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 <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  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, flags;
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                 if (S_ISDIR(state->inode->i_mode))
2254                         mask = NFS4_ACCESS_LOOKUP;
2255                 else
2256                         mask = NFS4_ACCESS_EXECUTE;
2257         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2258                 mask = NFS4_ACCESS_READ;
2259
2260         cache.cred = cred;
2261         cache.jiffies = jiffies;
2262         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2263         nfs_access_add_cache(state->inode, &cache);
2264
2265         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2266         if ((mask & ~cache.mask & flags) == 0)
2267                 return 0;
2268
2269         return -EACCES;
2270 }
2271
2272 /*
2273  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2274  */
2275 static int _nfs4_proc_open(struct nfs4_opendata *data)
2276 {
2277         struct inode *dir = d_inode(data->dir);
2278         struct nfs_server *server = NFS_SERVER(dir);
2279         struct nfs_openargs *o_arg = &data->o_arg;
2280         struct nfs_openres *o_res = &data->o_res;
2281         int status;
2282
2283         status = nfs4_run_open_task(data, 0);
2284         if (!data->rpc_done)
2285                 return status;
2286         if (status != 0) {
2287                 if (status == -NFS4ERR_BADNAME &&
2288                                 !(o_arg->open_flags & O_CREAT))
2289                         return -ENOENT;
2290                 return status;
2291         }
2292
2293         nfs_fattr_map_and_free_names(server, &data->f_attr);
2294
2295         if (o_arg->open_flags & O_CREAT) {
2296                 if (o_arg->open_flags & O_EXCL)
2297                         data->file_created = true;
2298                 else if (o_res->cinfo.before != o_res->cinfo.after)
2299                         data->file_created = true;
2300                 if (data->file_created || dir->i_version != o_res->cinfo.after)
2301                         update_changeattr(dir, &o_res->cinfo,
2302                                         o_res->f_attr->time_start);
2303         }
2304         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2305                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2306         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2307                 status = _nfs4_proc_open_confirm(data);
2308                 if (status != 0)
2309                         return status;
2310         }
2311         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2312                 nfs4_sequence_free_slot(&o_res->seq_res);
2313                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2314         }
2315         return 0;
2316 }
2317
2318 /*
2319  * OPEN_EXPIRED:
2320  *      reclaim state on the server after a network partition.
2321  *      Assumes caller holds the appropriate lock
2322  */
2323 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2324 {
2325         struct nfs4_opendata *opendata;
2326         int ret;
2327
2328         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2329                         NFS4_OPEN_CLAIM_FH);
2330         if (IS_ERR(opendata))
2331                 return PTR_ERR(opendata);
2332         ret = nfs4_open_recover(opendata, state);
2333         if (ret == -ESTALE)
2334                 d_drop(ctx->dentry);
2335         nfs4_opendata_put(opendata);
2336         return ret;
2337 }
2338
2339 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2340 {
2341         struct nfs_server *server = NFS_SERVER(state->inode);
2342         struct nfs4_exception exception = { };
2343         int err;
2344
2345         do {
2346                 err = _nfs4_open_expired(ctx, state);
2347                 trace_nfs4_open_expired(ctx, 0, err);
2348                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2349                         continue;
2350                 switch (err) {
2351                 default:
2352                         goto out;
2353                 case -NFS4ERR_GRACE:
2354                 case -NFS4ERR_DELAY:
2355                         nfs4_handle_exception(server, err, &exception);
2356                         err = 0;
2357                 }
2358         } while (exception.retry);
2359 out:
2360         return err;
2361 }
2362
2363 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2364 {
2365         struct nfs_open_context *ctx;
2366         int ret;
2367
2368         ctx = nfs4_state_find_open_context(state);
2369         if (IS_ERR(ctx))
2370                 return -EAGAIN;
2371         ret = nfs4_do_open_expired(ctx, state);
2372         put_nfs_open_context(ctx);
2373         return ret;
2374 }
2375
2376 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2377                 const nfs4_stateid *stateid)
2378 {
2379         nfs_remove_bad_delegation(state->inode, stateid);
2380         write_seqlock(&state->seqlock);
2381         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2382         write_sequnlock(&state->seqlock);
2383         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2384 }
2385
2386 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2387 {
2388         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2389                 nfs_finish_clear_delegation_stateid(state, NULL);
2390 }
2391
2392 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2393 {
2394         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2395         nfs40_clear_delegation_stateid(state);
2396         return nfs4_open_expired(sp, state);
2397 }
2398
2399 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2400                 nfs4_stateid *stateid,
2401                 struct rpc_cred *cred)
2402 {
2403         return -NFS4ERR_BAD_STATEID;
2404 }
2405
2406 #if defined(CONFIG_NFS_V4_1)
2407 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2408                 nfs4_stateid *stateid,
2409                 struct rpc_cred *cred)
2410 {
2411         int status;
2412
2413         switch (stateid->type) {
2414         default:
2415                 break;
2416         case NFS4_INVALID_STATEID_TYPE:
2417         case NFS4_SPECIAL_STATEID_TYPE:
2418                 return -NFS4ERR_BAD_STATEID;
2419         case NFS4_REVOKED_STATEID_TYPE:
2420                 goto out_free;
2421         }
2422
2423         status = nfs41_test_stateid(server, stateid, cred);
2424         switch (status) {
2425         case -NFS4ERR_EXPIRED:
2426         case -NFS4ERR_ADMIN_REVOKED:
2427         case -NFS4ERR_DELEG_REVOKED:
2428                 break;
2429         default:
2430                 return status;
2431         }
2432 out_free:
2433         /* Ack the revoked state to the server */
2434         nfs41_free_stateid(server, stateid, cred, true);
2435         return -NFS4ERR_EXPIRED;
2436 }
2437
2438 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2439 {
2440         struct nfs_server *server = NFS_SERVER(state->inode);
2441         nfs4_stateid stateid;
2442         struct nfs_delegation *delegation;
2443         struct rpc_cred *cred;
2444         int status;
2445
2446         /* Get the delegation credential for use by test/free_stateid */
2447         rcu_read_lock();
2448         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2449         if (delegation == NULL) {
2450                 rcu_read_unlock();
2451                 return;
2452         }
2453
2454         nfs4_stateid_copy(&stateid, &delegation->stateid);
2455         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2456                 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2457                         &delegation->flags)) {
2458                 rcu_read_unlock();
2459                 nfs_finish_clear_delegation_stateid(state, &stateid);
2460                 return;
2461         }
2462
2463         cred = get_rpccred(delegation->cred);
2464         rcu_read_unlock();
2465         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2466         trace_nfs4_test_delegation_stateid(state, NULL, status);
2467         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2468                 nfs_finish_clear_delegation_stateid(state, &stateid);
2469
2470         put_rpccred(cred);
2471 }
2472
2473 /**
2474  * nfs41_check_expired_locks - possibly free a lock stateid
2475  *
2476  * @state: NFSv4 state for an inode
2477  *
2478  * Returns NFS_OK if recovery for this stateid is now finished.
2479  * Otherwise a negative NFS4ERR value is returned.
2480  */
2481 static int nfs41_check_expired_locks(struct nfs4_state *state)
2482 {
2483         int status, ret = NFS_OK;
2484         struct nfs4_lock_state *lsp, *prev = NULL;
2485         struct nfs_server *server = NFS_SERVER(state->inode);
2486
2487         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2488                 goto out;
2489
2490         spin_lock(&state->state_lock);
2491         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2492                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2493                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2494
2495                         atomic_inc(&lsp->ls_count);
2496                         spin_unlock(&state->state_lock);
2497
2498                         nfs4_put_lock_state(prev);
2499                         prev = lsp;
2500
2501                         status = nfs41_test_and_free_expired_stateid(server,
2502                                         &lsp->ls_stateid,
2503                                         cred);
2504                         trace_nfs4_test_lock_stateid(state, lsp, status);
2505                         if (status == -NFS4ERR_EXPIRED ||
2506                             status == -NFS4ERR_BAD_STATEID) {
2507                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2508                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2509                                 if (!recover_lost_locks)
2510                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2511                         } else if (status != NFS_OK) {
2512                                 ret = status;
2513                                 nfs4_put_lock_state(prev);
2514                                 goto out;
2515                         }
2516                         spin_lock(&state->state_lock);
2517                 }
2518         }
2519         spin_unlock(&state->state_lock);
2520         nfs4_put_lock_state(prev);
2521 out:
2522         return ret;
2523 }
2524
2525 /**
2526  * nfs41_check_open_stateid - possibly free an open stateid
2527  *
2528  * @state: NFSv4 state for an inode
2529  *
2530  * Returns NFS_OK if recovery for this stateid is now finished.
2531  * Otherwise a negative NFS4ERR value is returned.
2532  */
2533 static int nfs41_check_open_stateid(struct nfs4_state *state)
2534 {
2535         struct nfs_server *server = NFS_SERVER(state->inode);
2536         nfs4_stateid *stateid = &state->open_stateid;
2537         struct rpc_cred *cred = state->owner->so_cred;
2538         int status;
2539
2540         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2541                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2542                         if (nfs4_have_delegation(state->inode, state->state))
2543                                 return NFS_OK;
2544                         return -NFS4ERR_OPENMODE;
2545                 }
2546                 return -NFS4ERR_BAD_STATEID;
2547         }
2548         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2549         trace_nfs4_test_open_stateid(state, NULL, status);
2550         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2551                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2552                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2553                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2554                 clear_bit(NFS_OPEN_STATE, &state->flags);
2555                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2556         }
2557         if (status != NFS_OK)
2558                 return status;
2559         if (nfs_open_stateid_recover_openmode(state))
2560                 return -NFS4ERR_OPENMODE;
2561         return NFS_OK;
2562 }
2563
2564 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2565 {
2566         int status;
2567
2568         nfs41_check_delegation_stateid(state);
2569         status = nfs41_check_expired_locks(state);
2570         if (status != NFS_OK)
2571                 return status;
2572         status = nfs41_check_open_stateid(state);
2573         if (status != NFS_OK)
2574                 status = nfs4_open_expired(sp, state);
2575         return status;
2576 }
2577 #endif
2578
2579 /*
2580  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2581  * fields corresponding to attributes that were used to store the verifier.
2582  * Make sure we clobber those fields in the later setattr call
2583  */
2584 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2585                                 struct iattr *sattr, struct nfs4_label **label)
2586 {
2587         const u32 *attrset = opendata->o_res.attrset;
2588
2589         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2590             !(sattr->ia_valid & ATTR_ATIME_SET))
2591                 sattr->ia_valid |= ATTR_ATIME;
2592
2593         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2594             !(sattr->ia_valid & ATTR_MTIME_SET))
2595                 sattr->ia_valid |= ATTR_MTIME;
2596
2597         /* Except MODE, it seems harmless of setting twice. */
2598         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2599                 (attrset[1] & FATTR4_WORD1_MODE ||
2600                  attrset[2] & FATTR4_WORD2_MODE_UMASK))
2601                 sattr->ia_valid &= ~ATTR_MODE;
2602
2603         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2604                 *label = NULL;
2605 }
2606
2607 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2608                 fmode_t fmode,
2609                 int flags,
2610                 struct nfs_open_context *ctx)
2611 {
2612         struct nfs4_state_owner *sp = opendata->owner;
2613         struct nfs_server *server = sp->so_server;
2614         struct dentry *dentry;
2615         struct nfs4_state *state;
2616         unsigned int seq;
2617         int ret;
2618
2619         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2620
2621         ret = _nfs4_proc_open(opendata);
2622         if (ret != 0)
2623                 goto out;
2624
2625         state = nfs4_opendata_to_nfs4_state(opendata);
2626         ret = PTR_ERR(state);
2627         if (IS_ERR(state))
2628                 goto out;
2629         ctx->state = state;
2630         if (server->caps & NFS_CAP_POSIX_LOCK)
2631                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2632         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2633                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2634
2635         dentry = opendata->dentry;
2636         if (d_really_is_negative(dentry)) {
2637                 struct dentry *alias;
2638                 d_drop(dentry);
2639                 alias = d_exact_alias(dentry, state->inode);
2640                 if (!alias)
2641                         alias = d_splice_alias(igrab(state->inode), dentry);
2642                 /* d_splice_alias() can't fail here - it's a non-directory */
2643                 if (alias) {
2644                         dput(ctx->dentry);
2645                         ctx->dentry = dentry = alias;
2646                 }
2647                 nfs_set_verifier(dentry,
2648                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2649         }
2650
2651         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2652         if (ret != 0)
2653                 goto out;
2654
2655         if (d_inode(dentry) == state->inode) {
2656                 nfs_inode_attach_open_context(ctx);
2657                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2658                         nfs4_schedule_stateid_recovery(server, state);
2659         }
2660 out:
2661         return ret;
2662 }
2663
2664 /*
2665  * Returns a referenced nfs4_state
2666  */
2667 static int _nfs4_do_open(struct inode *dir,
2668                         struct nfs_open_context *ctx,
2669                         int flags,
2670                         struct iattr *sattr,
2671                         struct nfs4_label *label,
2672                         int *opened)
2673 {
2674         struct nfs4_state_owner  *sp;
2675         struct nfs4_state     *state = NULL;
2676         struct nfs_server       *server = NFS_SERVER(dir);
2677         struct nfs4_opendata *opendata;
2678         struct dentry *dentry = ctx->dentry;
2679         struct rpc_cred *cred = ctx->cred;
2680         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2681         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2682         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2683         struct nfs4_label *olabel = NULL;
2684         int status;
2685
2686         /* Protect against reboot recovery conflicts */
2687         status = -ENOMEM;
2688         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2689         if (sp == NULL) {
2690                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2691                 goto out_err;
2692         }
2693         status = nfs4_client_recover_expired_lease(server->nfs_client);
2694         if (status != 0)
2695                 goto err_put_state_owner;
2696         if (d_really_is_positive(dentry))
2697                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2698         status = -ENOMEM;
2699         if (d_really_is_positive(dentry))
2700                 claim = NFS4_OPEN_CLAIM_FH;
2701         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2702                         label, claim, GFP_KERNEL);
2703         if (opendata == NULL)
2704                 goto err_put_state_owner;
2705
2706         if (label) {
2707                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2708                 if (IS_ERR(olabel)) {
2709                         status = PTR_ERR(olabel);
2710                         goto err_opendata_put;
2711                 }
2712         }
2713
2714         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2715                 if (!opendata->f_attr.mdsthreshold) {
2716                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2717                         if (!opendata->f_attr.mdsthreshold)
2718                                 goto err_free_label;
2719                 }
2720                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2721         }
2722         if (d_really_is_positive(dentry))
2723                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2724
2725         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2726         if (status != 0)
2727                 goto err_free_label;
2728         state = ctx->state;
2729
2730         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2731             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2732                 nfs4_exclusive_attrset(opendata, sattr, &label);
2733                 /*
2734                  * send create attributes which was not set by open
2735                  * with an extra setattr.
2736                  */
2737                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2738                         nfs_fattr_init(opendata->o_res.f_attr);
2739                         status = nfs4_do_setattr(state->inode, cred,
2740                                         opendata->o_res.f_attr, sattr,
2741                                         ctx, label, olabel);
2742                         if (status == 0) {
2743                                 nfs_setattr_update_inode(state->inode, sattr,
2744                                                 opendata->o_res.f_attr);
2745                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2746                         }
2747                 }
2748         }
2749         if (opened && opendata->file_created)
2750                 *opened |= FILE_CREATED;
2751
2752         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2753                 *ctx_th = opendata->f_attr.mdsthreshold;
2754                 opendata->f_attr.mdsthreshold = NULL;
2755         }
2756
2757         nfs4_label_free(olabel);
2758
2759         nfs4_opendata_put(opendata);
2760         nfs4_put_state_owner(sp);
2761         return 0;
2762 err_free_label:
2763         nfs4_label_free(olabel);
2764 err_opendata_put:
2765         nfs4_opendata_put(opendata);
2766 err_put_state_owner:
2767         nfs4_put_state_owner(sp);
2768 out_err:
2769         return status;
2770 }
2771
2772
2773 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2774                                         struct nfs_open_context *ctx,
2775                                         int flags,
2776                                         struct iattr *sattr,
2777                                         struct nfs4_label *label,
2778                                         int *opened)
2779 {
2780         struct nfs_server *server = NFS_SERVER(dir);
2781         struct nfs4_exception exception = { };
2782         struct nfs4_state *res;
2783         int status;
2784
2785         do {
2786                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2787                 res = ctx->state;
2788                 trace_nfs4_open_file(ctx, flags, status);
2789                 if (status == 0)
2790                         break;
2791                 /* NOTE: BAD_SEQID means the server and client disagree about the
2792                  * book-keeping w.r.t. state-changing operations
2793                  * (OPEN/CLOSE/LOCK/LOCKU...)
2794                  * It is actually a sign of a bug on the client or on the server.
2795                  *
2796                  * If we receive a BAD_SEQID error in the particular case of
2797                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2798                  * have unhashed the old state_owner for us, and that we can
2799                  * therefore safely retry using a new one. We should still warn
2800                  * the user though...
2801                  */
2802                 if (status == -NFS4ERR_BAD_SEQID) {
2803                         pr_warn_ratelimited("NFS: v4 server %s "
2804                                         " returned a bad sequence-id error!\n",
2805                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2806                         exception.retry = 1;
2807                         continue;
2808                 }
2809                 /*
2810                  * BAD_STATEID on OPEN means that the server cancelled our
2811                  * state before it received the OPEN_CONFIRM.
2812                  * Recover by retrying the request as per the discussion
2813                  * on Page 181 of RFC3530.
2814                  */
2815                 if (status == -NFS4ERR_BAD_STATEID) {
2816                         exception.retry = 1;
2817                         continue;
2818                 }
2819                 if (status == -EAGAIN) {
2820                         /* We must have found a delegation */
2821                         exception.retry = 1;
2822                         continue;
2823                 }
2824                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2825                         continue;
2826                 res = ERR_PTR(nfs4_handle_exception(server,
2827                                         status, &exception));
2828         } while (exception.retry);
2829         return res;
2830 }
2831
2832 static int _nfs4_do_setattr(struct inode *inode,
2833                             struct nfs_setattrargs *arg,
2834                             struct nfs_setattrres *res,
2835                             struct rpc_cred *cred,
2836                             struct nfs_open_context *ctx)
2837 {
2838         struct nfs_server *server = NFS_SERVER(inode);
2839         struct rpc_message msg = {
2840                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2841                 .rpc_argp       = arg,
2842                 .rpc_resp       = res,
2843                 .rpc_cred       = cred,
2844         };
2845         struct rpc_cred *delegation_cred = NULL;
2846         unsigned long timestamp = jiffies;
2847         fmode_t fmode;
2848         bool truncate;
2849         int status;
2850
2851         nfs_fattr_init(res->fattr);
2852
2853         /* Servers should only apply open mode checks for file size changes */
2854         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2855         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2856
2857         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2858                 /* Use that stateid */
2859         } else if (truncate && ctx != NULL) {
2860                 struct nfs_lock_context *l_ctx;
2861                 if (!nfs4_valid_open_stateid(ctx->state))
2862                         return -EBADF;
2863                 l_ctx = nfs_get_lock_context(ctx);
2864                 if (IS_ERR(l_ctx))
2865                         return PTR_ERR(l_ctx);
2866                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2867                                                 &arg->stateid, &delegation_cred);
2868                 nfs_put_lock_context(l_ctx);
2869                 if (status == -EIO)
2870                         return -EBADF;
2871         } else
2872                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2873         if (delegation_cred)
2874                 msg.rpc_cred = delegation_cred;
2875
2876         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2877
2878         put_rpccred(delegation_cred);
2879         if (status == 0 && ctx != NULL)
2880                 renew_lease(server, timestamp);
2881         trace_nfs4_setattr(inode, &arg->stateid, status);
2882         return status;
2883 }
2884
2885 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2886                            struct nfs_fattr *fattr, struct iattr *sattr,
2887                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2888                            struct nfs4_label *olabel)
2889 {
2890         struct nfs_server *server = NFS_SERVER(inode);
2891         struct nfs4_state *state = ctx ? ctx->state : NULL;
2892         struct nfs_setattrargs  arg = {
2893                 .fh             = NFS_FH(inode),
2894                 .iap            = sattr,
2895                 .server         = server,
2896                 .bitmask = server->attr_bitmask,
2897                 .label          = ilabel,
2898         };
2899         struct nfs_setattrres  res = {
2900                 .fattr          = fattr,
2901                 .label          = olabel,
2902                 .server         = server,
2903         };
2904         struct nfs4_exception exception = {
2905                 .state = state,
2906                 .inode = inode,
2907                 .stateid = &arg.stateid,
2908         };
2909         int err;
2910
2911         arg.bitmask = nfs4_bitmask(server, ilabel);
2912         if (ilabel)
2913                 arg.bitmask = nfs4_bitmask(server, olabel);
2914
2915         do {
2916                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
2917                 switch (err) {
2918                 case -NFS4ERR_OPENMODE:
2919                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2920                                 pr_warn_once("NFSv4: server %s is incorrectly "
2921                                                 "applying open mode checks to "
2922                                                 "a SETATTR that is not "
2923                                                 "changing file size.\n",
2924                                                 server->nfs_client->cl_hostname);
2925                         }
2926                         if (state && !(state->state & FMODE_WRITE)) {
2927                                 err = -EBADF;
2928                                 if (sattr->ia_valid & ATTR_OPEN)
2929                                         err = -EACCES;
2930                                 goto out;
2931                         }
2932                 }
2933                 err = nfs4_handle_exception(server, err, &exception);
2934         } while (exception.retry);
2935 out:
2936         return err;
2937 }
2938
2939 static bool
2940 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2941 {
2942         if (inode == NULL || !nfs_have_layout(inode))
2943                 return false;
2944
2945         return pnfs_wait_on_layoutreturn(inode, task);
2946 }
2947
2948 struct nfs4_closedata {
2949         struct inode *inode;
2950         struct nfs4_state *state;
2951         struct nfs_closeargs arg;
2952         struct nfs_closeres res;
2953         struct {
2954                 struct nfs4_layoutreturn_args arg;
2955                 struct nfs4_layoutreturn_res res;
2956                 struct nfs4_xdr_opaque_data ld_private;
2957                 u32 roc_barrier;
2958                 bool roc;
2959         } lr;
2960         struct nfs_fattr fattr;
2961         unsigned long timestamp;
2962 };
2963
2964 static void nfs4_free_closedata(void *data)
2965 {
2966         struct nfs4_closedata *calldata = data;
2967         struct nfs4_state_owner *sp = calldata->state->owner;
2968         struct super_block *sb = calldata->state->inode->i_sb;
2969
2970         if (calldata->lr.roc)
2971                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
2972                                 calldata->res.lr_ret);
2973         nfs4_put_open_state(calldata->state);
2974         nfs_free_seqid(calldata->arg.seqid);
2975         nfs4_put_state_owner(sp);
2976         nfs_sb_deactive(sb);
2977         kfree(calldata);
2978 }
2979
2980 static void nfs4_close_done(struct rpc_task *task, void *data)
2981 {
2982         struct nfs4_closedata *calldata = data;
2983         struct nfs4_state *state = calldata->state;
2984         struct nfs_server *server = NFS_SERVER(calldata->inode);
2985         nfs4_stateid *res_stateid = NULL;
2986
2987         dprintk("%s: begin!\n", __func__);
2988         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2989                 return;
2990         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2991
2992         /* Handle Layoutreturn errors */
2993         if (calldata->arg.lr_args && task->tk_status != 0) {
2994                 switch (calldata->res.lr_ret) {
2995                 default:
2996                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
2997                         break;
2998                 case 0:
2999                         calldata->arg.lr_args = NULL;
3000                         calldata->res.lr_res = NULL;
3001                         break;
3002                 case -NFS4ERR_ADMIN_REVOKED:
3003                 case -NFS4ERR_DELEG_REVOKED:
3004                 case -NFS4ERR_EXPIRED:
3005                 case -NFS4ERR_BAD_STATEID:
3006                 case -NFS4ERR_OLD_STATEID:
3007                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3008                 case -NFS4ERR_WRONG_CRED:
3009                         calldata->arg.lr_args = NULL;
3010                         calldata->res.lr_res = NULL;
3011                         calldata->res.lr_ret = 0;
3012                         rpc_restart_call_prepare(task);
3013                         return;
3014                 }
3015         }
3016
3017         /* hmm. we are done with the inode, and in the process of freeing
3018          * the state_owner. we keep this around to process errors
3019          */
3020         switch (task->tk_status) {
3021                 case 0:
3022                         res_stateid = &calldata->res.stateid;
3023                         renew_lease(server, calldata->timestamp);
3024                         break;
3025                 case -NFS4ERR_ACCESS:
3026                         if (calldata->arg.bitmask != NULL) {
3027                                 calldata->arg.bitmask = NULL;
3028                                 calldata->res.fattr = NULL;
3029                                 task->tk_status = 0;
3030                                 rpc_restart_call_prepare(task);
3031                                 goto out_release;
3032
3033                         }
3034                         break;
3035                 case -NFS4ERR_ADMIN_REVOKED:
3036                 case -NFS4ERR_STALE_STATEID:
3037                 case -NFS4ERR_EXPIRED:
3038                         nfs4_free_revoked_stateid(server,
3039                                         &calldata->arg.stateid,
3040                                         task->tk_msg.rpc_cred);
3041                 case -NFS4ERR_OLD_STATEID:
3042                 case -NFS4ERR_BAD_STATEID:
3043                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3044                                                 &state->open_stateid)) {
3045                                 rpc_restart_call_prepare(task);
3046                                 goto out_release;
3047                         }
3048                         if (calldata->arg.fmode == 0)
3049                                 break;
3050                 default:
3051                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3052                                 rpc_restart_call_prepare(task);
3053                                 goto out_release;
3054                         }
3055         }
3056         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3057                         res_stateid, calldata->arg.fmode);
3058 out_release:
3059         nfs_release_seqid(calldata->arg.seqid);
3060         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3061         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3062 }
3063
3064 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3065 {
3066         struct nfs4_closedata *calldata = data;
3067         struct nfs4_state *state = calldata->state;
3068         struct inode *inode = calldata->inode;
3069         bool is_rdonly, is_wronly, is_rdwr;
3070         int call_close = 0;
3071
3072         dprintk("%s: begin!\n", __func__);
3073         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3074                 goto out_wait;
3075
3076         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3077         spin_lock(&state->owner->so_lock);
3078         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3079         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3080         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3081         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3082         /* Calculate the change in open mode */
3083         calldata->arg.fmode = 0;
3084         if (state->n_rdwr == 0) {
3085                 if (state->n_rdonly == 0)
3086                         call_close |= is_rdonly;
3087                 else if (is_rdonly)
3088                         calldata->arg.fmode |= FMODE_READ;
3089                 if (state->n_wronly == 0)
3090                         call_close |= is_wronly;
3091                 else if (is_wronly)
3092                         calldata->arg.fmode |= FMODE_WRITE;
3093                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3094                         call_close |= is_rdwr;
3095         } else if (is_rdwr)
3096                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3097
3098         if (!nfs4_valid_open_stateid(state) ||
3099             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3100                 call_close = 0;
3101         spin_unlock(&state->owner->so_lock);
3102
3103         if (!call_close) {
3104                 /* Note: exit _without_ calling nfs4_close_done */
3105                 goto out_no_action;
3106         }
3107
3108         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3109                 nfs_release_seqid(calldata->arg.seqid);
3110                 goto out_wait;
3111         }
3112
3113         if (calldata->arg.fmode == 0)
3114                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3115
3116         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3117                 /* Close-to-open cache consistency revalidation */
3118                 if (!nfs4_have_delegation(inode, FMODE_READ))
3119                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3120                 else
3121                         calldata->arg.bitmask = NULL;
3122         }
3123
3124         calldata->arg.share_access =
3125                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3126                                 calldata->arg.fmode, 0);
3127
3128         if (calldata->res.fattr == NULL)
3129                 calldata->arg.bitmask = NULL;
3130         else if (calldata->arg.bitmask == NULL)
3131                 calldata->res.fattr = NULL;
3132         calldata->timestamp = jiffies;
3133         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3134                                 &calldata->arg.seq_args,
3135                                 &calldata->res.seq_res,
3136                                 task) != 0)
3137                 nfs_release_seqid(calldata->arg.seqid);
3138         dprintk("%s: done!\n", __func__);
3139         return;
3140 out_no_action:
3141         task->tk_action = NULL;
3142 out_wait:
3143         nfs4_sequence_done(task, &calldata->res.seq_res);
3144 }
3145
3146 static const struct rpc_call_ops nfs4_close_ops = {
3147         .rpc_call_prepare = nfs4_close_prepare,
3148         .rpc_call_done = nfs4_close_done,
3149         .rpc_release = nfs4_free_closedata,
3150 };
3151
3152 /*
3153  * It is possible for data to be read/written from a mem-mapped file
3154  * after the sys_close call (which hits the vfs layer as a flush).
