fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/file.h>
  42 #include <linux/string.h>
  43 #include <linux/ratelimit.h>
  44 #include <linux/printk.h>
  45 #include <linux/slab.h>
  46 #include <linux/sunrpc/clnt.h>
  47 #include <linux/nfs.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfs_fs.h>
  50 #include <linux/nfs_page.h>
  51 #include <linux/nfs_mount.h>
  52 #include <linux/namei.h>
  53 #include <linux/mount.h>
  54 #include <linux/module.h>
  55 #include <linux/xattr.h>
  56 #include <linux/utsname.h>
  57 #include <linux/freezer.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "nfs4idmap.h"
  67 #include "nfs4session.h"
  68 #include "fscache.h"
  69
  70 #include "nfs4trace.h"
  71
  72 #define NFSDBG_FACILITY         NFSDBG_PROC
  73
  74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  76
  77 /* file attributes which can be mapped to nfs attributes */
  78 #define NFS4_VALID_ATTRS (ATTR_MODE \
  79         | ATTR_UID \
  80         | ATTR_GID \
  81         | ATTR_SIZE \
  82         | ATTR_ATIME \
  83         | ATTR_MTIME \
  84         | ATTR_CTIME \
  85         | ATTR_ATIME_SET \
  86         | ATTR_MTIME_SET)
  87
  88 struct nfs4_opendata;
  89 static int _nfs4_proc_open(struct nfs4_opendata *data);
  90 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  91 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  92 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  93 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  94 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  95 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  96                             struct nfs_fattr *fattr, struct iattr *sattr,
  97                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  98                             struct nfs4_label *olabel);
  99 #ifdef CONFIG_NFS_V4_1
 100 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 101                 struct rpc_cred *);
 102 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 103                 struct rpc_cred *, bool);
 104 #endif
 105
 106 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 107 static inline struct nfs4_label *
 108 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 109         struct iattr *sattr, struct nfs4_label *label)
 110 {
 111         int err;
 112
 113         if (label == NULL)
 114                 return NULL;
 115
 116         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 117                 return NULL;
 118
 119         err = security_dentry_init_security(dentry, sattr->ia_mode,
 120                                 &dentry->d_name, (void **)&label->label, &label->len);
 121         if (err == 0)
 122                 return label;
 123
 124         return NULL;
 125 }
 126 static inline void
 127 nfs4_label_release_security(struct nfs4_label *label)
 128 {
 129         if (label)
 130                 security_release_secctx(label->label, label->len);
 131 }
 132 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 133 {
 134         if (label)
 135                 return server->attr_bitmask;
 136
 137         return server->attr_bitmask_nl;
 138 }
 139 #else
 140 static inline struct nfs4_label *
 141 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 142         struct iattr *sattr, struct nfs4_label *l)
 143 { return NULL; }
 144 static inline void
 145 nfs4_label_release_security(struct nfs4_label *label)
 146 { return; }
 147 static inline u32 *
 148 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 149 { return server->attr_bitmask; }
 150 #endif
 151
 152 /* Prevent leaks of NFSv4 errors into userland */
 153 static int nfs4_map_errors(int err)
 154 {
 155         if (err >= -1000)
 156                 return err;
 157         switch (err) {
 158         case -NFS4ERR_RESOURCE:
 159         case -NFS4ERR_LAYOUTTRYLATER:
 160         case -NFS4ERR_RECALLCONFLICT:
 161                 return -EREMOTEIO;
 162         case -NFS4ERR_WRONGSEC:
 163         case -NFS4ERR_WRONG_CRED:
 164                 return -EPERM;
 165         case -NFS4ERR_BADOWNER:
 166         case -NFS4ERR_BADNAME:
 167                 return -EINVAL;
 168         case -NFS4ERR_SHARE_DENIED:
 169                 return -EACCES;
 170         case -NFS4ERR_MINOR_VERS_MISMATCH:
 171                 return -EPROTONOSUPPORT;
 172         case -NFS4ERR_FILE_OPEN:
 173                 return -EBUSY;
 174         default:
 175                 dprintk("%s could not handle NFSv4 error %d\n",
 176                                 __func__, -err);
 177                 break;
 178         }
 179         return -EIO;
 180 }
 181
 182 /*
 183  * This is our standard bitmap for GETATTR requests.
 184  */
 185 const u32 nfs4_fattr_bitmap[3] = {
 186         FATTR4_WORD0_TYPE
 187         | FATTR4_WORD0_CHANGE
 188         | FATTR4_WORD0_SIZE
 189         | FATTR4_WORD0_FSID
 190         | FATTR4_WORD0_FILEID,
 191         FATTR4_WORD1_MODE
 192         | FATTR4_WORD1_NUMLINKS
 193         | FATTR4_WORD1_OWNER
 194         | FATTR4_WORD1_OWNER_GROUP
 195         | FATTR4_WORD1_RAWDEV
 196         | FATTR4_WORD1_SPACE_USED
 197         | FATTR4_WORD1_TIME_ACCESS
 198         | FATTR4_WORD1_TIME_METADATA
 199         | FATTR4_WORD1_TIME_MODIFY
 200         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 201 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 202         FATTR4_WORD2_SECURITY_LABEL
 203 #endif
 204 };
 205
 206 static const u32 nfs4_pnfs_open_bitmap[3] = {
 207         FATTR4_WORD0_TYPE
 208         | FATTR4_WORD0_CHANGE
 209         | FATTR4_WORD0_SIZE
 210         | FATTR4_WORD0_FSID
 211         | FATTR4_WORD0_FILEID,
 212         FATTR4_WORD1_MODE
 213         | FATTR4_WORD1_NUMLINKS
 214         | FATTR4_WORD1_OWNER
 215         | FATTR4_WORD1_OWNER_GROUP
 216         | FATTR4_WORD1_RAWDEV
 217         | FATTR4_WORD1_SPACE_USED
 218         | FATTR4_WORD1_TIME_ACCESS
 219         | FATTR4_WORD1_TIME_METADATA
 220         | FATTR4_WORD1_TIME_MODIFY,
 221         FATTR4_WORD2_MDSTHRESHOLD
 222 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 223         | FATTR4_WORD2_SECURITY_LABEL
 224 #endif
 225 };
 226
 227 static const u32 nfs4_open_noattr_bitmap[3] = {
 228         FATTR4_WORD0_TYPE
 229         | FATTR4_WORD0_FILEID,
 230 };
 231
 232 const u32 nfs4_statfs_bitmap[3] = {
 233         FATTR4_WORD0_FILES_AVAIL
 234         | FATTR4_WORD0_FILES_FREE
 235         | FATTR4_WORD0_FILES_TOTAL,
 236         FATTR4_WORD1_SPACE_AVAIL
 237         | FATTR4_WORD1_SPACE_FREE
 238         | FATTR4_WORD1_SPACE_TOTAL
 239 };
 240
 241 const u32 nfs4_pathconf_bitmap[3] = {
 242         FATTR4_WORD0_MAXLINK
 243         | FATTR4_WORD0_MAXNAME,
 244         0
 245 };
 246
 247 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 248                         | FATTR4_WORD0_MAXREAD
 249                         | FATTR4_WORD0_MAXWRITE
 250                         | FATTR4_WORD0_LEASE_TIME,
 251                         FATTR4_WORD1_TIME_DELTA
 252                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 253                         FATTR4_WORD2_LAYOUT_BLKSIZE
 254                         | FATTR4_WORD2_CLONE_BLKSIZE
 255 };
 256
 257 const u32 nfs4_fs_locations_bitmap[3] = {
 258         FATTR4_WORD0_TYPE
 259         | FATTR4_WORD0_CHANGE
 260         | FATTR4_WORD0_SIZE
 261         | FATTR4_WORD0_FSID
 262         | FATTR4_WORD0_FILEID
 263         | FATTR4_WORD0_FS_LOCATIONS,
 264         FATTR4_WORD1_MODE
 265         | FATTR4_WORD1_NUMLINKS
 266         | FATTR4_WORD1_OWNER
 267         | FATTR4_WORD1_OWNER_GROUP
 268         | FATTR4_WORD1_RAWDEV
 269         | FATTR4_WORD1_SPACE_USED
 270         | FATTR4_WORD1_TIME_ACCESS
 271         | FATTR4_WORD1_TIME_METADATA
 272         | FATTR4_WORD1_TIME_MODIFY
 273         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 274 };
 275
 276 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 277                 struct nfs4_readdir_arg *readdir)
 278 {
 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(FATTR4_WORD0_FILEID);             /* bitmap */
 310                 *p++ = htonl(8);              /* attribute buffer length */
 311                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 312         }
 313
 314         *p++ = xdr_one;                                  /* next */
 315         *p++ = xdr_zero;                   /* cookie, first word */
 316         *p++ = xdr_two;                   /* cookie, second word */
 317         *p++ = xdr_two;                             /* entry len */
 318         memcpy(p, "..\0\0", 4);                         /* entry */
 319         p++;
 320         *p++ = xdr_one;                         /* bitmap length */
 321         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 322         *p++ = htonl(8);              /* attribute buffer length */
 323         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 324
 325         readdir->pgbase = (char *)p - (char *)start;
 326         readdir->count -= readdir->pgbase;
 327         kunmap_atomic(start);
 328 }
 329
 330 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 331                 nfs4_stateid *stateid,
 332                 struct rpc_cred *cred)
 333 {
 334         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 335
 336         ops->test_and_free_expired(server, stateid, cred);
 337 }
 338
 339 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 340                 nfs4_stateid *stateid,
 341                 struct rpc_cred *cred)
 342 {
 343         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 344         nfs4_test_and_free_stateid(server, stateid, cred);
 345 }
 346
 347 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 348                 const nfs4_stateid *stateid,
 349                 struct rpc_cred *cred)
 350 {
 351         nfs4_stateid tmp;
 352
 353         nfs4_stateid_copy(&tmp, stateid);
 354         __nfs4_free_revoked_stateid(server, &tmp, cred);
 355 }
 356
 357 static long nfs4_update_delay(long *timeout)
 358 {
 359         long ret;
 360         if (!timeout)
 361                 return NFS4_POLL_RETRY_MAX;
 362         if (*timeout <= 0)
 363                 *timeout = NFS4_POLL_RETRY_MIN;
 364         if (*timeout > NFS4_POLL_RETRY_MAX)
 365                 *timeout = NFS4_POLL_RETRY_MAX;
 366         ret = *timeout;
 367         *timeout <<= 1;
 368         return ret;
 369 }
 370
 371 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 372 {
 373         int res = 0;
 374
 375         might_sleep();
 376
 377         freezable_schedule_timeout_killable_unsafe(
 378                 nfs4_update_delay(timeout));
 379         if (fatal_signal_pending(current))
 380                 res = -ERESTARTSYS;
 381         return res;
 382 }
 383
 384 /* This is the error handling routine for processes that are allowed
 385  * to sleep.
 386  */
 387 static int nfs4_do_handle_exception(struct nfs_server *server,
 388                 int errorcode, struct nfs4_exception *exception)
 389 {
 390         struct nfs_client *clp = server->nfs_client;
 391         struct nfs4_state *state = exception->state;
 392         const nfs4_stateid *stateid = exception->stateid;
 393         struct inode *inode = exception->inode;
 394         int ret = errorcode;
 395
 396         exception->delay = 0;
 397         exception->recovering = 0;
 398         exception->retry = 0;
 399
 400         if (stateid == NULL && state != NULL)
 401                 stateid = &state->stateid;
 402
 403         switch(errorcode) {
 404                 case 0:
 405                         return 0;
 406                 case -NFS4ERR_DELEG_REVOKED:
 407                 case -NFS4ERR_ADMIN_REVOKED:
 408                 case -NFS4ERR_EXPIRED:
 409                 case -NFS4ERR_BAD_STATEID:
 410                         if (inode != NULL && stateid != NULL) {
 411                                 nfs_inode_find_state_and_recover(inode,
 412                                                 stateid);
 413                                 goto wait_on_recovery;
 414                         }
 415                 case -NFS4ERR_OPENMODE:
 416                         if (inode) {
 417                                 int err;
 418
 419                                 err = nfs_async_inode_return_delegation(inode,
 420                                                 stateid);
 421                                 if (err == 0)
 422                                         goto wait_on_recovery;
 423                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 424                                         exception->retry = 1;
 425                                         break;
 426                                 }
 427                         }
 428                         if (state == NULL)
 429                                 break;
 430                         ret = nfs4_schedule_stateid_recovery(server, state);
 431                         if (ret < 0)
 432                                 break;
 433                         goto wait_on_recovery;
 434                 case -NFS4ERR_STALE_STATEID:
 435                 case -NFS4ERR_STALE_CLIENTID:
 436                         nfs4_schedule_lease_recovery(clp);
 437                         goto wait_on_recovery;
 438                 case -NFS4ERR_MOVED:
 439                         ret = nfs4_schedule_migration_recovery(server);
 440                         if (ret < 0)
 441                                 break;
 442                         goto wait_on_recovery;
 443                 case -NFS4ERR_LEASE_MOVED:
 444                         nfs4_schedule_lease_moved_recovery(clp);
 445                         goto wait_on_recovery;
 446 #if defined(CONFIG_NFS_V4_1)
 447                 case -NFS4ERR_BADSESSION:
 448                 case -NFS4ERR_BADSLOT:
 449                 case -NFS4ERR_BAD_HIGH_SLOT:
 450                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 451                 case -NFS4ERR_DEADSESSION:
 452                 case -NFS4ERR_SEQ_FALSE_RETRY:
 453                 case -NFS4ERR_SEQ_MISORDERED:
 454                         dprintk("%s ERROR: %d Reset session\n", __func__,
 455                                 errorcode);
 456                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 457                         goto wait_on_recovery;
 458 #endif /* defined(CONFIG_NFS_V4_1) */
 459                 case -NFS4ERR_FILE_OPEN:
 460                         if (exception->timeout > HZ) {
 461                                 /* We have retried a decent amount, time to
 462                                  * fail
 463                                  */
 464                                 ret = -EBUSY;
 465                                 break;
 466                         }
 467                 case -NFS4ERR_DELAY:
 468                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 469                 case -NFS4ERR_GRACE:
 470                 case -NFS4ERR_LAYOUTTRYLATER:
 471                 case -NFS4ERR_RECALLCONFLICT:
 472                         exception->delay = 1;
 473                         return 0;
 474
 475                 case -NFS4ERR_RETRY_UNCACHED_REP:
 476                 case -NFS4ERR_OLD_STATEID:
 477                         exception->retry = 1;
 478                         break;
 479                 case -NFS4ERR_BADOWNER:
 480                         /* The following works around a Linux server bug! */
 481                 case -NFS4ERR_BADNAME:
 482                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 483                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 484                                 exception->retry = 1;
 485                                 printk(KERN_WARNING "NFS: v4 server %s "
 486                                                 "does not accept raw "
 487                                                 "uid/gids. "
 488                                                 "Reenabling the idmapper.\n",
 489                                                 server->nfs_client->cl_hostname);
 490                         }
 491         }
 492         /* We failed to handle the error */
 493         return nfs4_map_errors(ret);
 494 wait_on_recovery:
 495         exception->recovering = 1;
 496         return 0;
 497 }
 498
 499 /* This is the error handling routine for processes that are allowed
 500  * to sleep.
 501  */
 502 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 503 {
 504         struct nfs_client *clp = server->nfs_client;
 505         int ret;
 506
 507         ret = nfs4_do_handle_exception(server, errorcode, exception);
 508         if (exception->delay) {
 509                 ret = nfs4_delay(server->client, &exception->timeout);
 510                 goto out_retry;
 511         }
 512         if (exception->recovering) {
 513                 ret = nfs4_wait_clnt_recover(clp);
 514                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 515                         return -EIO;
 516                 goto out_retry;
 517         }
 518         return ret;
 519 out_retry:
 520         if (ret == 0)
 521                 exception->retry = 1;
 522         return ret;
 523 }
 524
 525 static int
 526 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 527                 int errorcode, struct nfs4_exception *exception)
 528 {
 529         struct nfs_client *clp = server->nfs_client;
 530         int ret;
 531
 532         ret = nfs4_do_handle_exception(server, errorcode, exception);
 533         if (exception->delay) {
 534                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 535                 goto out_retry;
 536         }
 537         if (exception->recovering) {
 538                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 539                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 540                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 541                 goto out_retry;
 542         }
 543         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 544                 ret = -EIO;
 545         return ret;
 546 out_retry:
 547         if (ret == 0)
 548                 exception->retry = 1;
 549         return ret;
 550 }
 551
 552 static int
 553 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 554                         struct nfs4_state *state, long *timeout)
 555 {
 556         struct nfs4_exception exception = {
 557                 .state = state,
 558         };
 559
 560         if (task->tk_status >= 0)
 561                 return 0;
 562         if (timeout)
 563                 exception.timeout = *timeout;
 564         task->tk_status = nfs4_async_handle_exception(task, server,
 565                         task->tk_status,
 566                         &exception);
 567         if (exception.delay && timeout)
 568                 *timeout = exception.timeout;
 569         if (exception.retry)
 570                 return -EAGAIN;
 571         return 0;
 572 }
 573
 574 /*
 575  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 576  * or 'false' otherwise.
 577  */
 578 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 579 {
 580         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 581
 582         if (flavor == RPC_AUTH_GSS_KRB5I ||
 583             flavor == RPC_AUTH_GSS_KRB5P)
 584                 return true;
 585
 586         return false;
 587 }
 588
 589 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 590 {
 591         spin_lock(&clp->cl_lock);
 592         if (time_before(clp->cl_last_renewal,timestamp))
 593                 clp->cl_last_renewal = timestamp;
 594         spin_unlock(&clp->cl_lock);
 595 }
 596
 597 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 598 {
 599         struct nfs_client *clp = server->nfs_client;
 600
 601         if (!nfs4_has_session(clp))
 602                 do_renew_lease(clp, timestamp);
 603 }
 604
 605 struct nfs4_call_sync_data {
 606         const struct nfs_server *seq_server;
 607         struct nfs4_sequence_args *seq_args;
 608         struct nfs4_sequence_res *seq_res;
 609 };
 610
 611 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 612                         struct nfs4_sequence_res *res, int cache_reply)
 613 {
 614         args->sa_slot = NULL;
 615         args->sa_cache_this = cache_reply;
 616         args->sa_privileged = 0;
 617
 618         res->sr_slot = NULL;
 619 }
 620
 621 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 622 {
 623         args->sa_privileged = 1;
 624 }
 625
 626 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 627                          struct nfs4_sequence_args *args,
 628                          struct nfs4_sequence_res *res,
 629                          struct rpc_task *task)
 630 {
 631         struct nfs4_slot *slot;
 632
 633         /* slot already allocated? */
 634         if (res->sr_slot != NULL)
 635                 goto out_start;
 636
 637         spin_lock(&tbl->slot_tbl_lock);
 638         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 639                 goto out_sleep;
 640
 641         slot = nfs4_alloc_slot(tbl);
 642         if (IS_ERR(slot)) {
 643                 if (slot == ERR_PTR(-ENOMEM))
 644                         task->tk_timeout = HZ >> 2;
 645                 goto out_sleep;
 646         }
 647         spin_unlock(&tbl->slot_tbl_lock);
 648
 649         slot->privileged = args->sa_privileged ? 1 : 0;
 650         args->sa_slot = slot;
 651         res->sr_slot = slot;
 652
 653 out_start:
 654         rpc_call_start(task);
 655         return 0;
 656
 657 out_sleep:
 658         if (args->sa_privileged)
 659                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 660                                 NULL, RPC_PRIORITY_PRIVILEGED);
 661         else
 662                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 663         spin_unlock(&tbl->slot_tbl_lock);
 664         return -EAGAIN;
 665 }
 666 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 667
 668 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 669 {
 670         struct nfs4_slot *slot = res->sr_slot;
 671         struct nfs4_slot_table *tbl;
 672
 673         tbl = slot->table;
 674         spin_lock(&tbl->slot_tbl_lock);
 675         if (!nfs41_wake_and_assign_slot(tbl, slot))
 676                 nfs4_free_slot(tbl, slot);
 677         spin_unlock(&tbl->slot_tbl_lock);
 678
 679         res->sr_slot = NULL;
 680 }
 681
 682 static int nfs40_sequence_done(struct rpc_task *task,
 683                                struct nfs4_sequence_res *res)
 684 {
 685         if (res->sr_slot != NULL)
 686                 nfs40_sequence_free_slot(res);
 687         return 1;
 688 }
 689
 690 #if defined(CONFIG_NFS_V4_1)
 691
 692 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 693 {
 694         struct nfs4_session *session;
 695         struct nfs4_slot_table *tbl;
 696         struct nfs4_slot *slot = res->sr_slot;
 697         bool send_new_highest_used_slotid = false;
 698
 699         tbl = slot->table;
 700         session = tbl->session;
 701
 702         /* Bump the slot sequence number */
 703         if (slot->seq_done)
 704                 slot->seq_nr++;
 705         slot->seq_done = 0;
 706
 707         spin_lock(&tbl->slot_tbl_lock);
 708         /* Be nice to the server: try to ensure that the last transmitted
 709          * value for highest_user_slotid <= target_highest_slotid
 710          */
 711         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 712                 send_new_highest_used_slotid = true;
 713
 714         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 715                 send_new_highest_used_slotid = false;
 716                 goto out_unlock;
 717         }
 718         nfs4_free_slot(tbl, slot);
 719
 720         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 721                 send_new_highest_used_slotid = false;
 722 out_unlock:
 723         spin_unlock(&tbl->slot_tbl_lock);
 724         res->sr_slot = NULL;
 725         if (send_new_highest_used_slotid)
 726                 nfs41_notify_server(session->clp);
 727         if (waitqueue_active(&tbl->slot_waitq))
 728                 wake_up_all(&tbl->slot_waitq);
 729 }
 730
 731 static int nfs41_sequence_process(struct rpc_task *task,
 732                 struct nfs4_sequence_res *res)
 733 {
 734         struct nfs4_session *session;
 735         struct nfs4_slot *slot = res->sr_slot;
 736         struct nfs_client *clp;
 737         bool interrupted = false;
 738         int ret = 1;
 739
 740         if (slot == NULL)
 741                 goto out_noaction;
 742         /* don't increment the sequence number if the task wasn't sent */
 743         if (!RPC_WAS_SENT(task))
 744                 goto out;
 745
 746         session = slot->table->session;
 747
 748         if (slot->interrupted) {
 749                 slot->interrupted = 0;
 750                 interrupted = true;
 751         }
 752
 753         trace_nfs4_sequence_done(session, res);
 754         /* Check the SEQUENCE operation status */
 755         switch (res->sr_status) {
 756         case 0:
 757                 /* If previous op on slot was interrupted and we reused
 758                  * the seq# and got a reply from the cache, then retry
 759                  */
 760                 if (task->tk_status == -EREMOTEIO && interrupted) {
 761                         ++slot->seq_nr;
 762                         goto retry_nowait;
 763                 }
 764                 /* Update the slot's sequence and clientid lease timer */
 765                 slot->seq_done = 1;
 766                 clp = session->clp;
 767                 do_renew_lease(clp, res->sr_timestamp);
 768                 /* Check sequence flags */
 769                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 770                                 !!slot->privileged);
 771                 nfs41_update_target_slotid(slot->table, slot, res);
 772                 break;
 773         case 1:
 774                 /*
 775                  * sr_status remains 1 if an RPC level error occurred.
 776                  * The server may or may not have processed the sequence
 777                  * operation..
 778                  * Mark the slot as having hosted an interrupted RPC call.
 779                  */
 780                 slot->interrupted = 1;
 781                 goto out;
 782         case -NFS4ERR_DELAY:
 783                 /* The server detected a resend of the RPC call and
 784                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 785                  * of RFC5661.
 786                  */
 787                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 788                         __func__,
 789                         slot->slot_nr,
 790                         slot->seq_nr);
 791                 goto out_retry;
 792         case -NFS4ERR_BADSLOT:
 793                 /*
 794                  * The slot id we used was probably retired. Try again
 795                  * using a different slot id.
 796                  */
 797                 goto retry_nowait;
 798         case -NFS4ERR_SEQ_MISORDERED:
 799                 /*
 800                  * Was the last operation on this sequence interrupted?
 801                  * If so, retry after bumping the sequence number.
 802                  */
 803                 if (interrupted) {
 804                         ++slot->seq_nr;
 805                         goto retry_nowait;
 806                 }
 807                 /*
 808                  * Could this slot have been previously retired?
 809                  * If so, then the server may be expecting seq_nr = 1!
 810                  */
 811                 if (slot->seq_nr != 1) {
 812                         slot->seq_nr = 1;
 813                         goto retry_nowait;
 814                 }
 815                 break;
 816         case -NFS4ERR_SEQ_FALSE_RETRY:
 817                 ++slot->seq_nr;
 818                 goto retry_nowait;
 819         case -NFS4ERR_DEADSESSION:
 820         case -NFS4ERR_BADSESSION:
 821                 nfs4_schedule_session_recovery(session, res->sr_status);
 822                 goto retry_nowait;
 823         default:
 824                 /* Just update the slot sequence no. */
 825                 slot->seq_done = 1;
 826         }
 827 out:
 828         /* The session may be reset by one of the error handlers. */
 829         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 830 out_noaction:
 831         return ret;
 832 retry_nowait:
 833         if (rpc_restart_call_prepare(task)) {
 834                 nfs41_sequence_free_slot(res);
 835                 task->tk_status = 0;
 836                 ret = 0;
 837         }
 838         goto out;
 839 out_retry:
 840         if (!rpc_restart_call(task))
 841                 goto out;
 842         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 843         return 0;
 844 }
 845
 846 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 847 {
 848         if (!nfs41_sequence_process(task, res))
 849                 return 0;
 850         if (res->sr_slot != NULL)
 851                 nfs41_sequence_free_slot(res);
 852         return 1;
 853
 854 }
 855 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 856
 857 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 858 {
 859         if (res->sr_slot == NULL)
 860                 return 1;
 861         if (res->sr_slot->table->session != NULL)
 862                 return nfs41_sequence_process(task, res);
 863         return nfs40_sequence_done(task, res);
 864 }
 865
 866 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 867 {
 868         if (res->sr_slot != NULL) {
 869                 if (res->sr_slot->table->session != NULL)
 870                         nfs41_sequence_free_slot(res);
 871                 else
 872                         nfs40_sequence_free_slot(res);
 873         }
 874 }
 875
 876 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 877 {
 878         if (res->sr_slot == NULL)
 879                 return 1;
 880         if (!res->sr_slot->table->session)
 881                 return nfs40_sequence_done(task, res);
 882         return nfs41_sequence_done(task, res);
 883 }
 884 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 885
 886 int nfs41_setup_sequence(struct nfs4_session *session,
 887                                 struct nfs4_sequence_args *args,
 888                                 struct nfs4_sequence_res *res,
 889                                 struct rpc_task *task)
 890 {
 891         struct nfs4_slot *slot;
 892         struct nfs4_slot_table *tbl;
 893
 894         dprintk("--> %s\n", __func__);
 895         /* slot already allocated? */
 896         if (res->sr_slot != NULL)
 897                 goto out_success;
 898
 899         tbl = &session->fc_slot_table;
 900
 901         task->tk_timeout = 0;
 902
 903         spin_lock(&tbl->slot_tbl_lock);
 904         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 905             !args->sa_privileged) {
 906                 /* The state manager will wait until the slot table is empty */
 907                 dprintk("%s session is draining\n", __func__);
 908                 goto out_sleep;
 909         }
 910
 911         slot = nfs4_alloc_slot(tbl);
 912         if (IS_ERR(slot)) {
 913                 /* If out of memory, try again in 1/4 second */
 914                 if (slot == ERR_PTR(-ENOMEM))
 915                         task->tk_timeout = HZ >> 2;
 916                 dprintk("<-- %s: no free slots\n", __func__);
 917                 goto out_sleep;
 918         }
 919         spin_unlock(&tbl->slot_tbl_lock);
 920
 921         slot->privileged = args->sa_privileged ? 1 : 0;
 922         args->sa_slot = slot;
 923
 924         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 925                         slot->slot_nr, slot->seq_nr);
 926
 927         res->sr_slot = slot;
 928         res->sr_timestamp = jiffies;
 929         res->sr_status_flags = 0;
 930         /*
 931          * sr_status is only set in decode_sequence, and so will remain
 932          * set to 1 if an rpc level failure occurs.
