOT:
8994         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8995         case -NFS4ERR_DEADSESSION:
8996                 nfs4_do_handle_exception(server, err, exception);
8997         }
8998 }
8999
9000 /**
9001  * nfs41_test_stateid - perform a TEST_STATEID operation
9002  *
9003  * @server: server / transport on which to perform the operation
9004  * @stateid: state ID to test
9005  * @cred: credential
9006  *
9007  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9008  * Otherwise a negative NFS4ERR value is returned if the operation
9009  * failed or the state ID is not currently valid.
9010  */
9011 static int nfs41_test_stateid(struct nfs_server *server,
9012                 nfs4_stateid *stateid,
9013                 struct rpc_cred *cred)
9014 {
9015         struct nfs4_exception exception = { };
9016         int err;
9017         do {
9018                 err = _nfs41_test_stateid(server, stateid, cred);
9019                 nfs4_handle_delay_or_session_error(server, err, &exception);
9020         } while (exception.retry);
9021         return err;
9022 }
9023
9024 struct nfs_free_stateid_data {
9025         struct nfs_server *server;
9026         struct nfs41_free_stateid_args args;
9027         struct nfs41_free_stateid_res res;
9028 };
9029
9030 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9031 {
9032         struct nfs_free_stateid_data *data = calldata;
9033         nfs4_setup_sequence(data->server->nfs_client,
9034                         &data->args.seq_args,
9035                         &data->res.seq_res,
9036                         task);
9037 }
9038
9039 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9040 {
9041         struct nfs_free_stateid_data *data = calldata;
9042
9043         nfs41_sequence_done(task, &data->res.seq_res);
9044
9045         switch (task->tk_status) {
9046         case -NFS4ERR_DELAY:
9047                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9048                         rpc_restart_call_prepare(task);
9049         }
9050 }
9051
9052 static void nfs41_free_stateid_release(void *calldata)
9053 {
9054         kfree(calldata);
9055 }
9056
9057 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9058         .rpc_call_prepare = nfs41_free_stateid_prepare,
9059         .rpc_call_done = nfs41_free_stateid_done,
9060         .rpc_release = nfs41_free_stateid_release,
9061 };
9062
9063 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9064                 const nfs4_stateid *stateid,
9065                 struct rpc_cred *cred,
9066                 bool privileged)
9067 {
9068         struct rpc_message msg = {
9069                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9070                 .rpc_cred = cred,
9071         };
9072         struct rpc_task_setup task_setup = {
9073                 .rpc_client = server->client,
9074                 .rpc_message = &msg,
9075                 .callback_ops = &nfs41_free_stateid_ops,
9076                 .flags = RPC_TASK_ASYNC,
9077         };
9078         struct nfs_free_stateid_data *data;
9079
9080         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9081                 &task_setup.rpc_client, &msg);
9082
9083         dprintk("NFS call  free_stateid %p\n", stateid);
9084         data = kmalloc(sizeof(*data), GFP_NOFS);
9085         if (!data)
9086                 return ERR_PTR(-ENOMEM);
9087         data->server = server;
9088         nfs4_stateid_copy(&data->args.stateid, stateid);
9089
9090         task_setup.callback_data = data;
9091
9092         msg.rpc_argp = &data->args;
9093         msg.rpc_resp = &data->res;
9094         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9095         if (privileged)
9096                 nfs4_set_sequence_privileged(&data->args.seq_args);
9097
9098         return rpc_run_task(&task_setup);
9099 }
9100
9101 /**
9102  * nfs41_free_stateid - perform a FREE_STATEID operation
9103  *
9104  * @server: server / transport on which to perform the operation
9105  * @stateid: state ID to release
9106  * @cred: credential
9107  * @is_recovery: set to true if this call needs to be privileged
9108  *
9109  * Note: this function is always asynchronous.