3155  * This means that we can't safely call nfsv4 close on a file until
3156  * the inode is cleared. This in turn means that we are not good
3157  * NFSv4 citizens - we do not indicate to the server to update the file's
3158  * share state even when we are done with one of the three share
3159  * stateid's in the inode.
3160  *
3161  * NOTE: Caller must be holding the sp->so_owner semaphore!
3162  */
3163 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3164 {
3165         struct nfs_server *server = NFS_SERVER(state->inode);
3166         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3167         struct nfs4_closedata *calldata;
3168         struct nfs4_state_owner *sp = state->owner;
3169         struct rpc_task *task;
3170         struct rpc_message msg = {
3171                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3172                 .rpc_cred = state->owner->so_cred,
3173         };
3174         struct rpc_task_setup task_setup_data = {
3175                 .rpc_client = server->client,
3176                 .rpc_message = &msg,
3177                 .callback_ops = &nfs4_close_ops,
3178                 .workqueue = nfsiod_workqueue,
3179                 .flags = RPC_TASK_ASYNC,
3180         };
3181         int status = -ENOMEM;
3182
3183         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3184                 &task_setup_data.rpc_client, &msg);
3185
3186         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3187         if (calldata == NULL)
3188                 goto out;
3189         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3190         calldata->inode = state->inode;
3191         calldata->state = state;
3192         calldata->arg.fh = NFS_FH(state->inode);
3193         /* Serialization for the sequence id */
3194         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3195         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3196         if (IS_ERR(calldata->arg.seqid))
3197                 goto out_free_calldata;
3198         nfs_fattr_init(&calldata->fattr);
3199         calldata->arg.fmode = 0;
3200         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3201         calldata->res.fattr = &calldata->fattr;
3202         calldata->res.seqid = calldata->arg.seqid;
3203         calldata->res.server = server;
3204         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3205         calldata->lr.roc = pnfs_roc(state->inode,
3206                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3207         if (calldata->lr.roc) {
3208                 calldata->arg.lr_args = &calldata->lr.arg;
3209                 calldata->res.lr_res = &calldata->lr.res;
3210         }
3211         nfs_sb_active(calldata->inode->i_sb);
3212
3213         msg.rpc_argp = &calldata->arg;
3214         msg.rpc_resp = &calldata->res;
3215         task_setup_data.callback_data = calldata;
3216         task = rpc_run_task(&task_setup_data);
3217         if (IS_ERR(task))
3218                 return PTR_ERR(task);
3219         status = 0;
3220         if (wait)
3221                 status = rpc_wait_for_completion_task(task);
3222         rpc_put_task(task);
3223         return status;
3224 out_free_calldata:
3225         kfree(calldata);
3226 out:
3227         nfs4_put_open_state(state);
3228         nfs4_put_state_owner(sp);
3229         return status;
3230 }
3231
3232 static struct inode *
3233 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3234                 int open_flags, struct iattr *attr, int *opened)
3235 {
3236         struct nfs4_state *state;
3237         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3238
3239         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3240
3241         /* Protect against concurrent sillydeletes */
3242         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3243
3244         nfs4_label_release_security(label);
3245
3246         if (IS_ERR(state))
3247                 return ERR_CAST(state);
3248         return state->inode;
3249 }
3250
3251 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3252 {
3253         if (ctx->state == NULL)
3254                 return;
3255         if (is_sync)
3256                 nfs4_close_sync(ctx->state, ctx->mode);
3257         else
3258                 nfs4_close_state(ctx->state, ctx->mode);
3259 }
3260
3261 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3262 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3263 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3264
3265 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3266 {
3267         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3268         struct nfs4_server_caps_arg args = {
3269                 .fhandle = fhandle,
3270                 .bitmask = bitmask,
3271         };
3272         struct nfs4_server_caps_res res = {};
3273         struct rpc_message msg = {
3274                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3275                 .rpc_argp = &args,
3276                 .rpc_resp = &res,
3277         };
3278         int status;
3279         int i;
3280
3281         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3282                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3283                      FATTR4_WORD0_LINK_SUPPORT |
3284                      FATTR4_WORD0_SYMLINK_SUPPORT |
3285                      FATTR4_WORD0_ACLSUPPORT;
3286         if (minorversion)
3287                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3288
3289         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3290         if (status == 0) {
3291                 /* Sanity check the server answers */
3292                 switch (minorversion) {
3293                 case 0:
3294                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3295                         res.attr_bitmask[2] = 0;
3296                         break;
3297                 case 1:
3298                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3299                         break;
3300                 case 2:
3301                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3302                 }
3303                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3304                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3305                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3306                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3307                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3308                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3309                                 NFS_CAP_SECURITY_LABEL);
3310                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3311                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3312                         server->caps |= NFS_CAP_ACLS;
3313                 if (res.has_links != 0)
3314                         server->caps |= NFS_CAP_HARDLINKS;
3315                 if (res.has_symlinks != 0)
3316                         server->caps |= NFS_CAP_SYMLINKS;
3317                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3318                         server->caps |= NFS_CAP_FILEID;
3319                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3320                         server->caps |= NFS_CAP_MODE;
3321                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3322                         server->caps |= NFS_CAP_NLINK;
3323                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3324                         server->caps |= NFS_CAP_OWNER;
3325                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3326                         server->caps |= NFS_CAP_OWNER_GROUP;
3327                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3328                         server->caps |= NFS_CAP_ATIME;
3329                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3330                         server->caps |= NFS_CAP_CTIME;
3331                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3332                         server->caps |= NFS_CAP_MTIME;
3333 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3334                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3335                         server->caps |= NFS_CAP_SECURITY_LABEL;
3336 #endif
3337                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3338                                 sizeof(server->attr_bitmask));
3339                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3340
3341                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3342                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3343                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3344                 server->cache_consistency_bitmask[2] = 0;
3345
3346                 /* Avoid a regression due to buggy server */
3347                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3348                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3349                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3350                         sizeof(server->exclcreat_bitmask));
3351
3352                 server->acl_bitmask = res.acl_bitmask;
3353                 server->fh_expire_type = res.fh_expire_type;
3354         }
3355
3356         return status;
3357 }
3358
3359 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3360 {
3361         struct nfs4_exception exception = { };
3362         int err;
3363         do {
3364                 err = nfs4_handle_exception(server,
3365                                 _nfs4_server_capabilities(server, fhandle),
3366                                 &exception);
3367         } while (exception.retry);
3368         return err;
3369 }
3370
3371 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3372                 struct nfs_fsinfo *info)
3373 {
3374         u32 bitmask[3];
3375         struct nfs4_lookup_root_arg args = {
3376                 .bitmask = bitmask,
3377         };
3378         struct nfs4_lookup_res res = {
3379                 .server = server,
3380                 .fattr = info->fattr,
3381                 .fh = fhandle,
3382         };
3383         struct rpc_message msg = {
3384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3385                 .rpc_argp = &args,
3386                 .rpc_resp = &res,
3387         };
3388
3389         bitmask[0] = nfs4_fattr_bitmap[0];
3390         bitmask[1] = nfs4_fattr_bitmap[1];
3391         /*
3392          * Process the label in the upcoming getfattr
3393          */
3394         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3395
3396         nfs_fattr_init(info->fattr);
3397         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3398 }
3399
3400 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3401                 struct nfs_fsinfo *info)
3402 {
3403         struct nfs4_exception exception = { };
3404         int err;
3405         do {
3406                 err = _nfs4_lookup_root(server, fhandle, info);
3407                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3408                 switch (err) {
3409                 case 0:
3410                 case -NFS4ERR_WRONGSEC:
3411                         goto out;
3412                 default:
3413                         err = nfs4_handle_exception(server, err, &exception);
3414                 }
3415         } while (exception.retry);
3416 out:
3417         return err;
3418 }
3419
3420 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3421                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3422 {
3423         struct rpc_auth_create_args auth_args = {
3424                 .pseudoflavor = flavor,
3425         };
3426         struct rpc_auth *auth;
3427
3428         auth = rpcauth_create(&auth_args, server->client);
3429         if (IS_ERR(auth))
3430                 return -EACCES;
3431         return nfs4_lookup_root(server, fhandle, info);
3432 }
3433
3434 /*
3435  * Retry pseudoroot lookup with various security flavors.  We do this when:
3436  *
3437  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3438  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3439  *
3440  * Returns zero on success, or a negative NFS4ERR value, or a
3441  * negative errno value.
3442  */
3443 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3444                               struct nfs_fsinfo *info)
3445 {
3446         /* Per 3530bis 15.33.5 */
3447         static const rpc_authflavor_t flav_array[] = {
3448                 RPC_AUTH_GSS_KRB5P,
3449                 RPC_AUTH_GSS_KRB5I,
3450                 RPC_AUTH_GSS_KRB5,
3451                 RPC_AUTH_UNIX,                  /* courtesy */
3452                 RPC_AUTH_NULL,
3453         };
3454         int status = -EPERM;
3455         size_t i;
3456
3457         if (server->auth_info.flavor_len > 0) {
3458                 /* try each flavor specified by user */
3459                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3460                         status = nfs4_lookup_root_sec(server, fhandle, info,
3461                                                 server->auth_info.flavors[i]);
3462                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3463                                 continue;
3464                         break;
3465                 }
3466         } else {
3467                 /* no flavors specified by user, try default list */
3468                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3469                         status = nfs4_lookup_root_sec(server, fhandle, info,
3470                                                       flav_array[i]);
3471                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3472                                 continue;
3473                         break;
3474                 }
3475         }
3476
3477         /*
3478          * -EACCESS could mean that the user doesn't have correct permissions
3479          * to access the mount.  It could also mean that we tried to mount
3480          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3481          * existing mount programs don't handle -EACCES very well so it should
3482          * be mapped to -EPERM instead.
3483          */
3484         if (status == -EACCES)
3485                 status = -EPERM;
3486         return status;
3487 }
3488
3489 /**
3490  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3491  * @server: initialized nfs_server handle
3492  * @fhandle: we fill in the pseudo-fs root file handle
3493  * @info: we fill in an FSINFO struct
3494  * @auth_probe: probe the auth flavours
3495  *
3496  * Returns zero on success, or a negative errno.
3497  */
3498 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3499                          struct nfs_fsinfo *info,
3500                          bool auth_probe)
3501 {
3502         int status = 0;
3503
3504         if (!auth_probe)
3505                 status = nfs4_lookup_root(server, fhandle, info);
3506
3507         if (auth_probe || status == NFS4ERR_WRONGSEC)
3508                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3509                                 fhandle, info);
3510
3511         if (status == 0)
3512                 status = nfs4_server_capabilities(server, fhandle);
3513         if (status == 0)
3514                 status = nfs4_do_fsinfo(server, fhandle, info);
3515
3516         return nfs4_map_errors(status);
3517 }
3518
3519 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3520                               struct nfs_fsinfo *info)
3521 {
3522         int error;
3523         struct nfs_fattr *fattr = info->fattr;
3524         struct nfs4_label *label = NULL;
3525
3526         error = nfs4_server_capabilities(server, mntfh);
3527         if (error < 0) {
3528                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3529                 return error;
3530         }
3531
3532         label = nfs4_label_alloc(server, GFP_KERNEL);
3533         if (IS_ERR(label))
3534                 return PTR_ERR(label);
3535
3536         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3537         if (error < 0) {
3538                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3539                 goto err_free_label;
3540         }
3541
3542         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3543             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3544                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3545
3546 err_free_label:
3547         nfs4_label_free(label);
3548
3549         return error;
3550 }
3551
3552 /*
3553  * Get locations and (maybe) other attributes of a referral.
3554  * Note that we'll actually follow the referral later when
3555  * we detect fsid mismatch in inode revalidation
3556  */
3557 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3558                              const struct qstr *name, struct nfs_fattr *fattr,
3559                              struct nfs_fh *fhandle)
3560 {
3561         int status = -ENOMEM;
3562         struct page *page = NULL;
3563         struct nfs4_fs_locations *locations = NULL;
3564
3565         page = alloc_page(GFP_KERNEL);
3566         if (page == NULL)
3567                 goto out;
3568         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3569         if (locations == NULL)
3570                 goto out;
3571
3572         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3573         if (status != 0)
3574                 goto out;
3575
3576         /*
3577          * If the fsid didn't change, this is a migration event, not a
3578          * referral.  Cause us to drop into the exception handler, which
3579          * will kick off migration recovery.
3580          */
3581         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3582                 dprintk("%s: server did not return a different fsid for"
3583                         " a referral at %s\n", __func__, name->name);
3584                 status = -NFS4ERR_MOVED;
3585                 goto out;
3586         }
3587         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3588         nfs_fixup_referral_attributes(&locations->fattr);
3589
3590         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3591         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3592         memset(fhandle, 0, sizeof(struct nfs_fh));
3593 out:
3594         if (page)
3595                 __free_page(page);
3596         kfree(locations);
3597         return status;
3598 }
3599
3600 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3601                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3602 {
3603         struct nfs4_getattr_arg args = {
3604                 .fh = fhandle,
3605                 .bitmask = server->attr_bitmask,
3606         };
3607         struct nfs4_getattr_res res = {
3608                 .fattr = fattr,
3609                 .label = label,
3610                 .server = server,
3611         };
3612         struct rpc_message msg = {
3613                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3614                 .rpc_argp = &args,
3615                 .rpc_resp = &res,
3616         };
3617
3618         args.bitmask = nfs4_bitmask(server, label);
3619
3620         nfs_fattr_init(fattr);
3621         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3622 }
3623
3624 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3625                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3626 {
3627         struct nfs4_exception exception = { };
3628         int err;
3629         do {
3630                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3631                 trace_nfs4_getattr(server, fhandle, fattr, err);
3632                 err = nfs4_handle_exception(server, err,
3633                                 &exception);
3634         } while (exception.retry);
3635         return err;
3636 }
3637
3638 /*
3639  * The file is not closed if it is opened due to the a request to change
3640  * the size of the file. The open call will not be needed once the
3641  * VFS layer lookup-intents are implemented.
3642  *
3643  * Close is called when the inode is destroyed.
3644  * If we haven't opened the file for O_WRONLY, we
3645  * need to in the size_change case to obtain a stateid.
3646  *
3647  * Got race?
3648  * Because OPEN is always done by name in nfsv4, it is
3649  * possible that we opened a different file by the same
3650  * name.  We can recognize this race condition, but we
3651  * can't do anything about it besides returning an error.
3652  *
3653  * This will be fixed with VFS changes (lookup-intent).
3654  */
3655 static int
3656 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3657                   struct iattr *sattr)
3658 {
3659         struct inode *inode = d_inode(dentry);
3660         struct rpc_cred *cred = NULL;
3661         struct nfs_open_context *ctx = NULL;
3662         struct nfs4_label *label = NULL;
3663         int status;
3664
3665         if (pnfs_ld_layoutret_on_setattr(inode) &&
3666             sattr->ia_valid & ATTR_SIZE &&
3667             sattr->ia_size < i_size_read(inode))
3668                 pnfs_commit_and_return_layout(inode);
3669
3670         nfs_fattr_init(fattr);
3671
3672         /* Deal with open(O_TRUNC) */
3673         if (sattr->ia_valid & ATTR_OPEN)
3674                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3675
3676         /* Optimization: if the end result is no change, don't RPC */
3677         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3678                 return 0;
3679
3680         /* Search for an existing open(O_WRITE) file */
3681         if (sattr->ia_valid & ATTR_FILE) {
3682
3683                 ctx = nfs_file_open_context(sattr->ia_file);
3684                 if (ctx)
3685                         cred = ctx->cred;
3686         }
3687
3688         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3689         if (IS_ERR(label))
3690                 return PTR_ERR(label);
3691
3692         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3693         if (status == 0) {
3694                 nfs_setattr_update_inode(inode, sattr, fattr);
3695                 nfs_setsecurity(inode, fattr, label);
3696         }
3697         nfs4_label_free(label);
3698         return status;
3699 }
3700
3701 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3702                 const struct qstr *name, struct nfs_fh *fhandle,
3703                 struct nfs_fattr *fattr, struct nfs4_label *label)
3704 {
3705         struct nfs_server *server = NFS_SERVER(dir);
3706         int                    status;
3707         struct nfs4_lookup_arg args = {
3708                 .bitmask = server->attr_bitmask,
3709                 .dir_fh = NFS_FH(dir),
3710                 .name = name,
3711         };
3712         struct nfs4_lookup_res res = {
3713                 .server = server,
3714                 .fattr = fattr,
3715                 .label = label,
3716                 .fh = fhandle,
3717         };
3718         struct rpc_message msg = {
3719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3720                 .rpc_argp = &args,
3721                 .rpc_resp = &res,
3722         };
3723
3724         args.bitmask = nfs4_bitmask(server, label);
3725
3726         nfs_fattr_init(fattr);
3727
3728         dprintk("NFS call  lookup %s\n", name->name);
3729         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3730         dprintk("NFS reply lookup: %d\n", status);
3731         return status;
3732 }
3733
3734 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3735 {
3736         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3737                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3738         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3739         fattr->nlink = 2;
3740 }
3741
3742 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3743                                    const struct qstr *name, struct nfs_fh *fhandle,
3744                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3745 {
3746         struct nfs4_exception exception = { };
3747         struct rpc_clnt *client = *clnt;
3748         int err;
3749         do {
3750                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3751                 trace_nfs4_lookup(dir, name, err);
3752                 switch (err) {
3753                 case -NFS4ERR_BADNAME:
3754                         err = -ENOENT;
3755                         goto out;
3756                 case -NFS4ERR_MOVED:
3757                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3758                         if (err == -NFS4ERR_MOVED)
3759                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3760                         goto out;
3761                 case -NFS4ERR_WRONGSEC:
3762                         err = -EPERM;
3763                         if (client != *clnt)
3764                                 goto out;
3765                         client = nfs4_negotiate_security(client, dir, name);
3766                         if (IS_ERR(client))
3767                                 return PTR_ERR(client);
3768
3769                         exception.retry = 1;
3770                         break;
3771                 default:
3772                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3773                 }
3774         } while (exception.retry);
3775
3776 out:
3777         if (err == 0)
3778                 *clnt = client;
3779         else if (client != *clnt)
3780                 rpc_shutdown_client(client);
3781
3782         return err;
3783 }
3784
3785 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3786                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3787                             struct nfs4_label *label)
3788 {
3789         int status;
3790         struct rpc_clnt *client = NFS_CLIENT(dir);
3791
3792         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3793         if (client != NFS_CLIENT(dir)) {
3794                 rpc_shutdown_client(client);
3795                 nfs_fixup_secinfo_attributes(fattr);
3796         }
3797         return status;
3798 }
3799
3800 struct rpc_clnt *
3801 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3802                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3803 {
3804         struct rpc_clnt *client = NFS_CLIENT(dir);
3805         int status;
3806
3807         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3808         if (status < 0)
3809                 return ERR_PTR(status);
3810         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3811 }
3812
3813 static int _nfs4_proc_lookupp(struct inode *inode,
3814                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3815                 struct nfs4_label *label)
3816 {
3817         struct rpc_clnt *clnt = NFS_CLIENT(inode);
3818         struct nfs_server *server = NFS_SERVER(inode);
3819         int                    status;
3820         struct nfs4_lookupp_arg args = {
3821                 .bitmask = server->attr_bitmask,
3822                 .fh = NFS_FH(inode),
3823         };
3824         struct nfs4_lookupp_res res = {
3825                 .server = server,
3826                 .fattr = fattr,
3827                 .label = label,
3828                 .fh = fhandle,
3829         };
3830         struct rpc_message msg = {
3831                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
3832                 .rpc_argp = &args,
3833                 .rpc_resp = &res,
3834         };
3835
3836         args.bitmask = nfs4_bitmask(server, label);
3837
3838         nfs_fattr_init(fattr);
3839
3840         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
3841         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
3842                                 &res.seq_res, 0);
3843         dprintk("NFS reply lookupp: %d\n", status);
3844         return status;
3845 }
3846
3847 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
3848                              struct nfs_fattr *fattr, struct nfs4_label *label)
3849 {
3850         struct nfs4_exception exception = { };
3851         int err;
3852         do {
3853                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
3854                 trace_nfs4_lookupp(inode, err);
3855                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3856                                 &exception);
3857         } while (exception.retry);
3858         return err;
3859 }
3860
3861 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3862 {
3863         struct nfs_server *server = NFS_SERVER(inode);
3864         struct nfs4_accessargs args = {
3865                 .fh = NFS_FH(inode),
3866                 .bitmask = server->cache_consistency_bitmask,
3867         };
3868         struct nfs4_accessres res = {
3869                 .server = server,
3870         };
3871         struct rpc_message msg = {
3872                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3873                 .rpc_argp = &args,
3874                 .rpc_resp = &res,
3875                 .rpc_cred = entry->cred,
3876         };
3877         int mode = entry->mask;
3878         int status = 0;
3879
3880         /*
3881          * Determine which access bits we want to ask for...
3882          */
3883         if (mode & MAY_READ)
3884                 args.access |= NFS4_ACCESS_READ;
3885         if (S_ISDIR(inode->i_mode)) {
3886                 if (mode & MAY_WRITE)
3887                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3888                 if (mode & MAY_EXEC)
3889                         args.access |= NFS4_ACCESS_LOOKUP;
3890         } else {
3891                 if (mode & MAY_WRITE)
3892                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3893                 if (mode & MAY_EXEC)
3894                         args.access |= NFS4_ACCESS_EXECUTE;
3895         }
3896
3897         res.fattr = nfs_alloc_fattr();
3898         if (res.fattr == NULL)
3899                 return -ENOMEM;
3900
3901         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3902         if (!status) {
3903                 nfs_access_set_mask(entry, res.access);
3904                 nfs_refresh_inode(inode, res.fattr);
3905         }
3906         nfs_free_fattr(res.fattr);
3907         return status;
3908 }
3909
3910 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3911 {
3912         struct nfs4_exception exception = { };
3913         int err;
3914         do {
3915                 err = _nfs4_proc_access(inode, entry);
3916                 trace_nfs4_access(inode, err);
3917                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3918                                 &exception);
3919         } while (exception.retry);
3920         return err;
3921 }
3922
3923 /*
3924  * TODO: For the time being, we don't try to get any attributes
3925  * along with any of the zero-copy operations READ, READDIR,
3926  * READLINK, WRITE.
3927  *
3928  * In the case of the first three, we want to put the GETATTR
3929  * after the read-type operation -- this is because it is hard
3930  * to predict the length of a GETATTR response in v4, and thus
3931  * align the READ data correctly.  This means that the GETATTR
3932  * may end up partially falling into the page cache, and we should
3933  * shift it into the 'tail' of the xdr_buf before processing.
3934  * To do this efficiently, we need to know the total length
3935  * of data received, which doesn't seem to be available outside
3936  * of the RPC layer.
3937  *
3938  * In the case of WRITE, we also want to put the GETATTR after
3939  * the operation -- in this case because we want to make sure
3940  * we get the post-operation mtime and size.
3941  *
3942  * Both of these changes to the XDR layer would in fact be quite
3943  * minor, but I decided to leave them for a subsequent patch.