 933          */
 934         res->sr_status = 1;
 935         trace_nfs4_setup_sequence(session, args);
 936 out_success:
 937         rpc_call_start(task);
 938         return 0;
 939 out_sleep:
 940         /* Privileged tasks are queued with top priority */
 941         if (args->sa_privileged)
 942                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 943                                 NULL, RPC_PRIORITY_PRIVILEGED);
 944         else
 945                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 946         spin_unlock(&tbl->slot_tbl_lock);
 947         return -EAGAIN;
 948 }
 949 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 950
 951 static int nfs4_setup_sequence(const struct nfs_server *server,
 952                                struct nfs4_sequence_args *args,
 953                                struct nfs4_sequence_res *res,
 954                                struct rpc_task *task)
 955 {
 956         struct nfs4_session *session = nfs4_get_session(server);
 957         int ret = 0;
 958
 959         if (!session)
 960                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 961                                             args, res, task);
 962
 963         dprintk("--> %s clp %p session %p sr_slot %u\n",
 964                 __func__, session->clp, session, res->sr_slot ?
 965                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 966
 967         ret = nfs41_setup_sequence(session, args, res, task);
 968
 969         dprintk("<-- %s status=%d\n", __func__, ret);
 970         return ret;
 971 }
 972
 973 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 974 {
 975         struct nfs4_call_sync_data *data = calldata;
 976         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 977
 978         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 979
 980         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 981 }
 982
 983 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 984 {
 985         struct nfs4_call_sync_data *data = calldata;
 986
 987         nfs41_sequence_done(task, data->seq_res);
 988 }
 989
 990 static const struct rpc_call_ops nfs41_call_sync_ops = {
 991         .rpc_call_prepare = nfs41_call_sync_prepare,
 992         .rpc_call_done = nfs41_call_sync_done,
 993 };
 994
 995 #else   /* !CONFIG_NFS_V4_1 */
 996
 997 static int nfs4_setup_sequence(const struct nfs_server *server,
 998                                struct nfs4_sequence_args *args,
 999                                struct nfs4_sequence_res *res,
1000                                struct rpc_task *task)
1001 {
1002         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
1003                                     args, res, task);
1004 }
1005
1006 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1007 {
1008         return nfs40_sequence_done(task, res);
1009 }
1010
1011 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1012 {
1013         if (res->sr_slot != NULL)
1014                 nfs40_sequence_free_slot(res);
1015 }
1016
1017 int nfs4_sequence_done(struct rpc_task *task,
1018                        struct nfs4_sequence_res *res)
1019 {
1020         return nfs40_sequence_done(task, res);
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1023
1024 #endif  /* !CONFIG_NFS_V4_1 */
1025
1026 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1027 {
1028         struct nfs4_call_sync_data *data = calldata;
1029         nfs4_setup_sequence(data->seq_server,
1030                                 data->seq_args, data->seq_res, task);
1031 }
1032
1033 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1034 {
1035         struct nfs4_call_sync_data *data = calldata;
1036         nfs4_sequence_done(task, data->seq_res);
1037 }
1038
1039 static const struct rpc_call_ops nfs40_call_sync_ops = {
1040         .rpc_call_prepare = nfs40_call_sync_prepare,
1041         .rpc_call_done = nfs40_call_sync_done,
1042 };
1043
1044 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1045                                    struct nfs_server *server,
1046                                    struct rpc_message *msg,
1047                                    struct nfs4_sequence_args *args,
1048                                    struct nfs4_sequence_res *res)
1049 {
1050         int ret;
1051         struct rpc_task *task;
1052         struct nfs_client *clp = server->nfs_client;
1053         struct nfs4_call_sync_data data = {
1054                 .seq_server = server,
1055                 .seq_args = args,
1056                 .seq_res = res,
1057         };
1058         struct rpc_task_setup task_setup = {
1059                 .rpc_client = clnt,
1060                 .rpc_message = msg,
1061                 .callback_ops = clp->cl_mvops->call_sync_ops,
1062                 .callback_data = &data
1063         };
1064
1065         task = rpc_run_task(&task_setup);
1066         if (IS_ERR(task))
1067                 ret = PTR_ERR(task);
1068         else {
1069                 ret = task->tk_status;
1070                 rpc_put_task(task);
1071         }
1072         return ret;
1073 }
1074
1075 int nfs4_call_sync(struct rpc_clnt *clnt,
1076                    struct nfs_server *server,
1077                    struct rpc_message *msg,
1078                    struct nfs4_sequence_args *args,
1079                    struct nfs4_sequence_res *res,
1080                    int cache_reply)
1081 {
1082         nfs4_init_sequence(args, res, cache_reply);
1083         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1084 }
1085
1086 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1087 {
1088         struct nfs_inode *nfsi = NFS_I(dir);
1089
1090         spin_lock(&dir->i_lock);
1091         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1092         if (cinfo->atomic && cinfo->before == dir->i_version) {
1093                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1094                 nfsi->attrtimeo_timestamp = jiffies;
1095         } else {
1096                 nfs_force_lookup_revalidate(dir);
1097                 if (cinfo->before != dir->i_version)
1098                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1099                                 NFS_INO_INVALID_ACL;
1100         }
1101         dir->i_version = cinfo->after;
1102         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1103         nfs_fscache_invalidate(dir);
1104         spin_unlock(&dir->i_lock);
1105 }
1106
1107 struct nfs4_opendata {
1108         struct kref kref;
1109         struct nfs_openargs o_arg;
1110         struct nfs_openres o_res;
1111         struct nfs_open_confirmargs c_arg;
1112         struct nfs_open_confirmres c_res;
1113         struct nfs4_string owner_name;
1114         struct nfs4_string group_name;
1115         struct nfs4_label *a_label;
1116         struct nfs_fattr f_attr;
1117         struct nfs4_label *f_label;
1118         struct dentry *dir;
1119         struct dentry *dentry;
1120         struct nfs4_state_owner *owner;
1121         struct nfs4_state *state;
1122         struct iattr attrs;
1123         unsigned long timestamp;
1124         unsigned int rpc_done : 1;
1125         unsigned int file_created : 1;
1126         unsigned int is_recover : 1;
1127         int rpc_status;
1128         int cancelled;
1129 };
1130
1131 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1132                 int err, struct nfs4_exception *exception)
1133 {
1134         if (err != -EINVAL)
1135                 return false;
1136         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1137                 return false;
1138         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1139         exception->retry = 1;
1140         return true;
1141 }
1142
1143 static u32
1144 nfs4_map_atomic_open_share(struct nfs_server *server,
1145                 fmode_t fmode, int openflags)
1146 {
1147         u32 res = 0;
1148
1149         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1150         case FMODE_READ:
1151                 res = NFS4_SHARE_ACCESS_READ;
1152                 break;
1153         case FMODE_WRITE:
1154                 res = NFS4_SHARE_ACCESS_WRITE;
1155                 break;
1156         case FMODE_READ|FMODE_WRITE:
1157                 res = NFS4_SHARE_ACCESS_BOTH;
1158         }
1159         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1160                 goto out;
1161         /* Want no delegation if we're using O_DIRECT */
1162         if (openflags & O_DIRECT)
1163                 res |= NFS4_SHARE_WANT_NO_DELEG;
1164 out:
1165         return res;
1166 }
1167
1168 static enum open_claim_type4
1169 nfs4_map_atomic_open_claim(struct nfs_server *server,
1170                 enum open_claim_type4 claim)
1171 {
1172         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1173                 return claim;
1174         switch (claim) {
1175         default:
1176                 return claim;
1177         case NFS4_OPEN_CLAIM_FH:
1178                 return NFS4_OPEN_CLAIM_NULL;
1179         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1180                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1181         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1182                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1183         }
1184 }
1185
1186 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1187 {
1188         p->o_res.f_attr = &p->f_attr;
1189         p->o_res.f_label = p->f_label;
1190         p->o_res.seqid = p->o_arg.seqid;
1191         p->c_res.seqid = p->c_arg.seqid;
1192         p->o_res.server = p->o_arg.server;
1193         p->o_res.access_request = p->o_arg.access;
1194         nfs_fattr_init(&p->f_attr);
1195         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1196 }
1197
1198 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1199                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1200                 const struct iattr *attrs,
1201                 struct nfs4_label *label,
1202                 enum open_claim_type4 claim,
1203                 gfp_t gfp_mask)
1204 {
1205         struct dentry *parent = dget_parent(dentry);
1206         struct inode *dir = d_inode(parent);
1207         struct nfs_server *server = NFS_SERVER(dir);
1208         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1209         struct nfs4_opendata *p;
1210
1211         p = kzalloc(sizeof(*p), gfp_mask);
1212         if (p == NULL)
1213                 goto err;
1214
1215         p->f_label = nfs4_label_alloc(server, gfp_mask);
1216         if (IS_ERR(p->f_label))
1217                 goto err_free_p;
1218
1219         p->a_label = nfs4_label_alloc(server, gfp_mask);
1220         if (IS_ERR(p->a_label))
1221                 goto err_free_f;
1222
1223         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1224         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1225         if (IS_ERR(p->o_arg.seqid))
1226                 goto err_free_label;
1227         nfs_sb_active(dentry->d_sb);
1228         p->dentry = dget(dentry);
1229         p->dir = parent;
1230         p->owner = sp;
1231         atomic_inc(&sp->so_count);
1232         p->o_arg.open_flags = flags;
1233         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1234         p->o_arg.umask = current_umask();
1235         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1236         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1237                         fmode, flags);
1238         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1239          * will return permission denied for all bits until close */
1240         if (!(flags & O_EXCL)) {
1241                 /* ask server to check for all possible rights as results
1242                  * are cached */
1243                 switch (p->o_arg.claim) {
1244                 default:
1245                         break;
1246                 case NFS4_OPEN_CLAIM_NULL:
1247                 case NFS4_OPEN_CLAIM_FH:
1248                         p->o_arg.access = NFS4_ACCESS_READ |
1249                                 NFS4_ACCESS_MODIFY |
1250                                 NFS4_ACCESS_EXTEND |
1251                                 NFS4_ACCESS_EXECUTE;
1252                 }
1253         }
1254         p->o_arg.clientid = server->nfs_client->cl_clientid;
1255         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1256         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1257         p->o_arg.name = &dentry->d_name;
1258         p->o_arg.server = server;
1259         p->o_arg.bitmask = nfs4_bitmask(server, label);
1260         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1261         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1262         switch (p->o_arg.claim) {
1263         case NFS4_OPEN_CLAIM_NULL:
1264         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1265         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1266                 p->o_arg.fh = NFS_FH(dir);
1267                 break;
1268         case NFS4_OPEN_CLAIM_PREVIOUS:
1269         case NFS4_OPEN_CLAIM_FH:
1270         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1271         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1272                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1273         }
1274         if (attrs != NULL && attrs->ia_valid != 0) {
1275                 __u32 verf[2];
1276
1277                 p->o_arg.u.attrs = &p->attrs;
1278                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1279
1280                 verf[0] = jiffies;
1281                 verf[1] = current->pid;
1282                 memcpy(p->o_arg.u.verifier.data, verf,
1283                                 sizeof(p->o_arg.u.verifier.data));
1284         }
1285         p->c_arg.fh = &p->o_res.fh;
1286         p->c_arg.stateid = &p->o_res.stateid;
1287         p->c_arg.seqid = p->o_arg.seqid;
1288         nfs4_init_opendata_res(p);
1289         kref_init(&p->kref);
1290         return p;
1291
1292 err_free_label:
1293         nfs4_label_free(p->a_label);
1294 err_free_f:
1295         nfs4_label_free(p->f_label);
1296 err_free_p:
1297         kfree(p);
1298 err:
1299         dput(parent);
1300         return NULL;
1301 }
1302
1303 static void nfs4_opendata_free(struct kref *kref)
1304 {
1305         struct nfs4_opendata *p = container_of(kref,
1306                         struct nfs4_opendata, kref);
1307         struct super_block *sb = p->dentry->d_sb;
1308
1309         nfs_free_seqid(p->o_arg.seqid);
1310         nfs4_sequence_free_slot(&p->o_res.seq_res);
1311         if (p->state != NULL)
1312                 nfs4_put_open_state(p->state);
1313         nfs4_put_state_owner(p->owner);
1314
1315         nfs4_label_free(p->a_label);
1316         nfs4_label_free(p->f_label);
1317
1318         dput(p->dir);
1319         dput(p->dentry);
1320         nfs_sb_deactive(sb);
1321         nfs_fattr_free_names(&p->f_attr);
1322         kfree(p->f_attr.mdsthreshold);
1323         kfree(p);
1324 }
1325
1326 static void nfs4_opendata_put(struct nfs4_opendata *p)
1327 {
1328         if (p != NULL)
1329                 kref_put(&p->kref, nfs4_opendata_free);
1330 }
1331
1332 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1333 {
1334         int ret;
1335
1336         ret = rpc_wait_for_completion_task(task);
1337         return ret;
1338 }
1339
1340 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1341                 fmode_t fmode)
1342 {
1343         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1344         case FMODE_READ|FMODE_WRITE:
1345                 return state->n_rdwr != 0;
1346         case FMODE_WRITE:
1347                 return state->n_wronly != 0;
1348         case FMODE_READ:
1349                 return state->n_rdonly != 0;
1350         }
1351         WARN_ON_ONCE(1);
1352         return false;
1353 }
1354
1355 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1356 {
1357         int ret = 0;
1358
1359         if (open_mode & (O_EXCL|O_TRUNC))
1360                 goto out;
1361         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1362                 case FMODE_READ:
1363                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1364                                 && state->n_rdonly != 0;
1365                         break;
1366                 case FMODE_WRITE:
1367                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1368                                 && state->n_wronly != 0;
1369                         break;
1370                 case FMODE_READ|FMODE_WRITE:
1371                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1372                                 && state->n_rdwr != 0;
1373         }
1374 out:
1375         return ret;
1376 }
1377
1378 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1379                 enum open_claim_type4 claim)
1380 {
1381         if (delegation == NULL)
1382                 return 0;
1383         if ((delegation->type & fmode) != fmode)
1384                 return 0;
1385         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1386                 return 0;
1387         switch (claim) {
1388         case NFS4_OPEN_CLAIM_NULL:
1389         case NFS4_OPEN_CLAIM_FH:
1390                 break;
1391         case NFS4_OPEN_CLAIM_PREVIOUS:
1392                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1393                         break;
1394         default:
1395                 return 0;
1396         }
1397         nfs_mark_delegation_referenced(delegation);
1398         return 1;
1399 }
1400
1401 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1402 {
1403         switch (fmode) {
1404                 case FMODE_WRITE:
1405                         state->n_wronly++;
1406                         break;
1407                 case FMODE_READ:
1408                         state->n_rdonly++;
1409                         break;
1410                 case FMODE_READ|FMODE_WRITE:
1411                         state->n_rdwr++;
1412         }
1413         nfs4_state_set_mode_locked(state, state->state | fmode);
1414 }
1415
1416 #ifdef CONFIG_NFS_V4_1
1417 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1418 {
1419         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1420                 return true;
1421         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1422                 return true;
1423         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1424                 return true;
1425         return false;
1426 }
1427 #endif /* CONFIG_NFS_V4_1 */
1428
1429 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1430 {
1431         struct nfs_client *clp = state->owner->so_server->nfs_client;
1432         bool need_recover = false;
1433
1434         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1435                 need_recover = true;
1436         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1437                 need_recover = true;
1438         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1439                 need_recover = true;
1440         if (need_recover)
1441                 nfs4_state_mark_reclaim_nograce(clp, state);
1442 }
1443
1444 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1445                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1446 {
1447         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1448                 return true;
1449         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1450                 nfs4_stateid_copy(freeme, &state->open_stateid);
1451                 nfs_test_and_clear_all_open_stateid(state);
1452                 return true;
1453         }
1454         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1455                 return true;
1456         return false;
1457 }
1458
1459 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1460 {
1461         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1462                 return;
1463         if (state->n_wronly)
1464                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1465         if (state->n_rdonly)
1466                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1467         if (state->n_rdwr)
1468                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1469         set_bit(NFS_OPEN_STATE, &state->flags);
1470 }
1471
1472 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1473                 nfs4_stateid *stateid, fmode_t fmode)
1474 {
1475         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1476         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1477         case FMODE_WRITE:
1478                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1479                 break;
1480         case FMODE_READ:
1481                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1482                 break;
1483         case 0:
1484                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1485                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1486                 clear_bit(NFS_OPEN_STATE, &state->flags);
1487         }
1488         if (stateid == NULL)
1489                 return;
1490         /* Handle OPEN+OPEN_DOWNGRADE races */
1491         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1492             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1493                 nfs_resync_open_stateid_locked(state);
1494                 return;
1495         }
1496         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1497                 nfs4_stateid_copy(&state->stateid, stateid);
1498         nfs4_stateid_copy(&state->open_stateid, stateid);
1499 }
1500
1501 static void nfs_clear_open_stateid(struct nfs4_state *state,
1502         nfs4_stateid *arg_stateid,
1503         nfs4_stateid *stateid, fmode_t fmode)
1504 {
1505         write_seqlock(&state->seqlock);
1506         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1507         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1508                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1509         write_sequnlock(&state->seqlock);
1510         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1511                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1512 }
1513
1514 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1515                 const nfs4_stateid *stateid, fmode_t fmode,
1516                 nfs4_stateid *freeme)
1517 {
1518         switch (fmode) {
1519                 case FMODE_READ:
1520                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1521                         break;
1522                 case FMODE_WRITE:
1523                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1524                         break;
1525                 case FMODE_READ|FMODE_WRITE:
1526                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1527         }
1528         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1529                 return;
1530         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1531                 nfs4_stateid_copy(&state->stateid, stateid);
1532         nfs4_stateid_copy(&state->open_stateid, stateid);
1533 }
1534
1535 static void __update_open_stateid(struct nfs4_state *state,
1536                 const nfs4_stateid *open_stateid,
1537                 const nfs4_stateid *deleg_stateid,
1538                 fmode_t fmode,
1539                 nfs4_stateid *freeme)
1540 {
1541         /*
1542          * Protect the call to nfs4_state_set_mode_locked and
1543          * serialise the stateid update
1544          */
1545         spin_lock(&state->owner->so_lock);
1546         write_seqlock(&state->seqlock);
1547         if (deleg_stateid != NULL) {
1548                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1549                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1550         }
1551         if (open_stateid != NULL)
1552                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1553         write_sequnlock(&state->seqlock);
1554         update_open_stateflags(state, fmode);
1555         spin_unlock(&state->owner->so_lock);
1556 }
1557
1558 static int update_open_stateid(struct nfs4_state *state,
1559                 const nfs4_stateid *open_stateid,
1560                 const nfs4_stateid *delegation,
1561                 fmode_t fmode)
1562 {
1563         struct nfs_server *server = NFS_SERVER(state->inode);
1564         struct nfs_client *clp = server->nfs_client;
1565         struct nfs_inode *nfsi = NFS_I(state->inode);
1566         struct nfs_delegation *deleg_cur;
1567         nfs4_stateid freeme = { };
1568         int ret = 0;
1569
1570         fmode &= (FMODE_READ|FMODE_WRITE);
1571
1572         rcu_read_lock();
1573         deleg_cur = rcu_dereference(nfsi->delegation);
1574         if (deleg_cur == NULL)
1575                 goto no_delegation;
1576
1577         spin_lock(&deleg_cur->lock);
1578         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1579            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1580             (deleg_cur->type & fmode) != fmode)
1581                 goto no_delegation_unlock;
1582
1583         if (delegation == NULL)
1584                 delegation = &deleg_cur->stateid;
1585         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1586                 goto no_delegation_unlock;
1587
1588         nfs_mark_delegation_referenced(deleg_cur);
1589         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1590                         fmode, &freeme);
1591         ret = 1;
1592 no_delegation_unlock:
1593         spin_unlock(&deleg_cur->lock);
1594 no_delegation:
1595         rcu_read_unlock();
1596
1597         if (!ret && open_stateid != NULL) {
1598                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1599                 ret = 1;
1600         }
1601         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1602                 nfs4_schedule_state_manager(clp);
1603         if (freeme.type != 0)
1604                 nfs4_test_and_free_stateid(server, &freeme,
1605                                 state->owner->so_cred);
1606
1607         return ret;
1608 }
1609
1610 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1611                 const nfs4_stateid *stateid)
1612 {
1613         struct nfs4_state *state = lsp->ls_state;
1614         bool ret = false;
1615
1616         spin_lock(&state->state_lock);
1617         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1618                 goto out_noupdate;
1619         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1620                 goto out_noupdate;
1621         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1622         ret = true;
1623 out_noupdate:
1624         spin_unlock(&state->state_lock);
1625         return ret;
1626 }
1627
1628 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1629 {
1630         struct nfs_delegation *delegation;
1631
1632         rcu_read_lock();
1633         delegation = rcu_dereference(NFS_I(inode)->delegation);
1634         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1635                 rcu_read_unlock();
1636                 return;
1637         }
1638         rcu_read_unlock();
1639         nfs4_inode_return_delegation(inode);
1640 }
1641
1642 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1643 {
1644         struct nfs4_state *state = opendata->state;
1645         struct nfs_inode *nfsi = NFS_I(state->inode);
1646         struct nfs_delegation *delegation;
1647         int open_mode = opendata->o_arg.open_flags;
1648         fmode_t fmode = opendata->o_arg.fmode;
1649         enum open_claim_type4 claim = opendata->o_arg.claim;
1650         nfs4_stateid stateid;
1651         int ret = -EAGAIN;
1652
1653         for (;;) {
1654                 spin_lock(&state->owner->so_lock);
1655                 if (can_open_cached(state, fmode, open_mode)) {
1656                         update_open_stateflags(state, fmode);
1657                         spin_unlock(&state->owner->so_lock);
1658                         goto out_return_state;
1659                 }
1660                 spin_unlock(&state->owner->so_lock);
1661                 rcu_read_lock();
1662                 delegation = rcu_dereference(nfsi->delegation);
1663                 if (!can_open_delegated(delegation, fmode, claim)) {
1664                         rcu_read_unlock();
1665                         break;
1666                 }
1667                 /* Save the delegation */
1668                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1669                 rcu_read_unlock();
1670                 nfs_release_seqid(opendata->o_arg.seqid);
1671                 if (!opendata->is_recover) {
1672                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1673                         if (ret != 0)
1674                                 goto out;
1675                 }
1676                 ret = -EAGAIN;
1677
1678                 /* Try to update the stateid using the delegation */
1679                 if (update_open_stateid(state, NULL, &stateid, fmode))
1680                         goto out_return_state;
1681         }
1682 out:
1683         return ERR_PTR(ret);
1684 out_return_state:
1685         atomic_inc(&state->count);
1686         return state;
1687 }
1688
1689 static void
1690 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1691 {
1692         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1693         struct nfs_delegation *delegation;
1694         int delegation_flags = 0;
1695
1696         rcu_read_lock();
1697         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1698         if (delegation)
1699                 delegation_flags = delegation->flags;
1700         rcu_read_unlock();
1701         switch (data->o_arg.claim) {
1702         default:
1703                 break;
1704         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1705         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1706                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1707                                    "returning a delegation for "
1708                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1709                                    clp->cl_hostname);
1710                 return;
1711         }
1712         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1713                 nfs_inode_set_delegation(state->inode,
1714                                          data->owner->so_cred,
1715                                          &data->o_res);
1716         else
1717                 nfs_inode_reclaim_delegation(state->inode,
1718                                              data->owner->so_cred,
1719                                              &data->o_res);
1720 }
1721
1722 /*
1723  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1724  * and update the nfs4_state.