9110  */
9111 static int nfs41_free_stateid(struct nfs_server *server,
9112                 const nfs4_stateid *stateid,
9113                 struct rpc_cred *cred,
9114                 bool is_recovery)
9115 {
9116         struct rpc_task *task;
9117
9118         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9119         if (IS_ERR(task))
9120                 return PTR_ERR(task);
9121         rpc_put_task(task);
9122         return 0;
9123 }
9124
9125 static void
9126 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9127 {
9128         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9129
9130         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9131         nfs4_free_lock_state(server, lsp);
9132 }
9133
9134 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9135                 const nfs4_stateid *s2)
9136 {
9137         if (s1->type != s2->type)
9138                 return false;
9139
9140         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9141                 return false;
9142
9143         if (s1->seqid == s2->seqid)
9144                 return true;
9145
9146         return s1->seqid == 0 || s2->seqid == 0;
9147 }
9148
9149 #endif /* CONFIG_NFS_V4_1 */
9150
9151 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9152                 const nfs4_stateid *s2)
9153 {
9154         return nfs4_stateid_match(s1, s2);
9155 }
9156
9157
9158 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9159         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9160         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9161         .recover_open   = nfs4_open_reclaim,
9162         .recover_lock   = nfs4_lock_reclaim,
9163         .establish_clid = nfs4_init_clientid,
9164         .detect_trunking = nfs40_discover_server_trunking,
9165 };
9166
9167 #if defined(CONFIG_NFS_V4_1)
9168 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9169         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9170         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9171         .recover_open   = nfs4_open_reclaim,
9172         .recover_lock   = nfs4_lock_reclaim,
9173         .establish_clid = nfs41_init_clientid,
9174         .reclaim_complete = nfs41_proc_reclaim_complete,
9175         .detect_trunking = nfs41_discover_server_trunking,
9176 };
9177 #endif /* CONFIG_NFS_V4_1 */
9178
9179 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9180         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9181         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9182         .recover_open   = nfs40_open_expired,
9183         .recover_lock   = nfs4_lock_expired,
9184         .establish_clid = nfs4_init_clientid,
9185 };
9186
9187 #if defined(CONFIG_NFS_V4_1)
9188 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9189         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9190         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9191         .recover_open   = nfs41_open_expired,
9192         .recover_lock   = nfs41_lock_expired,
9193         .establish_clid = nfs41_init_clientid,
9194 };
9195 #endif /* CONFIG_NFS_V4_1 */
9196
9197 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9198         .sched_state_renewal = nfs4_proc_async_renew,
9199         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9200         .renew_lease = nfs4_proc_renew,
9201 };
9202
9203 #if defined(CONFIG_NFS_V4_1)
9204 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9205         .sched_state_renewal = nfs41_proc_async_sequence,
9206         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9207         .renew_lease = nfs4_proc_sequence,
9208 };
9209 #endif
9210
9211 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9212         .get_locations = _nfs40_proc_get_locations,
9213         .fsid_present = _nfs40_proc_fsid_present,
9214 };
9215
9216 #if defined(CONFIG_NFS_V4_1)
9217 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9218         .get_locations = _nfs41_proc_get_locations,
9219         .fsid_present = _nfs41_proc_fsid_present,
9220 };
9221 #endif  /* CONFIG_NFS_V4_1 */
9222
9223 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9224         .minor_version = 0,
9225         .init_caps = NFS_CAP_READDIRPLUS