3944  */
3945 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3946                 unsigned int pgbase, unsigned int pglen)
3947 {
3948         struct nfs4_readlink args = {
3949                 .fh       = NFS_FH(inode),
3950                 .pgbase   = pgbase,
3951                 .pglen    = pglen,
3952                 .pages    = &page,
3953         };
3954         struct nfs4_readlink_res res;
3955         struct rpc_message msg = {
3956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3957                 .rpc_argp = &args,
3958                 .rpc_resp = &res,
3959         };
3960
3961         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3962 }
3963
3964 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3965                 unsigned int pgbase, unsigned int pglen)
3966 {
3967         struct nfs4_exception exception = { };
3968         int err;
3969         do {
3970                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3971                 trace_nfs4_readlink(inode, err);
3972                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3973                                 &exception);
3974         } while (exception.retry);
3975         return err;
3976 }
3977
3978 /*
3979  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3980  */
3981 static int
3982 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3983                  int flags)
3984 {
3985         struct nfs_server *server = NFS_SERVER(dir);
3986         struct nfs4_label l, *ilabel = NULL;
3987         struct nfs_open_context *ctx;
3988         struct nfs4_state *state;
3989         int status = 0;
3990
3991         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
3992         if (IS_ERR(ctx))
3993                 return PTR_ERR(ctx);
3994
3995         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3996
3997         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
3998                 sattr->ia_mode &= ~current_umask();
3999         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4000         if (IS_ERR(state)) {
4001                 status = PTR_ERR(state);
4002                 goto out;
4003         }
4004 out:
4005         nfs4_label_release_security(ilabel);
4006         put_nfs_open_context(ctx);
4007         return status;
4008 }
4009
4010 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4011 {
4012         struct nfs_server *server = NFS_SERVER(dir);
4013         struct nfs_removeargs args = {
4014                 .fh = NFS_FH(dir),
4015                 .name = *name,
4016         };
4017         struct nfs_removeres res = {
4018                 .server = server,
4019         };
4020         struct rpc_message msg = {
4021                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4022                 .rpc_argp = &args,
4023                 .rpc_resp = &res,
4024         };
4025         unsigned long timestamp = jiffies;
4026         int status;
4027
4028         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4029         if (status == 0)
4030                 update_changeattr(dir, &res.cinfo, timestamp);
4031         return status;
4032 }
4033
4034 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4035 {
4036         struct nfs4_exception exception = { };
4037         int err;
4038         do {
4039                 err = _nfs4_proc_remove(dir, name);
4040                 trace_nfs4_remove(dir, name, err);
4041                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4042                                 &exception);
4043         } while (exception.retry);
4044         return err;
4045 }
4046
4047 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4048 {
4049         struct nfs_server *server = NFS_SERVER(dir);
4050         struct nfs_removeargs *args = msg->rpc_argp;
4051         struct nfs_removeres *res = msg->rpc_resp;
4052
4053         res->server = server;
4054         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4055         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4056
4057         nfs_fattr_init(res->dir_attr);
4058 }
4059
4060 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4061 {
4062         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4063                         &data->args.seq_args,
4064                         &data->res.seq_res,
4065                         task);
4066 }
4067
4068 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4069 {
4070         struct nfs_unlinkdata *data = task->tk_calldata;
4071         struct nfs_removeres *res = &data->res;
4072
4073         if (!nfs4_sequence_done(task, &res->seq_res))
4074                 return 0;
4075         if (nfs4_async_handle_error(task, res->server, NULL,
4076                                     &data->timeout) == -EAGAIN)
4077                 return 0;
4078         if (task->tk_status == 0)
4079                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4080         return 1;
4081 }
4082
4083 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4084 {
4085         struct nfs_server *server = NFS_SERVER(dir);
4086         struct nfs_renameargs *arg = msg->rpc_argp;
4087         struct nfs_renameres *res = msg->rpc_resp;
4088
4089         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4090         res->server = server;
4091         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4092 }
4093
4094 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4095 {
4096         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4097                         &data->args.seq_args,
4098                         &data->res.seq_res,
4099                         task);
4100 }
4101
4102 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4103                                  struct inode *new_dir)
4104 {
4105         struct nfs_renamedata *data = task->tk_calldata;
4106         struct nfs_renameres *res = &data->res;
4107
4108         if (!nfs4_sequence_done(task, &res->seq_res))
4109                 return 0;
4110         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4111                 return 0;
4112
4113         if (task->tk_status == 0) {
4114                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4115                 if (new_dir != old_dir)
4116                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4117         }
4118         return 1;
4119 }
4120
4121 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4122 {
4123         struct nfs_server *server = NFS_SERVER(inode);
4124         struct nfs4_link_arg arg = {
4125                 .fh     = NFS_FH(inode),
4126                 .dir_fh = NFS_FH(dir),
4127                 .name   = name,
4128                 .bitmask = server->attr_bitmask,
4129         };
4130         struct nfs4_link_res res = {
4131                 .server = server,
4132                 .label = NULL,
4133         };
4134         struct rpc_message msg = {
4135                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4136                 .rpc_argp = &arg,
4137                 .rpc_resp = &res,
4138         };
4139         int status = -ENOMEM;
4140
4141         res.fattr = nfs_alloc_fattr();
4142         if (res.fattr == NULL)
4143                 goto out;
4144
4145         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4146         if (IS_ERR(res.label)) {
4147                 status = PTR_ERR(res.label);
4148                 goto out;
4149         }
4150         arg.bitmask = nfs4_bitmask(server, res.label);
4151
4152         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4153         if (!status) {
4154                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4155                 status = nfs_post_op_update_inode(inode, res.fattr);
4156                 if (!status)
4157                         nfs_setsecurity(inode, res.fattr, res.label);
4158         }
4159
4160
4161         nfs4_label_free(res.label);
4162
4163 out:
4164         nfs_free_fattr(res.fattr);
4165         return status;
4166 }
4167
4168 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4169 {
4170         struct nfs4_exception exception = { };
4171         int err;
4172         do {
4173                 err = nfs4_handle_exception(NFS_SERVER(inode),
4174                                 _nfs4_proc_link(inode, dir, name),
4175                                 &exception);
4176         } while (exception.retry);
4177         return err;
4178 }
4179
4180 struct nfs4_createdata {
4181         struct rpc_message msg;
4182         struct nfs4_create_arg arg;
4183         struct nfs4_create_res res;
4184         struct nfs_fh fh;
4185         struct nfs_fattr fattr;
4186         struct nfs4_label *label;
4187 };
4188
4189 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4190                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4191 {
4192         struct nfs4_createdata *data;
4193
4194         data = kzalloc(sizeof(*data), GFP_KERNEL);
4195         if (data != NULL) {
4196                 struct nfs_server *server = NFS_SERVER(dir);
4197
4198                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4199                 if (IS_ERR(data->label))
4200                         goto out_free;
4201
4202                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4203                 data->msg.rpc_argp = &data->arg;
4204                 data->msg.rpc_resp = &data->res;
4205                 data->arg.dir_fh = NFS_FH(dir);
4206                 data->arg.server = server;
4207                 data->arg.name = name;
4208                 data->arg.attrs = sattr;
4209                 data->arg.ftype = ftype;
4210                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4211                 data->arg.umask = current_umask();
4212                 data->res.server = server;
4213                 data->res.fh = &data->fh;
4214                 data->res.fattr = &data->fattr;
4215                 data->res.label = data->label;
4216                 nfs_fattr_init(data->res.fattr);
4217         }
4218         return data;
4219 out_free:
4220         kfree(data);
4221         return NULL;
4222 }
4223
4224 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4225 {
4226         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4227                                     &data->arg.seq_args, &data->res.seq_res, 1);
4228         if (status == 0) {
4229                 update_changeattr(dir, &data->res.dir_cinfo,
4230                                 data->res.fattr->time_start);
4231                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4232         }
4233         return status;
4234 }
4235
4236 static void nfs4_free_createdata(struct nfs4_createdata *data)
4237 {
4238         nfs4_label_free(data->label);
4239         kfree(data);
4240 }
4241
4242 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4243                 struct page *page, unsigned int len, struct iattr *sattr,
4244                 struct nfs4_label *label)
4245 {
4246         struct nfs4_createdata *data;
4247         int status = -ENAMETOOLONG;
4248
4249         if (len > NFS4_MAXPATHLEN)
4250                 goto out;
4251
4252         status = -ENOMEM;
4253         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4254         if (data == NULL)
4255                 goto out;
4256
4257         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4258         data->arg.u.symlink.pages = &page;
4259         data->arg.u.symlink.len = len;
4260         data->arg.label = label;
4261
4262         status = nfs4_do_create(dir, dentry, data);
4263
4264         nfs4_free_createdata(data);
4265 out:
4266         return status;
4267 }
4268
4269 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4270                 struct page *page, unsigned int len, struct iattr *sattr)
4271 {
4272         struct nfs4_exception exception = { };
4273         struct nfs4_label l, *label = NULL;
4274         int err;
4275
4276         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4277
4278         do {
4279                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4280                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4281                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4282                                 &exception);
4283         } while (exception.retry);
4284
4285         nfs4_label_release_security(label);
4286         return err;
4287 }
4288
4289 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4290                 struct iattr *sattr, struct nfs4_label *label)
4291 {
4292         struct nfs4_createdata *data;
4293         int status = -ENOMEM;
4294
4295         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4296         if (data == NULL)
4297                 goto out;
4298
4299         data->arg.label = label;
4300         status = nfs4_do_create(dir, dentry, data);
4301
4302         nfs4_free_createdata(data);
4303 out:
4304         return status;
4305 }
4306
4307 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4308                 struct iattr *sattr)
4309 {
4310         struct nfs_server *server = NFS_SERVER(dir);
4311         struct nfs4_exception exception = { };
4312         struct nfs4_label l, *label = NULL;
4313         int err;
4314
4315         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4316
4317         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4318                 sattr->ia_mode &= ~current_umask();
4319         do {
4320                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4321                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4322                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4323                                 &exception);
4324         } while (exception.retry);
4325         nfs4_label_release_security(label);
4326
4327         return err;
4328 }
4329
4330 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4331                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4332 {
4333         struct inode            *dir = d_inode(dentry);
4334         struct nfs4_readdir_arg args = {
4335                 .fh = NFS_FH(dir),
4336                 .pages = pages,
4337                 .pgbase = 0,
4338                 .count = count,
4339                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4340                 .plus = plus,
4341         };
4342         struct nfs4_readdir_res res;
4343         struct rpc_message msg = {
4344                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4345                 .rpc_argp = &args,
4346                 .rpc_resp = &res,
4347                 .rpc_cred = cred,
4348         };
4349         int                     status;
4350
4351         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4352                         dentry,
4353                         (unsigned long long)cookie);
4354         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4355         res.pgbase = args.pgbase;
4356         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4357         if (status >= 0) {
4358                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4359                 status += args.pgbase;
4360         }
4361
4362         nfs_invalidate_atime(dir);
4363
4364         dprintk("%s: returns %d\n", __func__, status);
4365         return status;
4366 }
4367
4368 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4369                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4370 {
4371         struct nfs4_exception exception = { };
4372         int err;
4373         do {
4374                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4375                                 pages, count, plus);
4376                 trace_nfs4_readdir(d_inode(dentry), err);
4377                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4378                                 &exception);
4379         } while (exception.retry);
4380         return err;
4381 }
4382
4383 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4384                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4385 {
4386         struct nfs4_createdata *data;
4387         int mode = sattr->ia_mode;
4388         int status = -ENOMEM;
4389
4390         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4391         if (data == NULL)
4392                 goto out;
4393
4394         if (S_ISFIFO(mode))
4395                 data->arg.ftype = NF4FIFO;
4396         else if (S_ISBLK(mode)) {
4397                 data->arg.ftype = NF4BLK;
4398                 data->arg.u.device.specdata1 = MAJOR(rdev);
4399                 data->arg.u.device.specdata2 = MINOR(rdev);
4400         }
4401         else if (S_ISCHR(mode)) {
4402                 data->arg.ftype = NF4CHR;
4403                 data->arg.u.device.specdata1 = MAJOR(rdev);
4404                 data->arg.u.device.specdata2 = MINOR(rdev);
4405         } else if (!S_ISSOCK(mode)) {
4406                 status = -EINVAL;
4407                 goto out_free;
4408         }
4409
4410         data->arg.label = label;
4411         status = nfs4_do_create(dir, dentry, data);
4412 out_free:
4413         nfs4_free_createdata(data);
4414 out:
4415         return status;
4416 }
4417
4418 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4419                 struct iattr *sattr, dev_t rdev)
4420 {
4421         struct nfs_server *server = NFS_SERVER(dir);
4422         struct nfs4_exception exception = { };
4423         struct nfs4_label l, *label = NULL;
4424         int err;
4425
4426         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4427
4428         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4429                 sattr->ia_mode &= ~current_umask();
4430         do {
4431                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4432                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4433                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4434                                 &exception);
4435         } while (exception.retry);
4436
4437         nfs4_label_release_security(label);
4438
4439         return err;
4440 }
4441
4442 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4443                  struct nfs_fsstat *fsstat)
4444 {
4445         struct nfs4_statfs_arg args = {
4446                 .fh = fhandle,
4447                 .bitmask = server->attr_bitmask,
4448         };
4449         struct nfs4_statfs_res res = {
4450                 .fsstat = fsstat,
4451         };
4452         struct rpc_message msg = {
4453                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4454                 .rpc_argp = &args,
4455                 .rpc_resp = &res,
4456         };
4457
4458         nfs_fattr_init(fsstat->fattr);
4459         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4460 }
4461
4462 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4463 {
4464         struct nfs4_exception exception = { };
4465         int err;
4466         do {
4467                 err = nfs4_handle_exception(server,
4468                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4469                                 &exception);
4470         } while (exception.retry);
4471         return err;
4472 }
4473
4474 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4475                 struct nfs_fsinfo *fsinfo)
4476 {
4477         struct nfs4_fsinfo_arg args = {
4478                 .fh = fhandle,
4479                 .bitmask = server->attr_bitmask,
4480         };
4481         struct nfs4_fsinfo_res res = {
4482                 .fsinfo = fsinfo,
4483         };
4484         struct rpc_message msg = {
4485                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4486                 .rpc_argp = &args,
4487                 .rpc_resp = &res,
4488         };
4489
4490         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4491 }
4492
4493 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4494 {
4495         struct nfs4_exception exception = { };
4496         unsigned long now = jiffies;
4497         int err;
4498
4499         do {
4500                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4501                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4502                 if (err == 0) {
4503                         nfs4_set_lease_period(server->nfs_client,
4504                                         fsinfo->lease_time * HZ,
4505                                         now);
4506                         break;
4507                 }
4508                 err = nfs4_handle_exception(server, err, &exception);
4509         } while (exception.retry);
4510         return err;
4511 }
4512
4513 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4514 {
4515         int error;
4516
4517         nfs_fattr_init(fsinfo->fattr);
4518         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4519         if (error == 0) {
4520                 /* block layout checks this! */
4521                 server->pnfs_blksize = fsinfo->blksize;
4522                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4523         }
4524
4525         return error;
4526 }
4527
4528 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4529                 struct nfs_pathconf *pathconf)
4530 {
4531         struct nfs4_pathconf_arg args = {
4532                 .fh = fhandle,
4533                 .bitmask = server->attr_bitmask,
4534         };
4535         struct nfs4_pathconf_res res = {
4536                 .pathconf = pathconf,
4537         };
4538         struct rpc_message msg = {
4539                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4540                 .rpc_argp = &args,
4541                 .rpc_resp = &res,
4542         };
4543
4544         /* None of the pathconf attributes are mandatory to implement */
4545         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4546                 memset(pathconf, 0, sizeof(*pathconf));
4547                 return 0;
4548         }
4549
4550         nfs_fattr_init(pathconf->fattr);
4551         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4552 }
4553
4554 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4555                 struct nfs_pathconf *pathconf)
4556 {
4557         struct nfs4_exception exception = { };
4558         int err;
4559
4560         do {
4561                 err = nfs4_handle_exception(server,
4562                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4563                                 &exception);
4564         } while (exception.retry);
4565         return err;
4566 }
4567
4568 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4569                 const struct nfs_open_context *ctx,
4570                 const struct nfs_lock_context *l_ctx,
4571                 fmode_t fmode)
4572 {
4573         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4574 }
4575 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4576
4577 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4578                 const struct nfs_open_context *ctx,
4579                 const struct nfs_lock_context *l_ctx,
4580                 fmode_t fmode)
4581 {
4582         nfs4_stateid current_stateid;
4583
4584         /* If the current stateid represents a lost lock, then exit */
4585         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4586                 return true;
4587         return nfs4_stateid_match(stateid, &current_stateid);
4588 }
4589
4590 static bool nfs4_error_stateid_expired(int err)
4591 {
4592         switch (err) {
4593         case -NFS4ERR_DELEG_REVOKED:
4594         case -NFS4ERR_ADMIN_REVOKED:
4595         case -NFS4ERR_BAD_STATEID:
4596         case -NFS4ERR_STALE_STATEID:
4597         case -NFS4ERR_OLD_STATEID:
4598         case -NFS4ERR_OPENMODE:
4599         case -NFS4ERR_EXPIRED:
4600                 return true;
4601         }
4602         return false;
4603 }
4604
4605 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4606 {
4607         struct nfs_server *server = NFS_SERVER(hdr->inode);
4608
4609         trace_nfs4_read(hdr, task->tk_status);
4610         if (task->tk_status < 0) {
4611                 struct nfs4_exception exception = {
4612                         .inode = hdr->inode,
4613                         .state = hdr->args.context->state,
4614                         .stateid = &hdr->args.stateid,
4615                 };
4616                 task->tk_status = nfs4_async_handle_exception(task,
4617                                 server, task->tk_status, &exception);
4618                 if (exception.retry) {
4619                         rpc_restart_call_prepare(task);
4620                         return -EAGAIN;
4621                 }
4622         }
4623
4624         if (task->tk_status > 0)
4625                 renew_lease(server, hdr->timestamp);
4626         return 0;
4627 }
4628
4629 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4630                 struct nfs_pgio_args *args)
4631 {
4632
4633         if (!nfs4_error_stateid_expired(task->tk_status) ||
4634                 nfs4_stateid_is_current(&args->stateid,
4635                                 args->context,
4636                                 args->lock_context,
4637                                 FMODE_READ))
4638                 return false;
4639         rpc_restart_call_prepare(task);
4640         return true;
4641 }
4642
4643 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4644 {
4645
4646         dprintk("--> %s\n", __func__);
4647
4648         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4649                 return -EAGAIN;
4650         if (nfs4_read_stateid_changed(task, &hdr->args))
4651                 return -EAGAIN;
4652         if (task->tk_status > 0)
4653                 nfs_invalidate_atime(hdr->inode);
4654         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4655                                     nfs4_read_done_cb(task, hdr);
4656 }
4657
4658 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4659                                  struct rpc_message *msg)
4660 {
4661         hdr->timestamp   = jiffies;
4662         if (!hdr->pgio_done_cb)
4663                 hdr->pgio_done_cb = nfs4_read_done_cb;
4664         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4665         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4666 }
4667
4668 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4669                                       struct nfs_pgio_header *hdr)
4670 {
4671         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4672                         &hdr->args.seq_args,
4673                         &hdr->res.seq_res,
4674                         task))
4675                 return 0;
4676         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4677                                 hdr->args.lock_context,
4678                                 hdr->rw_mode) == -EIO)
4679                 return -EIO;
4680         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4681                 return -EIO;
4682         return 0;
4683 }
4684
4685 static int nfs4_write_done_cb(struct rpc_task *task,
4686                               struct nfs_pgio_header *hdr)
4687 {
4688         struct inode *inode = hdr->inode;
4689
4690         trace_nfs4_write(hdr, task->tk_status);
4691         if (task->tk_status < 0) {
4692                 struct nfs4_exception exception = {
4693                         .inode = hdr->inode,
4694                         .state = hdr->args.context->state,
4695                         .stateid = &hdr->args.stateid,
4696                 };
4697                 task->tk_status = nfs4_async_handle_exception(task,
4698                                 NFS_SERVER(inode), task->tk_status,
4699                                 &exception);
4700                 if (exception.retry) {
4701                         rpc_restart_call_prepare(task);
4702                         return -EAGAIN;
4703                 }
4704         }
4705         if (task->tk_status >= 0) {
4706                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4707                 nfs_writeback_update_inode(hdr);
4708         }
4709         return 0;
4710 }
4711
4712 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4713                 struct nfs_pgio_args *args)
4714 {
4715
4716         if (!nfs4_error_stateid_expired(task->tk_status) ||
4717                 nfs4_stateid_is_current(&args->stateid,
4718                                 args->context,
4719                                 args->lock_context,
4720                                 FMODE_WRITE))
4721                 return false;
4722         rpc_restart_call_prepare(task);
4723         return true;
4724 }
4725
4726 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4727 {
4728         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4729                 return -EAGAIN;
4730         if (nfs4_write_stateid_changed(task, &hdr->args))
4731                 return -EAGAIN;
4732         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4733                 nfs4_write_done_cb(task, hdr);
4734 }
4735
4736 static
4737 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4738 {
4739         /* Don't request attributes for pNFS or O_DIRECT writes */
4740         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4741                 return false;
4742         /* Otherwise, request attributes if and only if we don't hold
4743          * a delegation
4744          */
4745         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4746 }
4747
4748 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4749                                   struct rpc_message *msg)
4750 {
4751         struct nfs_server *server = NFS_SERVER(hdr->inode);
4752
4753         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4754                 hdr->args.bitmask = NULL;
4755                 hdr->res.fattr = NULL;
4756         } else
4757                 hdr->args.bitmask = server->cache_consistency_bitmask;
4758
4759         if (!hdr->pgio_done_cb)
4760                 hdr->pgio_done_cb = nfs4_write_done_cb;
4761         hdr->res.server = server;
4762         hdr->timestamp   = jiffies;
4763
4764         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4765         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4766 }
4767
4768 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4769 {
4770         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
4771                         &data->args.seq_args,
4772                         &data->res.seq_res,
4773                         task);
4774 }
4775
4776 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4777 {
4778         struct inode *inode = data->inode;
4779
4780         trace_nfs4_commit(data, task->tk_status);
4781         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4782                                     NULL, NULL) == -EAGAIN) {
4783                 rpc_restart_call_prepare(task);
4784                 return -EAGAIN;
4785         }
4786         return 0;
4787 }
4788
4789 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4790 {
4791         if (!nfs4_sequence_done(task, &data->res.seq_res))
4792                 return -EAGAIN;
4793         return data->commit_done_cb(task, data);
4794 }
4795
4796 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4797 {
4798         struct nfs_server *server = NFS_SERVER(data->inode);
4799
4800         if (data->commit_done_cb == NULL)
4801                 data->commit_done_cb = nfs4_commit_done_cb;
4802         data->res.server = server;
4803         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4804         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4805 }
4806
4807 struct nfs4_renewdata {
4808         struct nfs_client       *client;
4809         unsigned long           timestamp;
4810 };
4811
4812 /*
4813  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4814  * standalone procedure for queueing an asynchronous RENEW.
4815  */
4816 static void nfs4_renew_release(void *calldata)
4817 {
4818         struct nfs4_renewdata *data = calldata;
4819         struct nfs_client *clp = data->client;
4820
4821         if (atomic_read(&clp->cl_count) > 1)
4822                 nfs4_schedule_state_renewal(clp);
4823         nfs_put_client(clp);
4824         kfree(data);
4825 }
4826
4827 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4828 {
4829         struct nfs4_renewdata *data = calldata;
4830         struct nfs_client *clp = data->client;
4831         unsigned long timestamp = data->timestamp;
4832
4833         trace_nfs4_renew_async(clp, task->tk_status);
4834         switch (task->tk_status) {
4835         case 0:
4836                 break;
4837         case -NFS4ERR_LEASE_MOVED:
4838                 nfs4_schedule_lease_moved_recovery(clp);
4839                 break;
4840         default:
4841                 /* Unless we're shutting down, schedule state recovery! */
4842                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4843                         return;
4844                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4845                         nfs4_schedule_lease_recovery(clp);
4846                         return;
4847                 }
4848                 nfs4_schedule_path_down_recovery(clp);
4849         }
4850         do_renew_lease(clp, timestamp);
4851 }
4852
4853 static const struct rpc_call_ops nfs4_renew_ops = {
4854         .rpc_call_done = nfs4_renew_done,
4855         .rpc_release = nfs4_renew_release,
4856 };
4857
4858 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4859 {
4860         struct rpc_message msg = {
4861                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4862                 .rpc_argp       = clp,
4863                 .rpc_cred       = cred,
4864         };
4865         struct nfs4_renewdata *data;
4866
4867         if (renew_flags == 0)
4868                 return 0;
4869         if (!atomic_inc_not_zero(&clp->cl_count))
4870                 return -EIO;
4871         data = kmalloc(sizeof(*data), GFP_NOFS);
4872         if (data == NULL) {
4873                 nfs_put_client(clp);
4874                 return -ENOMEM;
4875         }
4876         data->client = clp;
4877         data->timestamp = jiffies;
4878         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4879                         &nfs4_renew_ops, data);
4880 }
4881
4882 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4883 {
4884         struct rpc_message msg = {
4885                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4886                 .rpc_argp       = clp,
4887                 .rpc_cred       = cred,
4888         };
4889         unsigned long now = jiffies;
4890         int status;
4891
4892         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4893         if (status < 0)
4894                 return status;
4895         do_renew_lease(clp, now);
4896         return 0;
4897 }
4898
4899 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4900 {
4901         return server->caps & NFS_CAP_ACLS;
4902 }
4903
4904 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4905  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4906  * the stack.
4907  */
4908 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4909
4910 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4911                 struct page **pages)
4912 {
4913         struct page *newpage, **spages;
4914         int rc = 0;
4915         size_t len;
4916         spages = pages;
4917
4918         do {
4919                 len = min_t(size_t, PAGE_SIZE, buflen);
4920                 newpage = alloc_page(GFP_KERNEL);
4921
4922                 if (newpage == NULL)
4923                         goto unwind;
4924                 memcpy(page_address(newpage), buf, len);
4925                 buf += len;
4926                 buflen -= len;
4927                 *pages++ = newpage;
4928                 rc++;
4929         } while (buflen != 0);
4930
4931         return rc;
4932
4933 unwind:
4934         for(; rc > 0; rc--)
4935                 __free_page(spages[rc-1]);
4936         return -ENOMEM;
4937 }
4938
4939 struct nfs4_cached_acl {
4940         int cached;
4941         size_t len;
4942         char data[0];
4943 };
4944
4945 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4946 {
4947         struct nfs_inode *nfsi = NFS_I(inode);
4948
4949         spin_lock(&inode->i_lock);
4950         kfree(nfsi->nfs4_acl);
4951         nfsi->nfs4_acl = acl;
4952         spin_unlock(&inode->i_lock);
4953 }
4954
4955 static void nfs4_zap_acl_attr(struct inode *inode)
4956 {
4957         nfs4_set_cached_acl(inode, NULL);
4958 }
4959
4960 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4961 {
4962         struct nfs_inode *nfsi = NFS_I(inode);
4963         struct nfs4_cached_acl *acl;
4964         int ret = -ENOENT;
4965
4966         spin_lock(&inode->i_lock);
4967         acl = nfsi->nfs4_acl;
4968         if (acl == NULL)
4969                 goto out;
4970         if (buf == NULL) /* user is just asking for length */
4971                 goto out_len;
4972         if (acl->cached == 0)
4973                 goto out;
4974         ret = -ERANGE; /* see getxattr(2) man page */
4975         if (acl->len > buflen)
4976                 goto out;
4977         memcpy(buf, acl->data, acl->len);
4978 out_len:
4979         ret = acl->len;
4980 out:
4981         spin_unlock(&inode->i_lock);
4982         return ret;
4983 }
4984
4985 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4986 {
4987         struct nfs4_cached_acl *acl;
4988         size_t buflen = sizeof(*acl) + acl_len;
4989
4990         if (buflen <= PAGE_SIZE) {
4991                 acl = kmalloc(buflen, GFP_KERNEL);
4992                 if (acl == NULL)
4993                         goto out;
4994                 acl->cached = 1;
4995                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4996         } else {
4997                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4998                 if (acl == NULL)
4999                         goto out;
5000                 acl->cached = 0;
5001         }
5002         acl->len = acl_len;
5003 out:
5004         nfs4_set_cached_acl(inode, acl);
5005 }
5006
5007 /*
5008  * The getxattr API returns the required buffer length when called with a
5009  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5010  * the required buf.  On a NULL buf, we send a page of data to the server
5011  * guessing that the ACL request can be serviced by a page. If so, we cache
5012  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5013  * the cache. If not so, we throw away the page, and cache the required
5014  * length. The next getxattr call will then produce another round trip to
5015  * the server, this time with the input buf of the required size.