1725  */
1726 static struct nfs4_state *
1727 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1728 {
1729         struct inode *inode = data->state->inode;
1730         struct nfs4_state *state = data->state;
1731         int ret;
1732
1733         if (!data->rpc_done) {
1734                 if (data->rpc_status) {
1735                         ret = data->rpc_status;
1736                         goto err;
1737                 }
1738                 /* cached opens have already been processed */
1739                 goto update;
1740         }
1741
1742         ret = nfs_refresh_inode(inode, &data->f_attr);
1743         if (ret)
1744                 goto err;
1745
1746         if (data->o_res.delegation_type != 0)
1747                 nfs4_opendata_check_deleg(data, state);
1748 update:
1749         update_open_stateid(state, &data->o_res.stateid, NULL,
1750                             data->o_arg.fmode);
1751         atomic_inc(&state->count);
1752
1753         return state;
1754 err:
1755         return ERR_PTR(ret);
1756
1757 }
1758
1759 static struct nfs4_state *
1760 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1761 {
1762         struct inode *inode;
1763         struct nfs4_state *state = NULL;
1764         int ret;
1765
1766         if (!data->rpc_done) {
1767                 state = nfs4_try_open_cached(data);
1768                 trace_nfs4_cached_open(data->state);
1769                 goto out;
1770         }
1771
1772         ret = -EAGAIN;
1773         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1774                 goto err;
1775         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1776         ret = PTR_ERR(inode);
1777         if (IS_ERR(inode))
1778                 goto err;
1779         ret = -ENOMEM;
1780         state = nfs4_get_open_state(inode, data->owner);
1781         if (state == NULL)
1782                 goto err_put_inode;
1783         if (data->o_res.delegation_type != 0)
1784                 nfs4_opendata_check_deleg(data, state);
1785         update_open_stateid(state, &data->o_res.stateid, NULL,
1786                         data->o_arg.fmode);
1787         iput(inode);
1788 out:
1789         nfs_release_seqid(data->o_arg.seqid);
1790         return state;
1791 err_put_inode:
1792         iput(inode);
1793 err:
1794         return ERR_PTR(ret);
1795 }
1796
1797 static struct nfs4_state *
1798 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1799 {
1800         struct nfs4_state *ret;
1801
1802         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1803                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1804         else
1805                 ret = _nfs4_opendata_to_nfs4_state(data);
1806         nfs4_sequence_free_slot(&data->o_res.seq_res);
1807         return ret;
1808 }
1809
1810 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1811 {
1812         struct nfs_inode *nfsi = NFS_I(state->inode);
1813         struct nfs_open_context *ctx;
1814
1815         spin_lock(&state->inode->i_lock);
1816         list_for_each_entry(ctx, &nfsi->open_files, list) {
1817                 if (ctx->state != state)
1818                         continue;
1819                 get_nfs_open_context(ctx);
1820                 spin_unlock(&state->inode->i_lock);
1821                 return ctx;
1822         }
1823         spin_unlock(&state->inode->i_lock);
1824         return ERR_PTR(-ENOENT);
1825 }
1826
1827 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1828                 struct nfs4_state *state, enum open_claim_type4 claim)
1829 {
1830         struct nfs4_opendata *opendata;
1831
1832         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1833                         NULL, NULL, claim, GFP_NOFS);
1834         if (opendata == NULL)
1835                 return ERR_PTR(-ENOMEM);
1836         opendata->state = state;
1837         atomic_inc(&state->count);
1838         return opendata;
1839 }
1840
1841 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1842                 fmode_t fmode)
1843 {
1844         struct nfs4_state *newstate;
1845         int ret;
1846
1847         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1848                 return 0;
1849         opendata->o_arg.open_flags = 0;
1850         opendata->o_arg.fmode = fmode;
1851         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1852                         NFS_SB(opendata->dentry->d_sb),
1853                         fmode, 0);
1854         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1855         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1856         nfs4_init_opendata_res(opendata);
1857         ret = _nfs4_recover_proc_open(opendata);
1858         if (ret != 0)
1859                 return ret;
1860         newstate = nfs4_opendata_to_nfs4_state(opendata);
1861         if (IS_ERR(newstate))
1862                 return PTR_ERR(newstate);
1863         if (newstate != opendata->state)
1864                 ret = -ESTALE;
1865         nfs4_close_state(newstate, fmode);
1866         return ret;
1867 }
1868
1869 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1870 {
1871         int ret;
1872
1873         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1874         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1875         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1876         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1877         /* memory barrier prior to reading state->n_* */
1878         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1879         clear_bit(NFS_OPEN_STATE, &state->flags);
1880         smp_rmb();
1881         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1882         if (ret != 0)
1883                 return ret;
1884         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1885         if (ret != 0)
1886                 return ret;
1887         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1888         if (ret != 0)
1889                 return ret;
1890         /*
1891          * We may have performed cached opens for all three recoveries.
1892          * Check if we need to update the current stateid.
1893          */
1894         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1895             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1896                 write_seqlock(&state->seqlock);
1897                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1898                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1899                 write_sequnlock(&state->seqlock);
1900         }
1901         return 0;
1902 }
1903
1904 /*
1905  * OPEN_RECLAIM:
1906  *      reclaim state on the server after a reboot.
1907  */
1908 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1909 {
1910         struct nfs_delegation *delegation;
1911         struct nfs4_opendata *opendata;
1912         fmode_t delegation_type = 0;
1913         int status;
1914
1915         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1916                         NFS4_OPEN_CLAIM_PREVIOUS);
1917         if (IS_ERR(opendata))
1918                 return PTR_ERR(opendata);
1919         rcu_read_lock();
1920         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1921         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1922                 delegation_type = delegation->type;
1923         rcu_read_unlock();
1924         opendata->o_arg.u.delegation_type = delegation_type;
1925         status = nfs4_open_recover(opendata, state);
1926         nfs4_opendata_put(opendata);
1927         return status;
1928 }
1929
1930 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1931 {
1932         struct nfs_server *server = NFS_SERVER(state->inode);
1933         struct nfs4_exception exception = { };
1934         int err;
1935         do {
1936                 err = _nfs4_do_open_reclaim(ctx, state);
1937                 trace_nfs4_open_reclaim(ctx, 0, err);
1938                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1939                         continue;
1940                 if (err != -NFS4ERR_DELAY)
1941                         break;
1942                 nfs4_handle_exception(server, err, &exception);
1943         } while (exception.retry);
1944         return err;
1945 }
1946
1947 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1948 {
1949         struct nfs_open_context *ctx;
1950         int ret;
1951
1952         ctx = nfs4_state_find_open_context(state);
1953         if (IS_ERR(ctx))
1954                 return -EAGAIN;
1955         ret = nfs4_do_open_reclaim(ctx, state);
1956         put_nfs_open_context(ctx);
1957         return ret;
1958 }
1959
1960 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1961 {
1962         switch (err) {
1963                 default:
1964                         printk(KERN_ERR "NFS: %s: unhandled error "
1965                                         "%d.\n", __func__, err);
1966                 case 0:
1967                 case -ENOENT:
1968                 case -EAGAIN:
1969                 case -ESTALE:
1970                         break;
1971                 case -NFS4ERR_BADSESSION:
1972                 case -NFS4ERR_BADSLOT:
1973                 case -NFS4ERR_BAD_HIGH_SLOT:
1974                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1975                 case -NFS4ERR_DEADSESSION:
1976                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1977                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1978                         return -EAGAIN;
1979                 case -NFS4ERR_STALE_CLIENTID:
1980                 case -NFS4ERR_STALE_STATEID:
1981                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1982                         /* Don't recall a delegation if it was lost */
1983                         nfs4_schedule_lease_recovery(server->nfs_client);
1984                         return -EAGAIN;
1985                 case -NFS4ERR_MOVED:
1986                         nfs4_schedule_migration_recovery(server);
1987                         return -EAGAIN;
1988                 case -NFS4ERR_LEASE_MOVED:
1989                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1990                         return -EAGAIN;
1991                 case -NFS4ERR_DELEG_REVOKED:
1992                 case -NFS4ERR_ADMIN_REVOKED:
1993                 case -NFS4ERR_EXPIRED:
1994                 case -NFS4ERR_BAD_STATEID:
1995                 case -NFS4ERR_OPENMODE:
1996                         nfs_inode_find_state_and_recover(state->inode,
1997                                         stateid);
1998                         nfs4_schedule_stateid_recovery(server, state);
1999                         return -EAGAIN;
2000                 case -NFS4ERR_DELAY:
2001                 case -NFS4ERR_GRACE:
2002                         set_bit(NFS_DELEGATED_STATE, &state->flags);
2003                         ssleep(1);
2004                         return -EAGAIN;
2005                 case -ENOMEM:
2006                 case -NFS4ERR_DENIED:
2007                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
2008                         return 0;
2009         }
2010         return err;
2011 }
2012
2013 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2014                 struct nfs4_state *state, const nfs4_stateid *stateid,
2015                 fmode_t type)
2016 {
2017         struct nfs_server *server = NFS_SERVER(state->inode);
2018         struct nfs4_opendata *opendata;
2019         int err = 0;
2020
2021         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2022                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2023         if (IS_ERR(opendata))
2024                 return PTR_ERR(opendata);
2025         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2026         write_seqlock(&state->seqlock);
2027         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2028         write_sequnlock(&state->seqlock);
2029         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2030         switch (type & (FMODE_READ|FMODE_WRITE)) {
2031         case FMODE_READ|FMODE_WRITE:
2032         case FMODE_WRITE:
2033                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2034                 if (err)
2035                         break;
2036                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2037                 if (err)
2038                         break;
2039         case FMODE_READ:
2040                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2041         }
2042         nfs4_opendata_put(opendata);
2043         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
2044 }
2045
2046 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2047 {
2048         struct nfs4_opendata *data = calldata;
2049
2050         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
2051                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
2052 }
2053
2054 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2055 {
2056         struct nfs4_opendata *data = calldata;
2057
2058         nfs40_sequence_done(task, &data->c_res.seq_res);
2059
2060         data->rpc_status = task->tk_status;
2061         if (data->rpc_status == 0) {
2062                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2063                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2064                 renew_lease(data->o_res.server, data->timestamp);
2065                 data->rpc_done = 1;
2066         }
2067 }
2068
2069 static void nfs4_open_confirm_release(void *calldata)
2070 {
2071         struct nfs4_opendata *data = calldata;
2072         struct nfs4_state *state = NULL;
2073
2074         /* If this request hasn't been cancelled, do nothing */
2075         if (data->cancelled == 0)
2076                 goto out_free;
2077         /* In case of error, no cleanup! */
2078         if (!data->rpc_done)
2079                 goto out_free;
2080         state = nfs4_opendata_to_nfs4_state(data);
2081         if (!IS_ERR(state))
2082                 nfs4_close_state(state, data->o_arg.fmode);
2083 out_free:
2084         nfs4_opendata_put(data);
2085 }
2086
2087 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2088         .rpc_call_prepare = nfs4_open_confirm_prepare,
2089         .rpc_call_done = nfs4_open_confirm_done,
2090         .rpc_release = nfs4_open_confirm_release,
2091 };
2092
2093 /*
2094  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2095  */
2096 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2097 {
2098         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2099         struct rpc_task *task;
2100         struct  rpc_message msg = {
2101                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2102                 .rpc_argp = &data->c_arg,
2103                 .rpc_resp = &data->c_res,
2104                 .rpc_cred = data->owner->so_cred,
2105         };
2106         struct rpc_task_setup task_setup_data = {
2107                 .rpc_client = server->client,
2108                 .rpc_message = &msg,
2109                 .callback_ops = &nfs4_open_confirm_ops,
2110                 .callback_data = data,
2111                 .workqueue = nfsiod_workqueue,
2112                 .flags = RPC_TASK_ASYNC,
2113         };
2114         int status;
2115
2116         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2117         kref_get(&data->kref);
2118         data->rpc_done = 0;
2119         data->rpc_status = 0;
2120         data->timestamp = jiffies;
2121         if (data->is_recover)
2122                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2123         task = rpc_run_task(&task_setup_data);
2124         if (IS_ERR(task))
2125                 return PTR_ERR(task);
2126         status = nfs4_wait_for_completion_rpc_task(task);
2127         if (status != 0) {
2128                 data->cancelled = 1;
2129                 smp_wmb();
2130         } else
2131                 status = data->rpc_status;
2132         rpc_put_task(task);
2133         return status;
2134 }
2135
2136 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2137 {
2138         struct nfs4_opendata *data = calldata;
2139         struct nfs4_state_owner *sp = data->owner;
2140         struct nfs_client *clp = sp->so_server->nfs_client;
2141         enum open_claim_type4 claim = data->o_arg.claim;
2142
2143         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2144                 goto out_wait;
2145         /*
2146          * Check if we still need to send an OPEN call, or if we can use
2147          * a delegation instead.
2148          */
2149         if (data->state != NULL) {
2150                 struct nfs_delegation *delegation;
2151
2152                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2153                         goto out_no_action;
2154                 rcu_read_lock();
2155                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2156                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2157                         goto unlock_no_action;
2158                 rcu_read_unlock();
2159         }
2160         /* Update client id. */
2161         data->o_arg.clientid = clp->cl_clientid;
2162         switch (claim) {
2163         default:
2164                 break;
2165         case NFS4_OPEN_CLAIM_PREVIOUS:
2166         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2167         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2168                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2169         case NFS4_OPEN_CLAIM_FH:
2170                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2171                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2172         }
2173         data->timestamp = jiffies;
2174         if (nfs4_setup_sequence(data->o_arg.server,
2175                                 &data->o_arg.seq_args,
2176                                 &data->o_res.seq_res,
2177                                 task) != 0)
2178                 nfs_release_seqid(data->o_arg.seqid);
2179
2180         /* Set the create mode (note dependency on the session type) */
2181         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2182         if (data->o_arg.open_flags & O_EXCL) {
2183                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2184                 if (nfs4_has_persistent_session(clp))
2185                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2186                 else if (clp->cl_mvops->minor_version > 0)
2187                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2188         }
2189         return;
2190 unlock_no_action:
2191         trace_nfs4_cached_open(data->state);
2192         rcu_read_unlock();
2193 out_no_action:
2194         task->tk_action = NULL;
2195 out_wait:
2196         nfs4_sequence_done(task, &data->o_res.seq_res);
2197 }
2198
2199 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2200 {
2201         struct nfs4_opendata *data = calldata;
2202
2203         data->rpc_status = task->tk_status;
2204
2205         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2206                 return;
2207
2208         if (task->tk_status == 0) {
2209                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2210                         switch (data->o_res.f_attr->mode & S_IFMT) {
2211                         case S_IFREG:
2212                                 break;
2213                         case S_IFLNK:
2214                                 data->rpc_status = -ELOOP;
2215                                 break;
2216                         case S_IFDIR:
2217                                 data->rpc_status = -EISDIR;
2218                                 break;
2219                         default:
2220                                 data->rpc_status = -ENOTDIR;
2221                         }
2222                 }
2223                 renew_lease(data->o_res.server, data->timestamp);
2224                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2225                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2226         }
2227         data->rpc_done = 1;
2228 }
2229
2230 static void nfs4_open_release(void *calldata)
2231 {
2232         struct nfs4_opendata *data = calldata;
2233         struct nfs4_state *state = NULL;
2234
2235         /* If this request hasn't been cancelled, do nothing */
2236         if (data->cancelled == 0)
2237                 goto out_free;
2238         /* In case of error, no cleanup! */
2239         if (data->rpc_status != 0 || !data->rpc_done)
2240                 goto out_free;
2241         /* In case we need an open_confirm, no cleanup! */
2242         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2243                 goto out_free;
2244         state = nfs4_opendata_to_nfs4_state(data);
2245         if (!IS_ERR(state))
2246                 nfs4_close_state(state, data->o_arg.fmode);
2247 out_free:
2248         nfs4_opendata_put(data);
2249 }
2250
2251 static const struct rpc_call_ops nfs4_open_ops = {
2252         .rpc_call_prepare = nfs4_open_prepare,
2253         .rpc_call_done = nfs4_open_done,
2254         .rpc_release = nfs4_open_release,
2255 };
2256
2257 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2258 {
2259         struct inode *dir = d_inode(data->dir);
2260         struct nfs_server *server = NFS_SERVER(dir);
2261         struct nfs_openargs *o_arg = &data->o_arg;
2262         struct nfs_openres *o_res = &data->o_res;
2263         struct rpc_task *task;
2264         struct rpc_message msg = {
2265                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2266                 .rpc_argp = o_arg,
2267                 .rpc_resp = o_res,
2268                 .rpc_cred = data->owner->so_cred,
2269         };
2270         struct rpc_task_setup task_setup_data = {
2271                 .rpc_client = server->client,
2272                 .rpc_message = &msg,
2273                 .callback_ops = &nfs4_open_ops,
2274                 .callback_data = data,
2275                 .workqueue = nfsiod_workqueue,
2276                 .flags = RPC_TASK_ASYNC,
2277         };
2278         int status;
2279
2280         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2281         kref_get(&data->kref);
2282         data->rpc_done = 0;
2283         data->rpc_status = 0;
2284         data->cancelled = 0;
2285         data->is_recover = 0;
2286         if (isrecover) {
2287                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2288                 data->is_recover = 1;
2289         }
2290         task = rpc_run_task(&task_setup_data);
2291         if (IS_ERR(task))
2292                 return PTR_ERR(task);
2293         status = nfs4_wait_for_completion_rpc_task(task);
2294         if (status != 0) {
2295                 data->cancelled = 1;
2296                 smp_wmb();
2297         } else
2298                 status = data->rpc_status;
2299         rpc_put_task(task);
2300
2301         return status;
2302 }
2303
2304 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2305 {
2306         struct inode *dir = d_inode(data->dir);
2307         struct nfs_openres *o_res = &data->o_res;
2308         int status;
2309
2310         status = nfs4_run_open_task(data, 1);
2311         if (status != 0 || !data->rpc_done)
2312                 return status;
2313
2314         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2315
2316         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2317                 status = _nfs4_proc_open_confirm(data);
2318                 if (status != 0)
2319                         return status;
2320         }
2321
2322         return status;
2323 }
2324
2325 /*
2326  * Additional permission checks in order to distinguish between an
2327  * open for read, and an open for execute. This works around the
2328  * fact that NFSv4 OPEN treats read and execute permissions as being
2329  * the same.
2330  * Note that in the non-execute case, we want to turn off permission
2331  * checking if we just created a new file (POSIX open() semantics).
2332  */
2333 static int nfs4_opendata_access(struct rpc_cred *cred,
2334                                 struct nfs4_opendata *opendata,
2335                                 struct nfs4_state *state, fmode_t fmode,
2336                                 int openflags)
2337 {
2338         struct nfs_access_entry cache;
2339         u32 mask;
2340
2341         /* access call failed or for some reason the server doesn't
2342          * support any access modes -- defer access call until later */
2343         if (opendata->o_res.access_supported == 0)
2344                 return 0;
2345
2346         mask = 0;
2347         /*
2348          * Use openflags to check for exec, because fmode won't
2349          * always have FMODE_EXEC set when file open for exec.
2350          */
2351         if (openflags & __FMODE_EXEC) {
2352                 /* ONLY check for exec rights */
2353                 mask = MAY_EXEC;
2354         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2355                 mask = MAY_READ;
2356
2357         cache.cred = cred;
2358         cache.jiffies = jiffies;
2359         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2360         nfs_access_add_cache(state->inode, &cache);
2361
2362         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2363                 return 0;
2364
2365         /* even though OPEN succeeded, access is denied. Close the file */
2366         nfs4_close_state(state, fmode);
2367         return -EACCES;
2368 }
2369
2370 /*
2371  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2372  */
2373 static int _nfs4_proc_open(struct nfs4_opendata *data)
2374 {
2375         struct inode *dir = d_inode(data->dir);
2376         struct nfs_server *server = NFS_SERVER(dir);
2377         struct nfs_openargs *o_arg = &data->o_arg;
2378         struct nfs_openres *o_res = &data->o_res;
2379         int status;
2380
2381         status = nfs4_run_open_task(data, 0);
2382         if (!data->rpc_done)
2383                 return status;
2384         if (status != 0) {
2385                 if (status == -NFS4ERR_BADNAME &&
2386                                 !(o_arg->open_flags & O_CREAT))
2387                         return -ENOENT;
2388                 return status;
2389         }
2390
2391         nfs_fattr_map_and_free_names(server, &data->f_attr);
2392
2393         if (o_arg->open_flags & O_CREAT) {
2394                 update_changeattr(dir, &o_res->cinfo);
2395                 if (o_arg->open_flags & O_EXCL)
2396                         data->file_created = 1;
2397                 else if (o_res->cinfo.before != o_res->cinfo.after)
2398                         data->file_created = 1;
2399         }
2400         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2401                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2402         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2403                 status = _nfs4_proc_open_confirm(data);
2404                 if (status != 0)
2405                         return status;
2406         }
2407         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2408                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2409         return 0;
2410 }
2411
2412 static int nfs4_recover_expired_lease(struct nfs_server *server)
2413 {
2414         return nfs4_client_recover_expired_lease(server->nfs_client);
2415 }
2416
2417 /*
2418  * OPEN_EXPIRED:
2419  *      reclaim state on the server after a network partition.
2420  *      Assumes caller holds the appropriate lock
2421  */
2422 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2423 {
2424         struct nfs4_opendata *opendata;
2425         int ret;
2426
2427         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2428                         NFS4_OPEN_CLAIM_FH);
2429         if (IS_ERR(opendata))
2430                 return PTR_ERR(opendata);
2431         ret = nfs4_open_recover(opendata, state);
2432         if (ret == -ESTALE)
2433                 d_drop(ctx->dentry);
2434         nfs4_opendata_put(opendata);
2435         return ret;
2436 }
2437
2438 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2439 {
2440         struct nfs_server *server = NFS_SERVER(state->inode);
2441         struct nfs4_exception exception = { };
2442         int err;
2443
2444         do {
2445                 err = _nfs4_open_expired(ctx, state);
2446                 trace_nfs4_open_expired(ctx, 0, err);
2447                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2448                         continue;
2449                 switch (err) {
2450                 default:
2451                         goto out;
2452                 case -NFS4ERR_GRACE:
2453                 case -NFS4ERR_DELAY:
2454                         nfs4_handle_exception(server, err, &exception);
2455                         err = 0;
2456                 }
2457         } while (exception.retry);
2458 out:
2459         return err;
2460 }
2461
2462 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2463 {
2464         struct nfs_open_context *ctx;
2465         int ret;
2466
2467         ctx = nfs4_state_find_open_context(state);
2468         if (IS_ERR(ctx))
2469                 return -EAGAIN;
2470         ret = nfs4_do_open_expired(ctx, state);
2471         put_nfs_open_context(ctx);
2472         return ret;
2473 }
2474
2475 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2476                 const nfs4_stateid *stateid)
2477 {
2478         nfs_remove_bad_delegation(state->inode, stateid);
2479         write_seqlock(&state->seqlock);
2480         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2481         write_sequnlock(&state->seqlock);
2482         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2483 }
2484
2485 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2486 {
2487         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2488                 nfs_finish_clear_delegation_stateid(state, NULL);
2489 }
2490
2491 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2492 {
2493         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2494         nfs40_clear_delegation_stateid(state);
2495         return nfs4_open_expired(sp, state);
2496 }
2497
2498 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2499                 nfs4_stateid *stateid,
2500                 struct rpc_cred *cred)
2501 {
2502         return -NFS4ERR_BAD_STATEID;
2503 }
2504
2505 #if defined(CONFIG_NFS_V4_1)
2506 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2507                 nfs4_stateid *stateid,
2508                 struct rpc_cred *cred)
2509 {
2510         int status;
2511
2512         switch (stateid->type) {
2513         default:
2514                 break;
2515         case NFS4_INVALID_STATEID_TYPE:
2516         case NFS4_SPECIAL_STATEID_TYPE:
2517                 return -NFS4ERR_BAD_STATEID;
2518         case NFS4_REVOKED_STATEID_TYPE:
2519                 goto out_free;
2520         }
2521
2522         status = nfs41_test_stateid(server, stateid, cred);
2523         switch (status) {
2524         case -NFS4ERR_EXPIRED:
2525         case -NFS4ERR_ADMIN_REVOKED:
2526         case -NFS4ERR_DELEG_REVOKED:
2527                 break;
2528         default:
2529                 return status;
2530         }
2531 out_free:
2532         /* Ack the revoked state to the server */
2533         nfs41_free_stateid(server, stateid, cred, true);
2534         return -NFS4ERR_EXPIRED;
2535 }
2536
2537 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2538 {
2539         struct nfs_server *server = NFS_SERVER(state->inode);
2540         nfs4_stateid stateid;
2541         struct nfs_delegation *delegation;
2542         struct rpc_cred *cred;
2543         int status;
2544
2545         /* Get the delegation credential for use by test/free_stateid */
2546         rcu_read_lock();
2547         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2548         if (delegation == NULL) {
2549                 rcu_read_unlock();
2550                 return;
2551         }
2552
2553         nfs4_stateid_copy(&stateid, &delegation->stateid);
2554         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2555                 rcu_read_unlock();
2556                 nfs_finish_clear_delegation_stateid(state, &stateid);
2557                 return;
2558         }
2559
2560         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags)) {
2561                 rcu_read_unlock();
2562                 return;
2563         }
2564
2565         cred = get_rpccred(delegation->cred);
2566         rcu_read_unlock();
2567         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2568         trace_nfs4_test_delegation_stateid(state, NULL, status);
2569         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2570                 nfs_finish_clear_delegation_stateid(state, &stateid);
2571
2572         put_rpccred(cred);
2573 }
2574
2575 /**
2576  * nfs41_check_expired_locks - possibly free a lock stateid
2577  *
2578  * @state: NFSv4 state for an inode
2579  *
2580  * Returns NFS_OK if recovery for this stateid is now finished.
2581  * Otherwise a negative NFS4ERR value is returned.
2582  */
2583 static int nfs41_check_expired_locks(struct nfs4_state *state)
2584 {
2585         int status, ret = NFS_OK;
2586         struct nfs4_lock_state *lsp, *prev = NULL;
2587         struct nfs_server *server = NFS_SERVER(state->inode);
2588
2589         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2590                 goto out;
2591
2592         spin_lock(&state->state_lock);
2593         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2594                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2595                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2596
2597                         atomic_inc(&lsp->ls_count);
2598                         spin_unlock(&state->state_lock);
2599
2600                         nfs4_put_lock_state(prev);
2601                         prev = lsp;
2602
2603                         status = nfs41_test_and_free_expired_stateid(server,
2604                                         &lsp->ls_stateid,
2605                                         cred);
2606                         trace_nfs4_test_lock_stateid(state, lsp, status);
2607                         if (status == -NFS4ERR_EXPIRED ||
2608                             status == -NFS4ERR_BAD_STATEID) {
2609                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2610                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2611                                 if (!recover_lost_locks)
2612                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2613                         } else if (status != NFS_OK) {
2614                                 ret = status;
2615                                 nfs4_put_lock_state(prev);
2616                                 goto out;
2617                         }
2618                         spin_lock(&state->state_lock);
2619                 }
2620         }
2621         spin_unlock(&state->state_lock);
2622         nfs4_put_lock_state(prev);
2623 out:
2624         return ret;
2625 }
2626
2627 /**
2628  * nfs41_check_open_stateid - possibly free an open stateid
2629  *
2630  * @state: NFSv4 state for an inode
2631  *
2632  * Returns NFS_OK if recovery for this stateid is now finished.
2633  * Otherwise a negative NFS4ERR value is returned.
2634  */
2635 static int nfs41_check_open_stateid(struct nfs4_state *state)
2636 {
2637         struct nfs_server *server = NFS_SERVER(state->inode);
2638         nfs4_stateid *stateid = &state->open_stateid;
2639         struct rpc_cred *cred = state->owner->so_cred;
2640         int status;
2641
2642         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2643                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2644                         if (nfs4_have_delegation(state->inode, state->state))
2645                                 return NFS_OK;
2646                         return -NFS4ERR_OPENMODE;
2647                 }
2648                 return -NFS4ERR_BAD_STATEID;
2649         }
2650         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2651         trace_nfs4_test_open_stateid(state, NULL, status);
2652         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2653                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2654                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2655                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2656                 clear_bit(NFS_OPEN_STATE, &state->flags);
2657                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2658         }
2659         if (status != NFS_OK)
2660                 return status;
2661         if (nfs_open_stateid_recover_openmode(state))
2662                 return -NFS4ERR_OPENMODE;
2663         return NFS_OK;
2664 }
2665
2666 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2667 {
2668         int status;
2669
2670         nfs41_check_delegation_stateid(state);
2671         status = nfs41_check_expired_locks(state);
2672         if (status != NFS_OK)
2673                 return status;
2674         status = nfs41_check_open_stateid(state);
2675         if (status != NFS_OK)
2676                 status = nfs4_open_expired(sp, state);
2677         return status;
2678 }
2679 #endif
2680
2681 /*
2682  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2683  * fields corresponding to attributes that were used to store the verifier.