9226                 | NFS_CAP_ATOMIC_OPEN
9227                 | NFS_CAP_POSIX_LOCK,
9228         .init_client = nfs40_init_client,
9229         .shutdown_client = nfs40_shutdown_client,
9230         .match_stateid = nfs4_match_stateid,
9231         .find_root_sec = nfs4_find_root_sec,
9232         .free_lock_state = nfs4_release_lockowner,
9233         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9234         .alloc_seqid = nfs_alloc_seqid,
9235         .call_sync_ops = &nfs40_call_sync_ops,
9236         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9237         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9238         .state_renewal_ops = &nfs40_state_renewal_ops,
9239         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9240 };
9241
9242 #if defined(CONFIG_NFS_V4_1)
9243 static struct nfs_seqid *
9244 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9245 {
9246         return NULL;
9247 }
9248
9249 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9250         .minor_version = 1,
9251         .init_caps = NFS_CAP_READDIRPLUS
9252                 | NFS_CAP_ATOMIC_OPEN
9253                 | NFS_CAP_POSIX_LOCK
9254                 | NFS_CAP_STATEID_NFSV41
9255                 | NFS_CAP_ATOMIC_OPEN_V1,
9256         .init_client = nfs41_init_client,
9257         .shutdown_client = nfs41_shutdown_client,
9258         .match_stateid = nfs41_match_stateid,
9259         .find_root_sec = nfs41_find_root_sec,
9260         .free_lock_state = nfs41_free_lock_state,
9261         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9262         .alloc_seqid = nfs_alloc_no_seqid,
9263         .session_trunk = nfs4_test_session_trunk,
9264         .call_sync_ops = &nfs41_call_sync_ops,
9265         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9266         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9267         .state_renewal_ops = &nfs41_state_renewal_ops,
9268         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9269 };
9270 #endif
9271
9272 #if defined(CONFIG_NFS_V4_2)
9273 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9274         .minor_version = 2,
9275         .init_caps = NFS_CAP_READDIRPLUS
9276                 | NFS_CAP_ATOMIC_OPEN
9277                 | NFS_CAP_POSIX_LOCK
9278                 | NFS_CAP_STATEID_NFSV41
9279                 | NFS_CAP_ATOMIC_OPEN_V1
9280                 | NFS_CAP_ALLOCATE
9281                 | NFS_CAP_COPY
9282                 | NFS_CAP_DEALLOCATE
9283                 | NFS_CAP_SEEK
9284                 | NFS_CAP_LAYOUTSTATS
9285                 | NFS_CAP_CLONE,
9286         .init_client = nfs41_init_client,
9287         .shutdown_client = nfs41_shutdown_client,
9288         .match_stateid = nfs41_match_stateid,
9289         .find_root_sec = nfs41_find_root_sec,
9290         .free_lock_state = nfs41_free_lock_state,
9291         .call_sync_ops = &nfs41_call_sync_ops,
9292         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9293         .alloc_seqid = nfs_alloc_no_seqid,
9294         .session_trunk = nfs4_test_session_trunk,
9295         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9296         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9297         .state_renewal_ops = &nfs41_state_renewal_ops,
9298         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9299 };
9300 #endif
9301
9302 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9303         [0] = &nfs_v4_0_minor_ops,
9304 #if defined(CONFIG_NFS_V4_1)
9305         [1] = &nfs_v4_1_minor_ops,
9306 #endif
9307 #if defined(CONFIG_NFS_V4_2)
9308         [2] = &nfs_v4_2_minor_ops,
9309 #endif
9310 };
9311
9312 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9313 {
9314         ssize_t error, error2;
9315
9316         error = generic_listxattr(dentry, list, size);
9317         if (error < 0)
9318                 return error;
9319         if (list) {
9320                 list += error;
9321                 size -= error;
9322         }
9323
9324         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9325         if (error2 < 0)
9326                 return error2;
9327         return error + error2;
9328 }
9329
9330 static const struct inode_operations nfs4_dir_inode_operations = {
9331         .create         = nfs_create,