5016  */
5017 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5018 {
5019         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5020         struct nfs_getaclargs args = {
5021                 .fh = NFS_FH(inode),
5022                 .acl_pages = pages,
5023                 .acl_len = buflen,
5024         };
5025         struct nfs_getaclres res = {
5026                 .acl_len = buflen,
5027         };
5028         struct rpc_message msg = {
5029                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5030                 .rpc_argp = &args,
5031                 .rpc_resp = &res,
5032         };
5033         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5034         int ret = -ENOMEM, i;
5035
5036         if (npages > ARRAY_SIZE(pages))
5037                 return -ERANGE;
5038
5039         for (i = 0; i < npages; i++) {
5040                 pages[i] = alloc_page(GFP_KERNEL);
5041                 if (!pages[i])
5042                         goto out_free;
5043         }
5044
5045         /* for decoding across pages */
5046         res.acl_scratch = alloc_page(GFP_KERNEL);
5047         if (!res.acl_scratch)
5048                 goto out_free;
5049
5050         args.acl_len = npages * PAGE_SIZE;
5051
5052         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5053                 __func__, buf, buflen, npages, args.acl_len);
5054         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5055                              &msg, &args.seq_args, &res.seq_res, 0);
5056         if (ret)
5057                 goto out_free;
5058
5059         /* Handle the case where the passed-in buffer is too short */
5060         if (res.acl_flags & NFS4_ACL_TRUNC) {
5061                 /* Did the user only issue a request for the acl length? */
5062                 if (buf == NULL)
5063                         goto out_ok;
5064                 ret = -ERANGE;
5065                 goto out_free;
5066         }
5067         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5068         if (buf) {
5069                 if (res.acl_len > buflen) {
5070                         ret = -ERANGE;
5071                         goto out_free;
5072                 }
5073                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5074         }
5075 out_ok:
5076         ret = res.acl_len;
5077 out_free:
5078         for (i = 0; i < npages; i++)
5079                 if (pages[i])
5080                         __free_page(pages[i]);
5081         if (res.acl_scratch)
5082                 __free_page(res.acl_scratch);
5083         return ret;
5084 }
5085
5086 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5087 {
5088         struct nfs4_exception exception = { };
5089         ssize_t ret;
5090         do {
5091                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5092                 trace_nfs4_get_acl(inode, ret);
5093                 if (ret >= 0)
5094                         break;
5095                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5096         } while (exception.retry);
5097         return ret;
5098 }
5099
5100 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5101 {
5102         struct nfs_server *server = NFS_SERVER(inode);
5103         int ret;
5104
5105         if (!nfs4_server_supports_acls(server))
5106                 return -EOPNOTSUPP;
5107         ret = nfs_revalidate_inode(server, inode);
5108         if (ret < 0)
5109                 return ret;
5110         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5111                 nfs_zap_acl_cache(inode);
5112         ret = nfs4_read_cached_acl(inode, buf, buflen);
5113         if (ret != -ENOENT)
5114                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5115                  * but no cached acl data, just the acl length */
5116                 return ret;
5117         return nfs4_get_acl_uncached(inode, buf, buflen);
5118 }
5119
5120 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5121 {
5122         struct nfs_server *server = NFS_SERVER(inode);
5123         struct page *pages[NFS4ACL_MAXPAGES];
5124         struct nfs_setaclargs arg = {
5125                 .fh             = NFS_FH(inode),
5126                 .acl_pages      = pages,
5127                 .acl_len        = buflen,
5128         };
5129         struct nfs_setaclres res;
5130         struct rpc_message msg = {
5131                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5132                 .rpc_argp       = &arg,
5133                 .rpc_resp       = &res,
5134         };
5135         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5136         int ret, i;
5137
5138         if (!nfs4_server_supports_acls(server))
5139                 return -EOPNOTSUPP;
5140         if (npages > ARRAY_SIZE(pages))
5141                 return -ERANGE;
5142         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5143         if (i < 0)
5144                 return i;
5145         nfs4_inode_return_delegation(inode);
5146         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5147
5148         /*
5149          * Free each page after tx, so the only ref left is
5150          * held by the network stack
5151          */
5152         for (; i > 0; i--)
5153                 put_page(pages[i-1]);
5154
5155         /*
5156          * Acl update can result in inode attribute update.
5157          * so mark the attribute cache invalid.
5158          */
5159         spin_lock(&inode->i_lock);
5160         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5161         spin_unlock(&inode->i_lock);
5162         nfs_access_zap_cache(inode);
5163         nfs_zap_acl_cache(inode);
5164         return ret;
5165 }
5166
5167 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5168 {
5169         struct nfs4_exception exception = { };
5170         int err;
5171         do {
5172                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5173                 trace_nfs4_set_acl(inode, err);
5174                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5175                                 &exception);
5176         } while (exception.retry);
5177         return err;
5178 }
5179
5180 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5181 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5182                                         size_t buflen)
5183 {
5184         struct nfs_server *server = NFS_SERVER(inode);
5185         struct nfs_fattr fattr;
5186         struct nfs4_label label = {0, 0, buflen, buf};
5187
5188         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5189         struct nfs4_getattr_arg arg = {
5190                 .fh             = NFS_FH(inode),
5191                 .bitmask        = bitmask,
5192         };
5193         struct nfs4_getattr_res res = {
5194                 .fattr          = &fattr,
5195                 .label          = &label,
5196                 .server         = server,
5197         };
5198         struct rpc_message msg = {
5199                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5200                 .rpc_argp       = &arg,
5201                 .rpc_resp       = &res,
5202         };
5203         int ret;
5204
5205         nfs_fattr_init(&fattr);
5206
5207         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5208         if (ret)
5209                 return ret;
5210         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5211                 return -ENOENT;
5212         if (buflen < label.len)
5213                 return -ERANGE;
5214         return 0;
5215 }
5216
5217 static int nfs4_get_security_label(struct inode *inode, void *buf,
5218                                         size_t buflen)
5219 {
5220         struct nfs4_exception exception = { };
5221         int err;
5222
5223         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5224                 return -EOPNOTSUPP;
5225
5226         do {
5227                 err = _nfs4_get_security_label(inode, buf, buflen);
5228                 trace_nfs4_get_security_label(inode, err);
5229                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5230                                 &exception);
5231         } while (exception.retry);
5232         return err;
5233 }
5234
5235 static int _nfs4_do_set_security_label(struct inode *inode,
5236                 struct nfs4_label *ilabel,
5237                 struct nfs_fattr *fattr,
5238                 struct nfs4_label *olabel)
5239 {
5240
5241         struct iattr sattr = {0};
5242         struct nfs_server *server = NFS_SERVER(inode);
5243         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5244         struct nfs_setattrargs arg = {
5245                 .fh             = NFS_FH(inode),
5246                 .iap            = &sattr,
5247                 .server         = server,
5248                 .bitmask        = bitmask,
5249                 .label          = ilabel,
5250         };
5251         struct nfs_setattrres res = {
5252                 .fattr          = fattr,
5253                 .label          = olabel,
5254                 .server         = server,
5255         };
5256         struct rpc_message msg = {
5257                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5258                 .rpc_argp       = &arg,
5259                 .rpc_resp       = &res,
5260         };
5261         int status;
5262
5263         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5264
5265         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5266         if (status)
5267                 dprintk("%s failed: %d\n", __func__, status);
5268
5269         return status;
5270 }
5271
5272 static int nfs4_do_set_security_label(struct inode *inode,
5273                 struct nfs4_label *ilabel,
5274                 struct nfs_fattr *fattr,
5275                 struct nfs4_label *olabel)
5276 {
5277         struct nfs4_exception exception = { };
5278         int err;
5279
5280         do {
5281                 err = _nfs4_do_set_security_label(inode, ilabel,
5282                                 fattr, olabel);
5283                 trace_nfs4_set_security_label(inode, err);
5284                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5285                                 &exception);
5286         } while (exception.retry);
5287         return err;
5288 }
5289
5290 static int
5291 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5292 {
5293         struct nfs4_label ilabel, *olabel = NULL;
5294         struct nfs_fattr fattr;
5295         struct rpc_cred *cred;
5296         int status;
5297
5298         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5299                 return -EOPNOTSUPP;
5300
5301         nfs_fattr_init(&fattr);
5302
5303         ilabel.pi = 0;
5304         ilabel.lfs = 0;
5305         ilabel.label = (char *)buf;
5306         ilabel.len = buflen;
5307
5308         cred = rpc_lookup_cred();
5309         if (IS_ERR(cred))
5310                 return PTR_ERR(cred);
5311
5312         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5313         if (IS_ERR(olabel)) {
5314                 status = -PTR_ERR(olabel);
5315                 goto out;
5316         }
5317
5318         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5319         if (status == 0)
5320                 nfs_setsecurity(inode, &fattr, olabel);
5321
5322         nfs4_label_free(olabel);
5323 out:
5324         put_rpccred(cred);
5325         return status;
5326 }
5327 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5328
5329
5330 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5331                                     nfs4_verifier *bootverf)
5332 {
5333         __be32 verf[2];
5334
5335         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5336                 /* An impossible timestamp guarantees this value
5337                  * will never match a generated boot time. */
5338                 verf[0] = cpu_to_be32(U32_MAX);
5339                 verf[1] = cpu_to_be32(U32_MAX);
5340         } else {
5341                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5342                 u64 ns = ktime_to_ns(nn->boot_time);
5343
5344                 verf[0] = cpu_to_be32(ns >> 32);
5345                 verf[1] = cpu_to_be32(ns);
5346         }
5347         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5348 }
5349
5350 static int
5351 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5352 {
5353         size_t len;
5354         char *str;
5355
5356         if (clp->cl_owner_id != NULL)
5357                 return 0;
5358
5359         rcu_read_lock();
5360         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5361                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5362                 1 +
5363                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5364                 1;
5365         rcu_read_unlock();
5366
5367         if (len > NFS4_OPAQUE_LIMIT + 1)
5368                 return -EINVAL;
5369
5370         /*
5371          * Since this string is allocated at mount time, and held until the
5372          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5373          * about a memory-reclaim deadlock.
5374          */
5375         str = kmalloc(len, GFP_KERNEL);
5376         if (!str)
5377                 return -ENOMEM;
5378
5379         rcu_read_lock();
5380         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5381                         clp->cl_ipaddr,
5382                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5383                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5384         rcu_read_unlock();
5385
5386         clp->cl_owner_id = str;
5387         return 0;
5388 }
5389
5390 static int
5391 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5392 {
5393         size_t len;
5394         char *str;
5395
5396         len = 10 + 10 + 1 + 10 + 1 +
5397                 strlen(nfs4_client_id_uniquifier) + 1 +
5398                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5399
5400         if (len > NFS4_OPAQUE_LIMIT + 1)
5401                 return -EINVAL;
5402
5403         /*
5404          * Since this string is allocated at mount time, and held until the
5405          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5406          * about a memory-reclaim deadlock.
5407          */
5408         str = kmalloc(len, GFP_KERNEL);
5409         if (!str)
5410                 return -ENOMEM;
5411
5412         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5413                         clp->rpc_ops->version, clp->cl_minorversion,
5414                         nfs4_client_id_uniquifier,
5415                         clp->cl_rpcclient->cl_nodename);
5416         clp->cl_owner_id = str;
5417         return 0;
5418 }
5419
5420 static int
5421 nfs4_init_uniform_client_string(struct nfs_client *clp)
5422 {
5423         size_t len;
5424         char *str;
5425
5426         if (clp->cl_owner_id != NULL)
5427                 return 0;
5428
5429         if (nfs4_client_id_uniquifier[0] != '\0')
5430                 return nfs4_init_uniquifier_client_string(clp);
5431
5432         len = 10 + 10 + 1 + 10 + 1 +
5433                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5434
5435         if (len > NFS4_OPAQUE_LIMIT + 1)
5436                 return -EINVAL;
5437
5438         /*
5439          * Since this string is allocated at mount time, and held until the
5440          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5441          * about a memory-reclaim deadlock.
5442          */
5443         str = kmalloc(len, GFP_KERNEL);
5444         if (!str)
5445                 return -ENOMEM;
5446
5447         scnprintf(str, len, "Linux NFSv%u.%u %s",
5448                         clp->rpc_ops->version, clp->cl_minorversion,
5449                         clp->cl_rpcclient->cl_nodename);
5450         clp->cl_owner_id = str;
5451         return 0;
5452 }
5453
5454 /*
5455  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5456  * services.  Advertise one based on the address family of the
5457  * clientaddr.
5458  */
5459 static unsigned int
5460 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5461 {
5462         if (strchr(clp->cl_ipaddr, ':') != NULL)
5463                 return scnprintf(buf, len, "tcp6");
5464         else
5465                 return scnprintf(buf, len, "tcp");
5466 }
5467
5468 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5469 {
5470         struct nfs4_setclientid *sc = calldata;
5471
5472         if (task->tk_status == 0)
5473                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5474 }
5475
5476 static const struct rpc_call_ops nfs4_setclientid_ops = {
5477         .rpc_call_done = nfs4_setclientid_done,
5478 };
5479
5480 /**
5481  * nfs4_proc_setclientid - Negotiate client ID
5482  * @clp: state data structure
5483  * @program: RPC program for NFSv4 callback service
5484  * @port: IP port number for NFS4 callback service
5485  * @cred: RPC credential to use for this call
5486  * @res: where to place the result
5487  *
5488  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5489  */
5490 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5491                 unsigned short port, struct rpc_cred *cred,
5492                 struct nfs4_setclientid_res *res)
5493 {
5494         nfs4_verifier sc_verifier;
5495         struct nfs4_setclientid setclientid = {
5496                 .sc_verifier = &sc_verifier,
5497                 .sc_prog = program,
5498                 .sc_clnt = clp,
5499         };
5500         struct rpc_message msg = {
5501                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5502                 .rpc_argp = &setclientid,
5503                 .rpc_resp = res,
5504                 .rpc_cred = cred,
5505         };
5506         struct rpc_task *task;
5507         struct rpc_task_setup task_setup_data = {
5508                 .rpc_client = clp->cl_rpcclient,
5509                 .rpc_message = &msg,
5510                 .callback_ops = &nfs4_setclientid_ops,
5511                 .callback_data = &setclientid,
5512                 .flags = RPC_TASK_TIMEOUT,
5513         };
5514         int status;
5515
5516         /* nfs_client_id4 */
5517         nfs4_init_boot_verifier(clp, &sc_verifier);
5518
5519         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5520                 status = nfs4_init_uniform_client_string(clp);
5521         else
5522                 status = nfs4_init_nonuniform_client_string(clp);
5523
5524         if (status)
5525                 goto out;
5526
5527         /* cb_client4 */
5528         setclientid.sc_netid_len =
5529                                 nfs4_init_callback_netid(clp,
5530                                                 setclientid.sc_netid,
5531                                                 sizeof(setclientid.sc_netid));
5532         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5533                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5534                                 clp->cl_ipaddr, port >> 8, port & 255);
5535
5536         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5537                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5538                 clp->cl_owner_id);
5539         task = rpc_run_task(&task_setup_data);
5540         if (IS_ERR(task)) {
5541                 status = PTR_ERR(task);
5542                 goto out;
5543         }
5544         status = task->tk_status;
5545         if (setclientid.sc_cred) {
5546                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5547                 put_rpccred(setclientid.sc_cred);
5548         }
5549         rpc_put_task(task);
5550 out:
5551         trace_nfs4_setclientid(clp, status);
5552         dprintk("NFS reply setclientid: %d\n", status);
5553         return status;
5554 }
5555
5556 /**
5557  * nfs4_proc_setclientid_confirm - Confirm client ID
5558  * @clp: state data structure
5559  * @res: result of a previous SETCLIENTID
5560  * @cred: RPC credential to use for this call
5561  *
5562  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5563  */
5564 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5565                 struct nfs4_setclientid_res *arg,
5566                 struct rpc_cred *cred)
5567 {
5568         struct rpc_message msg = {
5569                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5570                 .rpc_argp = arg,
5571                 .rpc_cred = cred,
5572         };
5573         int status;
5574
5575         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5576                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5577                 clp->cl_clientid);
5578         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5579         trace_nfs4_setclientid_confirm(clp, status);
5580         dprintk("NFS reply setclientid_confirm: %d\n", status);
5581         return status;
5582 }
5583
5584 struct nfs4_delegreturndata {
5585         struct nfs4_delegreturnargs args;
5586         struct nfs4_delegreturnres res;
5587         struct nfs_fh fh;
5588         nfs4_stateid stateid;
5589         unsigned long timestamp;
5590         struct {
5591                 struct nfs4_layoutreturn_args arg;
5592                 struct nfs4_layoutreturn_res res;
5593                 struct nfs4_xdr_opaque_data ld_private;
5594                 u32 roc_barrier;
5595                 bool roc;
5596         } lr;
5597         struct nfs_fattr fattr;
5598         int rpc_status;
5599         struct inode *inode;
5600 };
5601
5602 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5603 {
5604         struct nfs4_delegreturndata *data = calldata;
5605
5606         if (!nfs4_sequence_done(task, &data->res.seq_res))
5607                 return;
5608
5609         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5610
5611         /* Handle Layoutreturn errors */
5612         if (data->args.lr_args && task->tk_status != 0) {
5613                 switch(data->res.lr_ret) {
5614                 default:
5615                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5616                         break;
5617                 case 0:
5618                         data->args.lr_args = NULL;
5619                         data->res.lr_res = NULL;
5620                         break;
5621                 case -NFS4ERR_ADMIN_REVOKED:
5622                 case -NFS4ERR_DELEG_REVOKED:
5623                 case -NFS4ERR_EXPIRED:
5624                 case -NFS4ERR_BAD_STATEID:
5625                 case -NFS4ERR_OLD_STATEID:
5626                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5627                 case -NFS4ERR_WRONG_CRED:
5628                         data->args.lr_args = NULL;
5629                         data->res.lr_res = NULL;
5630                         data->res.lr_ret = 0;
5631                         rpc_restart_call_prepare(task);
5632                         return;
5633                 }
5634         }
5635
5636         switch (task->tk_status) {
5637         case 0:
5638                 renew_lease(data->res.server, data->timestamp);
5639                 break;
5640         case -NFS4ERR_ADMIN_REVOKED:
5641         case -NFS4ERR_DELEG_REVOKED:
5642         case -NFS4ERR_EXPIRED:
5643                 nfs4_free_revoked_stateid(data->res.server,
5644                                 data->args.stateid,
5645                                 task->tk_msg.rpc_cred);
5646         case -NFS4ERR_BAD_STATEID:
5647         case -NFS4ERR_OLD_STATEID:
5648         case -NFS4ERR_STALE_STATEID:
5649                 task->tk_status = 0;
5650                 break;
5651         case -NFS4ERR_ACCESS:
5652                 if (data->args.bitmask) {
5653                         data->args.bitmask = NULL;
5654                         data->res.fattr = NULL;
5655                         task->tk_status = 0;
5656                         rpc_restart_call_prepare(task);
5657                         return;
5658                 }
5659         default:
5660                 if (nfs4_async_handle_error(task, data->res.server,
5661                                             NULL, NULL) == -EAGAIN) {
5662                         rpc_restart_call_prepare(task);
5663                         return;
5664                 }
5665         }
5666         data->rpc_status = task->tk_status;
5667 }
5668
5669 static void nfs4_delegreturn_release(void *calldata)
5670 {
5671         struct nfs4_delegreturndata *data = calldata;
5672         struct inode *inode = data->inode;
5673
5674         if (inode) {
5675                 if (data->lr.roc)
5676                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5677                                         data->res.lr_ret);
5678                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5679                 nfs_iput_and_deactive(inode);
5680         }
5681         kfree(calldata);
5682 }
5683
5684 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5685 {
5686         struct nfs4_delegreturndata *d_data;
5687
5688         d_data = (struct nfs4_delegreturndata *)data;
5689
5690         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5691                 return;
5692
5693         nfs4_setup_sequence(d_data->res.server->nfs_client,
5694                         &d_data->args.seq_args,
5695                         &d_data->res.seq_res,
5696                         task);
5697 }
5698
5699 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5700         .rpc_call_prepare = nfs4_delegreturn_prepare,
5701         .rpc_call_done = nfs4_delegreturn_done,
5702         .rpc_release = nfs4_delegreturn_release,
5703 };
5704
5705 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5706 {
5707         struct nfs4_delegreturndata *data;
5708         struct nfs_server *server = NFS_SERVER(inode);
5709         struct rpc_task *task;
5710         struct rpc_message msg = {
5711                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5712                 .rpc_cred = cred,
5713         };
5714         struct rpc_task_setup task_setup_data = {
5715                 .rpc_client = server->client,
5716                 .rpc_message = &msg,
5717                 .callback_ops = &nfs4_delegreturn_ops,
5718                 .flags = RPC_TASK_ASYNC,
5719         };
5720         int status = 0;
5721
5722         data = kzalloc(sizeof(*data), GFP_NOFS);
5723         if (data == NULL)
5724                 return -ENOMEM;
5725         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5726
5727         nfs4_state_protect(server->nfs_client,
5728                         NFS_SP4_MACH_CRED_CLEANUP,
5729                         &task_setup_data.rpc_client, &msg);
5730
5731         data->args.fhandle = &data->fh;
5732         data->args.stateid = &data->stateid;
5733         data->args.bitmask = server->cache_consistency_bitmask;
5734         nfs_copy_fh(&data->fh, NFS_FH(inode));
5735         nfs4_stateid_copy(&data->stateid, stateid);
5736         data->res.fattr = &data->fattr;
5737         data->res.server = server;
5738         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5739         data->lr.arg.ld_private = &data->lr.ld_private;
5740         nfs_fattr_init(data->res.fattr);
5741         data->timestamp = jiffies;
5742         data->rpc_status = 0;
5743         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5744         data->inode = nfs_igrab_and_active(inode);
5745         if (data->inode) {
5746                 if (data->lr.roc) {
5747                         data->args.lr_args = &data->lr.arg;
5748                         data->res.lr_res = &data->lr.res;
5749                 }
5750         } else if (data->lr.roc) {
5751                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5752                 data->lr.roc = false;
5753         }
5754
5755         task_setup_data.callback_data = data;
5756         msg.rpc_argp = &data->args;
5757         msg.rpc_resp = &data->res;
5758         task = rpc_run_task(&task_setup_data);
5759         if (IS_ERR(task))
5760                 return PTR_ERR(task);
5761         if (!issync)
5762                 goto out;
5763         status = rpc_wait_for_completion_task(task);
5764         if (status != 0)
5765                 goto out;
5766         status = data->rpc_status;
5767 out:
5768         rpc_put_task(task);
5769         return status;
5770 }
5771
5772 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5773 {
5774         struct nfs_server *server = NFS_SERVER(inode);
5775         struct nfs4_exception exception = { };
5776         int err;
5777         do {
5778                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5779                 trace_nfs4_delegreturn(inode, stateid, err);
5780                 switch (err) {
5781                         case -NFS4ERR_STALE_STATEID:
5782                         case -NFS4ERR_EXPIRED:
5783                         case 0:
5784                                 return 0;
5785                 }
5786                 err = nfs4_handle_exception(server, err, &exception);
5787         } while (exception.retry);
5788         return err;
5789 }
5790
5791 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5792 {
5793         struct inode *inode = state->inode;
5794         struct nfs_server *server = NFS_SERVER(inode);
5795         struct nfs_client *clp = server->nfs_client;
5796         struct nfs_lockt_args arg = {
5797                 .fh = NFS_FH(inode),
5798                 .fl = request,
5799         };
5800         struct nfs_lockt_res res = {
5801                 .denied = request,
5802         };
5803         struct rpc_message msg = {
5804                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5805                 .rpc_argp       = &arg,
5806                 .rpc_resp       = &res,
5807                 .rpc_cred       = state->owner->so_cred,
5808         };
5809         struct nfs4_lock_state *lsp;
5810         int status;
5811
5812         arg.lock_owner.clientid = clp->cl_clientid;
5813         status = nfs4_set_lock_state(state, request);
5814         if (status != 0)
5815                 goto out;
5816         lsp = request->fl_u.nfs4_fl.owner;
5817         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5818         arg.lock_owner.s_dev = server->s_dev;
5819         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5820         switch (status) {
5821                 case 0:
5822                         request->fl_type = F_UNLCK;
5823                         break;
5824                 case -NFS4ERR_DENIED:
5825                         status = 0;
5826         }
5827         request->fl_ops->fl_release_private(request);
5828         request->fl_ops = NULL;
5829 out:
5830         return status;
5831 }
5832
5833 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5834 {
5835         struct nfs4_exception exception = { };
5836         int err;
5837
5838         do {
5839                 err = _nfs4_proc_getlk(state, cmd, request);
5840                 trace_nfs4_get_lock(request, state, cmd, err);
5841                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5842                                 &exception);
5843         } while (exception.retry);
5844         return err;
5845 }
5846
5847 struct nfs4_unlockdata {
5848         struct nfs_locku_args arg;
5849         struct nfs_locku_res res;
5850         struct nfs4_lock_state *lsp;
5851         struct nfs_open_context *ctx;
5852         struct nfs_lock_context *l_ctx;
5853         struct file_lock fl;
5854         struct nfs_server *server;
5855         unsigned long timestamp;
5856 };
5857
5858 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5859                 struct nfs_open_context *ctx,
5860                 struct nfs4_lock_state *lsp,
5861                 struct nfs_seqid *seqid)
5862 {
5863         struct nfs4_unlockdata *p;
5864         struct inode *inode = lsp->ls_state->inode;
5865
5866         p = kzalloc(sizeof(*p), GFP_NOFS);
5867         if (p == NULL)
5868                 return NULL;
5869         p->arg.fh = NFS_FH(inode);
5870         p->arg.fl = &p->fl;
5871         p->arg.seqid = seqid;
5872         p->res.seqid = seqid;
5873         p->lsp = lsp;
5874         atomic_inc(&lsp->ls_count);
5875         /* Ensure we don't close file until we're done freeing locks! */
5876         p->ctx = get_nfs_open_context(ctx);
5877         p->l_ctx = nfs_get_lock_context(ctx);
5878         memcpy(&p->fl, fl, sizeof(p->fl));
5879         p->server = NFS_SERVER(inode);
5880         return p;
5881 }
5882
5883 static void nfs4_locku_release_calldata(void *data)
5884 {
5885         struct nfs4_unlockdata *calldata = data;
5886         nfs_free_seqid(calldata->arg.seqid);
5887         nfs4_put_lock_state(calldata->lsp);
5888         nfs_put_lock_context(calldata->l_ctx);
5889         put_nfs_open_context(calldata->ctx);
5890         kfree(calldata);
5891 }
5892
5893 static void nfs4_locku_done(struct rpc_task *task, void *data)
5894 {
5895         struct nfs4_unlockdata *calldata = data;
5896
5897         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5898                 return;
5899         switch (task->tk_status) {
5900                 case 0:
5901                         renew_lease(calldata->server, calldata->timestamp);
5902                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5903                         if (nfs4_update_lock_stateid(calldata->lsp,
5904                                         &calldata->res.stateid))
5905                                 break;
5906                 case -NFS4ERR_ADMIN_REVOKED:
5907                 case -NFS4ERR_EXPIRED:
5908                         nfs4_free_revoked_stateid(calldata->server,
5909                                         &calldata->arg.stateid,
5910                                         task->tk_msg.rpc_cred);
5911                 case -NFS4ERR_BAD_STATEID:
5912                 case -NFS4ERR_OLD_STATEID:
5913                 case -NFS4ERR_STALE_STATEID:
5914                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5915                                                 &calldata->lsp->ls_stateid))
5916                                 rpc_restart_call_prepare(task);
5917                         break;
5918                 default:
5919                         if (nfs4_async_handle_error(task, calldata->server,
5920                                                     NULL, NULL) == -EAGAIN)
5921                                 rpc_restart_call_prepare(task);
5922         }
5923         nfs_release_seqid(calldata->arg.seqid);
5924 }
5925
5926 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5927 {
5928         struct nfs4_unlockdata *calldata = data;
5929
5930         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
5931                 nfs_async_iocounter_wait(task, calldata->l_ctx))
5932                 return;
5933
5934         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5935                 goto out_wait;
5936         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5937         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5938                 /* Note: exit _without_ running nfs4_locku_done */
5939                 goto out_no_action;
5940         }
5941         calldata->timestamp = jiffies;
5942         if (nfs4_setup_sequence(calldata->server->nfs_client,
5943                                 &calldata->arg.seq_args,
5944                                 &calldata->res.seq_res,
5945                                 task) != 0)
5946                 nfs_release_seqid(calldata->arg.seqid);
5947         return;
5948 out_no_action:
5949         task->tk_action = NULL;
5950 out_wait:
5951         nfs4_sequence_done(task, &calldata->res.seq_res);
5952 }
5953
5954 static const struct rpc_call_ops nfs4_locku_ops = {
5955         .rpc_call_prepare = nfs4_locku_prepare,
5956         .rpc_call_done = nfs4_locku_done,
5957         .rpc_release = nfs4_locku_release_calldata,
5958 };
5959
5960 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5961                 struct nfs_open_context *ctx,
5962                 struct nfs4_lock_state *lsp,
5963                 struct nfs_seqid *seqid)
5964 {
5965         struct nfs4_unlockdata *data;
5966         struct rpc_message msg = {
5967                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5968                 .rpc_cred = ctx->cred,
5969         };
5970         struct rpc_task_setup task_setup_data = {
5971                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5972                 .rpc_message = &msg,
5973                 .callback_ops = &nfs4_locku_ops,
5974                 .workqueue = nfsiod_workqueue,
5975                 .flags = RPC_TASK_ASYNC,
5976         };
5977
5978         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5979                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5980
5981         /* Ensure this is an unlock - when canceling a lock, the
5982          * canceled lock is passed in, and it won't be an unlock.