2684  * Make sure we clobber those fields in the later setattr call
2685  */
2686 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2687                                 struct iattr *sattr, struct nfs4_label **label)
2688 {
2689         const u32 *attrset = opendata->o_res.attrset;
2690
2691         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2692             !(sattr->ia_valid & ATTR_ATIME_SET))
2693                 sattr->ia_valid |= ATTR_ATIME;
2694
2695         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2696             !(sattr->ia_valid & ATTR_MTIME_SET))
2697                 sattr->ia_valid |= ATTR_MTIME;
2698
2699         /* Except MODE, it seems harmless of setting twice. */
2700         if ((attrset[1] & FATTR4_WORD1_MODE))
2701                 sattr->ia_valid &= ~ATTR_MODE;
2702
2703         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2704                 *label = NULL;
2705 }
2706
2707 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2708                 fmode_t fmode,
2709                 int flags,
2710                 struct nfs_open_context *ctx)
2711 {
2712         struct nfs4_state_owner *sp = opendata->owner;
2713         struct nfs_server *server = sp->so_server;
2714         struct dentry *dentry;
2715         struct nfs4_state *state;
2716         unsigned int seq;
2717         int ret;
2718
2719         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2720
2721         ret = _nfs4_proc_open(opendata);
2722         if (ret != 0)
2723                 goto out;
2724
2725         state = nfs4_opendata_to_nfs4_state(opendata);
2726         ret = PTR_ERR(state);
2727         if (IS_ERR(state))
2728                 goto out;
2729         if (server->caps & NFS_CAP_POSIX_LOCK)
2730                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2731         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2732                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2733
2734         dentry = opendata->dentry;
2735         if (d_really_is_negative(dentry)) {
2736                 struct dentry *alias;
2737                 d_drop(dentry);
2738                 alias = d_exact_alias(dentry, state->inode);
2739                 if (!alias)
2740                         alias = d_splice_alias(igrab(state->inode), dentry);
2741                 /* d_splice_alias() can't fail here - it's a non-directory */
2742                 if (alias) {
2743                         dput(ctx->dentry);
2744                         ctx->dentry = dentry = alias;
2745                 }
2746                 nfs_set_verifier(dentry,
2747                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2748         }
2749
2750         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2751         if (ret != 0)
2752                 goto out;
2753
2754         ctx->state = state;
2755         if (d_inode(dentry) == state->inode) {
2756                 nfs_inode_attach_open_context(ctx);
2757                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2758                         nfs4_schedule_stateid_recovery(server, state);
2759         }
2760 out:
2761         return ret;
2762 }
2763
2764 /*
2765  * Returns a referenced nfs4_state
2766  */
2767 static int _nfs4_do_open(struct inode *dir,
2768                         struct nfs_open_context *ctx,
2769                         int flags,
2770                         struct iattr *sattr,
2771                         struct nfs4_label *label,
2772                         int *opened)
2773 {
2774         struct nfs4_state_owner  *sp;
2775         struct nfs4_state     *state = NULL;
2776         struct nfs_server       *server = NFS_SERVER(dir);
2777         struct nfs4_opendata *opendata;
2778         struct dentry *dentry = ctx->dentry;
2779         struct rpc_cred *cred = ctx->cred;
2780         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2781         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2782         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2783         struct nfs4_label *olabel = NULL;
2784         int status;
2785
2786         /* Protect against reboot recovery conflicts */
2787         status = -ENOMEM;
2788         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2789         if (sp == NULL) {
2790                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2791                 goto out_err;
2792         }
2793         status = nfs4_recover_expired_lease(server);
2794         if (status != 0)
2795                 goto err_put_state_owner;
2796         if (d_really_is_positive(dentry))
2797                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2798         status = -ENOMEM;
2799         if (d_really_is_positive(dentry))
2800                 claim = NFS4_OPEN_CLAIM_FH;
2801         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2802                         label, claim, GFP_KERNEL);
2803         if (opendata == NULL)
2804                 goto err_put_state_owner;
2805
2806         if (label) {
2807                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2808                 if (IS_ERR(olabel)) {
2809                         status = PTR_ERR(olabel);
2810                         goto err_opendata_put;
2811                 }
2812         }
2813
2814         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2815                 if (!opendata->f_attr.mdsthreshold) {
2816                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2817                         if (!opendata->f_attr.mdsthreshold)
2818                                 goto err_free_label;
2819                 }
2820                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2821         }
2822         if (d_really_is_positive(dentry))
2823                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2824
2825         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2826         if (status != 0)
2827                 goto err_free_label;
2828         state = ctx->state;
2829
2830         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2831             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2832                 nfs4_exclusive_attrset(opendata, sattr, &label);
2833                 /*
2834                  * send create attributes which was not set by open
2835                  * with an extra setattr.
2836                  */
2837                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2838                         nfs_fattr_init(opendata->o_res.f_attr);
2839                         status = nfs4_do_setattr(state->inode, cred,
2840                                         opendata->o_res.f_attr, sattr,
2841                                         ctx, label, olabel);
2842                         if (status == 0) {
2843                                 nfs_setattr_update_inode(state->inode, sattr,
2844                                                 opendata->o_res.f_attr);
2845                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2846                         }
2847                 }
2848         }
2849         if (opened && opendata->file_created)
2850                 *opened |= FILE_CREATED;
2851
2852         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2853                 *ctx_th = opendata->f_attr.mdsthreshold;
2854                 opendata->f_attr.mdsthreshold = NULL;
2855         }
2856
2857         nfs4_label_free(olabel);
2858
2859         nfs4_opendata_put(opendata);
2860         nfs4_put_state_owner(sp);
2861         return 0;
2862 err_free_label:
2863         nfs4_label_free(olabel);
2864 err_opendata_put:
2865         nfs4_opendata_put(opendata);
2866 err_put_state_owner:
2867         nfs4_put_state_owner(sp);
2868 out_err:
2869         return status;
2870 }
2871
2872
2873 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2874                                         struct nfs_open_context *ctx,
2875                                         int flags,
2876                                         struct iattr *sattr,
2877                                         struct nfs4_label *label,
2878                                         int *opened)
2879 {
2880         struct nfs_server *server = NFS_SERVER(dir);
2881         struct nfs4_exception exception = { };
2882         struct nfs4_state *res;
2883         int status;
2884
2885         do {
2886                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2887                 res = ctx->state;
2888                 trace_nfs4_open_file(ctx, flags, status);
2889                 if (status == 0)
2890                         break;
2891                 /* NOTE: BAD_SEQID means the server and client disagree about the
2892                  * book-keeping w.r.t. state-changing operations
2893                  * (OPEN/CLOSE/LOCK/LOCKU...)
2894                  * It is actually a sign of a bug on the client or on the server.
2895                  *
2896                  * If we receive a BAD_SEQID error in the particular case of
2897                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2898                  * have unhashed the old state_owner for us, and that we can
2899                  * therefore safely retry using a new one. We should still warn
2900                  * the user though...
2901                  */
2902                 if (status == -NFS4ERR_BAD_SEQID) {
2903                         pr_warn_ratelimited("NFS: v4 server %s "
2904                                         " returned a bad sequence-id error!\n",
2905                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2906                         exception.retry = 1;
2907                         continue;
2908                 }
2909                 /*
2910                  * BAD_STATEID on OPEN means that the server cancelled our
2911                  * state before it received the OPEN_CONFIRM.
2912                  * Recover by retrying the request as per the discussion
2913                  * on Page 181 of RFC3530.
2914                  */
2915                 if (status == -NFS4ERR_BAD_STATEID) {
2916                         exception.retry = 1;
2917                         continue;
2918                 }
2919                 if (status == -EAGAIN) {
2920                         /* We must have found a delegation */
2921                         exception.retry = 1;
2922                         continue;
2923                 }
2924                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2925                         continue;
2926                 res = ERR_PTR(nfs4_handle_exception(server,
2927                                         status, &exception));
2928         } while (exception.retry);
2929         return res;
2930 }
2931
2932 static int _nfs4_do_setattr(struct inode *inode,
2933                             struct nfs_setattrargs *arg,
2934                             struct nfs_setattrres *res,
2935                             struct rpc_cred *cred,
2936                             struct nfs_open_context *ctx)
2937 {
2938         struct nfs_server *server = NFS_SERVER(inode);
2939         struct rpc_message msg = {
2940                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2941                 .rpc_argp       = arg,
2942                 .rpc_resp       = res,
2943                 .rpc_cred       = cred,
2944         };
2945         struct rpc_cred *delegation_cred = NULL;
2946         unsigned long timestamp = jiffies;
2947         fmode_t fmode;
2948         bool truncate;
2949         int status;
2950
2951         nfs_fattr_init(res->fattr);
2952
2953         /* Servers should only apply open mode checks for file size changes */
2954         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2955         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2956
2957         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2958                 /* Use that stateid */
2959         } else if (truncate && ctx != NULL) {
2960                 struct nfs_lock_context *l_ctx;
2961                 if (!nfs4_valid_open_stateid(ctx->state))
2962                         return -EBADF;
2963                 l_ctx = nfs_get_lock_context(ctx);
2964                 if (IS_ERR(l_ctx))
2965                         return PTR_ERR(l_ctx);
2966                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2967                                                 &arg->stateid, &delegation_cred);
2968                 nfs_put_lock_context(l_ctx);
2969                 if (status == -EIO)
2970                         return -EBADF;
2971         } else
2972                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2973         if (delegation_cred)
2974                 msg.rpc_cred = delegation_cred;
2975
2976         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2977
2978         put_rpccred(delegation_cred);
2979         if (status == 0 && ctx != NULL)
2980                 renew_lease(server, timestamp);
2981         trace_nfs4_setattr(inode, &arg->stateid, status);
2982         return status;
2983 }
2984
2985 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2986                            struct nfs_fattr *fattr, struct iattr *sattr,
2987                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2988                            struct nfs4_label *olabel)
2989 {
2990         struct nfs_server *server = NFS_SERVER(inode);
2991         struct nfs4_state *state = ctx ? ctx->state : NULL;
2992         struct nfs_setattrargs  arg = {
2993                 .fh             = NFS_FH(inode),
2994                 .iap            = sattr,
2995                 .server         = server,
2996                 .bitmask = server->attr_bitmask,
2997                 .label          = ilabel,
2998         };
2999         struct nfs_setattrres  res = {
3000                 .fattr          = fattr,
3001                 .label          = olabel,
3002                 .server         = server,
3003         };
3004         struct nfs4_exception exception = {
3005                 .state = state,
3006                 .inode = inode,
3007                 .stateid = &arg.stateid,
3008         };
3009         int err;
3010
3011         arg.bitmask = nfs4_bitmask(server, ilabel);
3012         if (ilabel)
3013                 arg.bitmask = nfs4_bitmask(server, olabel);
3014
3015         do {
3016                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3017                 switch (err) {
3018                 case -NFS4ERR_OPENMODE:
3019                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3020                                 pr_warn_once("NFSv4: server %s is incorrectly "
3021                                                 "applying open mode checks to "
3022                                                 "a SETATTR that is not "
3023                                                 "changing file size.\n",
3024                                                 server->nfs_client->cl_hostname);
3025                         }
3026                         if (state && !(state->state & FMODE_WRITE)) {
3027                                 err = -EBADF;
3028                                 if (sattr->ia_valid & ATTR_OPEN)
3029                                         err = -EACCES;
3030                                 goto out;
3031                         }
3032                 }
3033                 err = nfs4_handle_exception(server, err, &exception);
3034         } while (exception.retry);
3035 out:
3036         return err;
3037 }
3038
3039 static bool
3040 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3041 {
3042         if (inode == NULL || !nfs_have_layout(inode))
3043                 return false;
3044
3045         return pnfs_wait_on_layoutreturn(inode, task);
3046 }
3047
3048 struct nfs4_closedata {
3049         struct inode *inode;
3050         struct nfs4_state *state;
3051         struct nfs_closeargs arg;
3052         struct nfs_closeres res;
3053         struct {
3054                 struct nfs4_layoutreturn_args arg;
3055                 struct nfs4_layoutreturn_res res;
3056                 struct nfs4_xdr_opaque_data ld_private;
3057                 u32 roc_barrier;
3058                 bool roc;
3059         } lr;
3060         struct nfs_fattr fattr;
3061         unsigned long timestamp;
3062 };
3063
3064 static void nfs4_free_closedata(void *data)
3065 {
3066         struct nfs4_closedata *calldata = data;
3067         struct nfs4_state_owner *sp = calldata->state->owner;
3068         struct super_block *sb = calldata->state->inode->i_sb;
3069
3070         if (calldata->lr.roc)
3071                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3072                                 calldata->res.lr_ret);
3073         nfs4_put_open_state(calldata->state);
3074         nfs_free_seqid(calldata->arg.seqid);
3075         nfs4_put_state_owner(sp);
3076         nfs_sb_deactive(sb);
3077         kfree(calldata);
3078 }
3079
3080 static void nfs4_close_done(struct rpc_task *task, void *data)
3081 {
3082         struct nfs4_closedata *calldata = data;
3083         struct nfs4_state *state = calldata->state;
3084         struct nfs_server *server = NFS_SERVER(calldata->inode);
3085         nfs4_stateid *res_stateid = NULL;
3086
3087         dprintk("%s: begin!\n", __func__);
3088         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3089                 return;
3090         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3091
3092         /* Handle Layoutreturn errors */
3093         if (calldata->arg.lr_args && task->tk_status != 0) {
3094                 switch (calldata->res.lr_ret) {
3095                 default:
3096                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3097                         break;
3098                 case 0:
3099                         calldata->arg.lr_args = NULL;
3100                         calldata->res.lr_res = NULL;
3101                         break;
3102                 case -NFS4ERR_ADMIN_REVOKED:
3103                 case -NFS4ERR_DELEG_REVOKED:
3104                 case -NFS4ERR_EXPIRED:
3105                 case -NFS4ERR_BAD_STATEID:
3106                 case -NFS4ERR_OLD_STATEID:
3107                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3108                 case -NFS4ERR_WRONG_CRED:
3109                         calldata->arg.lr_args = NULL;
3110                         calldata->res.lr_res = NULL;
3111                         calldata->res.lr_ret = 0;
3112                         rpc_restart_call_prepare(task);
3113                         return;
3114                 }
3115         }
3116
3117         /* hmm. we are done with the inode, and in the process of freeing
3118          * the state_owner. we keep this around to process errors
3119          */
3120         switch (task->tk_status) {
3121                 case 0:
3122                         res_stateid = &calldata->res.stateid;
3123                         renew_lease(server, calldata->timestamp);
3124                         break;
3125                 case -NFS4ERR_ADMIN_REVOKED:
3126                 case -NFS4ERR_STALE_STATEID:
3127                 case -NFS4ERR_EXPIRED:
3128                         nfs4_free_revoked_stateid(server,
3129                                         &calldata->arg.stateid,
3130                                         task->tk_msg.rpc_cred);
3131                 case -NFS4ERR_OLD_STATEID:
3132                 case -NFS4ERR_BAD_STATEID:
3133                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3134                                                 &state->open_stateid)) {
3135                                 rpc_restart_call_prepare(task);
3136                                 goto out_release;
3137                         }
3138                         if (calldata->arg.fmode == 0)
3139                                 break;
3140                 default:
3141                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3142                                 rpc_restart_call_prepare(task);
3143                                 goto out_release;
3144                         }
3145         }
3146         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3147                         res_stateid, calldata->arg.fmode);
3148 out_release:
3149         nfs_release_seqid(calldata->arg.seqid);
3150         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3151         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3152 }
3153
3154 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3155 {
3156         struct nfs4_closedata *calldata = data;
3157         struct nfs4_state *state = calldata->state;
3158         struct inode *inode = calldata->inode;
3159         bool is_rdonly, is_wronly, is_rdwr;
3160         int call_close = 0;
3161
3162         dprintk("%s: begin!\n", __func__);
3163         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3164                 goto out_wait;
3165
3166         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3167         spin_lock(&state->owner->so_lock);
3168         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3169         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3170         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3171         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3172         /* Calculate the change in open mode */
3173         calldata->arg.fmode = 0;
3174         if (state->n_rdwr == 0) {
3175                 if (state->n_rdonly == 0)
3176                         call_close |= is_rdonly;
3177                 else if (is_rdonly)
3178                         calldata->arg.fmode |= FMODE_READ;
3179                 if (state->n_wronly == 0)
3180                         call_close |= is_wronly;
3181                 else if (is_wronly)
3182                         calldata->arg.fmode |= FMODE_WRITE;
3183                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3184                         call_close |= is_rdwr;
3185         } else if (is_rdwr)
3186                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3187
3188         if (!nfs4_valid_open_stateid(state) ||
3189             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3190                 call_close = 0;
3191         spin_unlock(&state->owner->so_lock);
3192
3193         if (!call_close) {
3194                 /* Note: exit _without_ calling nfs4_close_done */
3195                 goto out_no_action;
3196         }
3197
3198         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3199                 nfs_release_seqid(calldata->arg.seqid);
3200                 goto out_wait;
3201         }
3202
3203         if (calldata->arg.fmode == 0)
3204                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3205
3206         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3207                 /* Close-to-open cache consistency revalidation */
3208                 if (!nfs4_have_delegation(inode, FMODE_READ))
3209                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3210                 else
3211                         calldata->arg.bitmask = NULL;
3212         }
3213
3214         calldata->arg.share_access =
3215                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3216                                 calldata->arg.fmode, 0);
3217
3218         if (calldata->res.fattr == NULL)
3219                 calldata->arg.bitmask = NULL;
3220         else if (calldata->arg.bitmask == NULL)
3221                 calldata->res.fattr = NULL;
3222         calldata->timestamp = jiffies;
3223         if (nfs4_setup_sequence(NFS_SERVER(inode),
3224                                 &calldata->arg.seq_args,
3225                                 &calldata->res.seq_res,
3226                                 task) != 0)
3227                 nfs_release_seqid(calldata->arg.seqid);
3228         dprintk("%s: done!\n", __func__);
3229         return;
3230 out_no_action:
3231         task->tk_action = NULL;
3232 out_wait:
3233         nfs4_sequence_done(task, &calldata->res.seq_res);
3234 }
3235
3236 static const struct rpc_call_ops nfs4_close_ops = {
3237         .rpc_call_prepare = nfs4_close_prepare,
3238         .rpc_call_done = nfs4_close_done,
3239         .rpc_release = nfs4_free_closedata,
3240 };
3241
3242 /*
3243  * It is possible for data to be read/written from a mem-mapped file
3244  * after the sys_close call (which hits the vfs layer as a flush).
3245  * This means that we can't safely call nfsv4 close on a file until
3246  * the inode is cleared. This in turn means that we are not good
3247  * NFSv4 citizens - we do not indicate to the server to update the file's
3248  * share state even when we are done with one of the three share
3249  * stateid's in the inode.
3250  *
3251  * NOTE: Caller must be holding the sp->so_owner semaphore!
3252  */
3253 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3254 {
3255         struct nfs_server *server = NFS_SERVER(state->inode);
3256         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3257         struct nfs4_closedata *calldata;
3258         struct nfs4_state_owner *sp = state->owner;
3259         struct rpc_task *task;
3260         struct rpc_message msg = {
3261                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3262                 .rpc_cred = state->owner->so_cred,
3263         };
3264         struct rpc_task_setup task_setup_data = {
3265                 .rpc_client = server->client,
3266                 .rpc_message = &msg,
3267                 .callback_ops = &nfs4_close_ops,
3268                 .workqueue = nfsiod_workqueue,
3269                 .flags = RPC_TASK_ASYNC,
3270         };
3271         int status = -ENOMEM;
3272
3273         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3274                 &task_setup_data.rpc_client, &msg);
3275
3276         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3277         if (calldata == NULL)
3278                 goto out;
3279         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3280         calldata->inode = state->inode;
3281         calldata->state = state;
3282         calldata->arg.fh = NFS_FH(state->inode);
3283         /* Serialization for the sequence id */
3284         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3285         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3286         if (IS_ERR(calldata->arg.seqid))
3287                 goto out_free_calldata;
3288         nfs_fattr_init(&calldata->fattr);
3289         calldata->arg.fmode = 0;
3290         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3291         calldata->res.fattr = &calldata->fattr;
3292         calldata->res.seqid = calldata->arg.seqid;
3293         calldata->res.server = server;
3294         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3295         calldata->lr.roc = pnfs_roc(state->inode,
3296                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3297         if (calldata->lr.roc) {
3298                 calldata->arg.lr_args = &calldata->lr.arg;
3299                 calldata->res.lr_res = &calldata->lr.res;
3300         }
3301         nfs_sb_active(calldata->inode->i_sb);
3302
3303         msg.rpc_argp = &calldata->arg;
3304         msg.rpc_resp = &calldata->res;
3305         task_setup_data.callback_data = calldata;
3306         task = rpc_run_task(&task_setup_data);
3307         if (IS_ERR(task))
3308                 return PTR_ERR(task);
3309         status = 0;
3310         if (wait)
3311                 status = rpc_wait_for_completion_task(task);
3312         rpc_put_task(task);
3313         return status;
3314 out_free_calldata:
3315         kfree(calldata);
3316 out:
3317         nfs4_put_open_state(state);
3318         nfs4_put_state_owner(sp);
3319         return status;
3320 }
3321
3322 static struct inode *
3323 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3324                 int open_flags, struct iattr *attr, int *opened)
3325 {
3326         struct nfs4_state *state;
3327         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3328
3329         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3330
3331         /* Protect against concurrent sillydeletes */
3332         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3333
3334         nfs4_label_release_security(label);
3335
3336         if (IS_ERR(state))
3337                 return ERR_CAST(state);
3338         return state->inode;
3339 }
3340
3341 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3342 {
3343         if (ctx->state == NULL)
3344                 return;
3345         if (is_sync)
3346                 nfs4_close_sync(ctx->state, ctx->mode);
3347         else
3348                 nfs4_close_state(ctx->state, ctx->mode);
3349 }
3350
3351 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3352 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3353 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3354
3355 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3356 {
3357         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3358         struct nfs4_server_caps_arg args = {
3359                 .fhandle = fhandle,
3360                 .bitmask = bitmask,
3361         };
3362         struct nfs4_server_caps_res res = {};
3363         struct rpc_message msg = {
3364                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3365                 .rpc_argp = &args,
3366                 .rpc_resp = &res,
3367         };
3368         int status;
3369
3370         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3371                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3372                      FATTR4_WORD0_LINK_SUPPORT |
3373                      FATTR4_WORD0_SYMLINK_SUPPORT |
3374                      FATTR4_WORD0_ACLSUPPORT;
3375         if (minorversion)
3376                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3377
3378         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3379         if (status == 0) {
3380                 /* Sanity check the server answers */
3381                 switch (minorversion) {
3382                 case 0:
3383                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3384                         res.attr_bitmask[2] = 0;
3385                         break;
3386                 case 1:
3387                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3388                         break;
3389                 case 2:
3390                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3391                 }
3392                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3393                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3394                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3395                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3396                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3397                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3398                                 NFS_CAP_SECURITY_LABEL);
3399                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3400                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3401                         server->caps |= NFS_CAP_ACLS;
3402                 if (res.has_links != 0)
3403                         server->caps |= NFS_CAP_HARDLINKS;
3404                 if (res.has_symlinks != 0)
3405                         server->caps |= NFS_CAP_SYMLINKS;
3406                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3407                         server->caps |= NFS_CAP_FILEID;
3408                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3409                         server->caps |= NFS_CAP_MODE;
3410                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3411                         server->caps |= NFS_CAP_NLINK;
3412                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3413                         server->caps |= NFS_CAP_OWNER;
3414                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3415                         server->caps |= NFS_CAP_OWNER_GROUP;
3416                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3417                         server->caps |= NFS_CAP_ATIME;
3418                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3419                         server->caps |= NFS_CAP_CTIME;
3420                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3421                         server->caps |= NFS_CAP_MTIME;
3422 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3423                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3424                         server->caps |= NFS_CAP_SECURITY_LABEL;
3425 #endif
3426                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3427                                 sizeof(server->attr_bitmask));
3428                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3429
3430                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3431                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3432                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3433                 server->cache_consistency_bitmask[2] = 0;
3434                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3435                         sizeof(server->exclcreat_bitmask));
3436                 server->acl_bitmask = res.acl_bitmask;
3437                 server->fh_expire_type = res.fh_expire_type;
3438         }
3439
3440         return status;
3441 }
3442
3443 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3444 {
3445         struct nfs4_exception exception = { };
3446         int err;
3447         do {
3448                 err = nfs4_handle_exception(server,
3449                                 _nfs4_server_capabilities(server, fhandle),
3450                                 &exception);
3451         } while (exception.retry);
3452         return err;
3453 }
3454
3455 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3456                 struct nfs_fsinfo *info)
3457 {
3458         u32 bitmask[3];
3459         struct nfs4_lookup_root_arg args = {
3460                 .bitmask = bitmask,
3461         };
3462         struct nfs4_lookup_res res = {
3463                 .server = server,
3464                 .fattr = info->fattr,
3465                 .fh = fhandle,
3466         };
3467         struct rpc_message msg = {
3468                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3469                 .rpc_argp = &args,
3470                 .rpc_resp = &res,
3471         };
3472
3473         bitmask[0] = nfs4_fattr_bitmap[0];
3474         bitmask[1] = nfs4_fattr_bitmap[1];
3475         /*
3476          * Process the label in the upcoming getfattr
3477          */
3478         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3479
3480         nfs_fattr_init(info->fattr);
3481         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3482 }
3483
3484 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3485                 struct nfs_fsinfo *info)
3486 {
3487         struct nfs4_exception exception = { };
3488         int err;
3489         do {
3490                 err = _nfs4_lookup_root(server, fhandle, info);
3491                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3492                 switch (err) {
3493                 case 0:
3494                 case -NFS4ERR_WRONGSEC:
3495                         goto out;
3496                 default:
3497                         err = nfs4_handle_exception(server, err, &exception);
3498                 }
3499         } while (exception.retry);
3500 out:
3501         return err;
3502 }
3503
3504 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3505                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3506 {
3507         struct rpc_auth_create_args auth_args = {
3508                 .pseudoflavor = flavor,
3509         };
3510         struct rpc_auth *auth;
3511         int ret;
3512
3513         auth = rpcauth_create(&auth_args, server->client);
3514         if (IS_ERR(auth)) {
3515                 ret = -EACCES;
3516                 goto out;
3517         }
3518         ret = nfs4_lookup_root(server, fhandle, info);
3519 out:
3520         return ret;
3521 }
3522
3523 /*
3524  * Retry pseudoroot lookup with various security flavors.  We do this when:
3525  *
3526  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3527  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3528  *
3529  * Returns zero on success, or a negative NFS4ERR value, or a
3530  * negative errno value.