9332         .lookup         = nfs_lookup,
9333         .atomic_open    = nfs_atomic_open,
9334         .link           = nfs_link,
9335         .unlink         = nfs_unlink,
9336         .symlink        = nfs_symlink,
9337         .mkdir          = nfs_mkdir,
9338         .rmdir          = nfs_rmdir,
9339         .mknod          = nfs_mknod,
9340         .rename         = nfs_rename,
9341         .permission     = nfs_permission,
9342         .getattr        = nfs_getattr,
9343         .setattr        = nfs_setattr,
9344         .listxattr      = nfs4_listxattr,
9345 };
9346
9347 static const struct inode_operations nfs4_file_inode_operations = {
9348         .permission     = nfs_permission,
9349         .getattr        = nfs_getattr,
9350         .setattr        = nfs_setattr,
9351         .listxattr      = nfs4_listxattr,
9352 };
9353
9354 const struct nfs_rpc_ops nfs_v4_clientops = {
9355         .version        = 4,                    /* protocol version */
9356         .dentry_ops     = &nfs4_dentry_operations,
9357         .dir_inode_ops  = &nfs4_dir_inode_operations,
9358         .file_inode_ops = &nfs4_file_inode_operations,
9359         .file_ops       = &nfs4_file_operations,
9360         .getroot        = nfs4_proc_get_root,
9361         .submount       = nfs4_submount,
9362         .try_mount      = nfs4_try_mount,
9363         .getattr        = nfs4_proc_getattr,
9364         .setattr        = nfs4_proc_setattr,
9365         .lookup         = nfs4_proc_lookup,
9366         .lookupp        = nfs4_proc_lookupp,
9367         .access         = nfs4_proc_access,
9368         .readlink       = nfs4_proc_readlink,
9369         .create         = nfs4_proc_create,
9370         .remove         = nfs4_proc_remove,
9371         .unlink_setup   = nfs4_proc_unlink_setup,
9372         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9373         .unlink_done    = nfs4_proc_unlink_done,
9374         .rename_setup   = nfs4_proc_rename_setup,
9375         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9376         .rename_done    = nfs4_proc_rename_done,
9377         .link           = nfs4_proc_link,
9378         .symlink        = nfs4_proc_symlink,
9379         .mkdir          = nfs4_proc_mkdir,
9380         .rmdir          = nfs4_proc_remove,
9381         .readdir        = nfs4_proc_readdir,
9382         .mknod          = nfs4_proc_mknod,
9383         .statfs         = nfs4_proc_statfs,
9384         .fsinfo         = nfs4_proc_fsinfo,
9385         .pathconf       = nfs4_proc_pathconf,
9386         .set_capabilities = nfs4_server_capabilities,
9387         .decode_dirent  = nfs4_decode_dirent,
9388         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9389         .read_setup     = nfs4_proc_read_setup,
9390         .read_done      = nfs4_read_done,
9391         .write_setup    = nfs4_proc_write_setup,
9392         .write_done     = nfs4_write_done,
9393         .commit_setup   = nfs4_proc_commit_setup,
9394         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9395         .commit_done    = nfs4_commit_done,
9396         .lock           = nfs4_proc_lock,
9397         .clear_acl_cache = nfs4_zap_acl_attr,
9398         .close_context  = nfs4_close_context,
9399         .open_context   = nfs4_atomic_open,
9400         .have_delegation = nfs4_have_delegation,
9401         .return_delegation = nfs4_inode_return_delegation,
9402         .alloc_client   = nfs4_alloc_client,
9403         .init_client    = nfs4_init_client,
9404         .free_client    = nfs4_free_client,
9405         .create_server  = nfs4_create_server,
9406         .clone_server   = nfs_clone_server,
9407 };
9408
9409 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9410         .name   = XATTR_NAME_NFSV4_ACL,
9411         .list   = nfs4_xattr_list_nfs4_acl,
9412         .get    = nfs4_xattr_get_nfs4_acl,
9413         .set    = nfs4_xattr_set_nfs4_acl,
9414 };
9415
9416 const struct xattr_handler *nfs4_xattr_handlers[] = {
9417         &nfs4_xattr_nfs4_acl_handler,
9418 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9419         &nfs4_xattr_nfs4_label_handler,
9420 #endif
9421         NULL
9422 };
9423
9424 /*
9425  * Local variables:
9426  *  c-basic-offset: 8
9427  * End:
9428  */