5983          */
5984         fl->fl_type = F_UNLCK;
5985         if (fl->fl_flags & FL_CLOSE)
5986                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
5987
5988         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5989         if (data == NULL) {
5990                 nfs_free_seqid(seqid);
5991                 return ERR_PTR(-ENOMEM);
5992         }
5993
5994         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5995         msg.rpc_argp = &data->arg;
5996         msg.rpc_resp = &data->res;
5997         task_setup_data.callback_data = data;
5998         return rpc_run_task(&task_setup_data);
5999 }
6000
6001 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6002 {
6003         struct inode *inode = state->inode;
6004         struct nfs4_state_owner *sp = state->owner;
6005         struct nfs_inode *nfsi = NFS_I(inode);
6006         struct nfs_seqid *seqid;
6007         struct nfs4_lock_state *lsp;
6008         struct rpc_task *task;
6009         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6010         int status = 0;
6011         unsigned char fl_flags = request->fl_flags;
6012
6013         status = nfs4_set_lock_state(state, request);
6014         /* Unlock _before_ we do the RPC call */
6015         request->fl_flags |= FL_EXISTS;
6016         /* Exclude nfs_delegation_claim_locks() */
6017         mutex_lock(&sp->so_delegreturn_mutex);
6018         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6019         down_read(&nfsi->rwsem);
6020         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6021                 up_read(&nfsi->rwsem);
6022                 mutex_unlock(&sp->so_delegreturn_mutex);
6023                 goto out;
6024         }
6025         up_read(&nfsi->rwsem);
6026         mutex_unlock(&sp->so_delegreturn_mutex);
6027         if (status != 0)
6028                 goto out;
6029         /* Is this a delegated lock? */
6030         lsp = request->fl_u.nfs4_fl.owner;
6031         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6032                 goto out;
6033         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6034         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6035         status = -ENOMEM;
6036         if (IS_ERR(seqid))
6037                 goto out;
6038         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6039         status = PTR_ERR(task);
6040         if (IS_ERR(task))
6041                 goto out;
6042         status = rpc_wait_for_completion_task(task);
6043         rpc_put_task(task);
6044 out:
6045         request->fl_flags = fl_flags;
6046         trace_nfs4_unlock(request, state, F_SETLK, status);
6047         return status;
6048 }
6049
6050 struct nfs4_lockdata {
6051         struct nfs_lock_args arg;
6052         struct nfs_lock_res res;
6053         struct nfs4_lock_state *lsp;
6054         struct nfs_open_context *ctx;
6055         struct file_lock fl;
6056         unsigned long timestamp;
6057         int rpc_status;
6058         int cancelled;
6059         struct nfs_server *server;
6060 };
6061
6062 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6063                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6064                 gfp_t gfp_mask)
6065 {
6066         struct nfs4_lockdata *p;
6067         struct inode *inode = lsp->ls_state->inode;
6068         struct nfs_server *server = NFS_SERVER(inode);
6069         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6070
6071         p = kzalloc(sizeof(*p), gfp_mask);
6072         if (p == NULL)
6073                 return NULL;
6074
6075         p->arg.fh = NFS_FH(inode);
6076         p->arg.fl = &p->fl;
6077         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6078         if (IS_ERR(p->arg.open_seqid))
6079                 goto out_free;
6080         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6081         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6082         if (IS_ERR(p->arg.lock_seqid))
6083                 goto out_free_seqid;
6084         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6085         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6086         p->arg.lock_owner.s_dev = server->s_dev;
6087         p->res.lock_seqid = p->arg.lock_seqid;
6088         p->lsp = lsp;
6089         p->server = server;
6090         atomic_inc(&lsp->ls_count);
6091         p->ctx = get_nfs_open_context(ctx);
6092         memcpy(&p->fl, fl, sizeof(p->fl));
6093         return p;
6094 out_free_seqid:
6095         nfs_free_seqid(p->arg.open_seqid);
6096 out_free:
6097         kfree(p);
6098         return NULL;
6099 }
6100
6101 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6102 {
6103         struct nfs4_lockdata *data = calldata;
6104         struct nfs4_state *state = data->lsp->ls_state;
6105
6106         dprintk("%s: begin!\n", __func__);
6107         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6108                 goto out_wait;
6109         /* Do we need to do an open_to_lock_owner? */
6110         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6111                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6112                         goto out_release_lock_seqid;
6113                 }
6114                 nfs4_stateid_copy(&data->arg.open_stateid,
6115                                 &state->open_stateid);
6116                 data->arg.new_lock_owner = 1;
6117                 data->res.open_seqid = data->arg.open_seqid;
6118         } else {
6119                 data->arg.new_lock_owner = 0;
6120                 nfs4_stateid_copy(&data->arg.lock_stateid,
6121                                 &data->lsp->ls_stateid);
6122         }
6123         if (!nfs4_valid_open_stateid(state)) {
6124                 data->rpc_status = -EBADF;
6125                 task->tk_action = NULL;
6126                 goto out_release_open_seqid;
6127         }
6128         data->timestamp = jiffies;
6129         if (nfs4_setup_sequence(data->server->nfs_client,
6130                                 &data->arg.seq_args,
6131                                 &data->res.seq_res,
6132                                 task) == 0)
6133                 return;
6134 out_release_open_seqid:
6135         nfs_release_seqid(data->arg.open_seqid);
6136 out_release_lock_seqid:
6137         nfs_release_seqid(data->arg.lock_seqid);
6138 out_wait:
6139         nfs4_sequence_done(task, &data->res.seq_res);
6140         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6141 }
6142
6143 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6144 {
6145         struct nfs4_lockdata *data = calldata;
6146         struct nfs4_lock_state *lsp = data->lsp;
6147
6148         dprintk("%s: begin!\n", __func__);
6149
6150         if (!nfs4_sequence_done(task, &data->res.seq_res))
6151                 return;
6152
6153         data->rpc_status = task->tk_status;
6154         switch (task->tk_status) {
6155         case 0:
6156                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6157                                 data->timestamp);
6158                 if (data->arg.new_lock) {
6159                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6160                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6161                                 rpc_restart_call_prepare(task);
6162                                 break;
6163                         }
6164                 }
6165                 if (data->arg.new_lock_owner != 0) {
6166                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6167                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6168                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6169                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6170                         rpc_restart_call_prepare(task);
6171                 break;
6172         case -NFS4ERR_BAD_STATEID:
6173         case -NFS4ERR_OLD_STATEID:
6174         case -NFS4ERR_STALE_STATEID:
6175         case -NFS4ERR_EXPIRED:
6176                 if (data->arg.new_lock_owner != 0) {
6177                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6178                                                 &lsp->ls_state->open_stateid))
6179                                 rpc_restart_call_prepare(task);
6180                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6181                                                 &lsp->ls_stateid))
6182                                 rpc_restart_call_prepare(task);
6183         }
6184         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6185 }
6186
6187 static void nfs4_lock_release(void *calldata)
6188 {
6189         struct nfs4_lockdata *data = calldata;
6190
6191         dprintk("%s: begin!\n", __func__);
6192         nfs_free_seqid(data->arg.open_seqid);
6193         if (data->cancelled) {
6194                 struct rpc_task *task;
6195                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6196                                 data->arg.lock_seqid);
6197                 if (!IS_ERR(task))
6198                         rpc_put_task_async(task);
6199                 dprintk("%s: cancelling lock!\n", __func__);
6200         } else
6201                 nfs_free_seqid(data->arg.lock_seqid);
6202         nfs4_put_lock_state(data->lsp);
6203         put_nfs_open_context(data->ctx);
6204         kfree(data);
6205         dprintk("%s: done!\n", __func__);
6206 }
6207
6208 static const struct rpc_call_ops nfs4_lock_ops = {
6209         .rpc_call_prepare = nfs4_lock_prepare,
6210         .rpc_call_done = nfs4_lock_done,
6211         .rpc_release = nfs4_lock_release,
6212 };
6213
6214 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6215 {
6216         switch (error) {
6217         case -NFS4ERR_ADMIN_REVOKED:
6218         case -NFS4ERR_EXPIRED:
6219         case -NFS4ERR_BAD_STATEID:
6220                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6221                 if (new_lock_owner != 0 ||
6222                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6223                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6224                 break;
6225         case -NFS4ERR_STALE_STATEID:
6226                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6227                 nfs4_schedule_lease_recovery(server->nfs_client);
6228         };
6229 }
6230
6231 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6232 {
6233         struct nfs4_lockdata *data;
6234         struct rpc_task *task;
6235         struct rpc_message msg = {
6236                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6237                 .rpc_cred = state->owner->so_cred,
6238         };
6239         struct rpc_task_setup task_setup_data = {
6240                 .rpc_client = NFS_CLIENT(state->inode),
6241                 .rpc_message = &msg,
6242                 .callback_ops = &nfs4_lock_ops,
6243                 .workqueue = nfsiod_workqueue,
6244                 .flags = RPC_TASK_ASYNC,
6245         };
6246         int ret;
6247
6248         dprintk("%s: begin!\n", __func__);
6249         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6250                         fl->fl_u.nfs4_fl.owner,
6251                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6252         if (data == NULL)
6253                 return -ENOMEM;
6254         if (IS_SETLKW(cmd))
6255                 data->arg.block = 1;
6256         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6257         msg.rpc_argp = &data->arg;
6258         msg.rpc_resp = &data->res;
6259         task_setup_data.callback_data = data;
6260         if (recovery_type > NFS_LOCK_NEW) {
6261                 if (recovery_type == NFS_LOCK_RECLAIM)
6262                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6263                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6264         } else
6265                 data->arg.new_lock = 1;
6266         task = rpc_run_task(&task_setup_data);
6267         if (IS_ERR(task))
6268                 return PTR_ERR(task);
6269         ret = rpc_wait_for_completion_task(task);
6270         if (ret == 0) {
6271                 ret = data->rpc_status;
6272                 if (ret)
6273                         nfs4_handle_setlk_error(data->server, data->lsp,
6274                                         data->arg.new_lock_owner, ret);
6275         } else
6276                 data->cancelled = true;
6277         rpc_put_task(task);
6278         dprintk("%s: done, ret = %d!\n", __func__, ret);
6279         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6280         return ret;
6281 }
6282
6283 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6284 {
6285         struct nfs_server *server = NFS_SERVER(state->inode);
6286         struct nfs4_exception exception = {
6287                 .inode = state->inode,
6288         };
6289         int err;
6290
6291         do {
6292                 /* Cache the lock if possible... */
6293                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6294                         return 0;
6295                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6296                 if (err != -NFS4ERR_DELAY)
6297                         break;
6298                 nfs4_handle_exception(server, err, &exception);
6299         } while (exception.retry);
6300         return err;
6301 }
6302
6303 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6304 {
6305         struct nfs_server *server = NFS_SERVER(state->inode);
6306         struct nfs4_exception exception = {
6307                 .inode = state->inode,
6308         };
6309         int err;
6310
6311         err = nfs4_set_lock_state(state, request);
6312         if (err != 0)
6313                 return err;
6314         if (!recover_lost_locks) {
6315                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6316                 return 0;
6317         }
6318         do {
6319                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6320                         return 0;
6321                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6322                 switch (err) {
6323                 default:
6324                         goto out;
6325                 case -NFS4ERR_GRACE:
6326                 case -NFS4ERR_DELAY:
6327                         nfs4_handle_exception(server, err, &exception);
6328                         err = 0;
6329                 }
6330         } while (exception.retry);
6331 out:
6332         return err;
6333 }
6334
6335 #if defined(CONFIG_NFS_V4_1)
6336 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6337 {
6338         struct nfs4_lock_state *lsp;
6339         int status;
6340
6341         status = nfs4_set_lock_state(state, request);
6342         if (status != 0)
6343                 return status;
6344         lsp = request->fl_u.nfs4_fl.owner;
6345         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6346             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6347                 return 0;
6348         return nfs4_lock_expired(state, request);
6349 }
6350 #endif
6351
6352 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6353 {
6354         struct nfs_inode *nfsi = NFS_I(state->inode);
6355         struct nfs4_state_owner *sp = state->owner;
6356         unsigned char fl_flags = request->fl_flags;
6357         int status;
6358
6359         request->fl_flags |= FL_ACCESS;
6360         status = locks_lock_inode_wait(state->inode, request);
6361         if (status < 0)
6362                 goto out;
6363         mutex_lock(&sp->so_delegreturn_mutex);
6364         down_read(&nfsi->rwsem);
6365         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6366                 /* Yes: cache locks! */
6367                 /* ...but avoid races with delegation recall... */
6368                 request->fl_flags = fl_flags & ~FL_SLEEP;
6369                 status = locks_lock_inode_wait(state->inode, request);
6370                 up_read(&nfsi->rwsem);
6371                 mutex_unlock(&sp->so_delegreturn_mutex);
6372                 goto out;
6373         }
6374         up_read(&nfsi->rwsem);
6375         mutex_unlock(&sp->so_delegreturn_mutex);
6376         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6377 out:
6378         request->fl_flags = fl_flags;
6379         return status;
6380 }
6381
6382 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6383 {
6384         struct nfs4_exception exception = {
6385                 .state = state,
6386                 .inode = state->inode,
6387         };
6388         int err;
6389
6390         do {
6391                 err = _nfs4_proc_setlk(state, cmd, request);
6392                 if (err == -NFS4ERR_DENIED)
6393                         err = -EAGAIN;
6394                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6395                                 err, &exception);
6396         } while (exception.retry);
6397         return err;
6398 }
6399
6400 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6401 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6402
6403 static int
6404 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6405                         struct file_lock *request)
6406 {
6407         int             status = -ERESTARTSYS;
6408         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6409
6410         while(!signalled()) {
6411                 status = nfs4_proc_setlk(state, cmd, request);
6412                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6413                         break;
6414                 freezable_schedule_timeout_interruptible(timeout);
6415                 timeout *= 2;
6416                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6417                 status = -ERESTARTSYS;
6418         }
6419         return status;
6420 }
6421
6422 #ifdef CONFIG_NFS_V4_1
6423 struct nfs4_lock_waiter {
6424         struct task_struct      *task;
6425         struct inode            *inode;
6426         struct nfs_lowner       *owner;
6427         bool                    notified;
6428 };
6429
6430 static int
6431 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6432 {
6433         int ret;
6434         struct cb_notify_lock_args *cbnl = key;
6435         struct nfs4_lock_waiter *waiter = wait->private;
6436         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6437                                 *wowner = waiter->owner;
6438
6439         /* Only wake if the callback was for the same owner */
6440         if (lowner->clientid != wowner->clientid ||
6441             lowner->id != wowner->id             ||
6442             lowner->s_dev != wowner->s_dev)
6443                 return 0;
6444
6445         /* Make sure it's for the right inode */
6446         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6447                 return 0;
6448
6449         waiter->notified = true;
6450
6451         /* override "private" so we can use default_wake_function */
6452         wait->private = waiter->task;
6453         ret = autoremove_wake_function(wait, mode, flags, key);
6454         wait->private = waiter;
6455         return ret;
6456 }
6457
6458 static int
6459 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6460 {
6461         int status = -ERESTARTSYS;
6462         unsigned long flags;
6463         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6464         struct nfs_server *server = NFS_SERVER(state->inode);
6465         struct nfs_client *clp = server->nfs_client;
6466         wait_queue_head_t *q = &clp->cl_lock_waitq;
6467         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6468                                     .id = lsp->ls_seqid.owner_id,
6469                                     .s_dev = server->s_dev };
6470         struct nfs4_lock_waiter waiter = { .task  = current,
6471                                            .inode = state->inode,
6472                                            .owner = &owner,
6473                                            .notified = false };
6474         wait_queue_entry_t wait;
6475
6476         /* Don't bother with waitqueue if we don't expect a callback */
6477         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6478                 return nfs4_retry_setlk_simple(state, cmd, request);
6479
6480         init_wait(&wait);
6481         wait.private = &waiter;
6482         wait.func = nfs4_wake_lock_waiter;
6483         add_wait_queue(q, &wait);
6484
6485         while(!signalled()) {
6486                 status = nfs4_proc_setlk(state, cmd, request);
6487                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6488                         break;
6489
6490                 status = -ERESTARTSYS;
6491                 spin_lock_irqsave(&q->lock, flags);
6492                 if (waiter.notified) {
6493                         spin_unlock_irqrestore(&q->lock, flags);
6494                         continue;
6495                 }
6496                 set_current_state(TASK_INTERRUPTIBLE);
6497                 spin_unlock_irqrestore(&q->lock, flags);
6498
6499                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6500         }
6501
6502         finish_wait(q, &wait);
6503         return status;
6504 }
6505 #else /* !CONFIG_NFS_V4_1 */
6506 static inline int
6507 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6508 {
6509         return nfs4_retry_setlk_simple(state, cmd, request);
6510 }
6511 #endif
6512
6513 static int
6514 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6515 {
6516         struct nfs_open_context *ctx;
6517         struct nfs4_state *state;
6518         int status;
6519
6520         /* verify open state */
6521         ctx = nfs_file_open_context(filp);
6522         state = ctx->state;
6523
6524         if (IS_GETLK(cmd)) {
6525                 if (state != NULL)
6526                         return nfs4_proc_getlk(state, F_GETLK, request);
6527                 return 0;
6528         }
6529
6530         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6531                 return -EINVAL;
6532
6533         if (request->fl_type == F_UNLCK) {
6534                 if (state != NULL)
6535                         return nfs4_proc_unlck(state, cmd, request);
6536                 return 0;
6537         }
6538
6539         if (state == NULL)
6540                 return -ENOLCK;
6541
6542         if ((request->fl_flags & FL_POSIX) &&
6543             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6544                 return -ENOLCK;
6545
6546         status = nfs4_set_lock_state(state, request);
6547         if (status != 0)
6548                 return status;
6549
6550         return nfs4_retry_setlk(state, cmd, request);
6551 }
6552
6553 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6554 {
6555         struct nfs_server *server = NFS_SERVER(state->inode);
6556         int err;
6557
6558         err = nfs4_set_lock_state(state, fl);
6559         if (err != 0)
6560                 return err;
6561         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6562         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6563 }
6564
6565 struct nfs_release_lockowner_data {
6566         struct nfs4_lock_state *lsp;
6567         struct nfs_server *server;
6568         struct nfs_release_lockowner_args args;
6569         struct nfs_release_lockowner_res res;
6570         unsigned long timestamp;
6571 };
6572
6573 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6574 {
6575         struct nfs_release_lockowner_data *data = calldata;
6576         struct nfs_server *server = data->server;
6577         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
6578                            &data->res.seq_res, task);
6579         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6580         data->timestamp = jiffies;
6581 }
6582
6583 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6584 {
6585         struct nfs_release_lockowner_data *data = calldata;
6586         struct nfs_server *server = data->server;
6587
6588         nfs40_sequence_done(task, &data->res.seq_res);
6589
6590         switch (task->tk_status) {
6591         case 0:
6592                 renew_lease(server, data->timestamp);
6593                 break;
6594         case -NFS4ERR_STALE_CLIENTID:
6595         case -NFS4ERR_EXPIRED:
6596                 nfs4_schedule_lease_recovery(server->nfs_client);
6597                 break;
6598         case -NFS4ERR_LEASE_MOVED:
6599         case -NFS4ERR_DELAY:
6600                 if (nfs4_async_handle_error(task, server,
6601                                             NULL, NULL) == -EAGAIN)
6602                         rpc_restart_call_prepare(task);
6603         }
6604 }
6605
6606 static void nfs4_release_lockowner_release(void *calldata)
6607 {
6608         struct nfs_release_lockowner_data *data = calldata;
6609         nfs4_free_lock_state(data->server, data->lsp);
6610         kfree(calldata);
6611 }
6612
6613 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6614         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6615         .rpc_call_done = nfs4_release_lockowner_done,
6616         .rpc_release = nfs4_release_lockowner_release,
6617 };
6618
6619 static void
6620 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6621 {
6622         struct nfs_release_lockowner_data *data;
6623         struct rpc_message msg = {
6624                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6625         };
6626
6627         if (server->nfs_client->cl_mvops->minor_version != 0)
6628                 return;
6629
6630         data = kmalloc(sizeof(*data), GFP_NOFS);
6631         if (!data)
6632                 return;
6633         data->lsp = lsp;
6634         data->server = server;
6635         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6636         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6637         data->args.lock_owner.s_dev = server->s_dev;
6638
6639         msg.rpc_argp = &data->args;
6640         msg.rpc_resp = &data->res;
6641         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6642         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6643 }
6644
6645 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6646
6647 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6648                                    struct dentry *unused, struct inode *inode,
6649                                    const char *key, const void *buf,
6650                                    size_t buflen, int flags)
6651 {
6652         return nfs4_proc_set_acl(inode, buf, buflen);
6653 }
6654
6655 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6656                                    struct dentry *unused, struct inode *inode,
6657                                    const char *key, void *buf, size_t buflen)
6658 {
6659         return nfs4_proc_get_acl(inode, buf, buflen);
6660 }
6661
6662 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6663 {
6664         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6665 }
6666
6667 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6668
6669 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6670                                      struct dentry *unused, struct inode *inode,
6671                                      const char *key, const void *buf,
6672                                      size_t buflen, int flags)
6673 {
6674         if (security_ismaclabel(key))
6675                 return nfs4_set_security_label(inode, buf, buflen);
6676
6677         return -EOPNOTSUPP;
6678 }
6679
6680 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6681                                      struct dentry *unused, struct inode *inode,
6682                                      const char *key, void *buf, size_t buflen)
6683 {
6684         if (security_ismaclabel(key))
6685                 return nfs4_get_security_label(inode, buf, buflen);
6686         return -EOPNOTSUPP;
6687 }
6688
6689 static ssize_t
6690 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6691 {
6692         int len = 0;
6693
6694         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6695                 len = security_inode_listsecurity(inode, list, list_len);
6696                 if (list_len && len > list_len)
6697                         return -ERANGE;
6698         }
6699         return len;
6700 }
6701
6702 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6703         .prefix = XATTR_SECURITY_PREFIX,
6704         .get    = nfs4_xattr_get_nfs4_label,
6705         .set    = nfs4_xattr_set_nfs4_label,
6706 };
6707
6708 #else
6709
6710 static ssize_t
6711 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6712 {
6713         return 0;
6714 }
6715
6716 #endif
6717
6718 /*
6719  * nfs_fhget will use either the mounted_on_fileid or the fileid
6720  */
6721 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6722 {
6723         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6724                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6725               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6726               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6727                 return;
6728
6729         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6730                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6731         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6732         fattr->nlink = 2;
6733 }
6734
6735 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6736                                    const struct qstr *name,
6737                                    struct nfs4_fs_locations *fs_locations,
6738                                    struct page *page)
6739 {
6740         struct nfs_server *server = NFS_SERVER(dir);
6741         u32 bitmask[3] = {
6742                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6743         };
6744         struct nfs4_fs_locations_arg args = {
6745                 .dir_fh = NFS_FH(dir),
6746                 .name = name,
6747                 .page = page,
6748                 .bitmask = bitmask,
6749         };
6750         struct nfs4_fs_locations_res res = {
6751                 .fs_locations = fs_locations,
6752         };
6753         struct rpc_message msg = {
6754                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6755                 .rpc_argp = &args,
6756                 .rpc_resp = &res,
6757         };
6758         int status;
6759
6760         dprintk("%s: start\n", __func__);
6761
6762         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6763          * is not supported */
6764         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6765                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6766         else
6767                 bitmask[0] |= FATTR4_WORD0_FILEID;
6768
6769         nfs_fattr_init(&fs_locations->fattr);
6770         fs_locations->server = server;
6771         fs_locations->nlocations = 0;
6772         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6773         dprintk("%s: returned status = %d\n", __func__, status);
6774         return status;
6775 }
6776
6777 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6778                            const struct qstr *name,
6779                            struct nfs4_fs_locations *fs_locations,
6780                            struct page *page)
6781 {
6782         struct nfs4_exception exception = { };
6783         int err;
6784         do {
6785                 err = _nfs4_proc_fs_locations(client, dir, name,
6786                                 fs_locations, page);
6787                 trace_nfs4_get_fs_locations(dir, name, err);
6788                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6789                                 &exception);
6790         } while (exception.retry);
6791         return err;
6792 }
6793
6794 /*
6795  * This operation also signals the server that this client is
6796  * performing migration recovery.  The server can stop returning
6797  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6798  * appended to this compound to identify the client ID which is
6799  * performing recovery.