3531  */
3532 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3533                               struct nfs_fsinfo *info)
3534 {
3535         /* Per 3530bis 15.33.5 */
3536         static const rpc_authflavor_t flav_array[] = {
3537                 RPC_AUTH_GSS_KRB5P,
3538                 RPC_AUTH_GSS_KRB5I,
3539                 RPC_AUTH_GSS_KRB5,
3540                 RPC_AUTH_UNIX,                  /* courtesy */
3541                 RPC_AUTH_NULL,
3542         };
3543         int status = -EPERM;
3544         size_t i;
3545
3546         if (server->auth_info.flavor_len > 0) {
3547                 /* try each flavor specified by user */
3548                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3549                         status = nfs4_lookup_root_sec(server, fhandle, info,
3550                                                 server->auth_info.flavors[i]);
3551                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3552                                 continue;
3553                         break;
3554                 }
3555         } else {
3556                 /* no flavors specified by user, try default list */
3557                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3558                         status = nfs4_lookup_root_sec(server, fhandle, info,
3559                                                       flav_array[i]);
3560                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3561                                 continue;
3562                         break;
3563                 }
3564         }
3565
3566         /*
3567          * -EACCESS could mean that the user doesn't have correct permissions
3568          * to access the mount.  It could also mean that we tried to mount
3569          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3570          * existing mount programs don't handle -EACCES very well so it should
3571          * be mapped to -EPERM instead.
3572          */
3573         if (status == -EACCES)
3574                 status = -EPERM;
3575         return status;
3576 }
3577
3578 /**
3579  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3580  * @server: initialized nfs_server handle
3581  * @fhandle: we fill in the pseudo-fs root file handle
3582  * @info: we fill in an FSINFO struct
3583  * @auth_probe: probe the auth flavours
3584  *
3585  * Returns zero on success, or a negative errno.
3586  */
3587 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3588                          struct nfs_fsinfo *info,
3589                          bool auth_probe)
3590 {
3591         int status = 0;
3592
3593         if (!auth_probe)
3594                 status = nfs4_lookup_root(server, fhandle, info);
3595
3596         if (auth_probe || status == NFS4ERR_WRONGSEC)
3597                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3598                                 fhandle, info);
3599
3600         if (status == 0)
3601                 status = nfs4_server_capabilities(server, fhandle);
3602         if (status == 0)
3603                 status = nfs4_do_fsinfo(server, fhandle, info);
3604
3605         return nfs4_map_errors(status);
3606 }
3607
3608 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3609                               struct nfs_fsinfo *info)
3610 {
3611         int error;
3612         struct nfs_fattr *fattr = info->fattr;
3613         struct nfs4_label *label = NULL;
3614
3615         error = nfs4_server_capabilities(server, mntfh);
3616         if (error < 0) {
3617                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3618                 return error;
3619         }
3620
3621         label = nfs4_label_alloc(server, GFP_KERNEL);
3622         if (IS_ERR(label))
3623                 return PTR_ERR(label);
3624
3625         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3626         if (error < 0) {
3627                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3628                 goto err_free_label;
3629         }
3630
3631         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3632             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3633                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3634
3635 err_free_label:
3636         nfs4_label_free(label);
3637
3638         return error;
3639 }
3640
3641 /*
3642  * Get locations and (maybe) other attributes of a referral.
3643  * Note that we'll actually follow the referral later when
3644  * we detect fsid mismatch in inode revalidation
3645  */
3646 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3647                              const struct qstr *name, struct nfs_fattr *fattr,
3648                              struct nfs_fh *fhandle)
3649 {
3650         int status = -ENOMEM;
3651         struct page *page = NULL;
3652         struct nfs4_fs_locations *locations = NULL;
3653
3654         page = alloc_page(GFP_KERNEL);
3655         if (page == NULL)
3656                 goto out;
3657         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3658         if (locations == NULL)
3659                 goto out;
3660
3661         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3662         if (status != 0)
3663                 goto out;
3664
3665         /*
3666          * If the fsid didn't change, this is a migration event, not a
3667          * referral.  Cause us to drop into the exception handler, which
3668          * will kick off migration recovery.
3669          */
3670         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3671                 dprintk("%s: server did not return a different fsid for"
3672                         " a referral at %s\n", __func__, name->name);
3673                 status = -NFS4ERR_MOVED;
3674                 goto out;
3675         }
3676         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3677         nfs_fixup_referral_attributes(&locations->fattr);
3678
3679         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3680         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3681         memset(fhandle, 0, sizeof(struct nfs_fh));
3682 out:
3683         if (page)
3684                 __free_page(page);
3685         kfree(locations);
3686         return status;
3687 }
3688
3689 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3690                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3691 {
3692         struct nfs4_getattr_arg args = {
3693                 .fh = fhandle,
3694                 .bitmask = server->attr_bitmask,
3695         };
3696         struct nfs4_getattr_res res = {
3697                 .fattr = fattr,
3698                 .label = label,
3699                 .server = server,
3700         };
3701         struct rpc_message msg = {
3702                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3703                 .rpc_argp = &args,
3704                 .rpc_resp = &res,
3705         };
3706
3707         args.bitmask = nfs4_bitmask(server, label);
3708
3709         nfs_fattr_init(fattr);
3710         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3711 }
3712
3713 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3714                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3715 {
3716         struct nfs4_exception exception = { };
3717         int err;
3718         do {
3719                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3720                 trace_nfs4_getattr(server, fhandle, fattr, err);
3721                 err = nfs4_handle_exception(server, err,
3722                                 &exception);
3723         } while (exception.retry);
3724         return err;
3725 }
3726
3727 /*
3728  * The file is not closed if it is opened due to the a request to change
3729  * the size of the file. The open call will not be needed once the
3730  * VFS layer lookup-intents are implemented.
3731  *
3732  * Close is called when the inode is destroyed.
3733  * If we haven't opened the file for O_WRONLY, we
3734  * need to in the size_change case to obtain a stateid.
3735  *
3736  * Got race?
3737  * Because OPEN is always done by name in nfsv4, it is
3738  * possible that we opened a different file by the same
3739  * name.  We can recognize this race condition, but we
3740  * can't do anything about it besides returning an error.
3741  *
3742  * This will be fixed with VFS changes (lookup-intent).
3743  */
3744 static int
3745 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3746                   struct iattr *sattr)
3747 {
3748         struct inode *inode = d_inode(dentry);
3749         struct rpc_cred *cred = NULL;
3750         struct nfs_open_context *ctx = NULL;
3751         struct nfs4_label *label = NULL;
3752         int status;
3753
3754         if (pnfs_ld_layoutret_on_setattr(inode) &&
3755             sattr->ia_valid & ATTR_SIZE &&
3756             sattr->ia_size < i_size_read(inode))
3757                 pnfs_commit_and_return_layout(inode);
3758
3759         nfs_fattr_init(fattr);
3760
3761         /* Deal with open(O_TRUNC) */
3762         if (sattr->ia_valid & ATTR_OPEN)
3763                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3764
3765         /* Optimization: if the end result is no change, don't RPC */
3766         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3767                 return 0;
3768
3769         /* Search for an existing open(O_WRITE) file */
3770         if (sattr->ia_valid & ATTR_FILE) {
3771
3772                 ctx = nfs_file_open_context(sattr->ia_file);
3773                 if (ctx)
3774                         cred = ctx->cred;
3775         }
3776
3777         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3778         if (IS_ERR(label))
3779                 return PTR_ERR(label);
3780
3781         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3782         if (status == 0) {
3783                 nfs_setattr_update_inode(inode, sattr, fattr);
3784                 nfs_setsecurity(inode, fattr, label);
3785         }
3786         nfs4_label_free(label);
3787         return status;
3788 }
3789
3790 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3791                 const struct qstr *name, struct nfs_fh *fhandle,
3792                 struct nfs_fattr *fattr, struct nfs4_label *label)
3793 {
3794         struct nfs_server *server = NFS_SERVER(dir);
3795         int                    status;
3796         struct nfs4_lookup_arg args = {
3797                 .bitmask = server->attr_bitmask,
3798                 .dir_fh = NFS_FH(dir),
3799                 .name = name,
3800         };
3801         struct nfs4_lookup_res res = {
3802                 .server = server,
3803                 .fattr = fattr,
3804                 .label = label,
3805                 .fh = fhandle,
3806         };
3807         struct rpc_message msg = {
3808                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3809                 .rpc_argp = &args,
3810                 .rpc_resp = &res,
3811         };
3812
3813         args.bitmask = nfs4_bitmask(server, label);
3814
3815         nfs_fattr_init(fattr);
3816
3817         dprintk("NFS call  lookup %s\n", name->name);
3818         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3819         dprintk("NFS reply lookup: %d\n", status);
3820         return status;
3821 }
3822
3823 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3824 {
3825         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3826                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3827         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3828         fattr->nlink = 2;
3829 }
3830
3831 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3832                                    const struct qstr *name, struct nfs_fh *fhandle,
3833                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3834 {
3835         struct nfs4_exception exception = { };
3836         struct rpc_clnt *client = *clnt;
3837         int err;
3838         do {
3839                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3840                 trace_nfs4_lookup(dir, name, err);
3841                 switch (err) {
3842                 case -NFS4ERR_BADNAME:
3843                         err = -ENOENT;
3844                         goto out;
3845                 case -NFS4ERR_MOVED:
3846                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3847                         if (err == -NFS4ERR_MOVED)
3848                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3849                         goto out;
3850                 case -NFS4ERR_WRONGSEC:
3851                         err = -EPERM;
3852                         if (client != *clnt)
3853                                 goto out;
3854                         client = nfs4_negotiate_security(client, dir, name);
3855                         if (IS_ERR(client))
3856                                 return PTR_ERR(client);
3857
3858                         exception.retry = 1;
3859                         break;
3860                 default:
3861                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3862                 }
3863         } while (exception.retry);
3864
3865 out:
3866         if (err == 0)
3867                 *clnt = client;
3868         else if (client != *clnt)
3869                 rpc_shutdown_client(client);
3870
3871         return err;
3872 }
3873
3874 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3875                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3876                             struct nfs4_label *label)
3877 {
3878         int status;
3879         struct rpc_clnt *client = NFS_CLIENT(dir);
3880
3881         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3882         if (client != NFS_CLIENT(dir)) {
3883                 rpc_shutdown_client(client);
3884                 nfs_fixup_secinfo_attributes(fattr);
3885         }
3886         return status;
3887 }
3888
3889 struct rpc_clnt *
3890 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3891                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3892 {
3893         struct rpc_clnt *client = NFS_CLIENT(dir);
3894         int status;
3895
3896         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3897         if (status < 0)
3898                 return ERR_PTR(status);
3899         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3900 }
3901
3902 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3903 {
3904         struct nfs_server *server = NFS_SERVER(inode);
3905         struct nfs4_accessargs args = {
3906                 .fh = NFS_FH(inode),
3907                 .bitmask = server->cache_consistency_bitmask,
3908         };
3909         struct nfs4_accessres res = {
3910                 .server = server,
3911         };
3912         struct rpc_message msg = {
3913                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3914                 .rpc_argp = &args,
3915                 .rpc_resp = &res,
3916                 .rpc_cred = entry->cred,
3917         };
3918         int mode = entry->mask;
3919         int status = 0;
3920
3921         /*
3922          * Determine which access bits we want to ask for...
3923          */
3924         if (mode & MAY_READ)
3925                 args.access |= NFS4_ACCESS_READ;
3926         if (S_ISDIR(inode->i_mode)) {
3927                 if (mode & MAY_WRITE)
3928                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3929                 if (mode & MAY_EXEC)
3930                         args.access |= NFS4_ACCESS_LOOKUP;
3931         } else {
3932                 if (mode & MAY_WRITE)
3933                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3934                 if (mode & MAY_EXEC)
3935                         args.access |= NFS4_ACCESS_EXECUTE;
3936         }
3937
3938         res.fattr = nfs_alloc_fattr();
3939         if (res.fattr == NULL)
3940                 return -ENOMEM;
3941
3942         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3943         if (!status) {
3944                 nfs_access_set_mask(entry, res.access);
3945                 nfs_refresh_inode(inode, res.fattr);
3946         }
3947         nfs_free_fattr(res.fattr);
3948         return status;
3949 }
3950
3951 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3952 {
3953         struct nfs4_exception exception = { };
3954         int err;
3955         do {
3956                 err = _nfs4_proc_access(inode, entry);
3957                 trace_nfs4_access(inode, err);
3958                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3959                                 &exception);
3960         } while (exception.retry);
3961         return err;
3962 }
3963
3964 /*
3965  * TODO: For the time being, we don't try to get any attributes
3966  * along with any of the zero-copy operations READ, READDIR,
3967  * READLINK, WRITE.
3968  *
3969  * In the case of the first three, we want to put the GETATTR
3970  * after the read-type operation -- this is because it is hard
3971  * to predict the length of a GETATTR response in v4, and thus
3972  * align the READ data correctly.  This means that the GETATTR
3973  * may end up partially falling into the page cache, and we should
3974  * shift it into the 'tail' of the xdr_buf before processing.
3975  * To do this efficiently, we need to know the total length
3976  * of data received, which doesn't seem to be available outside
3977  * of the RPC layer.
3978  *
3979  * In the case of WRITE, we also want to put the GETATTR after
3980  * the operation -- in this case because we want to make sure
3981  * we get the post-operation mtime and size.
3982  *
3983  * Both of these changes to the XDR layer would in fact be quite
3984  * minor, but I decided to leave them for a subsequent patch.
3985  */
3986 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3987                 unsigned int pgbase, unsigned int pglen)
3988 {
3989         struct nfs4_readlink args = {
3990                 .fh       = NFS_FH(inode),
3991                 .pgbase   = pgbase,
3992                 .pglen    = pglen,
3993                 .pages    = &page,
3994         };
3995         struct nfs4_readlink_res res;
3996         struct rpc_message msg = {
3997                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3998                 .rpc_argp = &args,
3999                 .rpc_resp = &res,
4000         };
4001
4002         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4003 }
4004
4005 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4006                 unsigned int pgbase, unsigned int pglen)
4007 {
4008         struct nfs4_exception exception = { };
4009         int err;
4010         do {
4011                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4012                 trace_nfs4_readlink(inode, err);
4013                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4014                                 &exception);
4015         } while (exception.retry);
4016         return err;
4017 }
4018
4019 /*
4020  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4021  */
4022 static int
4023 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4024                  int flags)
4025 {
4026         struct nfs_server *server = NFS_SERVER(dir);
4027         struct nfs4_label l, *ilabel = NULL;
4028         struct nfs_open_context *ctx;
4029         struct nfs4_state *state;
4030         int status = 0;
4031
4032         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4033         if (IS_ERR(ctx))
4034                 return PTR_ERR(ctx);
4035
4036         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4037
4038         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4039                 sattr->ia_mode &= ~current_umask();
4040         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4041         if (IS_ERR(state)) {
4042                 status = PTR_ERR(state);
4043                 goto out;
4044         }
4045 out:
4046         nfs4_label_release_security(ilabel);
4047         put_nfs_open_context(ctx);
4048         return status;
4049 }
4050
4051 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4052 {
4053         struct nfs_server *server = NFS_SERVER(dir);
4054         struct nfs_removeargs args = {
4055                 .fh = NFS_FH(dir),
4056                 .name = *name,
4057         };
4058         struct nfs_removeres res = {
4059                 .server = server,
4060         };
4061         struct rpc_message msg = {
4062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4063                 .rpc_argp = &args,
4064                 .rpc_resp = &res,
4065         };
4066         int status;
4067
4068         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4069         if (status == 0)
4070                 update_changeattr(dir, &res.cinfo);
4071         return status;
4072 }
4073
4074 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4075 {
4076         struct nfs4_exception exception = { };
4077         int err;
4078         do {
4079                 err = _nfs4_proc_remove(dir, name);
4080                 trace_nfs4_remove(dir, name, err);
4081                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4082                                 &exception);
4083         } while (exception.retry);
4084         return err;
4085 }
4086
4087 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4088 {
4089         struct nfs_server *server = NFS_SERVER(dir);
4090         struct nfs_removeargs *args = msg->rpc_argp;
4091         struct nfs_removeres *res = msg->rpc_resp;
4092
4093         res->server = server;
4094         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4095         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4096
4097         nfs_fattr_init(res->dir_attr);
4098 }
4099
4100 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4101 {
4102         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4103                         &data->args.seq_args,
4104                         &data->res.seq_res,
4105                         task);
4106 }
4107
4108 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4109 {
4110         struct nfs_unlinkdata *data = task->tk_calldata;
4111         struct nfs_removeres *res = &data->res;
4112
4113         if (!nfs4_sequence_done(task, &res->seq_res))
4114                 return 0;
4115         if (nfs4_async_handle_error(task, res->server, NULL,
4116                                     &data->timeout) == -EAGAIN)
4117                 return 0;
4118         update_changeattr(dir, &res->cinfo);
4119         return 1;
4120 }
4121
4122 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4123 {
4124         struct nfs_server *server = NFS_SERVER(dir);
4125         struct nfs_renameargs *arg = msg->rpc_argp;
4126         struct nfs_renameres *res = msg->rpc_resp;
4127
4128         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4129         res->server = server;
4130         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4131 }
4132
4133 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4134 {
4135         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4136                         &data->args.seq_args,
4137                         &data->res.seq_res,
4138                         task);
4139 }
4140
4141 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4142                                  struct inode *new_dir)
4143 {
4144         struct nfs_renamedata *data = task->tk_calldata;
4145         struct nfs_renameres *res = &data->res;
4146
4147         if (!nfs4_sequence_done(task, &res->seq_res))
4148                 return 0;
4149         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4150                 return 0;
4151
4152         update_changeattr(old_dir, &res->old_cinfo);
4153         update_changeattr(new_dir, &res->new_cinfo);
4154         return 1;
4155 }
4156
4157 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4158 {
4159         struct nfs_server *server = NFS_SERVER(inode);
4160         struct nfs4_link_arg arg = {
4161                 .fh     = NFS_FH(inode),
4162                 .dir_fh = NFS_FH(dir),
4163                 .name   = name,
4164                 .bitmask = server->attr_bitmask,
4165         };
4166         struct nfs4_link_res res = {
4167                 .server = server,
4168                 .label = NULL,
4169         };
4170         struct rpc_message msg = {
4171                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4172                 .rpc_argp = &arg,
4173                 .rpc_resp = &res,
4174         };
4175         int status = -ENOMEM;
4176
4177         res.fattr = nfs_alloc_fattr();
4178         if (res.fattr == NULL)
4179                 goto out;
4180
4181         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4182         if (IS_ERR(res.label)) {
4183                 status = PTR_ERR(res.label);
4184                 goto out;
4185         }
4186         arg.bitmask = nfs4_bitmask(server, res.label);
4187
4188         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4189         if (!status) {
4190                 update_changeattr(dir, &res.cinfo);
4191                 status = nfs_post_op_update_inode(inode, res.fattr);
4192                 if (!status)
4193                         nfs_setsecurity(inode, res.fattr, res.label);
4194         }
4195
4196
4197         nfs4_label_free(res.label);
4198
4199 out:
4200         nfs_free_fattr(res.fattr);
4201         return status;
4202 }
4203
4204 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4205 {
4206         struct nfs4_exception exception = { };
4207         int err;
4208         do {
4209                 err = nfs4_handle_exception(NFS_SERVER(inode),
4210                                 _nfs4_proc_link(inode, dir, name),
4211                                 &exception);
4212         } while (exception.retry);
4213         return err;
4214 }
4215
4216 struct nfs4_createdata {
4217         struct rpc_message msg;
4218         struct nfs4_create_arg arg;
4219         struct nfs4_create_res res;
4220         struct nfs_fh fh;
4221         struct nfs_fattr fattr;
4222         struct nfs4_label *label;
4223 };
4224
4225 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4226                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4227 {
4228         struct nfs4_createdata *data;
4229
4230         data = kzalloc(sizeof(*data), GFP_KERNEL);
4231         if (data != NULL) {
4232                 struct nfs_server *server = NFS_SERVER(dir);
4233
4234                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4235                 if (IS_ERR(data->label))
4236                         goto out_free;
4237
4238                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4239                 data->msg.rpc_argp = &data->arg;
4240                 data->msg.rpc_resp = &data->res;
4241                 data->arg.dir_fh = NFS_FH(dir);
4242                 data->arg.server = server;
4243                 data->arg.name = name;
4244                 data->arg.attrs = sattr;
4245                 data->arg.ftype = ftype;
4246                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4247                 data->arg.umask = current_umask();
4248                 data->res.server = server;
4249                 data->res.fh = &data->fh;
4250                 data->res.fattr = &data->fattr;
4251                 data->res.label = data->label;
4252                 nfs_fattr_init(data->res.fattr);
4253         }
4254         return data;
4255 out_free:
4256         kfree(data);
4257         return NULL;
4258 }
4259
4260 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4261 {
4262         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4263                                     &data->arg.seq_args, &data->res.seq_res, 1);
4264         if (status == 0) {
4265                 update_changeattr(dir, &data->res.dir_cinfo);
4266                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4267         }
4268         return status;
4269 }
4270
4271 static void nfs4_free_createdata(struct nfs4_createdata *data)
4272 {
4273         nfs4_label_free(data->label);
4274         kfree(data);
4275 }
4276
4277 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4278                 struct page *page, unsigned int len, struct iattr *sattr,
4279                 struct nfs4_label *label)
4280 {
4281         struct nfs4_createdata *data;
4282         int status = -ENAMETOOLONG;
4283
4284         if (len > NFS4_MAXPATHLEN)
4285                 goto out;
4286
4287         status = -ENOMEM;
4288         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4289         if (data == NULL)
4290                 goto out;
4291
4292         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4293         data->arg.u.symlink.pages = &page;
4294         data->arg.u.symlink.len = len;
4295         data->arg.label = label;
4296
4297         status = nfs4_do_create(dir, dentry, data);
4298
4299         nfs4_free_createdata(data);
4300 out:
4301         return status;
4302 }
4303
4304 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4305                 struct page *page, unsigned int len, struct iattr *sattr)
4306 {
4307         struct nfs4_exception exception = { };
4308         struct nfs4_label l, *label = NULL;
4309         int err;
4310
4311         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4312
4313         do {
4314                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4315                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4316                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4317                                 &exception);
4318         } while (exception.retry);
4319
4320         nfs4_label_release_security(label);
4321         return err;
4322 }
4323
4324 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4325                 struct iattr *sattr, struct nfs4_label *label)
4326 {
4327         struct nfs4_createdata *data;
4328         int status = -ENOMEM;
4329
4330         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4331         if (data == NULL)
4332                 goto out;
4333
4334         data->arg.label = label;
4335         status = nfs4_do_create(dir, dentry, data);
4336
4337         nfs4_free_createdata(data);
4338 out:
4339         return status;
4340 }
4341
4342 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4343                 struct iattr *sattr)
4344 {
4345         struct nfs_server *server = NFS_SERVER(dir);
4346         struct nfs4_exception exception = { };
4347         struct nfs4_label l, *label = NULL;
4348         int err;
4349
4350         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4351
4352         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4353                 sattr->ia_mode &= ~current_umask();
4354         do {
4355                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4356                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4357                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4358                                 &exception);
4359         } while (exception.retry);
4360         nfs4_label_release_security(label);
4361
4362         return err;
4363 }
4364
4365 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4366                 u64 cookie, struct page **pages, unsigned int count, int plus)
4367 {
4368         struct inode            *dir = d_inode(dentry);
4369         struct nfs4_readdir_arg args = {
4370                 .fh = NFS_FH(dir),
4371                 .pages = pages,
4372                 .pgbase = 0,
4373                 .count = count,
4374                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4375                 .plus = plus,
4376         };
4377         struct nfs4_readdir_res res;
4378         struct rpc_message msg = {
4379                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4380                 .rpc_argp = &args,
4381                 .rpc_resp = &res,
4382                 .rpc_cred = cred,
4383         };
4384         int                     status;
4385
4386         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4387                         dentry,
4388                         (unsigned long long)cookie);
4389         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4390         res.pgbase = args.pgbase;
4391         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4392         if (status >= 0) {
4393                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4394                 status += args.pgbase;
4395         }
4396
4397         nfs_invalidate_atime(dir);
4398
4399         dprintk("%s: returns %d\n", __func__, status);
4400         return status;
4401 }
4402
4403 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4404                 u64 cookie, struct page **pages, unsigned int count, int plus)
4405 {
4406         struct nfs4_exception exception = { };
4407         int err;
4408         do {
4409                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4410                                 pages, count, plus);
4411                 trace_nfs4_readdir(d_inode(dentry), err);
4412                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4413                                 &exception);
4414         } while (exception.retry);
4415         return err;
4416 }
4417
4418 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4419                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4420 {
4421         struct nfs4_createdata *data;
4422         int mode = sattr->ia_mode;
4423         int status = -ENOMEM;
4424
4425         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4426         if (data == NULL)
4427                 goto out;
4428
4429         if (S_ISFIFO(mode))
4430                 data->arg.ftype = NF4FIFO;
4431         else if (S_ISBLK(mode)) {
4432                 data->arg.ftype = NF4BLK;
4433                 data->arg.u.device.specdata1 = MAJOR(rdev);
4434                 data->arg.u.device.specdata2 = MINOR(rdev);
4435         }
4436         else if (S_ISCHR(mode)) {
4437                 data->arg.ftype = NF4CHR;
4438                 data->arg.u.device.specdata1 = MAJOR(rdev);
4439                 data->arg.u.device.specdata2 = MINOR(rdev);
4440         } else if (!S_ISSOCK(mode)) {
4441                 status = -EINVAL;
4442                 goto out_free;
4443         }
4444
4445         data->arg.label = label;
4446         status = nfs4_do_create(dir, dentry, data);
4447 out_free:
4448         nfs4_free_createdata(data);
4449 out:
4450         return status;
4451 }
4452
4453 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4454                 struct iattr *sattr, dev_t rdev)
4455 {
4456         struct nfs_server *server = NFS_SERVER(dir);
4457         struct nfs4_exception exception = { };
4458         struct nfs4_label l, *label = NULL;
4459         int err;
4460
4461         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4462
4463         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4464                 sattr->ia_mode &= ~current_umask();
4465         do {
4466                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4467                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4468                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4469                                 &exception);
4470         } while (exception.retry);
4471
4472         nfs4_label_release_security(label);
4473
4474         return err;
4475 }
4476
4477 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4478                  struct nfs_fsstat *fsstat)
4479 {
4480         struct nfs4_statfs_arg args = {
4481                 .fh = fhandle,
4482                 .bitmask = server->attr_bitmask,
4483         };
4484         struct nfs4_statfs_res res = {
4485                 .fsstat = fsstat,
4486         };
4487         struct rpc_message msg = {
4488                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4489                 .rpc_argp = &args,
4490                 .rpc_resp = &res,
4491         };
4492
4493         nfs_fattr_init(fsstat->fattr);
4494         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4495 }
4496
4497 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4498 {
4499         struct nfs4_exception exception = { };
4500         int err;
4501         do {
4502                 err = nfs4_handle_exception(server,
4503                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4504                                 &exception);
4505         } while (exception.retry);
4506         return err;
4507 }
4508
4509 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4510                 struct nfs_fsinfo *fsinfo)
4511 {
4512         struct nfs4_fsinfo_arg args = {
4513                 .fh = fhandle,
4514                 .bitmask = server->attr_bitmask,
4515         };
4516         struct nfs4_fsinfo_res res = {
4517                 .fsinfo = fsinfo,
4518         };
4519         struct rpc_message msg = {
4520                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4521                 .rpc_argp = &args,
4522                 .rpc_resp = &res,
4523         };
4524
4525         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4526 }
4527
4528 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4529 {
4530         struct nfs4_exception exception = { };
4531         unsigned long now = jiffies;
4532         int err;
4533
4534         do {
4535                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4536                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4537                 if (err == 0) {
4538                         nfs4_set_lease_period(server->nfs_client,
4539                                         fsinfo->lease_time * HZ,
4540                                         now);
4541                         break;
4542                 }
4543                 err = nfs4_handle_exception(server, err, &exception);
4544         } while (exception.retry);
4545         return err;
4546 }
4547
4548 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4549 {
4550         int error;
4551
4552         nfs_fattr_init(fsinfo->fattr);
4553         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4554         if (error == 0) {
4555                 /* block layout checks this! */
4556                 server->pnfs_blksize = fsinfo->blksize;
4557                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4558         }
4559
4560         return error;
4561 }
4562
4563 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4564                 struct nfs_pathconf *pathconf)
4565 {
4566         struct nfs4_pathconf_arg args = {
4567                 .fh = fhandle,
4568                 .bitmask = server->attr_bitmask,
4569         };
4570         struct nfs4_pathconf_res res = {
4571                 .pathconf = pathconf,
4572         };
4573         struct rpc_message msg = {
4574                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4575                 .rpc_argp = &args,
4576                 .rpc_resp = &res,
4577         };
4578
4579         /* None of the pathconf attributes are mandatory to implement */
4580         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4581                 memset(pathconf, 0, sizeof(*pathconf));
4582                 return 0;
4583         }
4584
4585         nfs_fattr_init(pathconf->fattr);
4586         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4587 }
4588
4589 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4590                 struct nfs_pathconf *pathconf)
4591 {
4592         struct nfs4_exception exception = { };
4593         int err;
4594
4595         do {
4596                 err = nfs4_handle_exception(server,
4597                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4598                                 &exception);
4599         } while (exception.retry);
4600         return err;
4601 }
4602
4603 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4604                 const struct nfs_open_context *ctx,
4605                 const struct nfs_lock_context *l_ctx,
4606                 fmode_t fmode)
4607 {
4608         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4609 }
4610 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4611
4612 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4613                 const struct nfs_open_context *ctx,
4614                 const struct nfs_lock_context *l_ctx,
4615                 fmode_t fmode)
4616 {
4617         nfs4_stateid current_stateid;
4618
4619         /* If the current stateid represents a lost lock, then exit */
4620         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4621                 return true;
4622         return nfs4_stateid_match(stateid, &current_stateid);
4623 }
4624
4625 static bool nfs4_error_stateid_expired(int err)
4626 {
4627         switch (err) {
4628         case -NFS4ERR_DELEG_REVOKED:
4629         case -NFS4ERR_ADMIN_REVOKED:
4630         case -NFS4ERR_BAD_STATEID:
4631         case -NFS4ERR_STALE_STATEID:
4632         case -NFS4ERR_OLD_STATEID:
4633         case -NFS4ERR_OPENMODE:
4634         case -NFS4ERR_EXPIRED:
4635                 return true;
4636         }
4637         return false;
4638 }
4639
4640 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4641 {
4642         struct nfs_server *server = NFS_SERVER(hdr->inode);
4643
4644         trace_nfs4_read(hdr, task->tk_status);
4645         if (task->tk_status < 0) {
4646                 struct nfs4_exception exception = {
4647                         .inode = hdr->inode,
4648                         .state = hdr->args.context->state,
4649                         .stateid = &hdr->args.stateid,
4650                 };
4651                 task->tk_status = nfs4_async_handle_exception(task,
4652                                 server, task->tk_status, &exception);
4653                 if (exception.retry) {
4654                         rpc_restart_call_prepare(task);
4655                         return -EAGAIN;
4656                 }
4657         }
4658
4659         if (task->tk_status > 0)
4660                 renew_lease(server, hdr->timestamp);
4661         return 0;
4662 }
4663
4664 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4665                 struct nfs_pgio_args *args)
4666 {
4667
4668         if (!nfs4_error_stateid_expired(task->tk_status) ||
4669                 nfs4_stateid_is_current(&args->stateid,
4670                                 args->context,
4671                                 args->lock_context,
4672                                 FMODE_READ))
4673                 return false;
4674         rpc_restart_call_prepare(task);
4675         return true;
4676 }
4677
4678 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4679 {
4680
4681         dprintk("--> %s\n", __func__);
4682
4683         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4684                 return -EAGAIN;
4685         if (nfs4_read_stateid_changed(task, &hdr->args))
4686                 return -EAGAIN;
4687         if (task->tk_status > 0)
4688                 nfs_invalidate_atime(hdr->inode);
4689         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4690                                     nfs4_read_done_cb(task, hdr);
4691 }
4692
4693 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4694                                  struct rpc_message *msg)
4695 {
4696         hdr->timestamp   = jiffies;
4697         if (!hdr->pgio_done_cb)
4698                 hdr->pgio_done_cb = nfs4_read_done_cb;
4699         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4700         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4701 }
4702
4703 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4704                                       struct nfs_pgio_header *hdr)
4705 {
4706         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4707                         &hdr->args.seq_args,
4708                         &hdr->res.seq_res,
4709                         task))
4710                 return 0;
4711         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4712                                 hdr->args.lock_context,
4713                                 hdr->rw_ops->rw_mode) == -EIO)
4714                 return -EIO;
4715         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4716                 return -EIO;
4717         return 0;
4718 }
4719
4720 static int nfs4_write_done_cb(struct rpc_task *task,
4721                               struct nfs_pgio_header *hdr)
4722 {
4723         struct inode *inode = hdr->inode;
4724
4725         trace_nfs4_write(hdr, task->tk_status);
4726         if (task->tk_status < 0) {
4727                 struct nfs4_exception exception = {
4728                         .inode = hdr->inode,
4729                         .state = hdr->args.context->state,
4730                         .stateid = &hdr->args.stateid,
4731                 };
4732                 task->tk_status = nfs4_async_handle_exception(task,
4733                                 NFS_SERVER(inode), task->tk_status,
4734                                 &exception);
4735                 if (exception.retry) {
4736                         rpc_restart_call_prepare(task);
4737                         return -EAGAIN;
4738                 }
4739         }
4740         if (task->tk_status >= 0) {
4741                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4742                 nfs_writeback_update_inode(hdr);
4743         }
4744         return 0;
4745 }
4746
4747 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4748                 struct nfs_pgio_args *args)
4749 {
4750
4751         if (!nfs4_error_stateid_expired(task->tk_status) ||
4752                 nfs4_stateid_is_current(&args->stateid,
4753                                 args->context,
4754                                 args->lock_context,
4755                                 FMODE_WRITE))
4756                 return false;
4757         rpc_restart_call_prepare(task);
4758         return true;
4759 }
4760
4761 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4762 {
4763         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4764                 return -EAGAIN;
4765         if (nfs4_write_stateid_changed(task, &hdr->args))
4766                 return -EAGAIN;
4767         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4768                 nfs4_write_done_cb(task, hdr);
4769 }
4770
4771 static
4772 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4773 {
4774         /* Don't request attributes for pNFS or O_DIRECT writes */
4775         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4776                 return false;
4777         /* Otherwise, request attributes if and only if we don't hold
4778          * a delegation
4779          */
4780         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4781 }
4782
4783 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4784                                   struct rpc_message *msg)
4785 {
4786         struct nfs_server *server = NFS_SERVER(hdr->inode);
4787
4788         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4789                 hdr->args.bitmask = NULL;
4790                 hdr->res.fattr = NULL;
4791         } else
4792                 hdr->args.bitmask = server->cache_consistency_bitmask;
4793
4794         if (!hdr->pgio_done_cb)
4795                 hdr->pgio_done_cb = nfs4_write_done_cb;
4796         hdr->res.server = server;
4797         hdr->timestamp   = jiffies;
4798
4799         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4800         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4801 }
4802
4803 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4804 {
4805         nfs4_setup_sequence(NFS_SERVER(data->inode),
4806                         &data->args.seq_args,
4807                         &data->res.seq_res,
4808                         task);
4809 }
4810
4811 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4812 {
4813         struct inode *inode = data->inode;
4814
4815         trace_nfs4_commit(data, task->tk_status);
4816         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4817                                     NULL, NULL) == -EAGAIN) {
4818                 rpc_restart_call_prepare(task);
4819                 return -EAGAIN;
4820         }
4821         return 0;
4822 }
4823
4824 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4825 {
4826         if (!nfs4_sequence_done(task, &data->res.seq_res))
4827                 return -EAGAIN;
4828         return data->commit_done_cb(task, data);
4829 }
4830
4831 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4832 {
4833         struct nfs_server *server = NFS_SERVER(data->inode);
4834
4835         if (data->commit_done_cb == NULL)
4836                 data->commit_done_cb = nfs4_commit_done_cb;
4837         data->res.server = server;
4838         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4839         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4840 }
4841
4842 struct nfs4_renewdata {
4843         struct nfs_client       *client;
4844         unsigned long           timestamp;
4845 };
4846
4847 /*
4848  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4849  * standalone procedure for queueing an asynchronous RENEW.
4850  */
4851 static void nfs4_renew_release(void *calldata)
4852 {
4853         struct nfs4_renewdata *data = calldata;
4854         struct nfs_client *clp = data->client;
4855
4856         if (atomic_read(&clp->cl_count) > 1)
4857                 nfs4_schedule_state_renewal(clp);
4858         nfs_put_client(clp);
4859         kfree(data);
4860 }
4861
4862 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4863 {
4864         struct nfs4_renewdata *data = calldata;
4865         struct nfs_client *clp = data->client;
4866         unsigned long timestamp = data->timestamp;
4867
4868         trace_nfs4_renew_async(clp, task->tk_status);
4869         switch (task->tk_status) {
4870         case 0:
4871                 break;
4872         case -NFS4ERR_LEASE_MOVED:
4873                 nfs4_schedule_lease_moved_recovery(clp);
4874                 break;
4875         default:
4876                 /* Unless we're shutting down, schedule state recovery! */
4877                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4878                         return;
4879                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4880                         nfs4_schedule_lease_recovery(clp);
4881                         return;
4882                 }
4883                 nfs4_schedule_path_down_recovery(clp);
4884         }
4885         do_renew_lease(clp, timestamp);
4886 }
4887
4888 static const struct rpc_call_ops nfs4_renew_ops = {
4889         .rpc_call_done = nfs4_renew_done,
4890         .rpc_release = nfs4_renew_release,
4891 };
4892
4893 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4894 {
4895         struct rpc_message msg = {
4896                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4897                 .rpc_argp       = clp,
4898                 .rpc_cred       = cred,
4899         };
4900         struct nfs4_renewdata *data;
4901
4902         if (renew_flags == 0)
4903                 return 0;
4904         if (!atomic_inc_not_zero(&clp->cl_count))
4905                 return -EIO;
4906         data = kmalloc(sizeof(*data), GFP_NOFS);
4907         if (data == NULL)
4908                 return -ENOMEM;
4909         data->client = clp;
4910         data->timestamp = jiffies;
4911         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4912                         &nfs4_renew_ops, data);
4913 }
4914
4915 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4916 {
4917         struct rpc_message msg = {
4918                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4919                 .rpc_argp       = clp,
4920                 .rpc_cred       = cred,
4921         };
4922         unsigned long now = jiffies;
4923         int status;
4924
4925         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4926         if (status < 0)
4927                 return status;
4928         do_renew_lease(clp, now);
4929         return 0;
4930 }
4931
4932 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4933 {
4934         return server->caps & NFS_CAP_ACLS;
4935 }
4936
4937 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4938  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4939  * the stack.
4940  */
4941 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4942
4943 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4944                 struct page **pages)
4945 {
4946         struct page *newpage, **spages;
4947         int rc = 0;
4948         size_t len;
4949         spages = pages;
4950
4951         do {
4952                 len = min_t(size_t, PAGE_SIZE, buflen);
4953                 newpage = alloc_page(GFP_KERNEL);
4954
4955                 if (newpage == NULL)
4956                         goto unwind;
4957                 memcpy(page_address(newpage), buf, len);
4958                 buf += len;
4959                 buflen -= len;
4960                 *pages++ = newpage;
4961                 rc++;
4962         } while (buflen != 0);
4963
4964         return rc;
4965
4966 unwind:
4967         for(; rc > 0; rc--)
4968                 __free_page(spages[rc-1]);
4969         return -ENOMEM;
4970 }
4971
4972 struct nfs4_cached_acl {
4973         int cached;
4974         size_t len;
4975         char data[0];
4976 };
4977
4978 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4979 {
4980         struct nfs_inode *nfsi = NFS_I(inode);
4981
4982         spin_lock(&inode->i_lock);
4983         kfree(nfsi->nfs4_acl);
4984         nfsi->nfs4_acl = acl;
4985         spin_unlock(&inode->i_lock);
4986 }
4987
4988 static void nfs4_zap_acl_attr(struct inode *inode)
4989 {
4990         nfs4_set_cached_acl(inode, NULL);
4991 }
4992
4993 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4994 {
4995         struct nfs_inode *nfsi = NFS_I(inode);
4996         struct nfs4_cached_acl *acl;
4997         int ret = -ENOENT;
4998
4999         spin_lock(&inode->i_lock);
5000         acl = nfsi->nfs4_acl;
5001         if (acl == NULL)
5002                 goto out;
5003         if (buf == NULL) /* user is just asking for length */
5004                 goto out_len;
5005         if (acl->cached == 0)
5006                 goto out;
5007         ret = -ERANGE; /* see getxattr(2) man page */
5008         if (acl->len > buflen)
5009                 goto out;
5010         memcpy(buf, acl->data, acl->len);
5011 out_len:
5012         ret = acl->len;
5013 out:
5014         spin_unlock(&inode->i_lock);
5015         return ret;
5016 }
5017
5018 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5019 {
5020         struct nfs4_cached_acl *acl;
5021         size_t buflen = sizeof(*acl) + acl_len;
5022
5023         if (buflen <= PAGE_SIZE) {
5024                 acl = kmalloc(buflen, GFP_KERNEL);
5025                 if (acl == NULL)
5026                         goto out;
5027                 acl->cached = 1;
5028                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5029         } else {
5030                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5031                 if (acl == NULL)
5032                         goto out;
5033                 acl->cached = 0;
5034         }
5035         acl->len = acl_len;
5036 out:
5037         nfs4_set_cached_acl(inode, acl);
5038 }
5039
5040 /*
5041  * The getxattr API returns the required buffer length when called with a
5042  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5043  * the required buf.  On a NULL buf, we send a page of data to the server
5044  * guessing that the ACL request can be serviced by a page. If so, we cache
5045  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5046  * the cache. If not so, we throw away the page, and cache the required
5047  * length. The next getxattr call will then produce another round trip to
5048  * the server, this time with the input buf of the required size.
5049  */
5050 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5051 {
5052         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
5053         struct nfs_getaclargs args = {
5054                 .fh = NFS_FH(inode),
5055                 .acl_pages = pages,
5056                 .acl_len = buflen,
5057         };
5058         struct nfs_getaclres res = {
5059                 .acl_len = buflen,
5060         };
5061         struct rpc_message msg = {
5062                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5063                 .rpc_argp = &args,
5064                 .rpc_resp = &res,
5065         };
5066         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5067         int ret = -ENOMEM, i;
5068
5069         /* As long as we're doing a round trip to the server anyway,
5070          * let's be prepared for a page of acl data. */
5071         if (npages == 0)
5072                 npages = 1;
5073         if (npages > ARRAY_SIZE(pages))
5074                 return -ERANGE;
5075
5076         for (i = 0; i < npages; i++) {
5077                 pages[i] = alloc_page(GFP_KERNEL);
5078                 if (!pages[i])
5079                         goto out_free;
5080         }
5081
5082         /* for decoding across pages */
5083         res.acl_scratch = alloc_page(GFP_KERNEL);
5084         if (!res.acl_scratch)
5085                 goto out_free;
5086
5087         args.acl_len = npages * PAGE_SIZE;
5088
5089         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5090                 __func__, buf, buflen, npages, args.acl_len);
5091         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5092                              &msg, &args.seq_args, &res.seq_res, 0);
5093         if (ret)
5094                 goto out_free;
5095
5096         /* Handle the case where the passed-in buffer is too short */
5097         if (res.acl_flags & NFS4_ACL_TRUNC) {
5098                 /* Did the user only issue a request for the acl length? */
5099                 if (buf == NULL)
5100                         goto out_ok;
5101                 ret = -ERANGE;
5102                 goto out_free;
5103         }
5104         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5105         if (buf) {
5106                 if (res.acl_len > buflen) {
5107                         ret = -ERANGE;
5108                         goto out_free;
5109                 }
5110                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5111         }
5112 out_ok:
5113         ret = res.acl_len;
5114 out_free:
5115         for (i = 0; i < npages; i++)
5116                 if (pages[i])
5117                         __free_page(pages[i]);
5118         if (res.acl_scratch)
5119                 __free_page(res.acl_scratch);
5120         return ret;
5121 }
5122
5123 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5124 {
5125         struct nfs4_exception exception = { };
5126         ssize_t ret;
5127         do {
5128                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5129                 trace_nfs4_get_acl(inode, ret);
5130                 if (ret >= 0)
5131                         break;
5132                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5133         } while (exception.retry);
5134         return ret;
5135 }
5136
5137 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5138 {
5139         struct nfs_server *server = NFS_SERVER(inode);
5140         int ret;
5141
5142         if (!nfs4_server_supports_acls(server))
5143                 return -EOPNOTSUPP;
5144         ret = nfs_revalidate_inode(server, inode);
5145         if (ret < 0)
5146                 return ret;
5147         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5148                 nfs_zap_acl_cache(inode);
5149         ret = nfs4_read_cached_acl(inode, buf, buflen);
5150         if (ret != -ENOENT)
5151                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5152                  * but no cached acl data, just the acl length */
5153                 return ret;
5154         return nfs4_get_acl_uncached(inode, buf, buflen);
5155 }
5156
5157 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5158 {
5159         struct nfs_server *server = NFS_SERVER(inode);
5160         struct page *pages[NFS4ACL_MAXPAGES];
5161         struct nfs_setaclargs arg = {
5162                 .fh             = NFS_FH(inode),
5163                 .acl_pages      = pages,
5164                 .acl_len        = buflen,
5165         };
5166         struct nfs_setaclres res;
5167         struct rpc_message msg = {
5168                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5169                 .rpc_argp       = &arg,
5170                 .rpc_resp       = &res,
5171         };
5172         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5173         int ret, i;
5174
5175         if (!nfs4_server_supports_acls(server))
5176                 return -EOPNOTSUPP;
5177         if (npages > ARRAY_SIZE(pages))
5178                 return -ERANGE;
5179         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5180         if (i < 0)
5181                 return i;
5182         nfs4_inode_return_delegation(inode);
5183         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5184
5185         /*
5186          * Free each page after tx, so the only ref left is
5187          * held by the network stack
5188          */
5189         for (; i > 0; i--)
5190                 put_page(pages[i-1]);
5191
5192         /*
5193          * Acl update can result in inode attribute update.
5194          * so mark the attribute cache invalid.
5195          */
5196         spin_lock(&inode->i_lock);
5197         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5198         spin_unlock(&inode->i_lock);
5199         nfs_access_zap_cache(inode);
5200         nfs_zap_acl_cache(inode);
5201         return ret;
5202 }
5203
5204 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5205 {
5206         struct nfs4_exception exception = { };
5207         int err;
5208         do {
5209                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5210                 trace_nfs4_set_acl(inode, err);
5211                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5212                                 &exception);
5213         } while (exception.retry);
5214         return err;
5215 }
5216
5217 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5218 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5219                                         size_t buflen)
5220 {
5221         struct nfs_server *server = NFS_SERVER(inode);
5222         struct nfs_fattr fattr;
5223         struct nfs4_label label = {0, 0, buflen, buf};
5224
5225         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5226         struct nfs4_getattr_arg arg = {
5227                 .fh             = NFS_FH(inode),
5228                 .bitmask        = bitmask,
5229         };
5230         struct nfs4_getattr_res res = {
5231                 .fattr          = &fattr,
5232                 .label          = &label,
5233                 .server         = server,
5234         };
5235         struct rpc_message msg = {
5236                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5237                 .rpc_argp       = &arg,
5238                 .rpc_resp       = &res,
5239         };
5240         int ret;
5241
5242         nfs_fattr_init(&fattr);
5243
5244         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5245         if (ret)
5246                 return ret;
5247         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5248                 return -ENOENT;
5249         if (buflen < label.len)
5250                 return -ERANGE;
5251         return 0;
5252 }
5253
5254 static int nfs4_get_security_label(struct inode *inode, void *buf,
5255                                         size_t buflen)
5256 {
5257         struct nfs4_exception exception = { };
5258         int err;
5259
5260         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5261                 return -EOPNOTSUPP;
5262
5263         do {
5264                 err = _nfs4_get_security_label(inode, buf, buflen);
5265                 trace_nfs4_get_security_label(inode, err);
5266                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5267                                 &exception);
5268         } while (exception.retry);
5269         return err;
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
5278         struct iattr sattr = {0};
5279         struct nfs_server *server = NFS_SERVER(inode);
5280         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5281         struct nfs_setattrargs arg = {
5282                 .fh             = NFS_FH(inode),
5283                 .iap            = &sattr,
5284                 .server         = server,
5285                 .bitmask        = bitmask,
5286                 .label          = ilabel,
5287         };
5288         struct nfs_setattrres res = {
5289                 .fattr          = fattr,
5290                 .label          = olabel,
5291                 .server         = server,
5292         };
5293         struct rpc_message msg = {
5294                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5295                 .rpc_argp       = &arg,
5296                 .rpc_resp       = &res,
5297         };
5298         int status;
5299
5300         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5301
5302         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5303         if (status)
5304                 dprintk("%s failed: %d\n", __func__, status);
5305
5306         return status;
5307 }
5308
5309 static int nfs4_do_set_security_label(struct inode *inode,
5310                 struct nfs4_label *ilabel,
5311                 struct nfs_fattr *fattr,
5312                 struct nfs4_label *olabel)
5313 {
5314         struct nfs4_exception exception = { };
5315         int err;
5316
5317         do {
5318                 err = _nfs4_do_set_security_label(inode, ilabel,
5319                                 fattr, olabel);
5320                 trace_nfs4_set_security_label(inode, err);
5321                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5322                                 &exception);
5323         } while (exception.retry);
5324         return err;
5325 }
5326
5327 static int
5328 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5329 {
5330         struct nfs4_label ilabel, *olabel = NULL;
5331         struct nfs_fattr fattr;
5332         struct rpc_cred *cred;
5333         int status;
5334
5335         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5336                 return -EOPNOTSUPP;
5337
5338         nfs_fattr_init(&fattr);
5339
5340         ilabel.pi = 0;
5341         ilabel.lfs = 0;
5342         ilabel.label = (char *)buf;
5343         ilabel.len = buflen;
5344
5345         cred = rpc_lookup_cred();
5346         if (IS_ERR(cred))
5347                 return PTR_ERR(cred);
5348
5349         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5350         if (IS_ERR(olabel)) {
5351                 status = -PTR_ERR(olabel);
5352                 goto out;
5353         }
5354
5355         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5356         if (status == 0)
5357                 nfs_setsecurity(inode, &fattr, olabel);
5358
5359         nfs4_label_free(olabel);
5360 out:
5361         put_rpccred(cred);
5362         return status;
5363 }
5364 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5365
5366
5367 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5368                                     nfs4_verifier *bootverf)
5369 {
5370         __be32 verf[2];
5371
5372         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5373                 /* An impossible timestamp guarantees this value
5374                  * will never match a generated boot time. */
5375                 verf[0] = cpu_to_be32(U32_MAX);
5376                 verf[1] = cpu_to_be32(U32_MAX);
5377         } else {
5378                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5379                 u64 ns = ktime_to_ns(nn->boot_time);
5380
5381                 verf[0] = cpu_to_be32(ns >> 32);
5382                 verf[1] = cpu_to_be32(ns);
5383         }
5384         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5385 }
5386
5387 static int
5388 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5389 {
5390         size_t len;
5391         char *str;
5392
5393         if (clp->cl_owner_id != NULL)
5394                 return 0;
5395
5396         rcu_read_lock();
5397         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5398                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5399                 1 +
5400                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5401                 1;
5402         rcu_read_unlock();
5403
5404         if (len > NFS4_OPAQUE_LIMIT + 1)
5405                 return -EINVAL;
5406
5407         /*
5408          * Since this string is allocated at mount time, and held until the
5409          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5410          * about a memory-reclaim deadlock.
5411          */
5412         str = kmalloc(len, GFP_KERNEL);
5413         if (!str)
5414                 return -ENOMEM;
5415
5416         rcu_read_lock();
5417         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5418                         clp->cl_ipaddr,
5419                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5420                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5421         rcu_read_unlock();
5422
5423         clp->cl_owner_id = str;
5424         return 0;
5425 }
5426
5427 static int
5428 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5429 {
5430         size_t len;
5431         char *str;
5432
5433         len = 10 + 10 + 1 + 10 + 1 +
5434                 strlen(nfs4_client_id_uniquifier) + 1 +
5435                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5436
5437         if (len > NFS4_OPAQUE_LIMIT + 1)
5438                 return -EINVAL;
5439
5440         /*
5441          * Since this string is allocated at mount time, and held until the
5442          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5443          * about a memory-reclaim deadlock.