6800  */
6801 static int _nfs40_proc_get_locations(struct inode *inode,
6802                                      struct nfs4_fs_locations *locations,
6803                                      struct page *page, struct rpc_cred *cred)
6804 {
6805         struct nfs_server *server = NFS_SERVER(inode);
6806         struct rpc_clnt *clnt = server->client;
6807         u32 bitmask[2] = {
6808                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6809         };
6810         struct nfs4_fs_locations_arg args = {
6811                 .clientid       = server->nfs_client->cl_clientid,
6812                 .fh             = NFS_FH(inode),
6813                 .page           = page,
6814                 .bitmask        = bitmask,
6815                 .migration      = 1,            /* skip LOOKUP */
6816                 .renew          = 1,            /* append RENEW */
6817         };
6818         struct nfs4_fs_locations_res res = {
6819                 .fs_locations   = locations,
6820                 .migration      = 1,
6821                 .renew          = 1,
6822         };
6823         struct rpc_message msg = {
6824                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6825                 .rpc_argp       = &args,
6826                 .rpc_resp       = &res,
6827                 .rpc_cred       = cred,
6828         };
6829         unsigned long now = jiffies;
6830         int status;
6831
6832         nfs_fattr_init(&locations->fattr);
6833         locations->server = server;
6834         locations->nlocations = 0;
6835
6836         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6837         nfs4_set_sequence_privileged(&args.seq_args);
6838         status = nfs4_call_sync_sequence(clnt, server, &msg,
6839                                         &args.seq_args, &res.seq_res);
6840         if (status)
6841                 return status;
6842
6843         renew_lease(server, now);
6844         return 0;
6845 }
6846
6847 #ifdef CONFIG_NFS_V4_1
6848
6849 /*
6850  * This operation also signals the server that this client is
6851  * performing migration recovery.  The server can stop asserting
6852  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6853  * performing this operation is identified in the SEQUENCE
6854  * operation in this compound.
6855  *
6856  * When the client supports GETATTR(fs_locations_info), it can
6857  * be plumbed in here.
6858  */
6859 static int _nfs41_proc_get_locations(struct inode *inode,
6860                                      struct nfs4_fs_locations *locations,
6861                                      struct page *page, struct rpc_cred *cred)
6862 {
6863         struct nfs_server *server = NFS_SERVER(inode);
6864         struct rpc_clnt *clnt = server->client;
6865         u32 bitmask[2] = {
6866                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6867         };
6868         struct nfs4_fs_locations_arg args = {
6869                 .fh             = NFS_FH(inode),
6870                 .page           = page,
6871                 .bitmask        = bitmask,
6872                 .migration      = 1,            /* skip LOOKUP */
6873         };
6874         struct nfs4_fs_locations_res res = {
6875                 .fs_locations   = locations,
6876                 .migration      = 1,
6877         };
6878         struct rpc_message msg = {
6879                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6880                 .rpc_argp       = &args,
6881                 .rpc_resp       = &res,
6882                 .rpc_cred       = cred,
6883         };
6884         int status;
6885
6886         nfs_fattr_init(&locations->fattr);
6887         locations->server = server;
6888         locations->nlocations = 0;
6889
6890         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6891         nfs4_set_sequence_privileged(&args.seq_args);
6892         status = nfs4_call_sync_sequence(clnt, server, &msg,
6893                                         &args.seq_args, &res.seq_res);
6894         if (status == NFS4_OK &&
6895             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6896                 status = -NFS4ERR_LEASE_MOVED;
6897         return status;
6898 }
6899
6900 #endif  /* CONFIG_NFS_V4_1 */
6901
6902 /**
6903  * nfs4_proc_get_locations - discover locations for a migrated FSID
6904  * @inode: inode on FSID that is migrating
6905  * @locations: result of query
6906  * @page: buffer
6907  * @cred: credential to use for this operation
6908  *
6909  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6910  * operation failed, or a negative errno if a local error occurred.
6911  *
6912  * On success, "locations" is filled in, but if the server has
6913  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6914  * asserted.
6915  *
6916  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6917  * from this client that require migration recovery.
6918  */
6919 int nfs4_proc_get_locations(struct inode *inode,
6920                             struct nfs4_fs_locations *locations,
6921                             struct page *page, struct rpc_cred *cred)
6922 {
6923         struct nfs_server *server = NFS_SERVER(inode);
6924         struct nfs_client *clp = server->nfs_client;
6925         const struct nfs4_mig_recovery_ops *ops =
6926                                         clp->cl_mvops->mig_recovery_ops;
6927         struct nfs4_exception exception = { };
6928         int status;
6929
6930         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6931                 (unsigned long long)server->fsid.major,
6932                 (unsigned long long)server->fsid.minor,
6933                 clp->cl_hostname);
6934         nfs_display_fhandle(NFS_FH(inode), __func__);
6935
6936         do {
6937                 status = ops->get_locations(inode, locations, page, cred);
6938                 if (status != -NFS4ERR_DELAY)
6939                         break;
6940                 nfs4_handle_exception(server, status, &exception);
6941         } while (exception.retry);
6942         return status;
6943 }
6944
6945 /*
6946  * This operation also signals the server that this client is
6947  * performing "lease moved" recovery.  The server can stop
6948  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6949  * is appended to this compound to identify the client ID which is
6950  * performing recovery.
6951  */
6952 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6953 {
6954         struct nfs_server *server = NFS_SERVER(inode);
6955         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6956         struct rpc_clnt *clnt = server->client;
6957         struct nfs4_fsid_present_arg args = {
6958                 .fh             = NFS_FH(inode),
6959                 .clientid       = clp->cl_clientid,
6960                 .renew          = 1,            /* append RENEW */
6961         };
6962         struct nfs4_fsid_present_res res = {
6963                 .renew          = 1,
6964         };
6965         struct rpc_message msg = {
6966                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6967                 .rpc_argp       = &args,
6968                 .rpc_resp       = &res,
6969                 .rpc_cred       = cred,
6970         };
6971         unsigned long now = jiffies;
6972         int status;
6973
6974         res.fh = nfs_alloc_fhandle();
6975         if (res.fh == NULL)
6976                 return -ENOMEM;
6977
6978         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6979         nfs4_set_sequence_privileged(&args.seq_args);
6980         status = nfs4_call_sync_sequence(clnt, server, &msg,
6981                                                 &args.seq_args, &res.seq_res);
6982         nfs_free_fhandle(res.fh);
6983         if (status)
6984                 return status;
6985
6986         do_renew_lease(clp, now);
6987         return 0;
6988 }
6989
6990 #ifdef CONFIG_NFS_V4_1
6991
6992 /*
6993  * This operation also signals the server that this client is
6994  * performing "lease moved" recovery.  The server can stop asserting
6995  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6996  * this operation is identified in the SEQUENCE operation in this
6997  * compound.
6998  */
6999 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7000 {
7001         struct nfs_server *server = NFS_SERVER(inode);
7002         struct rpc_clnt *clnt = server->client;
7003         struct nfs4_fsid_present_arg args = {
7004                 .fh             = NFS_FH(inode),
7005         };
7006         struct nfs4_fsid_present_res res = {
7007         };
7008         struct rpc_message msg = {
7009                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7010                 .rpc_argp       = &args,
7011                 .rpc_resp       = &res,
7012                 .rpc_cred       = cred,
7013         };
7014         int status;
7015
7016         res.fh = nfs_alloc_fhandle();
7017         if (res.fh == NULL)
7018                 return -ENOMEM;
7019
7020         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7021         nfs4_set_sequence_privileged(&args.seq_args);
7022         status = nfs4_call_sync_sequence(clnt, server, &msg,
7023                                                 &args.seq_args, &res.seq_res);
7024         nfs_free_fhandle(res.fh);
7025         if (status == NFS4_OK &&
7026             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7027                 status = -NFS4ERR_LEASE_MOVED;
7028         return status;
7029 }
7030
7031 #endif  /* CONFIG_NFS_V4_1 */
7032
7033 /**
7034  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7035  * @inode: inode on FSID to check
7036  * @cred: credential to use for this operation
7037  *
7038  * Server indicates whether the FSID is present, moved, or not
7039  * recognized.  This operation is necessary to clear a LEASE_MOVED
7040  * condition for this client ID.
7041  *
7042  * Returns NFS4_OK if the FSID is present on this server,
7043  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7044  *  NFS4ERR code if some error occurred on the server, or a
7045  *  negative errno if a local failure occurred.
7046  */
7047 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7048 {
7049         struct nfs_server *server = NFS_SERVER(inode);
7050         struct nfs_client *clp = server->nfs_client;
7051         const struct nfs4_mig_recovery_ops *ops =
7052                                         clp->cl_mvops->mig_recovery_ops;
7053         struct nfs4_exception exception = { };
7054         int status;
7055
7056         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7057                 (unsigned long long)server->fsid.major,
7058                 (unsigned long long)server->fsid.minor,
7059                 clp->cl_hostname);
7060         nfs_display_fhandle(NFS_FH(inode), __func__);
7061
7062         do {
7063                 status = ops->fsid_present(inode, cred);
7064                 if (status != -NFS4ERR_DELAY)
7065                         break;
7066                 nfs4_handle_exception(server, status, &exception);
7067         } while (exception.retry);
7068         return status;
7069 }
7070
7071 /**
7072  * If 'use_integrity' is true and the state managment nfs_client
7073  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7074  * and the machine credential as per RFC3530bis and RFC5661 Security
7075  * Considerations sections. Otherwise, just use the user cred with the
7076  * filesystem's rpc_client.
7077  */
7078 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7079 {
7080         int status;
7081         struct nfs4_secinfo_arg args = {
7082                 .dir_fh = NFS_FH(dir),
7083                 .name   = name,
7084         };
7085         struct nfs4_secinfo_res res = {
7086                 .flavors     = flavors,
7087         };
7088         struct rpc_message msg = {
7089                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7090                 .rpc_argp = &args,
7091                 .rpc_resp = &res,
7092         };
7093         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7094         struct rpc_cred *cred = NULL;
7095
7096         if (use_integrity) {
7097                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7098                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7099                 msg.rpc_cred = cred;
7100         }
7101
7102         dprintk("NFS call  secinfo %s\n", name->name);
7103
7104         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7105                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7106
7107         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7108                                 &res.seq_res, 0);
7109         dprintk("NFS reply  secinfo: %d\n", status);
7110
7111         if (cred)
7112                 put_rpccred(cred);
7113
7114         return status;
7115 }
7116
7117 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7118                       struct nfs4_secinfo_flavors *flavors)
7119 {
7120         struct nfs4_exception exception = { };
7121         int err;
7122         do {
7123                 err = -NFS4ERR_WRONGSEC;
7124
7125                 /* try to use integrity protection with machine cred */
7126                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7127                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7128
7129                 /*
7130                  * if unable to use integrity protection, or SECINFO with
7131                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7132                  * disallowed by spec, but exists in deployed servers) use
7133                  * the current filesystem's rpc_client and the user cred.
7134                  */
7135                 if (err == -NFS4ERR_WRONGSEC)
7136                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7137
7138                 trace_nfs4_secinfo(dir, name, err);
7139                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7140                                 &exception);
7141         } while (exception.retry);
7142         return err;
7143 }
7144
7145 #ifdef CONFIG_NFS_V4_1
7146 /*
7147  * Check the exchange flags returned by the server for invalid flags, having
7148  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7149  * DS flags set.
7150  */
7151 static int nfs4_check_cl_exchange_flags(u32 flags)
7152 {
7153         if (flags & ~EXCHGID4_FLAG_MASK_R)
7154                 goto out_inval;
7155         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7156             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7157                 goto out_inval;
7158         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7159                 goto out_inval;
7160         return NFS_OK;
7161 out_inval:
7162         return -NFS4ERR_INVAL;
7163 }
7164
7165 static bool
7166 nfs41_same_server_scope(struct nfs41_server_scope *a,
7167                         struct nfs41_server_scope *b)
7168 {
7169         if (a->server_scope_sz != b->server_scope_sz)
7170                 return false;
7171         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7172 }
7173
7174 static void
7175 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7176 {
7177 }
7178
7179 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7180         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7181 };
7182
7183 /*
7184  * nfs4_proc_bind_one_conn_to_session()
7185  *
7186  * The 4.1 client currently uses the same TCP connection for the
7187  * fore and backchannel.
7188  */
7189 static
7190 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7191                 struct rpc_xprt *xprt,
7192                 struct nfs_client *clp,
7193                 struct rpc_cred *cred)
7194 {
7195         int status;
7196         struct nfs41_bind_conn_to_session_args args = {
7197                 .client = clp,
7198                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7199         };
7200         struct nfs41_bind_conn_to_session_res res;
7201         struct rpc_message msg = {
7202                 .rpc_proc =
7203                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7204                 .rpc_argp = &args,
7205                 .rpc_resp = &res,
7206                 .rpc_cred = cred,
7207         };
7208         struct rpc_task_setup task_setup_data = {
7209                 .rpc_client = clnt,
7210                 .rpc_xprt = xprt,
7211                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7212                 .rpc_message = &msg,
7213                 .flags = RPC_TASK_TIMEOUT,
7214         };
7215         struct rpc_task *task;
7216
7217         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7218         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7219                 args.dir = NFS4_CDFC4_FORE;
7220
7221         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7222         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7223                 args.dir = NFS4_CDFC4_FORE;
7224
7225         task = rpc_run_task(&task_setup_data);
7226         if (!IS_ERR(task)) {
7227                 status = task->tk_status;
7228                 rpc_put_task(task);
7229         } else
7230                 status = PTR_ERR(task);
7231         trace_nfs4_bind_conn_to_session(clp, status);
7232         if (status == 0) {
7233                 if (memcmp(res.sessionid.data,
7234                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7235                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7236                         return -EIO;
7237                 }
7238                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7239                         dprintk("NFS: %s: Unexpected direction from server\n",
7240                                 __func__);
7241                         return -EIO;
7242                 }
7243                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7244                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7245                                 __func__);
7246                         return -EIO;
7247                 }
7248         }
7249
7250         return status;
7251 }
7252
7253 struct rpc_bind_conn_calldata {
7254         struct nfs_client *clp;
7255         struct rpc_cred *cred;
7256 };
7257
7258 static int
7259 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7260                 struct rpc_xprt *xprt,
7261                 void *calldata)
7262 {
7263         struct rpc_bind_conn_calldata *p = calldata;
7264
7265         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7266 }
7267
7268 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7269 {
7270         struct rpc_bind_conn_calldata data = {
7271                 .clp = clp,
7272                 .cred = cred,
7273         };
7274         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7275                         nfs4_proc_bind_conn_to_session_callback, &data);
7276 }
7277
7278 /*
7279  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7280  * and operations we'd like to see to enable certain features in the allow map
7281  */
7282 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7283         .how = SP4_MACH_CRED,
7284         .enforce.u.words = {
7285                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7286                       1 << (OP_EXCHANGE_ID - 32) |
7287                       1 << (OP_CREATE_SESSION - 32) |
7288                       1 << (OP_DESTROY_SESSION - 32) |
7289                       1 << (OP_DESTROY_CLIENTID - 32)
7290         },
7291         .allow.u.words = {
7292                 [0] = 1 << (OP_CLOSE) |
7293                       1 << (OP_OPEN_DOWNGRADE) |
7294                       1 << (OP_LOCKU) |
7295                       1 << (OP_DELEGRETURN) |
7296                       1 << (OP_COMMIT),
7297                 [1] = 1 << (OP_SECINFO - 32) |
7298                       1 << (OP_SECINFO_NO_NAME - 32) |
7299                       1 << (OP_LAYOUTRETURN - 32) |
7300                       1 << (OP_TEST_STATEID - 32) |
7301                       1 << (OP_FREE_STATEID - 32) |
7302                       1 << (OP_WRITE - 32)
7303         }
7304 };
7305
7306 /*
7307  * Select the state protection mode for client `clp' given the server results
7308  * from exchange_id in `sp'.
7309  *
7310  * Returns 0 on success, negative errno otherwise.
7311  */
7312 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7313                                  struct nfs41_state_protection *sp)
7314 {
7315         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7316                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7317                       1 << (OP_EXCHANGE_ID - 32) |
7318                       1 << (OP_CREATE_SESSION - 32) |
7319                       1 << (OP_DESTROY_SESSION - 32) |
7320                       1 << (OP_DESTROY_CLIENTID - 32)
7321         };
7322         unsigned int i;
7323
7324         if (sp->how == SP4_MACH_CRED) {
7325                 /* Print state protect result */
7326                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7327                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7328                         if (test_bit(i, sp->enforce.u.longs))
7329                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7330                         if (test_bit(i, sp->allow.u.longs))
7331                                 dfprintk(MOUNT, "  allow op %d\n", i);
7332                 }
7333
7334                 /* make sure nothing is on enforce list that isn't supported */
7335                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7336                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7337                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7338                                 return -EINVAL;
7339                         }
7340                 }
7341
7342                 /*
7343                  * Minimal mode - state operations are allowed to use machine
7344                  * credential.  Note this already happens by default, so the
7345                  * client doesn't have to do anything more than the negotiation.
7346                  *
7347                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7348                  *       we're already using the machine cred for exchange_id
7349                  *       and will never use a different cred.
7350                  */
7351                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7352                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7353                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7354                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7355                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7356                         dfprintk(MOUNT, "  minimal mode enabled\n");
7357                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7358                 } else {
7359                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7360                         return -EINVAL;
7361                 }
7362
7363                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7364                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7365                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7366                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7367                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7368                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7369                 }
7370
7371                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7372                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7373                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7374                                 &clp->cl_sp4_flags);
7375                 }
7376
7377                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7378                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7379                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7380                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7381                 }
7382
7383                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7384                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7385                         dfprintk(MOUNT, "  stateid mode enabled\n");
7386                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7387                 }
7388
7389                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7390                         dfprintk(MOUNT, "  write mode enabled\n");
7391                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7392                 }
7393
7394                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7395                         dfprintk(MOUNT, "  commit mode enabled\n");
7396                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7397                 }
7398         }
7399
7400         return 0;
7401 }
7402
7403 struct nfs41_exchange_id_data {
7404         struct nfs41_exchange_id_res res;
7405         struct nfs41_exchange_id_args args;
7406         struct rpc_xprt *xprt;
7407         int rpc_status;
7408 };
7409
7410 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7411 {
7412         struct nfs41_exchange_id_data *cdata =
7413                                         (struct nfs41_exchange_id_data *)data;
7414         struct nfs_client *clp = cdata->args.client;
7415         int status = task->tk_status;
7416
7417         trace_nfs4_exchange_id(clp, status);
7418
7419         if (status == 0)
7420                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7421
7422         if (cdata->xprt && status == 0) {
7423                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7424                                                       cdata->xprt);
7425                 goto out;
7426         }
7427
7428         if (status  == 0)
7429                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7430
7431         if (status == 0) {
7432                 clp->cl_clientid = cdata->res.clientid;
7433                 clp->cl_exchange_flags = cdata->res.flags;
7434                 clp->cl_seqid = cdata->res.seqid;
7435                 /* Client ID is not confirmed */
7436                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R))
7437                         clear_bit(NFS4_SESSION_ESTABLISHED,
7438                                   &clp->cl_session->session_state);
7439
7440                 kfree(clp->cl_serverowner);
7441                 clp->cl_serverowner = cdata->res.server_owner;
7442                 cdata->res.server_owner = NULL;
7443
7444                 /* use the most recent implementation id */
7445                 kfree(clp->cl_implid);
7446                 clp->cl_implid = cdata->res.impl_id;
7447                 cdata->res.impl_id = NULL;
7448
7449                 if (clp->cl_serverscope != NULL &&
7450                     !nfs41_same_server_scope(clp->cl_serverscope,
7451                                         cdata->res.server_scope)) {
7452                         dprintk("%s: server_scope mismatch detected\n",
7453                                 __func__);
7454                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7455                         kfree(clp->cl_serverscope);
7456                         clp->cl_serverscope = NULL;
7457                 }
7458
7459                 if (clp->cl_serverscope == NULL) {
7460                         clp->cl_serverscope = cdata->res.server_scope;
7461                         cdata->res.server_scope = NULL;
7462                 }
7463                 /* Save the EXCHANGE_ID verifier session trunk tests */
7464                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7465                        sizeof(clp->cl_confirm.data));
7466         }
7467 out:
7468         cdata->rpc_status = status;
7469         return;
7470 }
7471
7472 static void nfs4_exchange_id_release(void *data)
7473 {
7474         struct nfs41_exchange_id_data *cdata =
7475                                         (struct nfs41_exchange_id_data *)data;
7476
7477         if (cdata->xprt) {
7478                 xprt_put(cdata->xprt);
7479                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7480         }
7481         nfs_put_client(cdata->args.client);
7482         kfree(cdata->res.impl_id);
7483         kfree(cdata->res.server_scope);
7484         kfree(cdata->res.server_owner);
7485         kfree(cdata);
7486 }
7487
7488 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7489         .rpc_call_done = nfs4_exchange_id_done,
7490         .rpc_release = nfs4_exchange_id_release,
7491 };
7492
7493 /*
7494  * _nfs4_proc_exchange_id()
7495  *
7496  * Wrapper for EXCHANGE_ID operation.
7497  */
7498 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7499                         u32 sp4_how, struct rpc_xprt *xprt)
7500 {
7501         nfs4_verifier verifier;
7502         struct rpc_message msg = {
7503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7504                 .rpc_cred = cred,
7505         };
7506         struct rpc_task_setup task_setup_data = {
7507                 .rpc_client = clp->cl_rpcclient,
7508                 .callback_ops = &nfs4_exchange_id_call_ops,
7509                 .rpc_message = &msg,
7510                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7511         };
7512         struct nfs41_exchange_id_data *calldata;
7513         struct rpc_task *task;
7514         int status;
7515
7516         if (!atomic_inc_not_zero(&clp->cl_count))
7517                 return -EIO;
7518
7519         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7520         if (!calldata) {
7521                 nfs_put_client(clp);
7522                 return -ENOMEM;
7523         }
7524
7525         if (!xprt)
7526                 nfs4_init_boot_verifier(clp, &verifier);
7527
7528         status = nfs4_init_uniform_client_string(clp);
7529         if (status)
7530                 goto out_calldata;
7531
7532         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7533                                                 GFP_NOFS);
7534         status = -ENOMEM;
7535         if (unlikely(calldata->res.server_owner == NULL))
7536                 goto out_calldata;
7537
7538         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7539                                         GFP_NOFS);
7540         if (unlikely(calldata->res.server_scope == NULL))
7541                 goto out_server_owner;
7542
7543         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7544         if (unlikely(calldata->res.impl_id == NULL))
7545                 goto out_server_scope;
7546
7547         switch (sp4_how) {
7548         case SP4_NONE:
7549                 calldata->args.state_protect.how = SP4_NONE;
7550                 break;
7551
7552         case SP4_MACH_CRED:
7553                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7554                 break;
7555
7556         default:
7557                 /* unsupported! */
7558                 WARN_ON_ONCE(1);
7559                 status = -EINVAL;
7560                 goto out_impl_id;
7561         }
7562         if (xprt) {
7563                 calldata->xprt = xprt;
7564                 task_setup_data.rpc_xprt = xprt;
7565                 task_setup_data.flags =
7566                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7567                 calldata->args.verifier = &clp->cl_confirm;
7568         } else {
7569                 calldata->args.verifier = &verifier;
7570         }
7571         calldata->args.client = clp;
7572 #ifdef CONFIG_NFS_V4_1_MIGRATION
7573         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7574         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7575         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7576 #else
7577         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7578         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7579 #endif
7580         msg.rpc_argp = &calldata->args;
7581         msg.rpc_resp = &calldata->res;
7582         task_setup_data.callback_data = calldata;
7583
7584         task = rpc_run_task(&task_setup_data);
7585         if (IS_ERR(task))
7586                 return PTR_ERR(task);
7587
7588         if (!xprt) {
7589                 status = rpc_wait_for_completion_task(task);
7590                 if (!status)
7591                         status = calldata->rpc_status;
7592         } else  /* session trunking test */
7593                 status = calldata->rpc_status;
7594
7595         rpc_put_task(task);
7596 out:
7597         return status;
7598
7599 out_impl_id:
7600         kfree(calldata->res.impl_id);
7601 out_server_scope:
7602         kfree(calldata->res.server_scope);
7603 out_server_owner:
7604         kfree(calldata->res.server_owner);
7605 out_calldata:
7606         kfree(calldata);
7607         nfs_put_client(clp);
7608         goto out;
7609 }
7610
7611 /*
7612  * nfs4_proc_exchange_id()
7613  *
7614  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7615  *
7616  * Since the clientid has expired, all compounds using sessions
7617  * associated with the stale clientid will be returning
7618  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7619  * be in some phase of session reset.
7620  *
7621  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7622  */
7623 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7624 {
7625         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7626         int status;
7627
7628         /* try SP4_MACH_CRED if krb5i/p */
7629         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7630             authflavor == RPC_AUTH_GSS_KRB5P) {
7631                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7632                 if (!status)
7633                         return 0;
7634         }
7635
7636         /* try SP4_NONE */
7637         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7638 }
7639
7640 /**
7641  * nfs4_test_session_trunk
7642  *
7643  * This is an add_xprt_test() test function called from
7644  * rpc_clnt_setup_test_and_add_xprt.
7645  *
7646  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7647  * and is dereferrenced in nfs4_exchange_id_release
7648  *
7649  * Upon success, add the new transport to the rpc_clnt
7650  *
7651  * @clnt: struct rpc_clnt to get new transport
7652  * @xprt: the rpc_xprt to test
7653  * @data: call data for _nfs4_proc_exchange_id.