5444          */
5445         str = kmalloc(len, GFP_KERNEL);
5446         if (!str)
5447                 return -ENOMEM;
5448
5449         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5450                         clp->rpc_ops->version, clp->cl_minorversion,
5451                         nfs4_client_id_uniquifier,
5452                         clp->cl_rpcclient->cl_nodename);
5453         clp->cl_owner_id = str;
5454         return 0;
5455 }
5456
5457 static int
5458 nfs4_init_uniform_client_string(struct nfs_client *clp)
5459 {
5460         size_t len;
5461         char *str;
5462
5463         if (clp->cl_owner_id != NULL)
5464                 return 0;
5465
5466         if (nfs4_client_id_uniquifier[0] != '\0')
5467                 return nfs4_init_uniquifier_client_string(clp);
5468
5469         len = 10 + 10 + 1 + 10 + 1 +
5470                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5471
5472         if (len > NFS4_OPAQUE_LIMIT + 1)
5473                 return -EINVAL;
5474
5475         /*
5476          * Since this string is allocated at mount time, and held until the
5477          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5478          * about a memory-reclaim deadlock.
5479          */
5480         str = kmalloc(len, GFP_KERNEL);
5481         if (!str)
5482                 return -ENOMEM;
5483
5484         scnprintf(str, len, "Linux NFSv%u.%u %s",
5485                         clp->rpc_ops->version, clp->cl_minorversion,
5486                         clp->cl_rpcclient->cl_nodename);
5487         clp->cl_owner_id = str;
5488         return 0;
5489 }
5490
5491 /*
5492  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5493  * services.  Advertise one based on the address family of the
5494  * clientaddr.
5495  */
5496 static unsigned int
5497 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5498 {
5499         if (strchr(clp->cl_ipaddr, ':') != NULL)
5500                 return scnprintf(buf, len, "tcp6");
5501         else
5502                 return scnprintf(buf, len, "tcp");
5503 }
5504
5505 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5506 {
5507         struct nfs4_setclientid *sc = calldata;
5508
5509         if (task->tk_status == 0)
5510                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5511 }
5512
5513 static const struct rpc_call_ops nfs4_setclientid_ops = {
5514         .rpc_call_done = nfs4_setclientid_done,
5515 };
5516
5517 /**
5518  * nfs4_proc_setclientid - Negotiate client ID
5519  * @clp: state data structure
5520  * @program: RPC program for NFSv4 callback service
5521  * @port: IP port number for NFS4 callback service
5522  * @cred: RPC credential to use for this call
5523  * @res: where to place the result
5524  *
5525  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5526  */
5527 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5528                 unsigned short port, struct rpc_cred *cred,
5529                 struct nfs4_setclientid_res *res)
5530 {
5531         nfs4_verifier sc_verifier;
5532         struct nfs4_setclientid setclientid = {
5533                 .sc_verifier = &sc_verifier,
5534                 .sc_prog = program,
5535                 .sc_clnt = clp,
5536         };
5537         struct rpc_message msg = {
5538                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5539                 .rpc_argp = &setclientid,
5540                 .rpc_resp = res,
5541                 .rpc_cred = cred,
5542         };
5543         struct rpc_task *task;
5544         struct rpc_task_setup task_setup_data = {
5545                 .rpc_client = clp->cl_rpcclient,
5546                 .rpc_message = &msg,
5547                 .callback_ops = &nfs4_setclientid_ops,
5548                 .callback_data = &setclientid,
5549                 .flags = RPC_TASK_TIMEOUT,
5550         };
5551         int status;
5552
5553         /* nfs_client_id4 */
5554         nfs4_init_boot_verifier(clp, &sc_verifier);
5555
5556         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5557                 status = nfs4_init_uniform_client_string(clp);
5558         else
5559                 status = nfs4_init_nonuniform_client_string(clp);
5560
5561         if (status)
5562                 goto out;
5563
5564         /* cb_client4 */
5565         setclientid.sc_netid_len =
5566                                 nfs4_init_callback_netid(clp,
5567                                                 setclientid.sc_netid,
5568                                                 sizeof(setclientid.sc_netid));
5569         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5570                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5571                                 clp->cl_ipaddr, port >> 8, port & 255);
5572
5573         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5574                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5575                 clp->cl_owner_id);
5576         task = rpc_run_task(&task_setup_data);
5577         if (IS_ERR(task)) {
5578                 status = PTR_ERR(task);
5579                 goto out;
5580         }
5581         status = task->tk_status;
5582         if (setclientid.sc_cred) {
5583                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5584                 put_rpccred(setclientid.sc_cred);
5585         }
5586         rpc_put_task(task);
5587 out:
5588         trace_nfs4_setclientid(clp, status);
5589         dprintk("NFS reply setclientid: %d\n", status);
5590         return status;
5591 }
5592
5593 /**
5594  * nfs4_proc_setclientid_confirm - Confirm client ID
5595  * @clp: state data structure
5596  * @res: result of a previous SETCLIENTID
5597  * @cred: RPC credential to use for this call
5598  *
5599  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5600  */
5601 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5602                 struct nfs4_setclientid_res *arg,
5603                 struct rpc_cred *cred)
5604 {
5605         struct rpc_message msg = {
5606                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5607                 .rpc_argp = arg,
5608                 .rpc_cred = cred,
5609         };
5610         int status;
5611
5612         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5613                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5614                 clp->cl_clientid);
5615         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5616         trace_nfs4_setclientid_confirm(clp, status);
5617         dprintk("NFS reply setclientid_confirm: %d\n", status);
5618         return status;
5619 }
5620
5621 struct nfs4_delegreturndata {
5622         struct nfs4_delegreturnargs args;
5623         struct nfs4_delegreturnres res;
5624         struct nfs_fh fh;
5625         nfs4_stateid stateid;
5626         unsigned long timestamp;
5627         struct {
5628                 struct nfs4_layoutreturn_args arg;
5629                 struct nfs4_layoutreturn_res res;
5630                 struct nfs4_xdr_opaque_data ld_private;
5631                 u32 roc_barrier;
5632                 bool roc;
5633         } lr;
5634         struct nfs_fattr fattr;
5635         int rpc_status;
5636         struct inode *inode;
5637 };
5638
5639 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5640 {
5641         struct nfs4_delegreturndata *data = calldata;
5642
5643         if (!nfs4_sequence_done(task, &data->res.seq_res))
5644                 return;
5645
5646         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5647
5648         /* Handle Layoutreturn errors */
5649         if (data->args.lr_args && task->tk_status != 0) {
5650                 switch(data->res.lr_ret) {
5651                 default:
5652                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5653                         break;
5654                 case 0:
5655                         data->args.lr_args = NULL;
5656                         data->res.lr_res = NULL;
5657                         break;
5658                 case -NFS4ERR_ADMIN_REVOKED:
5659                 case -NFS4ERR_DELEG_REVOKED:
5660                 case -NFS4ERR_EXPIRED:
5661                 case -NFS4ERR_BAD_STATEID:
5662                 case -NFS4ERR_OLD_STATEID:
5663                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5664                 case -NFS4ERR_WRONG_CRED:
5665                         data->args.lr_args = NULL;
5666                         data->res.lr_res = NULL;
5667                         data->res.lr_ret = 0;
5668                         rpc_restart_call_prepare(task);
5669                         return;
5670                 }
5671         }
5672
5673         switch (task->tk_status) {
5674         case 0:
5675                 renew_lease(data->res.server, data->timestamp);
5676                 break;
5677         case -NFS4ERR_ADMIN_REVOKED:
5678         case -NFS4ERR_DELEG_REVOKED:
5679         case -NFS4ERR_EXPIRED:
5680                 nfs4_free_revoked_stateid(data->res.server,
5681                                 data->args.stateid,
5682                                 task->tk_msg.rpc_cred);
5683         case -NFS4ERR_BAD_STATEID:
5684         case -NFS4ERR_OLD_STATEID:
5685         case -NFS4ERR_STALE_STATEID:
5686                 task->tk_status = 0;
5687                 break;
5688         default:
5689                 if (nfs4_async_handle_error(task, data->res.server,
5690                                             NULL, NULL) == -EAGAIN) {
5691                         rpc_restart_call_prepare(task);
5692                         return;
5693                 }
5694         }
5695         data->rpc_status = task->tk_status;
5696 }
5697
5698 static void nfs4_delegreturn_release(void *calldata)
5699 {
5700         struct nfs4_delegreturndata *data = calldata;
5701         struct inode *inode = data->inode;
5702
5703         if (inode) {
5704                 if (data->lr.roc)
5705                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5706                                         data->res.lr_ret);
5707                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5708                 nfs_iput_and_deactive(inode);
5709         }
5710         kfree(calldata);
5711 }
5712
5713 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5714 {
5715         struct nfs4_delegreturndata *d_data;
5716
5717         d_data = (struct nfs4_delegreturndata *)data;
5718
5719         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5720                 return;
5721
5722         nfs4_setup_sequence(d_data->res.server,
5723                         &d_data->args.seq_args,
5724                         &d_data->res.seq_res,
5725                         task);
5726 }
5727
5728 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5729         .rpc_call_prepare = nfs4_delegreturn_prepare,
5730         .rpc_call_done = nfs4_delegreturn_done,
5731         .rpc_release = nfs4_delegreturn_release,
5732 };
5733
5734 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5735 {
5736         struct nfs4_delegreturndata *data;
5737         struct nfs_server *server = NFS_SERVER(inode);
5738         struct rpc_task *task;
5739         struct rpc_message msg = {
5740                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5741                 .rpc_cred = cred,
5742         };
5743         struct rpc_task_setup task_setup_data = {
5744                 .rpc_client = server->client,
5745                 .rpc_message = &msg,
5746                 .callback_ops = &nfs4_delegreturn_ops,
5747                 .flags = RPC_TASK_ASYNC,
5748         };
5749         int status = 0;
5750
5751         data = kzalloc(sizeof(*data), GFP_NOFS);
5752         if (data == NULL)
5753                 return -ENOMEM;
5754         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5755
5756         nfs4_state_protect(server->nfs_client,
5757                         NFS_SP4_MACH_CRED_CLEANUP,
5758                         &task_setup_data.rpc_client, &msg);
5759
5760         data->args.fhandle = &data->fh;
5761         data->args.stateid = &data->stateid;
5762         data->args.bitmask = server->cache_consistency_bitmask;
5763         nfs_copy_fh(&data->fh, NFS_FH(inode));
5764         nfs4_stateid_copy(&data->stateid, stateid);
5765         data->res.fattr = &data->fattr;
5766         data->res.server = server;
5767         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5768         data->lr.arg.ld_private = &data->lr.ld_private;
5769         nfs_fattr_init(data->res.fattr);
5770         data->timestamp = jiffies;
5771         data->rpc_status = 0;
5772         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5773         data->inode = nfs_igrab_and_active(inode);
5774         if (data->inode) {
5775                 if (data->lr.roc) {
5776                         data->args.lr_args = &data->lr.arg;
5777                         data->res.lr_res = &data->lr.res;
5778                 }
5779         } else if (data->lr.roc) {
5780                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5781                 data->lr.roc = false;
5782         }
5783
5784         task_setup_data.callback_data = data;
5785         msg.rpc_argp = &data->args;
5786         msg.rpc_resp = &data->res;
5787         task = rpc_run_task(&task_setup_data);
5788         if (IS_ERR(task))
5789                 return PTR_ERR(task);
5790         if (!issync)
5791                 goto out;
5792         status = nfs4_wait_for_completion_rpc_task(task);
5793         if (status != 0)
5794                 goto out;
5795         status = data->rpc_status;
5796 out:
5797         rpc_put_task(task);
5798         return status;
5799 }
5800
5801 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5802 {
5803         struct nfs_server *server = NFS_SERVER(inode);
5804         struct nfs4_exception exception = { };
5805         int err;
5806         do {
5807                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5808                 trace_nfs4_delegreturn(inode, stateid, err);
5809                 switch (err) {
5810                         case -NFS4ERR_STALE_STATEID:
5811                         case -NFS4ERR_EXPIRED:
5812                         case 0:
5813                                 return 0;
5814                 }
5815                 err = nfs4_handle_exception(server, err, &exception);
5816         } while (exception.retry);
5817         return err;
5818 }
5819
5820 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5821 {
5822         struct inode *inode = state->inode;
5823         struct nfs_server *server = NFS_SERVER(inode);
5824         struct nfs_client *clp = server->nfs_client;
5825         struct nfs_lockt_args arg = {
5826                 .fh = NFS_FH(inode),
5827                 .fl = request,
5828         };
5829         struct nfs_lockt_res res = {
5830                 .denied = request,
5831         };
5832         struct rpc_message msg = {
5833                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5834                 .rpc_argp       = &arg,
5835                 .rpc_resp       = &res,
5836                 .rpc_cred       = state->owner->so_cred,
5837         };
5838         struct nfs4_lock_state *lsp;
5839         int status;
5840
5841         arg.lock_owner.clientid = clp->cl_clientid;
5842         status = nfs4_set_lock_state(state, request);
5843         if (status != 0)
5844                 goto out;
5845         lsp = request->fl_u.nfs4_fl.owner;
5846         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5847         arg.lock_owner.s_dev = server->s_dev;
5848         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5849         switch (status) {
5850                 case 0:
5851                         request->fl_type = F_UNLCK;
5852                         break;
5853                 case -NFS4ERR_DENIED:
5854                         status = 0;
5855         }
5856         request->fl_ops->fl_release_private(request);
5857         request->fl_ops = NULL;
5858 out:
5859         return status;
5860 }
5861
5862 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5863 {
5864         struct nfs4_exception exception = { };
5865         int err;
5866
5867         do {
5868                 err = _nfs4_proc_getlk(state, cmd, request);
5869                 trace_nfs4_get_lock(request, state, cmd, err);
5870                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5871                                 &exception);
5872         } while (exception.retry);
5873         return err;
5874 }
5875
5876 struct nfs4_unlockdata {
5877         struct nfs_locku_args arg;
5878         struct nfs_locku_res res;
5879         struct nfs4_lock_state *lsp;
5880         struct nfs_open_context *ctx;
5881         struct file_lock fl;
5882         struct nfs_server *server;
5883         unsigned long timestamp;
5884 };
5885
5886 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5887                 struct nfs_open_context *ctx,
5888                 struct nfs4_lock_state *lsp,
5889                 struct nfs_seqid *seqid)
5890 {
5891         struct nfs4_unlockdata *p;
5892         struct inode *inode = lsp->ls_state->inode;
5893
5894         p = kzalloc(sizeof(*p), GFP_NOFS);
5895         if (p == NULL)
5896                 return NULL;
5897         p->arg.fh = NFS_FH(inode);
5898         p->arg.fl = &p->fl;
5899         p->arg.seqid = seqid;
5900         p->res.seqid = seqid;
5901         p->lsp = lsp;
5902         atomic_inc(&lsp->ls_count);
5903         /* Ensure we don't close file until we're done freeing locks! */
5904         p->ctx = get_nfs_open_context(ctx);
5905         memcpy(&p->fl, fl, sizeof(p->fl));
5906         p->server = NFS_SERVER(inode);
5907         return p;
5908 }
5909
5910 static void nfs4_locku_release_calldata(void *data)
5911 {
5912         struct nfs4_unlockdata *calldata = data;
5913         nfs_free_seqid(calldata->arg.seqid);
5914         nfs4_put_lock_state(calldata->lsp);
5915         put_nfs_open_context(calldata->ctx);
5916         kfree(calldata);
5917 }
5918
5919 static void nfs4_locku_done(struct rpc_task *task, void *data)
5920 {
5921         struct nfs4_unlockdata *calldata = data;
5922
5923         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5924                 return;
5925         switch (task->tk_status) {
5926                 case 0:
5927                         renew_lease(calldata->server, calldata->timestamp);
5928                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5929                         if (nfs4_update_lock_stateid(calldata->lsp,
5930                                         &calldata->res.stateid))
5931                                 break;
5932                 case -NFS4ERR_ADMIN_REVOKED:
5933                 case -NFS4ERR_EXPIRED:
5934                         nfs4_free_revoked_stateid(calldata->server,
5935                                         &calldata->arg.stateid,
5936                                         task->tk_msg.rpc_cred);
5937                 case -NFS4ERR_BAD_STATEID:
5938                 case -NFS4ERR_OLD_STATEID:
5939                 case -NFS4ERR_STALE_STATEID:
5940                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5941                                                 &calldata->lsp->ls_stateid))
5942                                 rpc_restart_call_prepare(task);
5943                         break;
5944                 default:
5945                         if (nfs4_async_handle_error(task, calldata->server,
5946                                                     NULL, NULL) == -EAGAIN)
5947                                 rpc_restart_call_prepare(task);
5948         }
5949         nfs_release_seqid(calldata->arg.seqid);
5950 }
5951
5952 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5953 {
5954         struct nfs4_unlockdata *calldata = data;
5955
5956         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5957                 goto out_wait;
5958         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5959         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5960                 /* Note: exit _without_ running nfs4_locku_done */
5961                 goto out_no_action;
5962         }
5963         calldata->timestamp = jiffies;
5964         if (nfs4_setup_sequence(calldata->server,
5965                                 &calldata->arg.seq_args,
5966                                 &calldata->res.seq_res,
5967                                 task) != 0)
5968                 nfs_release_seqid(calldata->arg.seqid);
5969         return;
5970 out_no_action:
5971         task->tk_action = NULL;
5972 out_wait:
5973         nfs4_sequence_done(task, &calldata->res.seq_res);
5974 }
5975
5976 static const struct rpc_call_ops nfs4_locku_ops = {
5977         .rpc_call_prepare = nfs4_locku_prepare,
5978         .rpc_call_done = nfs4_locku_done,
5979         .rpc_release = nfs4_locku_release_calldata,
5980 };
5981
5982 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5983                 struct nfs_open_context *ctx,
5984                 struct nfs4_lock_state *lsp,
5985                 struct nfs_seqid *seqid)
5986 {
5987         struct nfs4_unlockdata *data;
5988         struct rpc_message msg = {
5989                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5990                 .rpc_cred = ctx->cred,
5991         };
5992         struct rpc_task_setup task_setup_data = {
5993                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5994                 .rpc_message = &msg,
5995                 .callback_ops = &nfs4_locku_ops,
5996                 .workqueue = nfsiod_workqueue,
5997                 .flags = RPC_TASK_ASYNC,
5998         };
5999
6000         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6001                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6002
6003         /* Ensure this is an unlock - when canceling a lock, the
6004          * canceled lock is passed in, and it won't be an unlock.
6005          */
6006         fl->fl_type = F_UNLCK;
6007
6008         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6009         if (data == NULL) {
6010                 nfs_free_seqid(seqid);
6011                 return ERR_PTR(-ENOMEM);
6012         }
6013
6014         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6015         msg.rpc_argp = &data->arg;
6016         msg.rpc_resp = &data->res;
6017         task_setup_data.callback_data = data;
6018         return rpc_run_task(&task_setup_data);
6019 }
6020
6021 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6022 {
6023         struct inode *inode = state->inode;
6024         struct nfs4_state_owner *sp = state->owner;
6025         struct nfs_inode *nfsi = NFS_I(inode);
6026         struct nfs_seqid *seqid;
6027         struct nfs4_lock_state *lsp;
6028         struct rpc_task *task;
6029         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6030         int status = 0;
6031         unsigned char fl_flags = request->fl_flags;
6032
6033         status = nfs4_set_lock_state(state, request);
6034         /* Unlock _before_ we do the RPC call */
6035         request->fl_flags |= FL_EXISTS;
6036         /* Exclude nfs_delegation_claim_locks() */
6037         mutex_lock(&sp->so_delegreturn_mutex);
6038         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6039         down_read(&nfsi->rwsem);
6040         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6041                 up_read(&nfsi->rwsem);
6042                 mutex_unlock(&sp->so_delegreturn_mutex);
6043                 goto out;
6044         }
6045         up_read(&nfsi->rwsem);
6046         mutex_unlock(&sp->so_delegreturn_mutex);
6047         if (status != 0)
6048                 goto out;
6049         /* Is this a delegated lock? */
6050         lsp = request->fl_u.nfs4_fl.owner;
6051         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6052                 goto out;
6053         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6054         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6055         status = -ENOMEM;
6056         if (IS_ERR(seqid))
6057                 goto out;
6058         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6059         status = PTR_ERR(task);
6060         if (IS_ERR(task))
6061                 goto out;
6062         status = nfs4_wait_for_completion_rpc_task(task);
6063         rpc_put_task(task);
6064 out:
6065         request->fl_flags = fl_flags;
6066         trace_nfs4_unlock(request, state, F_SETLK, status);
6067         return status;
6068 }
6069
6070 struct nfs4_lockdata {
6071         struct nfs_lock_args arg;
6072         struct nfs_lock_res res;
6073         struct nfs4_lock_state *lsp;
6074         struct nfs_open_context *ctx;
6075         struct file_lock fl;
6076         unsigned long timestamp;
6077         int rpc_status;
6078         int cancelled;
6079         struct nfs_server *server;
6080 };
6081
6082 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6083                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6084                 gfp_t gfp_mask)
6085 {
6086         struct nfs4_lockdata *p;
6087         struct inode *inode = lsp->ls_state->inode;
6088         struct nfs_server *server = NFS_SERVER(inode);
6089         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6090
6091         p = kzalloc(sizeof(*p), gfp_mask);
6092         if (p == NULL)
6093                 return NULL;
6094
6095         p->arg.fh = NFS_FH(inode);
6096         p->arg.fl = &p->fl;
6097         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6098         if (IS_ERR(p->arg.open_seqid))
6099                 goto out_free;
6100         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6101         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6102         if (IS_ERR(p->arg.lock_seqid))
6103                 goto out_free_seqid;
6104         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6105         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6106         p->arg.lock_owner.s_dev = server->s_dev;
6107         p->res.lock_seqid = p->arg.lock_seqid;
6108         p->lsp = lsp;
6109         p->server = server;
6110         atomic_inc(&lsp->ls_count);
6111         p->ctx = get_nfs_open_context(ctx);
6112         get_file(fl->fl_file);
6113         memcpy(&p->fl, fl, sizeof(p->fl));
6114         return p;
6115 out_free_seqid:
6116         nfs_free_seqid(p->arg.open_seqid);
6117 out_free:
6118         kfree(p);
6119         return NULL;
6120 }
6121
6122 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6123 {
6124         struct nfs4_lockdata *data = calldata;
6125         struct nfs4_state *state = data->lsp->ls_state;
6126
6127         dprintk("%s: begin!\n", __func__);
6128         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6129                 goto out_wait;
6130         /* Do we need to do an open_to_lock_owner? */
6131         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6132                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6133                         goto out_release_lock_seqid;
6134                 }
6135                 nfs4_stateid_copy(&data->arg.open_stateid,
6136                                 &state->open_stateid);
6137                 data->arg.new_lock_owner = 1;
6138                 data->res.open_seqid = data->arg.open_seqid;
6139         } else {
6140                 data->arg.new_lock_owner = 0;
6141                 nfs4_stateid_copy(&data->arg.lock_stateid,
6142                                 &data->lsp->ls_stateid);
6143         }
6144         if (!nfs4_valid_open_stateid(state)) {
6145                 data->rpc_status = -EBADF;
6146                 task->tk_action = NULL;
6147                 goto out_release_open_seqid;
6148         }
6149         data->timestamp = jiffies;
6150         if (nfs4_setup_sequence(data->server,
6151                                 &data->arg.seq_args,
6152                                 &data->res.seq_res,
6153                                 task) == 0)
6154                 return;
6155 out_release_open_seqid:
6156         nfs_release_seqid(data->arg.open_seqid);
6157 out_release_lock_seqid:
6158         nfs_release_seqid(data->arg.lock_seqid);
6159 out_wait:
6160         nfs4_sequence_done(task, &data->res.seq_res);
6161         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6162 }
6163
6164 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6165 {
6166         struct nfs4_lockdata *data = calldata;
6167         struct nfs4_lock_state *lsp = data->lsp;
6168
6169         dprintk("%s: begin!\n", __func__);
6170
6171         if (!nfs4_sequence_done(task, &data->res.seq_res))
6172                 return;
6173
6174         data->rpc_status = task->tk_status;
6175         switch (task->tk_status) {
6176         case 0:
6177                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6178                                 data->timestamp);
6179                 if (data->arg.new_lock) {
6180                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6181                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6182                                 rpc_restart_call_prepare(task);
6183                                 break;
6184                         }
6185                 }
6186                 if (data->arg.new_lock_owner != 0) {
6187                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6188                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6189                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6190                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6191                         rpc_restart_call_prepare(task);
6192                 break;
6193         case -NFS4ERR_BAD_STATEID:
6194         case -NFS4ERR_OLD_STATEID:
6195         case -NFS4ERR_STALE_STATEID:
6196         case -NFS4ERR_EXPIRED:
6197                 if (data->arg.new_lock_owner != 0) {
6198                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6199                                                 &lsp->ls_state->open_stateid))
6200                                 rpc_restart_call_prepare(task);
6201                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6202                                                 &lsp->ls_stateid))
6203                                 rpc_restart_call_prepare(task);
6204         }
6205         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6206 }
6207
6208 static void nfs4_lock_release(void *calldata)
6209 {
6210         struct nfs4_lockdata *data = calldata;
6211
6212         dprintk("%s: begin!\n", __func__);
6213         nfs_free_seqid(data->arg.open_seqid);
6214         if (data->cancelled != 0) {
6215                 struct rpc_task *task;
6216                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6217                                 data->arg.lock_seqid);
6218                 if (!IS_ERR(task))
6219                         rpc_put_task_async(task);
6220                 dprintk("%s: cancelling lock!\n", __func__);
6221         } else
6222                 nfs_free_seqid(data->arg.lock_seqid);
6223         nfs4_put_lock_state(data->lsp);
6224         put_nfs_open_context(data->ctx);
6225         fput(data->fl.fl_file);
6226         kfree(data);
6227         dprintk("%s: done!\n", __func__);
6228 }
6229
6230 static const struct rpc_call_ops nfs4_lock_ops = {
6231         .rpc_call_prepare = nfs4_lock_prepare,
6232         .rpc_call_done = nfs4_lock_done,
6233         .rpc_release = nfs4_lock_release,
6234 };
6235
6236 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6237 {
6238         switch (error) {
6239         case -NFS4ERR_ADMIN_REVOKED:
6240         case -NFS4ERR_EXPIRED:
6241         case -NFS4ERR_BAD_STATEID:
6242                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6243                 if (new_lock_owner != 0 ||
6244                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6245                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6246                 break;
6247         case -NFS4ERR_STALE_STATEID:
6248                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6249                 nfs4_schedule_lease_recovery(server->nfs_client);
6250         };
6251 }
6252
6253 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6254 {
6255         struct nfs4_lockdata *data;
6256         struct rpc_task *task;
6257         struct rpc_message msg = {
6258                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6259                 .rpc_cred = state->owner->so_cred,
6260         };
6261         struct rpc_task_setup task_setup_data = {
6262                 .rpc_client = NFS_CLIENT(state->inode),
6263                 .rpc_message = &msg,
6264                 .callback_ops = &nfs4_lock_ops,
6265                 .workqueue = nfsiod_workqueue,
6266                 .flags = RPC_TASK_ASYNC,
6267         };
6268         int ret;
6269
6270         dprintk("%s: begin!\n", __func__);
6271         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6272                         fl->fl_u.nfs4_fl.owner,
6273                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6274         if (data == NULL)
6275                 return -ENOMEM;
6276         if (IS_SETLKW(cmd))
6277                 data->arg.block = 1;