7654  */
7655 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7656                             void *data)
7657 {
7658         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7659         u32 sp4_how;
7660
7661         dprintk("--> %s try %s\n", __func__,
7662                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7663
7664         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7665
7666         /* Test connection for session trunking. Async exchange_id call */
7667         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7668 }
7669 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7670
7671 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7672                 struct rpc_cred *cred)
7673 {
7674         struct rpc_message msg = {
7675                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7676                 .rpc_argp = clp,
7677                 .rpc_cred = cred,
7678         };
7679         int status;
7680
7681         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7682         trace_nfs4_destroy_clientid(clp, status);
7683         if (status)
7684                 dprintk("NFS: Got error %d from the server %s on "
7685                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7686         return status;
7687 }
7688
7689 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7690                 struct rpc_cred *cred)
7691 {
7692         unsigned int loop;
7693         int ret;
7694
7695         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7696                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7697                 switch (ret) {
7698                 case -NFS4ERR_DELAY:
7699                 case -NFS4ERR_CLIENTID_BUSY:
7700                         ssleep(1);
7701                         break;
7702                 default:
7703                         return ret;
7704                 }
7705         }
7706         return 0;
7707 }
7708
7709 int nfs4_destroy_clientid(struct nfs_client *clp)
7710 {
7711         struct rpc_cred *cred;
7712         int ret = 0;
7713
7714         if (clp->cl_mvops->minor_version < 1)
7715                 goto out;
7716         if (clp->cl_exchange_flags == 0)
7717                 goto out;
7718         if (clp->cl_preserve_clid)
7719                 goto out;
7720         cred = nfs4_get_clid_cred(clp);
7721         ret = nfs4_proc_destroy_clientid(clp, cred);
7722         if (cred)
7723                 put_rpccred(cred);
7724         switch (ret) {
7725         case 0:
7726         case -NFS4ERR_STALE_CLIENTID:
7727                 clp->cl_exchange_flags = 0;
7728         }
7729 out:
7730         return ret;
7731 }
7732
7733 struct nfs4_get_lease_time_data {
7734         struct nfs4_get_lease_time_args *args;
7735         struct nfs4_get_lease_time_res *res;
7736         struct nfs_client *clp;
7737 };
7738
7739 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7740                                         void *calldata)
7741 {
7742         struct nfs4_get_lease_time_data *data =
7743                         (struct nfs4_get_lease_time_data *)calldata;
7744
7745         dprintk("--> %s\n", __func__);
7746         /* just setup sequence, do not trigger session recovery
7747            since we're invoked within one */
7748         nfs4_setup_sequence(data->clp,
7749                         &data->args->la_seq_args,
7750                         &data->res->lr_seq_res,
7751                         task);
7752         dprintk("<-- %s\n", __func__);
7753 }
7754
7755 /*
7756  * Called from nfs4_state_manager thread for session setup, so don't recover
7757  * from sequence operation or clientid errors.
7758  */
7759 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7760 {
7761         struct nfs4_get_lease_time_data *data =
7762                         (struct nfs4_get_lease_time_data *)calldata;
7763
7764         dprintk("--> %s\n", __func__);
7765         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7766                 return;
7767         switch (task->tk_status) {
7768         case -NFS4ERR_DELAY:
7769         case -NFS4ERR_GRACE:
7770                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7771                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7772                 task->tk_status = 0;
7773                 /* fall through */
7774         case -NFS4ERR_RETRY_UNCACHED_REP:
7775                 rpc_restart_call_prepare(task);
7776                 return;
7777         }
7778         dprintk("<-- %s\n", __func__);
7779 }
7780
7781 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7782         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7783         .rpc_call_done = nfs4_get_lease_time_done,
7784 };
7785
7786 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7787 {
7788         struct rpc_task *task;
7789         struct nfs4_get_lease_time_args args;
7790         struct nfs4_get_lease_time_res res = {
7791                 .lr_fsinfo = fsinfo,
7792         };
7793         struct nfs4_get_lease_time_data data = {
7794                 .args = &args,
7795                 .res = &res,
7796                 .clp = clp,
7797         };
7798         struct rpc_message msg = {
7799                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7800                 .rpc_argp = &args,
7801                 .rpc_resp = &res,
7802         };
7803         struct rpc_task_setup task_setup = {
7804                 .rpc_client = clp->cl_rpcclient,
7805                 .rpc_message = &msg,
7806                 .callback_ops = &nfs4_get_lease_time_ops,
7807                 .callback_data = &data,
7808                 .flags = RPC_TASK_TIMEOUT,
7809         };
7810         int status;
7811
7812         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7813         nfs4_set_sequence_privileged(&args.la_seq_args);
7814         task = rpc_run_task(&task_setup);
7815
7816         if (IS_ERR(task))
7817                 return PTR_ERR(task);
7818
7819         status = task->tk_status;
7820         rpc_put_task(task);
7821         return status;
7822 }
7823
7824 /*
7825  * Initialize the values to be used by the client in CREATE_SESSION
7826  * If nfs4_init_session set the fore channel request and response sizes,
7827  * use them.
7828  *
7829  * Set the back channel max_resp_sz_cached to zero to force the client to
7830  * always set csa_cachethis to FALSE because the current implementation
7831  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7832  */
7833 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7834                                     struct rpc_clnt *clnt)
7835 {
7836         unsigned int max_rqst_sz, max_resp_sz;
7837         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7838
7839         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7840         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7841
7842         /* Fore channel attributes */
7843         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7844         args->fc_attrs.max_resp_sz = max_resp_sz;
7845         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7846         args->fc_attrs.max_reqs = max_session_slots;
7847
7848         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7849                 "max_ops=%u max_reqs=%u\n",
7850                 __func__,
7851                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7852                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7853
7854         /* Back channel attributes */
7855         args->bc_attrs.max_rqst_sz = max_bc_payload;
7856         args->bc_attrs.max_resp_sz = max_bc_payload;
7857         args->bc_attrs.max_resp_sz_cached = 0;
7858         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7859         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7860
7861         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7862                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7863                 __func__,
7864                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7865                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7866                 args->bc_attrs.max_reqs);
7867 }
7868
7869 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7870                 struct nfs41_create_session_res *res)
7871 {
7872         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7873         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7874
7875         if (rcvd->max_resp_sz > sent->max_resp_sz)
7876                 return -EINVAL;
7877         /*
7878          * Our requested max_ops is the minimum we need; we're not
7879          * prepared to break up compounds into smaller pieces than that.
7880          * So, no point even trying to continue if the server won't
7881          * cooperate:
7882          */
7883         if (rcvd->max_ops < sent->max_ops)
7884                 return -EINVAL;
7885         if (rcvd->max_reqs == 0)
7886                 return -EINVAL;
7887         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7888                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7889         return 0;
7890 }
7891
7892 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7893                 struct nfs41_create_session_res *res)
7894 {
7895         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7896         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7897
7898         if (!(res->flags & SESSION4_BACK_CHAN))
7899                 goto out;
7900         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7901                 return -EINVAL;
7902         if (rcvd->max_resp_sz < sent->max_resp_sz)
7903                 return -EINVAL;
7904         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7905                 return -EINVAL;
7906         if (rcvd->max_ops > sent->max_ops)
7907                 return -EINVAL;
7908         if (rcvd->max_reqs > sent->max_reqs)
7909                 return -EINVAL;
7910 out:
7911         return 0;
7912 }
7913
7914 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7915                                      struct nfs41_create_session_res *res)
7916 {
7917         int ret;
7918
7919         ret = nfs4_verify_fore_channel_attrs(args, res);
7920         if (ret)
7921                 return ret;
7922         return nfs4_verify_back_channel_attrs(args, res);
7923 }
7924
7925 static void nfs4_update_session(struct nfs4_session *session,
7926                 struct nfs41_create_session_res *res)
7927 {
7928         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7929         /* Mark client id and session as being confirmed */
7930         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7931         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7932         session->flags = res->flags;
7933         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7934         if (res->flags & SESSION4_BACK_CHAN)
7935                 memcpy(&session->bc_attrs, &res->bc_attrs,
7936                                 sizeof(session->bc_attrs));
7937 }
7938
7939 static int _nfs4_proc_create_session(struct nfs_client *clp,
7940                 struct rpc_cred *cred)
7941 {
7942         struct nfs4_session *session = clp->cl_session;
7943         struct nfs41_create_session_args args = {
7944                 .client = clp,
7945                 .clientid = clp->cl_clientid,
7946                 .seqid = clp->cl_seqid,
7947                 .cb_program = NFS4_CALLBACK,
7948         };
7949         struct nfs41_create_session_res res;
7950
7951         struct rpc_message msg = {
7952                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7953                 .rpc_argp = &args,
7954                 .rpc_resp = &res,
7955                 .rpc_cred = cred,
7956         };
7957         int status;
7958
7959         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7960         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7961
7962         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7963         trace_nfs4_create_session(clp, status);
7964
7965         switch (status) {
7966         case -NFS4ERR_STALE_CLIENTID:
7967         case -NFS4ERR_DELAY:
7968         case -ETIMEDOUT:
7969         case -EACCES:
7970         case -EAGAIN:
7971                 goto out;
7972         };
7973
7974         clp->cl_seqid++;
7975         if (!status) {
7976                 /* Verify the session's negotiated channel_attrs values */
7977                 status = nfs4_verify_channel_attrs(&args, &res);
7978                 /* Increment the clientid slot sequence id */
7979                 if (status)
7980                         goto out;
7981                 nfs4_update_session(session, &res);
7982         }
7983 out:
7984         return status;
7985 }
7986
7987 /*
7988  * Issues a CREATE_SESSION operation to the server.
7989  * It is the responsibility of the caller to verify the session is
7990  * expired before calling this routine.
7991  */
7992 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7993 {
7994         int status;
7995         unsigned *ptr;
7996         struct nfs4_session *session = clp->cl_session;
7997
7998         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7999
8000         status = _nfs4_proc_create_session(clp, cred);
8001         if (status)
8002                 goto out;
8003
8004         /* Init or reset the session slot tables */
8005         status = nfs4_setup_session_slot_tables(session);
8006         dprintk("slot table setup returned %d\n", status);
8007         if (status)
8008                 goto out;
8009
8010         ptr = (unsigned *)&session->sess_id.data[0];
8011         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8012                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8013 out:
8014         dprintk("<-- %s\n", __func__);
8015         return status;
8016 }
8017
8018 /*
8019  * Issue the over-the-wire RPC DESTROY_SESSION.
8020  * The caller must serialize access to this routine.
8021  */
8022 int nfs4_proc_destroy_session(struct nfs4_session *session,
8023                 struct rpc_cred *cred)
8024 {
8025         struct rpc_message msg = {
8026                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8027                 .rpc_argp = session,
8028                 .rpc_cred = cred,
8029         };
8030         int status = 0;
8031
8032         dprintk("--> nfs4_proc_destroy_session\n");
8033
8034         /* session is still being setup */
8035         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8036                 return 0;
8037
8038         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8039         trace_nfs4_destroy_session(session->clp, status);
8040
8041         if (status)
8042                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8043                         "Session has been destroyed regardless...\n", status);
8044
8045         dprintk("<-- nfs4_proc_destroy_session\n");
8046         return status;
8047 }
8048
8049 /*
8050  * Renew the cl_session lease.
8051  */
8052 struct nfs4_sequence_data {
8053         struct nfs_client *clp;
8054         struct nfs4_sequence_args args;
8055         struct nfs4_sequence_res res;
8056 };
8057
8058 static void nfs41_sequence_release(void *data)
8059 {
8060         struct nfs4_sequence_data *calldata = data;
8061         struct nfs_client *clp = calldata->clp;
8062
8063         if (atomic_read(&clp->cl_count) > 1)
8064                 nfs4_schedule_state_renewal(clp);
8065         nfs_put_client(clp);
8066         kfree(calldata);
8067 }
8068
8069 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8070 {
8071         switch(task->tk_status) {
8072         case -NFS4ERR_DELAY:
8073                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8074                 return -EAGAIN;
8075         default:
8076                 nfs4_schedule_lease_recovery(clp);
8077         }
8078         return 0;
8079 }
8080
8081 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8082 {
8083         struct nfs4_sequence_data *calldata = data;
8084         struct nfs_client *clp = calldata->clp;
8085
8086         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8087                 return;
8088
8089         trace_nfs4_sequence(clp, task->tk_status);
8090         if (task->tk_status < 0) {
8091                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8092                 if (atomic_read(&clp->cl_count) == 1)
8093                         goto out;
8094
8095                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8096                         rpc_restart_call_prepare(task);
8097                         return;
8098                 }
8099         }
8100         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8101 out:
8102         dprintk("<-- %s\n", __func__);
8103 }
8104
8105 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8106 {
8107         struct nfs4_sequence_data *calldata = data;
8108         struct nfs_client *clp = calldata->clp;
8109         struct nfs4_sequence_args *args;
8110         struct nfs4_sequence_res *res;
8111
8112         args = task->tk_msg.rpc_argp;
8113         res = task->tk_msg.rpc_resp;
8114
8115         nfs4_setup_sequence(clp, args, res, task);
8116 }
8117
8118 static const struct rpc_call_ops nfs41_sequence_ops = {
8119         .rpc_call_done = nfs41_sequence_call_done,
8120         .rpc_call_prepare = nfs41_sequence_prepare,
8121         .rpc_release = nfs41_sequence_release,
8122 };
8123
8124 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8125                 struct rpc_cred *cred,
8126                 bool is_privileged)
8127 {
8128         struct nfs4_sequence_data *calldata;
8129         struct rpc_message msg = {
8130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8131                 .rpc_cred = cred,
8132         };
8133         struct rpc_task_setup task_setup_data = {
8134                 .rpc_client = clp->cl_rpcclient,
8135                 .rpc_message = &msg,
8136                 .callback_ops = &nfs41_sequence_ops,
8137                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8138         };
8139
8140         if (!atomic_inc_not_zero(&clp->cl_count))
8141                 return ERR_PTR(-EIO);
8142         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8143         if (calldata == NULL) {
8144                 nfs_put_client(clp);
8145                 return ERR_PTR(-ENOMEM);
8146         }
8147         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8148         if (is_privileged)
8149                 nfs4_set_sequence_privileged(&calldata->args);
8150         msg.rpc_argp = &calldata->args;
8151         msg.rpc_resp = &calldata->res;
8152         calldata->clp = clp;
8153         task_setup_data.callback_data = calldata;
8154
8155         return rpc_run_task(&task_setup_data);
8156 }
8157
8158 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8159 {
8160         struct rpc_task *task;
8161         int ret = 0;
8162
8163         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8164                 return -EAGAIN;
8165         task = _nfs41_proc_sequence(clp, cred, false);
8166         if (IS_ERR(task))
8167                 ret = PTR_ERR(task);
8168         else
8169                 rpc_put_task_async(task);
8170         dprintk("<-- %s status=%d\n", __func__, ret);
8171         return ret;
8172 }
8173
8174 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8175 {
8176         struct rpc_task *task;
8177         int ret;
8178
8179         task = _nfs41_proc_sequence(clp, cred, true);
8180         if (IS_ERR(task)) {
8181                 ret = PTR_ERR(task);
8182                 goto out;
8183         }
8184         ret = rpc_wait_for_completion_task(task);
8185         if (!ret)
8186                 ret = task->tk_status;
8187         rpc_put_task(task);
8188 out:
8189         dprintk("<-- %s status=%d\n", __func__, ret);
8190         return ret;
8191 }
8192
8193 struct nfs4_reclaim_complete_data {
8194         struct nfs_client *clp;
8195         struct nfs41_reclaim_complete_args arg;
8196         struct nfs41_reclaim_complete_res res;
8197 };
8198
8199 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8200 {
8201         struct nfs4_reclaim_complete_data *calldata = data;
8202
8203         nfs4_setup_sequence(calldata->clp,
8204                         &calldata->arg.seq_args,
8205                         &calldata->res.seq_res,
8206                         task);
8207 }
8208
8209 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8210 {
8211         switch(task->tk_status) {
8212         case 0:
8213         case -NFS4ERR_COMPLETE_ALREADY:
8214         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8215                 break;
8216         case -NFS4ERR_DELAY:
8217                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8218                 /* fall through */
8219         case -NFS4ERR_RETRY_UNCACHED_REP:
8220                 return -EAGAIN;
8221         case -NFS4ERR_BADSESSION:
8222         case -NFS4ERR_DEADSESSION:
8223         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8224                 nfs4_schedule_session_recovery(clp->cl_session,
8225                                 task->tk_status);
8226                 break;
8227         default:
8228                 nfs4_schedule_lease_recovery(clp);
8229         }
8230         return 0;
8231 }
8232
8233 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8234 {
8235         struct nfs4_reclaim_complete_data *calldata = data;
8236         struct nfs_client *clp = calldata->clp;
8237         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8238
8239         dprintk("--> %s\n", __func__);
8240         if (!nfs41_sequence_done(task, res))
8241                 return;
8242
8243         trace_nfs4_reclaim_complete(clp, task->tk_status);
8244         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8245                 rpc_restart_call_prepare(task);
8246                 return;
8247         }
8248         dprintk("<-- %s\n", __func__);
8249 }
8250
8251 static void nfs4_free_reclaim_complete_data(void *data)
8252 {
8253         struct nfs4_reclaim_complete_data *calldata = data;
8254
8255         kfree(calldata);
8256 }
8257
8258 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8259         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8260         .rpc_call_done = nfs4_reclaim_complete_done,
8261         .rpc_release = nfs4_free_reclaim_complete_data,
8262 };
8263
8264 /*
8265  * Issue a global reclaim complete.
8266  */
8267 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8268                 struct rpc_cred *cred)
8269 {
8270         struct nfs4_reclaim_complete_data *calldata;
8271         struct rpc_task *task;
8272         struct rpc_message msg = {
8273                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8274                 .rpc_cred = cred,
8275         };
8276         struct rpc_task_setup task_setup_data = {
8277                 .rpc_client = clp->cl_rpcclient,
8278                 .rpc_message = &msg,
8279                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8280                 .flags = RPC_TASK_ASYNC,
8281         };
8282         int status = -ENOMEM;
8283
8284         dprintk("--> %s\n", __func__);
8285         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8286         if (calldata == NULL)
8287                 goto out;
8288         calldata->clp = clp;
8289         calldata->arg.one_fs = 0;
8290
8291         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8292         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8293         msg.rpc_argp = &calldata->arg;
8294         msg.rpc_resp = &calldata->res;
8295         task_setup_data.callback_data = calldata;
8296         task = rpc_run_task(&task_setup_data);
8297         if (IS_ERR(task)) {
8298                 status = PTR_ERR(task);
8299                 goto out;
8300         }
8301         status = rpc_wait_for_completion_task(task);
8302         if (status == 0)
8303                 status = task->tk_status;
8304         rpc_put_task(task);
8305 out:
8306         dprintk("<-- %s status=%d\n", __func__, status);
8307         return status;
8308 }
8309
8310 static void
8311 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8312 {
8313         struct nfs4_layoutget *lgp = calldata;
8314         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8315
8316         dprintk("--> %s\n", __func__);
8317         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8318                                 &lgp->res.seq_res, task);
8319         dprintk("<-- %s\n", __func__);
8320 }
8321
8322 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8323 {
8324         struct nfs4_layoutget *lgp = calldata;
8325
8326         dprintk("--> %s\n", __func__);
8327         nfs41_sequence_process(task, &lgp->res.seq_res);
8328         dprintk("<-- %s\n", __func__);
8329 }
8330
8331 static int
8332 nfs4_layoutget_handle_exception(struct rpc_task *task,
8333                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8334 {
8335         struct inode *inode = lgp->args.inode;
8336         struct nfs_server *server = NFS_SERVER(inode);
8337         struct pnfs_layout_hdr *lo;
8338         int nfs4err = task->tk_status;
8339         int err, status = 0;
8340         LIST_HEAD(head);
8341
8342         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8343
8344         switch (nfs4err) {
8345         case 0:
8346                 goto out;
8347
8348         /*
8349          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8350          * on the file. set tk_status to -ENODATA to tell upper layer to
8351          * retry go inband.
8352          */
8353         case -NFS4ERR_LAYOUTUNAVAILABLE:
8354                 status = -ENODATA;
8355                 goto out;
8356         /*
8357          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8358          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8359          */
8360         case -NFS4ERR_BADLAYOUT:
8361                 status = -EOVERFLOW;
8362                 goto out;
8363         /*
8364          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8365          * (or clients) writing to the same RAID stripe except when
8366          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8367          *
8368          * Treat it like we would RECALLCONFLICT -- we retry for a little
8369          * while, and then eventually give up.
8370          */
8371         case -NFS4ERR_LAYOUTTRYLATER:
8372                 if (lgp->args.minlength == 0) {
8373                         status = -EOVERFLOW;
8374                         goto out;
8375                 }
8376                 status = -EBUSY;
8377                 break;
8378         case -NFS4ERR_RECALLCONFLICT:
8379                 status = -ERECALLCONFLICT;
8380                 break;
8381         case -NFS4ERR_DELEG_REVOKED:
8382         case -NFS4ERR_ADMIN_REVOKED:
8383         case -NFS4ERR_EXPIRED:
8384         case -NFS4ERR_BAD_STATEID:
8385                 exception->timeout = 0;
8386                 spin_lock(&inode->i_lock);
8387                 lo = NFS_I(inode)->layout;
8388                 /* If the open stateid was bad, then recover it. */
8389                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8390                     nfs4_stateid_match_other(&lgp->args.stateid,
8391                                         &lgp->args.ctx->state->stateid)) {
8392                         spin_unlock(&inode->i_lock);
8393                         exception->state = lgp->args.ctx->state;
8394                         exception->stateid = &lgp->args.stateid;
8395                         break;
8396                 }
8397
8398                 /*
8399                  * Mark the bad layout state as invalid, then retry
8400                  */
8401                 pnfs_mark_layout_stateid_invalid(lo, &head);
8402                 spin_unlock(&inode->i_lock);
8403                 nfs_commit_inode(inode, 0);
8404                 pnfs_free_lseg_list(&head);
8405                 status = -EAGAIN;
8406                 goto out;
8407         }
8408
8409         nfs4_sequence_free_slot(&lgp->res.seq_res);
8410         err = nfs4_handle_exception(server, nfs4err, exception);
8411         if (!status) {
8412                 if (exception->retry)
8413                         status = -EAGAIN;
8414                 else
8415                         status = err;
8416         }
8417 out:
8418         dprintk("<-- %s\n", __func__);
8419         return status;
8420 }
8421
8422 static size_t max_response_pages(struct nfs_server *server)
8423 {
8424         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8425         return nfs_page_array_len(0, max_resp_sz);
8426 }
8427
8428 static void nfs4_free_pages(struct page **pages, size_t size)
8429 {
8430         int i;
8431
8432         if (!pages)
8433                 return;
8434
8435         for (i = 0; i < size; i++) {
8436                 if (!pages[i])
8437                         break;
8438                 __free_page(pages[i]);
8439         }
8440         kfree(pages);
8441 }
8442
8443 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8444 {
8445         struct page **pages;
8446         int i;
8447
8448         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8449         if (!pages) {
8450                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8451                 return NULL;
8452         }
8453
8454         for (i = 0; i < size; i++) {
8455                 pages[i] = alloc_page(gfp_flags);
8456                 if (!pages[i]) {
8457                         dprintk("%s: failed to allocate page\n", __func__);
8458                         nfs4_free_pages(pages, size);
8459                         return NULL;
8460                 }
8461         }
8462
8463         return pages;
8464 }
8465
8466 static void nfs4_layoutget_release(void *calldata)
8467 {
8468         struct nfs4_layoutget *lgp = calldata;
8469         struct inode *inode = lgp->args.inode;
8470         struct nfs_server *server = NFS_SERVER(inode);
8471         size_t max_pages = max_response_pages(server);
8472
8473         dprintk("--> %s\n", __func__);
8474         nfs4_sequence_free_slot(&lgp->res.seq_res);
8475         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8476         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8477         put_nfs_open_context(lgp->args.ctx);
8478         kfree(calldata);
8479         dprintk("<-- %s\n", __func__);
8480 }
8481
8482 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8483         .rpc_call_prepare = nfs4_layoutget_prepare,
8484         .rpc_call_done = nfs4_layoutget_done,
8485         .rpc_release = nfs4_layoutget_release,
8486 };
8487
8488 struct pnfs_layout_segment *
8489 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8490 {
8491         struct inode *inode = lgp->args.inode;
8492         struct nfs_server *server = NFS_SERVER(inode);
8493         size_t max_pages = max_response_pages(server);
8494         struct rpc_task *task;
8495         struct rpc_message msg = {
8496                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8497                 .rpc_argp = &lgp->args,
8498                 .rpc_resp = &lgp->res,
8499                 .rpc_cred = lgp->cred,
8500         };
8501         struct rpc_task_setup task_setup_data = {
8502                 .rpc_client = server->client,
8503                 .rpc_message = &msg,
8504                 .callback_ops = &nfs4_layoutget_call_ops,
8505                 .callback_data = lgp,
8506                 .flags = RPC_TASK_ASYNC,
8507         };
8508         struct pnfs_layout_segment *lseg = NULL;
8509         struct nfs4_exception exception = {
8510                 .inode = inode,
8511                 .timeout = *timeout,
8512         };
8513         int status = 0;
8514
8515         dprintk("--> %s\n", __func__);
8516
8517         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8518         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8519
8520         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8521         if (!lgp->args.layout.pages) {
8522                 nfs4_layoutget_release(lgp);
8523                 return ERR_PTR(-ENOMEM);
8524         }
8525         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8526
8527         lgp->res.layoutp = &lgp->args.layout;
8528         lgp->res.seq_res.sr_slot = NULL;
8529         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8530
8531         task = rpc_run_task(&task_setup_data);
8532         if (IS_ERR(task))
8533                 return ERR_CAST(task);
8534         status = rpc_wait_for_completion_task(task);
8535         if (status == 0) {
8536                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8537                 *timeout = exception.timeout;
8538         }
8539
8540         trace_nfs4_layoutget(lgp->args.ctx,
8541                         &lgp->args.range,
8542                         &lgp->res.range,
8543                         &lgp->res.stateid,
8544                         status);
8545
8546         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8547         if (status == 0 && lgp->res.layoutp->len)
8548                 lseg = pnfs_layout_process(lgp);
8549         rpc_put_task(task);
8550         dprintk("<-- %s status=%d\n", __func__, status);
8551         if (status)
8552                 return ERR_PTR(status);
8553         return lseg;
8554 }
8555
8556 static void
8557 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8558 {
8559         struct nfs4_layoutreturn *lrp = calldata;
8560
8561         dprintk("--> %s\n", __func__);
8562         nfs4_setup_sequence(lrp->clp,
8563                         &lrp->args.seq_args,
8564                         &lrp->res.seq_res,
8565                         task);
8566 }
8567
8568 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8569 {
8570         struct nfs4_layoutreturn *lrp = calldata;
8571         struct nfs_server *server;
8572
8573         dprintk("--> %s\n", __func__);
8574
8575         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8576                 return;
8577
8578         server = NFS_SERVER(lrp->args.inode);
8579         switch (task->tk_status) {
8580         default:
8581                 task->tk_status = 0;
8582         case 0:
8583                 break;
8584         case -NFS4ERR_DELAY:
8585                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8586                         break;
8587                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8588                 rpc_restart_call_prepare(task);
8589                 return;
8590         }
8591         dprintk("<-- %s\n", __func__);
8592 }
8593
8594 static void nfs4_layoutreturn_release(void *calldata)
8595 {
8596         struct nfs4_layoutreturn *lrp = calldata;
8597         struct pnfs_layout_hdr *lo = lrp->args.layout;
8598
8599         dprintk("--> %s\n", __func__);
8600         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8601                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8602         nfs4_sequence_free_slot(&lrp->res.seq_res);
8603         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8604                 lrp->ld_private.ops->free(&lrp->ld_private);
8605         pnfs_put_layout_hdr(lrp->args.layout);
8606         nfs_iput_and_deactive(lrp->inode);
8607         kfree(calldata);
8608         dprintk("<-- %s\n", __func__);
8609 }
8610
8611 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8612         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8613         .rpc_call_done = nfs4_layoutreturn_done,
8614         .rpc_release = nfs4_layoutreturn_release,
8615 };
8616
8617 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8618 {
8619         struct rpc_task *task;
8620         struct rpc_message msg = {
8621                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8622                 .rpc_argp = &lrp->args,
8623                 .rpc_resp = &lrp->res,
8624                 .rpc_cred = lrp->cred,
8625         };
8626         struct rpc_task_setup task_setup_data = {
8627                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8628                 .rpc_message = &msg,
8629                 .callback_ops = &nfs4_layoutreturn_call_ops,
8630                 .callback_data = lrp,
8631         };
8632         int status = 0;
8633
8634         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8635                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8636                         &task_setup_data.rpc_client, &msg);
8637
8638         dprintk("--> %s\n", __func__);
8639         if (!sync) {
8640                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8641                 if (!lrp->inode) {
8642                         nfs4_layoutreturn_release(lrp);
8643                         return -EAGAIN;
8644                 }
8645                 task_setup_data.flags |= RPC_TASK_ASYNC;
8646         }
8647         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8648         task = rpc_run_task(&task_setup_data);
8649         if (IS_ERR(task))
8650                 return PTR_ERR(task);
8651         if (sync)
8652                 status = task->tk_status;
8653         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8654         dprintk("<-- %s status=%d\n", __func__, status);
8655         rpc_put_task(task);
8656         return status;
8657 }
8658
8659 static int
8660 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8661                 struct pnfs_device *pdev,
8662                 struct rpc_cred *cred)
8663 {
8664         struct nfs4_getdeviceinfo_args args = {
8665                 .pdev = pdev,
8666                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8667                         NOTIFY_DEVICEID4_DELETE,
8668         };
8669         struct nfs4_getdeviceinfo_res res = {
8670                 .pdev = pdev,
8671         };
8672         struct rpc_message msg = {
8673                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8674                 .rpc_argp = &args,
8675                 .rpc_resp = &res,
8676                 .rpc_cred = cred,
8677         };
8678         int status;
8679
8680         dprintk("--> %s\n", __func__);
8681         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8682         if (res.notification & ~args.notify_types)
8683                 dprintk("%s: unsupported notification\n", __func__);
8684         if (res.notification != args.notify_types)
8685                 pdev->nocache = 1;
8686
8687         dprintk("<-- %s status=%d\n", __func__, status);
8688
8689         return status;
8690 }
8691
8692 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8693                 struct pnfs_device *pdev,
8694                 struct rpc_cred *cred)
8695 {
8696         struct nfs4_exception exception = { };
8697         int err;
8698
8699         do {
8700                 err = nfs4_handle_exception(server,
8701                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8702                                         &exception);
8703         } while (exception.retry);
8704         return err;
8705 }
8706 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8707
8708 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8709 {
8710         struct nfs4_layoutcommit_data *data = calldata;
8711         struct nfs_server *server = NFS_SERVER(data->args.inode);
8712
8713         nfs4_setup_sequence(server->nfs_client,
8714                         &data->args.seq_args,
8715                         &data->res.seq_res,
8716                         task);
8717 }
8718
8719 static void
8720 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8721 {
8722         struct nfs4_layoutcommit_data *data = calldata;
8723         struct nfs_server *server = NFS_SERVER(data->args.inode);
8724
8725         if (!nfs41_sequence_done(task, &data->res.seq_res))
8726                 return;
8727
8728         switch (task->tk_status) { /* Just ignore these failures */
8729         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8730         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8731         case -NFS4ERR_BADLAYOUT:     /* no layout */
8732         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8733                 task->tk_status = 0;
8734         case 0:
8735                 break;
8736         default:
8737                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8738                         rpc_restart_call_prepare(task);
8739                         return;
8740                 }
8741         }
8742 }
8743
8744 static void nfs4_layoutcommit_release(void *calldata)
8745 {
8746         struct nfs4_layoutcommit_data *data = calldata;
8747
8748         pnfs_cleanup_layoutcommit(data);
8749         nfs_post_op_update_inode_force_wcc(data->args.inode,
8750                                            data->res.fattr);
8751         put_rpccred(data->cred);
8752         nfs_iput_and_deactive(data->inode);
8753         kfree(data);
8754 }
8755
8756 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8757         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8758         .rpc_call_done = nfs4_layoutcommit_done,
8759         .rpc_release = nfs4_layoutcommit_release,
8760 };
8761
8762 int
8763 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8764 {
8765         struct rpc_message msg = {
8766                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8767                 .rpc_argp = &data->args,
8768                 .rpc_resp = &data->res,
8769                 .rpc_cred = data->cred,
8770         };
8771         struct rpc_task_setup task_setup_data = {
8772                 .task = &data->task,
8773                 .rpc_client = NFS_CLIENT(data->args.inode),
8774                 .rpc_message = &msg,
8775                 .callback_ops = &nfs4_layoutcommit_ops,
8776                 .callback_data = data,
8777         };
8778         struct rpc_task *task;
8779         int status = 0;
8780
8781         dprintk("NFS: initiating layoutcommit call. sync %d "
8782                 "lbw: %llu inode %lu\n", sync,
8783                 data->args.lastbytewritten,
8784                 data->args.inode->i_ino);
8785
8786         if (!sync) {
8787                 data->inode = nfs_igrab_and_active(data->args.inode);
8788                 if (data->inode == NULL) {
8789                         nfs4_layoutcommit_release(data);
8790                         return -EAGAIN;
8791                 }
8792                 task_setup_data.flags = RPC_TASK_ASYNC;
8793         }
8794         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8795         task = rpc_run_task(&task_setup_data);
8796         if (IS_ERR(task))
8797                 return PTR_ERR(task);
8798         if (sync)
8799                 status = task->tk_status;
8800         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8801         dprintk("%s: status %d\n", __func__, status);
8802         rpc_put_task(task);
8803         return status;
8804 }
8805
8806 /**
8807  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8808  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8809  */
8810 static int
8811 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8812                     struct nfs_fsinfo *info,
8813                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8814 {
8815         struct nfs41_secinfo_no_name_args args = {
8816                 .style = SECINFO_STYLE_CURRENT_FH,
8817         };
8818         struct nfs4_secinfo_res res = {
8819                 .flavors = flavors,
8820         };
8821         struct rpc_message msg = {
8822                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8823                 .rpc_argp = &args,
8824                 .rpc_resp = &res,
8825         };
8826         struct rpc_clnt *clnt = server->client;
8827         struct rpc_cred *cred = NULL;
8828         int status;
8829
8830         if (use_integrity) {
8831                 clnt = server->nfs_client->cl_rpcclient;
8832                 cred = nfs4_get_clid_cred(server->nfs_client);
8833                 msg.rpc_cred = cred;
8834         }
8835
8836         dprintk("--> %s\n", __func__);
8837         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8838                                 &res.seq_res, 0);
8839         dprintk("<-- %s status=%d\n", __func__, status);
8840
8841         if (cred)
8842                 put_rpccred(cred);
8843
8844         return status;
8845 }
8846
8847 static int
8848 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8849                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8850 {
8851         struct nfs4_exception exception = { };
8852         int err;
8853         do {
8854                 /* first try using integrity protection */
8855                 err = -NFS4ERR_WRONGSEC;
8856
8857                 /* try to use integrity protection with machine cred */
8858                 if (_nfs4_is_integrity_protected(server->nfs_client))
8859                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8860                                                           flavors, true);
8861
8862                 /*
8863                  * if unable to use integrity protection, or SECINFO with
8864                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8865                  * disallowed by spec, but exists in deployed servers) use
8866                  * the current filesystem's rpc_client and the user cred.
8867                  */
8868                 if (err == -NFS4ERR_WRONGSEC)
8869                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8870                                                           flavors, false);
8871
8872                 switch (err) {
8873                 case 0:
8874                 case -NFS4ERR_WRONGSEC:
8875                 case -ENOTSUPP:
8876                         goto out;
8877                 default:
8878                         err = nfs4_handle_exception(server, err, &exception);
8879                 }
8880         } while (exception.retry);
8881 out:
8882         return err;
8883 }
8884
8885 static int
8886 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8887                     struct nfs_fsinfo *info)
8888 {
8889         int err;
8890         struct page *page;
8891         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8892         struct nfs4_secinfo_flavors *flavors;
8893         struct nfs4_secinfo4 *secinfo;
8894         int i;
8895
8896         page = alloc_page(GFP_KERNEL);
8897         if (!page) {
8898                 err = -ENOMEM;
8899                 goto out;
8900         }
8901
8902         flavors = page_address(page);
8903         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8904
8905         /*
8906          * Fall back on "guess and check" method if
8907          * the server doesn't support SECINFO_NO_NAME
8908          */
8909         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8910                 err = nfs4_find_root_sec(server, fhandle, info);
8911                 goto out_freepage;
8912         }
8913         if (err)
8914                 goto out_freepage;
8915
8916         for (i = 0; i < flavors->num_flavors; i++) {
8917                 secinfo = &flavors->flavors[i];
8918
8919                 switch (secinfo->flavor) {
8920                 case RPC_AUTH_NULL:
8921                 case RPC_AUTH_UNIX:
8922                 case RPC_AUTH_GSS:
8923                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8924                                         &secinfo->flavor_info);
8925                         break;
8926                 default:
8927                         flavor = RPC_AUTH_MAXFLAVOR;
8928                         break;
8929                 }
8930
8931                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8932                         flavor = RPC_AUTH_MAXFLAVOR;
8933
8934                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8935                         err = nfs4_lookup_root_sec(server, fhandle,
8936                                                    info, flavor);
8937                         if (!err)
8938                                 break;
8939                 }
8940         }
8941
8942         if (flavor == RPC_AUTH_MAXFLAVOR)
8943                 err = -EPERM;
8944
8945 out_freepage:
8946         put_page(page);
8947         if (err == -EACCES)
8948                 return -EPERM;
8949 out:
8950         return err;
8951 }
8952
8953 static int _nfs41_test_stateid(struct nfs_server *server,
8954                 nfs4_stateid *stateid,
8955                 struct rpc_cred *cred)
8956 {
8957         int status;
8958         struct nfs41_test_stateid_args args = {
8959                 .stateid = stateid,
8960         };
8961         struct nfs41_test_stateid_res res;
8962         struct rpc_message msg = {
8963                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8964                 .rpc_argp = &args,
8965                 .rpc_resp = &res,
8966                 .rpc_cred = cred,
8967         };
8968         struct rpc_clnt *rpc_client = server->client;
8969
8970         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8971                 &rpc_client, &msg);
8972
8973         dprintk("NFS call  test_stateid %p\n", stateid);
8974         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8975         nfs4_set_sequence_privileged(&args.seq_args);
8976         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8977                         &args.seq_args, &res.seq_res);
8978         if (status != NFS_OK) {
8979                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8980                 return status;
8981         }
8982         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8983         return -res.status;
8984 }
8985
8986 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8987                 int err, struct nfs4_exception *exception)
8988 {
8989         exception->retry = 0;
8990         switch(err) {
8991         case -NFS4ERR_DELAY:
8992         case -NFS4ERR_RETRY_UNCACHED_REP:
8993                 nfs4_handle_exception(server, err, exception);
8994                 break;
8995         case -NFS4ERR_BADSESSION:
8996         case -NFS4ERR_BADSLOT:
8997         case -NFS4ERR_BAD_HIGH_SLOT:
8998         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8999         case -NFS4ERR_DEADSESSION:
9000                 nfs4_do_handle_exception(server, err, exception);
9001         }
9002 }
9003
9004 /**
9005  * nfs41_test_stateid - perform a TEST_STATEID operation
9006  *
9007  * @server: server / transport on which to perform the operation
9008  * @stateid: state ID to test
9009  * @cred: credential
9010  *
9011  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9012  * Otherwise a negative NFS4ERR value is returned if the operation
9013  * failed or the state ID is not currently valid.
9014  */
9015 static int nfs41_test_stateid(struct nfs_server *server,
9016                 nfs4_stateid *stateid,
9017                 struct rpc_cred *cred)
9018 {
9019         struct nfs4_exception exception = { };
9020         int err;
9021         do {
9022                 err = _nfs41_test_stateid(server, stateid, cred);
9023                 nfs4_handle_delay_or_session_error(server, err, &exception);
9024         } while (exception.retry);
9025         return err;
9026 }
9027
9028 struct nfs_free_stateid_data {
9029         struct nfs_server *server;
9030         struct nfs41_free_stateid_args args;
9031         struct nfs41_free_stateid_res res;
9032 };
9033
9034 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9035 {
9036         struct nfs_free_stateid_data *data = calldata;
9037         nfs4_setup_sequence(data->server->nfs_client,
9038                         &data->args.seq_args,
9039                         &data->res.seq_res,
9040                         task);
9041 }
9042
9043 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9044 {
9045         struct nfs_free_stateid_data *data = calldata;
9046
9047         nfs41_sequence_done(task, &data->res.seq_res);
9048
9049         switch (task->tk_status) {
9050         case -NFS4ERR_DELAY:
9051                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9052                         rpc_restart_call_prepare(task);
9053         }
9054 }
9055
9056 static void nfs41_free_stateid_release(void *calldata)
9057 {
9058         kfree(calldata);
9059 }
9060
9061 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9062         .rpc_call_prepare = nfs41_free_stateid_prepare,
9063         .rpc_call_done = nfs41_free_stateid_done,
9064         .rpc_release = nfs41_free_stateid_release,
9065 };
9066
9067 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9068                 const nfs4_stateid *stateid,
9069                 struct rpc_cred *cred,
9070                 bool privileged)
9071 {
9072         struct rpc_message msg = {
9073                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9074                 .rpc_cred = cred,
9075         };
9076         struct rpc_task_setup task_setup = {
9077                 .rpc_client = server->client,
9078                 .rpc_message = &msg,
9079                 .callback_ops = &nfs41_free_stateid_ops,
9080                 .flags = RPC_TASK_ASYNC,
9081         };
9082         struct nfs_free_stateid_data *data;
9083
9084         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9085                 &task_setup.rpc_client, &msg);
9086
9087         dprintk("NFS call  free_stateid %p\n", stateid);
9088         data = kmalloc(sizeof(*data), GFP_NOFS);
9089         if (!data)
9090                 return ERR_PTR(-ENOMEM);
9091         data->server = server;
9092         nfs4_stateid_copy(&data->args.stateid, stateid);
9093
9094         task_setup.callback_data = data;
9095
9096         msg.rpc_argp = &data->args;
9097         msg.rpc_resp = &data->res;
9098         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9099         if (privileged)
9100                 nfs4_set_sequence_privileged(&data->args.seq_args);
9101
9102         return rpc_run_task(&task_setup);
9103 }
9104
9105 /**
9106  * nfs41_free_stateid - perform a FREE_STATEID operation
9107  *
9108  * @server: server / transport on which to perform the operation
9109  * @stateid: state ID to release
9110  * @cred: credential
9111  * @is_recovery: set to true if this call needs to be privileged
9112  *
9113  * Note: this function is always asynchronous.
9114  */
9115 static int nfs41_free_stateid(struct nfs_server *server,
9116                 const nfs4_stateid *stateid,
9117                 struct rpc_cred *cred,
9118                 bool is_recovery)
9119 {
9120         struct rpc_task *task;
9121
9122         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9123         if (IS_ERR(task))
9124                 return PTR_ERR(task);
9125         rpc_put_task(task);
9126         return 0;
9127 }
9128
9129 static void
9130 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9131 {
9132         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9133
9134         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9135         nfs4_free_lock_state(server, lsp);
9136 }
9137
9138 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9139                 const nfs4_stateid *s2)
9140 {
9141         if (s1->type != s2->type)
9142                 return false;
9143
9144         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9145                 return false;
9146
9147         if (s1->seqid == s2->seqid)
9148                 return true;
9149
9150         return s1->seqid == 0 || s2->seqid == 0;
9151 }
9152
9153 #endif /* CONFIG_NFS_V4_1 */
9154
9155 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9156                 const nfs4_stateid *s2)
9157 {
9158         return nfs4_stateid_match(s1, s2);
9159 }
9160
9161
9162 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9163         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9164         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9165         .recover_open   = nfs4_open_reclaim,
9166         .recover_lock   = nfs4_lock_reclaim,
9167         .establish_clid = nfs4_init_clientid,
9168         .detect_trunking = nfs40_discover_server_trunking,
9169 };
9170
9171 #if defined(CONFIG_NFS_V4_1)
9172 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9173         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9174         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9175         .recover_open   = nfs4_open_reclaim,
9176         .recover_lock   = nfs4_lock_reclaim,
9177         .establish_clid = nfs41_init_clientid,
9178         .reclaim_complete = nfs41_proc_reclaim_complete,
9179         .detect_trunking = nfs41_discover_server_trunking,
9180 };
9181 #endif /* CONFIG_NFS_V4_1 */
9182
9183 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9184         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9185         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9186         .recover_open   = nfs40_open_expired,
9187         .recover_lock   = nfs4_lock_expired,
9188         .establish_clid = nfs4_init_clientid,
9189 };
9190
9191 #if defined(CONFIG_NFS_V4_1)
9192 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9193         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9194         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9195         .recover_open   = nfs41_open_expired,
9196         .recover_lock   = nfs41_lock_expired,
9197         .establish_clid = nfs41_init_clientid,
9198 };
9199 #endif /* CONFIG_NFS_V4_1 */
9200
9201 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9202         .sched_state_renewal = nfs4_proc_async_renew,
9203         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9204         .renew_lease = nfs4_proc_renew,
9205 };
9206
9207 #if defined(CONFIG_NFS_V4_1)
9208 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9209         .sched_state_renewal = nfs41_proc_async_sequence,
9210         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9211         .renew_lease = nfs4_proc_sequence,
9212 };
9213 #endif
9214
9215 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9216         .get_locations = _nfs40_proc_get_locations,
9217         .fsid_present = _nfs40_proc_fsid_present,
9218 };
9219
9220 #if defined(CONFIG_NFS_V4_1)
9221 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9222         .get_locations = _nfs41_proc_get_locations,
9223         .fsid_present = _nfs41_proc_fsid_present,
9224 };
9225 #endif  /* CONFIG_NFS_V4_1 */
9226
9227 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9228         .minor_version = 0,
9229         .init_caps = NFS_CAP_READDIRPLUS
9230                 | NFS_CAP_ATOMIC_OPEN
9231                 | NFS_CAP_POSIX_LOCK,
9232         .init_client = nfs40_init_client,
9233         .shutdown_client = nfs40_shutdown_client,
9234         .match_stateid = nfs4_match_stateid,
9235         .find_root_sec = nfs4_find_root_sec,
9236         .free_lock_state = nfs4_release_lockowner,
9237         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9238         .alloc_seqid = nfs_alloc_seqid,
9239         .call_sync_ops = &nfs40_call_sync_ops,
9240         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9241         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9242         .state_renewal_ops = &nfs40_state_renewal_ops,
9243         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9244 };
9245
9246 #if defined(CONFIG_NFS_V4_1)
9247 static struct nfs_seqid *
9248 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9249 {
9250         return NULL;
9251 }
9252
9253 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9254         .minor_version = 1,
9255         .init_caps = NFS_CAP_READDIRPLUS
9256                 | NFS_CAP_ATOMIC_OPEN
9257                 | NFS_CAP_POSIX_LOCK
9258                 | NFS_CAP_STATEID_NFSV41
9259                 | NFS_CAP_ATOMIC_OPEN_V1,
9260         .init_client = nfs41_init_client,
9261         .shutdown_client = nfs41_shutdown_client,
9262         .match_stateid = nfs41_match_stateid,
9263         .find_root_sec = nfs41_find_root_sec,
9264         .free_lock_state = nfs41_free_lock_state,
9265         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9266         .alloc_seqid = nfs_alloc_no_seqid,
9267         .session_trunk = nfs4_test_session_trunk,
9268         .call_sync_ops = &nfs41_call_sync_ops,
9269         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9270         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9271         .state_renewal_ops = &nfs41_state_renewal_ops,
9272         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9273 };
9274 #endif
9275
9276 #if defined(CONFIG_NFS_V4_2)
9277 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9278         .minor_version = 2,
9279         .init_caps = NFS_CAP_READDIRPLUS
9280                 | NFS_CAP_ATOMIC_OPEN
9281                 | NFS_CAP_POSIX_LOCK
9282                 | NFS_CAP_STATEID_NFSV41
9283                 | NFS_CAP_ATOMIC_OPEN_V1
9284                 | NFS_CAP_ALLOCATE
9285                 | NFS_CAP_COPY
9286                 | NFS_CAP_DEALLOCATE
9287                 | NFS_CAP_SEEK
9288                 | NFS_CAP_LAYOUTSTATS
9289                 | NFS_CAP_CLONE,
9290         .init_client = nfs41_init_client,
9291         .shutdown_client = nfs41_shutdown_client,
9292         .match_stateid = nfs41_match_stateid,
9293         .find_root_sec = nfs41_find_root_sec,
9294         .free_lock_state = nfs41_free_lock_state,
9295         .call_sync_ops = &nfs41_call_sync_ops,
9296         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9297         .alloc_seqid = nfs_alloc_no_seqid,
9298         .session_trunk = nfs4_test_session_trunk,
9299         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9300         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9301         .state_renewal_ops = &nfs41_state_renewal_ops,
9302         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9303 };
9304 #endif
9305
9306 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9307         [0] = &nfs_v4_0_minor_ops,
9308 #if defined(CONFIG_NFS_V4_1)
9309         [1] = &nfs_v4_1_minor_ops,
9310 #endif
9311 #if defined(CONFIG_NFS_V4_2)
9312         [2] = &nfs_v4_2_minor_ops,
9313 #endif
9314 };
9315
9316 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9317 {
9318         ssize_t error, error2;
9319
9320         error = generic_listxattr(dentry, list, size);
9321         if (error < 0)
9322                 return error;
9323         if (list) {
9324                 list += error;
9325                 size -= error;
9326         }
9327
9328         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9329         if (error2 < 0)
9330                 return error2;
9331         return error + error2;
9332 }
9333
9334 static const struct inode_operations nfs4_dir_inode_operations = {
9335         .create         = nfs_create,
9336         .lookup         = nfs_lookup,
9337         .atomic_open    = nfs_atomic_open,
9338         .link           = nfs_link,
9339         .unlink         = nfs_unlink,
9340         .symlink        = nfs_symlink,
9341         .mkdir          = nfs_mkdir,
9342         .rmdir          = nfs_rmdir,
9343         .mknod          = nfs_mknod,
9344         .rename         = nfs_rename,
9345         .permission     = nfs_permission,
9346         .getattr        = nfs_getattr,
9347         .setattr        = nfs_setattr,
9348         .listxattr      = nfs4_listxattr,
9349 };
9350
9351 static const struct inode_operations nfs4_file_inode_operations = {
9352         .permission     = nfs_permission,
9353         .getattr        = nfs_getattr,
9354         .setattr        = nfs_setattr,
9355         .listxattr      = nfs4_listxattr,
9356 };
9357
9358 const struct nfs_rpc_ops nfs_v4_clientops = {
9359         .version        = 4,                    /* protocol version */
9360         .dentry_ops     = &nfs4_dentry_operations,
9361         .dir_inode_ops  = &nfs4_dir_inode_operations,
9362         .file_inode_ops = &nfs4_file_inode_operations,
9363         .file_ops       = &nfs4_file_operations,
9364         .getroot        = nfs4_proc_get_root,
9365         .submount       = nfs4_submount,
9366         .try_mount      = nfs4_try_mount,
9367         .getattr        = nfs4_proc_getattr,
9368         .setattr        = nfs4_proc_setattr,
9369         .lookup         = nfs4_proc_lookup,
9370         .lookupp        = nfs4_proc_lookupp,
9371         .access         = nfs4_proc_access,
9372         .readlink       = nfs4_proc_readlink,
9373         .create         = nfs4_proc_create,
9374         .remove         = nfs4_proc_remove,
9375         .unlink_setup   = nfs4_proc_unlink_setup,
9376         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9377         .unlink_done    = nfs4_proc_unlink_done,
9378         .rename_setup   = nfs4_proc_rename_setup,
9379         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9380         .rename_done    = nfs4_proc_rename_done,
9381         .link           = nfs4_proc_link,
9382         .symlink        = nfs4_proc_symlink,
9383         .mkdir          = nfs4_proc_mkdir,
9384         .rmdir          = nfs4_proc_remove,
9385         .readdir        = nfs4_proc_readdir,
9386         .mknod          = nfs4_proc_mknod,
9387         .statfs         = nfs4_proc_statfs,
9388         .fsinfo         = nfs4_proc_fsinfo,
9389         .pathconf       = nfs4_proc_pathconf,
9390         .set_capabilities = nfs4_server_capabilities,
9391         .decode_dirent  = nfs4_decode_dirent,
9392         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9393         .read_setup     = nfs4_proc_read_setup,
9394         .read_done      = nfs4_read_done,
9395         .write_setup    = nfs4_proc_write_setup,
9396         .write_done     = nfs4_write_done,
9397         .commit_setup   = nfs4_proc_commit_setup,
9398         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9399         .commit_done    = nfs4_commit_done,
9400         .lock           = nfs4_proc_lock,
9401         .clear_acl_cache = nfs4_zap_acl_attr,
9402         .close_context  = nfs4_close_context,
9403         .open_context   = nfs4_atomic_open,
9404         .have_delegation = nfs4_have_delegation,
9405         .return_delegation = nfs4_inode_return_delegation,
9406         .alloc_client   = nfs4_alloc_client,
9407         .init_client    = nfs4_init_client,
9408         .free_client    = nfs4_free_client,
9409         .create_server  = nfs4_create_server,
9410         .clone_server   = nfs_clone_server,
9411 };
9412
9413 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9414         .name   = XATTR_NAME_NFSV4_ACL,
9415         .list   = nfs4_xattr_list_nfs4_acl,
9416         .get    = nfs4_xattr_get_nfs4_acl,
9417         .set    = nfs4_xattr_set_nfs4_acl,
9418 };
9419
9420 const struct xattr_handler *nfs4_xattr_handlers[] = {
9421         &nfs4_xattr_nfs4_acl_handler,
9422 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9423         &nfs4_xattr_nfs4_label_handler,
9424 #endif
9425         NULL
9426 };
9427
9428 /*
9429  * Local variables:
9430  *  c-basic-offset: 8
9431  * End:
9432  */