6278         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6279         msg.rpc_argp = &data->arg;
6280         msg.rpc_resp = &data->res;
6281         task_setup_data.callback_data = data;
6282         if (recovery_type > NFS_LOCK_NEW) {
6283                 if (recovery_type == NFS_LOCK_RECLAIM)
6284                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6285                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6286         } else
6287                 data->arg.new_lock = 1;
6288         task = rpc_run_task(&task_setup_data);
6289         if (IS_ERR(task))
6290                 return PTR_ERR(task);
6291         ret = nfs4_wait_for_completion_rpc_task(task);
6292         if (ret == 0) {
6293                 ret = data->rpc_status;
6294                 if (ret)
6295                         nfs4_handle_setlk_error(data->server, data->lsp,
6296                                         data->arg.new_lock_owner, ret);
6297         } else
6298                 data->cancelled = 1;
6299         rpc_put_task(task);
6300         dprintk("%s: done, ret = %d!\n", __func__, ret);
6301         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6302         return ret;
6303 }
6304
6305 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6306 {
6307         struct nfs_server *server = NFS_SERVER(state->inode);
6308         struct nfs4_exception exception = {
6309                 .inode = state->inode,
6310         };
6311         int err;
6312
6313         do {
6314                 /* Cache the lock if possible... */
6315                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6316                         return 0;
6317                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6318                 if (err != -NFS4ERR_DELAY)
6319                         break;
6320                 nfs4_handle_exception(server, err, &exception);
6321         } while (exception.retry);
6322         return err;
6323 }
6324
6325 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6326 {
6327         struct nfs_server *server = NFS_SERVER(state->inode);
6328         struct nfs4_exception exception = {
6329                 .inode = state->inode,
6330         };
6331         int err;
6332
6333         err = nfs4_set_lock_state(state, request);
6334         if (err != 0)
6335                 return err;
6336         if (!recover_lost_locks) {
6337                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6338                 return 0;
6339         }
6340         do {
6341                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6342                         return 0;
6343                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6344                 switch (err) {
6345                 default:
6346                         goto out;
6347                 case -NFS4ERR_GRACE:
6348                 case -NFS4ERR_DELAY:
6349                         nfs4_handle_exception(server, err, &exception);
6350                         err = 0;
6351                 }
6352         } while (exception.retry);
6353 out:
6354         return err;
6355 }
6356
6357 #if defined(CONFIG_NFS_V4_1)
6358 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6359 {
6360         struct nfs4_lock_state *lsp;
6361         int status;
6362
6363         status = nfs4_set_lock_state(state, request);
6364         if (status != 0)
6365                 return status;
6366         lsp = request->fl_u.nfs4_fl.owner;
6367         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6368             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6369                 return 0;
6370         status = nfs4_lock_expired(state, request);
6371         return status;
6372 }
6373 #endif
6374
6375 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6376 {
6377         struct nfs_inode *nfsi = NFS_I(state->inode);
6378         struct nfs4_state_owner *sp = state->owner;
6379         unsigned char fl_flags = request->fl_flags;
6380         int status;
6381
6382         request->fl_flags |= FL_ACCESS;
6383         status = locks_lock_inode_wait(state->inode, request);
6384         if (status < 0)
6385                 goto out;
6386         mutex_lock(&sp->so_delegreturn_mutex);
6387         down_read(&nfsi->rwsem);
6388         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6389                 /* Yes: cache locks! */
6390                 /* ...but avoid races with delegation recall... */
6391                 request->fl_flags = fl_flags & ~FL_SLEEP;
6392                 status = locks_lock_inode_wait(state->inode, request);
6393                 up_read(&nfsi->rwsem);
6394                 mutex_unlock(&sp->so_delegreturn_mutex);
6395                 goto out;
6396         }
6397         up_read(&nfsi->rwsem);
6398         mutex_unlock(&sp->so_delegreturn_mutex);
6399         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6400 out:
6401         request->fl_flags = fl_flags;
6402         return status;
6403 }
6404
6405 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6406 {
6407         struct nfs4_exception exception = {
6408                 .state = state,
6409                 .inode = state->inode,
6410         };
6411         int err;
6412
6413         do {
6414                 err = _nfs4_proc_setlk(state, cmd, request);
6415                 if (err == -NFS4ERR_DENIED)
6416                         err = -EAGAIN;
6417                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6418                                 err, &exception);
6419         } while (exception.retry);
6420         return err;
6421 }
6422
6423 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6424 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6425
6426 static int
6427 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6428                         struct file_lock *request)
6429 {
6430         int             status = -ERESTARTSYS;
6431         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6432
6433         while(!signalled()) {
6434                 status = nfs4_proc_setlk(state, cmd, request);
6435                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6436                         break;
6437                 freezable_schedule_timeout_interruptible(timeout);
6438                 timeout *= 2;
6439                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6440                 status = -ERESTARTSYS;
6441         }
6442         return status;
6443 }
6444
6445 #ifdef CONFIG_NFS_V4_1
6446 struct nfs4_lock_waiter {
6447         struct task_struct      *task;
6448         struct inode            *inode;
6449         struct nfs_lowner       *owner;
6450         bool                    notified;
6451 };
6452
6453 static int
6454 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6455 {
6456         int ret;
6457         struct cb_notify_lock_args *cbnl = key;
6458         struct nfs4_lock_waiter *waiter = wait->private;
6459         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6460                                 *wowner = waiter->owner;
6461
6462         /* Only wake if the callback was for the same owner */
6463         if (lowner->clientid != wowner->clientid ||
6464             lowner->id != wowner->id             ||
6465             lowner->s_dev != wowner->s_dev)
6466                 return 0;
6467
6468         /* Make sure it's for the right inode */
6469         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6470                 return 0;
6471
6472         waiter->notified = true;
6473
6474         /* override "private" so we can use default_wake_function */
6475         wait->private = waiter->task;
6476         ret = autoremove_wake_function(wait, mode, flags, key);
6477         wait->private = waiter;
6478         return ret;
6479 }
6480
6481 static int
6482 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6483 {
6484         int status = -ERESTARTSYS;
6485         unsigned long flags;
6486         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6487         struct nfs_server *server = NFS_SERVER(state->inode);
6488         struct nfs_client *clp = server->nfs_client;
6489         wait_queue_head_t *q = &clp->cl_lock_waitq;
6490         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6491                                     .id = lsp->ls_seqid.owner_id,
6492                                     .s_dev = server->s_dev };
6493         struct nfs4_lock_waiter waiter = { .task  = current,
6494                                            .inode = state->inode,
6495                                            .owner = &owner,
6496                                            .notified = false };
6497         wait_queue_t wait;
6498
6499         /* Don't bother with waitqueue if we don't expect a callback */
6500         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6501                 return nfs4_retry_setlk_simple(state, cmd, request);
6502
6503         init_wait(&wait);
6504         wait.private = &waiter;
6505         wait.func = nfs4_wake_lock_waiter;
6506         add_wait_queue(q, &wait);
6507
6508         while(!signalled()) {
6509                 status = nfs4_proc_setlk(state, cmd, request);
6510                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6511                         break;
6512
6513                 status = -ERESTARTSYS;
6514                 spin_lock_irqsave(&q->lock, flags);
6515                 if (waiter.notified) {
6516                         spin_unlock_irqrestore(&q->lock, flags);
6517                         continue;
6518                 }
6519                 set_current_state(TASK_INTERRUPTIBLE);
6520                 spin_unlock_irqrestore(&q->lock, flags);
6521
6522                 freezable_schedule_timeout_interruptible(NFS4_LOCK_MAXTIMEOUT);
6523         }
6524
6525         finish_wait(q, &wait);
6526         return status;
6527 }
6528 #else /* !CONFIG_NFS_V4_1 */
6529 static inline int
6530 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6531 {
6532         return nfs4_retry_setlk_simple(state, cmd, request);
6533 }
6534 #endif
6535
6536 static int
6537 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6538 {
6539         struct nfs_open_context *ctx;
6540         struct nfs4_state *state;
6541         int status;
6542
6543         /* verify open state */
6544         ctx = nfs_file_open_context(filp);
6545         state = ctx->state;
6546
6547         if (request->fl_start < 0 || request->fl_end < 0)
6548                 return -EINVAL;
6549
6550         if (IS_GETLK(cmd)) {
6551                 if (state != NULL)
6552                         return nfs4_proc_getlk(state, F_GETLK, request);
6553                 return 0;
6554         }
6555
6556         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6557                 return -EINVAL;
6558
6559         if (request->fl_type == F_UNLCK) {
6560                 if (state != NULL)
6561                         return nfs4_proc_unlck(state, cmd, request);
6562                 return 0;
6563         }
6564
6565         if (state == NULL)
6566                 return -ENOLCK;
6567
6568         if ((request->fl_flags & FL_POSIX) &&
6569             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6570                 return -ENOLCK;
6571
6572         /*
6573          * Don't rely on the VFS having checked the file open mode,
6574          * since it won't do this for flock() locks.
6575          */
6576         switch (request->fl_type) {
6577         case F_RDLCK:
6578                 if (!(filp->f_mode & FMODE_READ))
6579                         return -EBADF;
6580                 break;
6581         case F_WRLCK:
6582                 if (!(filp->f_mode & FMODE_WRITE))
6583                         return -EBADF;
6584         }
6585
6586         status = nfs4_set_lock_state(state, request);
6587         if (status != 0)
6588                 return status;
6589
6590         return nfs4_retry_setlk(state, cmd, request);
6591 }
6592
6593 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6594 {
6595         struct nfs_server *server = NFS_SERVER(state->inode);
6596         int err;
6597
6598         err = nfs4_set_lock_state(state, fl);
6599         if (err != 0)
6600                 return err;
6601         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6602         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6603 }
6604
6605 struct nfs_release_lockowner_data {
6606         struct nfs4_lock_state *lsp;
6607         struct nfs_server *server;
6608         struct nfs_release_lockowner_args args;
6609         struct nfs_release_lockowner_res res;
6610         unsigned long timestamp;
6611 };
6612
6613 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6614 {
6615         struct nfs_release_lockowner_data *data = calldata;
6616         struct nfs_server *server = data->server;
6617         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6618                              &data->args.seq_args, &data->res.seq_res, task);
6619         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6620         data->timestamp = jiffies;
6621 }
6622
6623 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6624 {
6625         struct nfs_release_lockowner_data *data = calldata;
6626         struct nfs_server *server = data->server;
6627
6628         nfs40_sequence_done(task, &data->res.seq_res);
6629
6630         switch (task->tk_status) {
6631         case 0:
6632                 renew_lease(server, data->timestamp);
6633                 break;
6634         case -NFS4ERR_STALE_CLIENTID:
6635         case -NFS4ERR_EXPIRED:
6636                 nfs4_schedule_lease_recovery(server->nfs_client);
6637                 break;
6638         case -NFS4ERR_LEASE_MOVED:
6639         case -NFS4ERR_DELAY:
6640                 if (nfs4_async_handle_error(task, server,
6641                                             NULL, NULL) == -EAGAIN)
6642                         rpc_restart_call_prepare(task);
6643         }
6644 }
6645
6646 static void nfs4_release_lockowner_release(void *calldata)
6647 {
6648         struct nfs_release_lockowner_data *data = calldata;
6649         nfs4_free_lock_state(data->server, data->lsp);
6650         kfree(calldata);
6651 }
6652
6653 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6654         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6655         .rpc_call_done = nfs4_release_lockowner_done,
6656         .rpc_release = nfs4_release_lockowner_release,
6657 };
6658
6659 static void
6660 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6661 {
6662         struct nfs_release_lockowner_data *data;
6663         struct rpc_message msg = {
6664                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6665         };
6666
6667         if (server->nfs_client->cl_mvops->minor_version != 0)
6668                 return;
6669
6670         data = kmalloc(sizeof(*data), GFP_NOFS);
6671         if (!data)
6672                 return;
6673         data->lsp = lsp;
6674         data->server = server;
6675         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6676         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6677         data->args.lock_owner.s_dev = server->s_dev;
6678
6679         msg.rpc_argp = &data->args;
6680         msg.rpc_resp = &data->res;
6681         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6682         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6683 }
6684
6685 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6686
6687 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6688                                    struct dentry *unused, struct inode *inode,
6689                                    const char *key, const void *buf,
6690                                    size_t buflen, int flags)
6691 {
6692         return nfs4_proc_set_acl(inode, buf, buflen);
6693 }
6694
6695 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6696                                    struct dentry *unused, struct inode *inode,
6697                                    const char *key, void *buf, size_t buflen)
6698 {
6699         return nfs4_proc_get_acl(inode, buf, buflen);
6700 }
6701
6702 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6703 {
6704         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6705 }
6706
6707 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6708
6709 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6710                                      struct dentry *unused, struct inode *inode,
6711                                      const char *key, const void *buf,
6712                                      size_t buflen, int flags)
6713 {
6714         if (security_ismaclabel(key))
6715                 return nfs4_set_security_label(inode, buf, buflen);
6716
6717         return -EOPNOTSUPP;
6718 }
6719
6720 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6721                                      struct dentry *unused, struct inode *inode,
6722                                      const char *key, void *buf, size_t buflen)
6723 {
6724         if (security_ismaclabel(key))
6725                 return nfs4_get_security_label(inode, buf, buflen);
6726         return -EOPNOTSUPP;
6727 }
6728
6729 static ssize_t
6730 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6731 {
6732         int len = 0;
6733
6734         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6735                 len = security_inode_listsecurity(inode, list, list_len);
6736                 if (list_len && len > list_len)
6737                         return -ERANGE;
6738         }
6739         return len;
6740 }
6741
6742 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6743         .prefix = XATTR_SECURITY_PREFIX,
6744         .get    = nfs4_xattr_get_nfs4_label,
6745         .set    = nfs4_xattr_set_nfs4_label,
6746 };
6747
6748 #else
6749
6750 static ssize_t
6751 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6752 {
6753         return 0;
6754 }
6755
6756 #endif
6757
6758 /*
6759  * nfs_fhget will use either the mounted_on_fileid or the fileid
6760  */
6761 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6762 {
6763         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6764                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6765               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6766               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6767                 return;
6768
6769         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6770                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6771         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6772         fattr->nlink = 2;
6773 }
6774
6775 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6776                                    const struct qstr *name,
6777                                    struct nfs4_fs_locations *fs_locations,
6778                                    struct page *page)
6779 {
6780         struct nfs_server *server = NFS_SERVER(dir);
6781         u32 bitmask[3] = {
6782                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6783         };
6784         struct nfs4_fs_locations_arg args = {
6785                 .dir_fh = NFS_FH(dir),
6786                 .name = name,
6787                 .page = page,
6788                 .bitmask = bitmask,
6789         };
6790         struct nfs4_fs_locations_res res = {
6791                 .fs_locations = fs_locations,
6792         };
6793         struct rpc_message msg = {
6794                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6795                 .rpc_argp = &args,
6796                 .rpc_resp = &res,
6797         };
6798         int status;
6799
6800         dprintk("%s: start\n", __func__);
6801
6802         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6803          * is not supported */
6804         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6805                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6806         else
6807                 bitmask[0] |= FATTR4_WORD0_FILEID;
6808
6809         nfs_fattr_init(&fs_locations->fattr);
6810         fs_locations->server = server;
6811         fs_locations->nlocations = 0;
6812         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6813         dprintk("%s: returned status = %d\n", __func__, status);
6814         return status;
6815 }
6816
6817 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6818                            const struct qstr *name,
6819                            struct nfs4_fs_locations *fs_locations,
6820                            struct page *page)
6821 {
6822         struct nfs4_exception exception = { };
6823         int err;
6824         do {
6825                 err = _nfs4_proc_fs_locations(client, dir, name,
6826                                 fs_locations, page);
6827                 trace_nfs4_get_fs_locations(dir, name, err);
6828                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6829                                 &exception);
6830         } while (exception.retry);
6831         return err;
6832 }
6833
6834 /*
6835  * This operation also signals the server that this client is
6836  * performing migration recovery.  The server can stop returning
6837  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6838  * appended to this compound to identify the client ID which is
6839  * performing recovery.
6840  */
6841 static int _nfs40_proc_get_locations(struct inode *inode,
6842                                      struct nfs4_fs_locations *locations,
6843                                      struct page *page, struct rpc_cred *cred)
6844 {
6845         struct nfs_server *server = NFS_SERVER(inode);
6846         struct rpc_clnt *clnt = server->client;
6847         u32 bitmask[2] = {
6848                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6849         };
6850         struct nfs4_fs_locations_arg args = {
6851                 .clientid       = server->nfs_client->cl_clientid,
6852                 .fh             = NFS_FH(inode),
6853                 .page           = page,
6854                 .bitmask        = bitmask,
6855                 .migration      = 1,            /* skip LOOKUP */
6856                 .renew          = 1,            /* append RENEW */
6857         };
6858         struct nfs4_fs_locations_res res = {
6859                 .fs_locations   = locations,
6860                 .migration      = 1,
6861                 .renew          = 1,
6862         };
6863         struct rpc_message msg = {
6864                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6865                 .rpc_argp       = &args,
6866                 .rpc_resp       = &res,
6867                 .rpc_cred       = cred,
6868         };
6869         unsigned long now = jiffies;
6870         int status;
6871
6872         nfs_fattr_init(&locations->fattr);
6873         locations->server = server;
6874         locations->nlocations = 0;
6875
6876         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6877         nfs4_set_sequence_privileged(&args.seq_args);
6878         status = nfs4_call_sync_sequence(clnt, server, &msg,
6879                                         &args.seq_args, &res.seq_res);
6880         if (status)
6881                 return status;
6882
6883         renew_lease(server, now);
6884         return 0;
6885 }
6886
6887 #ifdef CONFIG_NFS_V4_1
6888
6889 /*
6890  * This operation also signals the server that this client is
6891  * performing migration recovery.  The server can stop asserting
6892  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6893  * performing this operation is identified in the SEQUENCE
6894  * operation in this compound.
6895  *
6896  * When the client supports GETATTR(fs_locations_info), it can
6897  * be plumbed in here.
6898  */
6899 static int _nfs41_proc_get_locations(struct inode *inode,
6900                                      struct nfs4_fs_locations *locations,
6901                                      struct page *page, struct rpc_cred *cred)
6902 {
6903         struct nfs_server *server = NFS_SERVER(inode);
6904         struct rpc_clnt *clnt = server->client;
6905         u32 bitmask[2] = {
6906                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6907         };
6908         struct nfs4_fs_locations_arg args = {
6909                 .fh             = NFS_FH(inode),
6910                 .page           = page,
6911                 .bitmask        = bitmask,
6912                 .migration      = 1,            /* skip LOOKUP */
6913         };
6914         struct nfs4_fs_locations_res res = {
6915                 .fs_locations   = locations,
6916                 .migration      = 1,
6917         };
6918         struct rpc_message msg = {
6919                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6920                 .rpc_argp       = &args,
6921                 .rpc_resp       = &res,
6922                 .rpc_cred       = cred,
6923         };
6924         int status;
6925
6926         nfs_fattr_init(&locations->fattr);
6927         locations->server = server;
6928         locations->nlocations = 0;
6929
6930         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6931         nfs4_set_sequence_privileged(&args.seq_args);
6932         status = nfs4_call_sync_sequence(clnt, server, &msg,
6933                                         &args.seq_args, &res.seq_res);
6934         if (status == NFS4_OK &&
6935             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6936                 status = -NFS4ERR_LEASE_MOVED;
6937         return status;
6938 }
6939
6940 #endif  /* CONFIG_NFS_V4_1 */
6941
6942 /**
6943  * nfs4_proc_get_locations - discover locations for a migrated FSID
6944  * @inode: inode on FSID that is migrating
6945  * @locations: result of query
6946  * @page: buffer
6947  * @cred: credential to use for this operation
6948  *
6949  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6950  * operation failed, or a negative errno if a local error occurred.
6951  *
6952  * On success, "locations" is filled in, but if the server has
6953  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6954  * asserted.
6955  *
6956  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6957  * from this client that require migration recovery.
6958  */
6959 int nfs4_proc_get_locations(struct inode *inode,
6960                             struct nfs4_fs_locations *locations,
6961                             struct page *page, struct rpc_cred *cred)
6962 {
6963         struct nfs_server *server = NFS_SERVER(inode);
6964         struct nfs_client *clp = server->nfs_client;
6965         const struct nfs4_mig_recovery_ops *ops =
6966                                         clp->cl_mvops->mig_recovery_ops;
6967         struct nfs4_exception exception = { };
6968         int status;
6969
6970         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6971                 (unsigned long long)server->fsid.major,
6972                 (unsigned long long)server->fsid.minor,
6973                 clp->cl_hostname);
6974         nfs_display_fhandle(NFS_FH(inode), __func__);
6975
6976         do {
6977                 status = ops->get_locations(inode, locations, page, cred);
6978                 if (status != -NFS4ERR_DELAY)
6979                         break;
6980                 nfs4_handle_exception(server, status, &exception);
6981         } while (exception.retry);
6982         return status;
6983 }
6984
6985 /*
6986  * This operation also signals the server that this client is
6987  * performing "lease moved" recovery.  The server can stop
6988  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6989  * is appended to this compound to identify the client ID which is
6990  * performing recovery.
6991  */
6992 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6993 {
6994         struct nfs_server *server = NFS_SERVER(inode);
6995         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6996         struct rpc_clnt *clnt = server->client;
6997         struct nfs4_fsid_present_arg args = {
6998                 .fh             = NFS_FH(inode),
6999                 .clientid       = clp->cl_clientid,
7000                 .renew          = 1,            /* append RENEW */
7001         };
7002         struct nfs4_fsid_present_res res = {
7003                 .renew          = 1,
7004         };
7005         struct rpc_message msg = {
7006                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7007                 .rpc_argp       = &args,
7008                 .rpc_resp       = &res,
7009                 .rpc_cred       = cred,
7010         };
7011         unsigned long now = jiffies;
7012         int status;
7013
7014         res.fh = nfs_alloc_fhandle();
7015         if (res.fh == NULL)
7016                 return -ENOMEM;
7017
7018         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7019         nfs4_set_sequence_privileged(&args.seq_args);
7020         status = nfs4_call_sync_sequence(clnt, server, &msg,
7021                                                 &args.seq_args, &res.seq_res);
7022         nfs_free_fhandle(res.fh);
7023         if (status)
7024                 return status;
7025
7026         do_renew_lease(clp, now);
7027         return 0;
7028 }
7029
7030 #ifdef CONFIG_NFS_V4_1
7031
7032 /*
7033  * This operation also signals the server that this client is
7034  * performing "lease moved" recovery.  The server can stop asserting
7035  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7036  * this operation is identified in the SEQUENCE operation in this
7037  * compound.
7038  */
7039 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7040 {
7041         struct nfs_server *server = NFS_SERVER(inode);
7042         struct rpc_clnt *clnt = server->client;
7043         struct nfs4_fsid_present_arg args = {
7044                 .fh             = NFS_FH(inode),
7045         };
7046         struct nfs4_fsid_present_res res = {
7047         };
7048         struct rpc_message msg = {
7049                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7050                 .rpc_argp       = &args,
7051                 .rpc_resp       = &res,
7052                 .rpc_cred       = cred,
7053         };
7054         int status;
7055
7056         res.fh = nfs_alloc_fhandle();
7057         if (res.fh == NULL)
7058                 return -ENOMEM;
7059
7060         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7061         nfs4_set_sequence_privileged(&args.seq_args);
7062         status = nfs4_call_sync_sequence(clnt, server, &msg,
7063                                                 &args.seq_args, &res.seq_res);
7064         nfs_free_fhandle(res.fh);
7065         if (status == NFS4_OK &&
7066             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7067                 status = -NFS4ERR_LEASE_MOVED;
7068         return status;
7069 }
7070
7071 #endif  /* CONFIG_NFS_V4_1 */
7072
7073 /**
7074  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7075  * @inode: inode on FSID to check
7076  * @cred: credential to use for this operation
7077  *
7078  * Server indicates whether the FSID is present, moved, or not
7079  * recognized.  This operation is necessary to clear a LEASE_MOVED
7080  * condition for this client ID.
7081  *
7082  * Returns NFS4_OK if the FSID is present on this server,
7083  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7084  *  NFS4ERR code if some error occurred on the server, or a
7085  *  negative errno if a local failure occurred.
7086  */
7087 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7088 {
7089         struct nfs_server *server = NFS_SERVER(inode);
7090         struct nfs_client *clp = server->nfs_client;
7091         const struct nfs4_mig_recovery_ops *ops =
7092                                         clp->cl_mvops->mig_recovery_ops;
7093         struct nfs4_exception exception = { };
7094         int status;
7095
7096         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7097                 (unsigned long long)server->fsid.major,
7098                 (unsigned long long)server->fsid.minor,
7099                 clp->cl_hostname);
7100         nfs_display_fhandle(NFS_FH(inode), __func__);
7101
7102         do {
7103                 status = ops->fsid_present(inode, cred);
7104                 if (status != -NFS4ERR_DELAY)
7105                         break;
7106                 nfs4_handle_exception(server, status, &exception);
7107         } while (exception.retry);
7108         return status;
7109 }
7110
7111 /**
7112  * If 'use_integrity' is true and the state managment nfs_client
7113  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7114  * and the machine credential as per RFC3530bis and RFC5661 Security
7115  * Considerations sections. Otherwise, just use the user cred with the
7116  * filesystem's rpc_client.
7117  */
7118 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7119 {
7120         int status;
7121         struct nfs4_secinfo_arg args = {
7122                 .dir_fh = NFS_FH(dir),
7123                 .name   = name,
7124         };
7125         struct nfs4_secinfo_res res = {
7126                 .flavors     = flavors,
7127         };
7128         struct rpc_message msg = {
7129                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7130                 .rpc_argp = &args,
7131                 .rpc_resp = &res,
7132         };
7133         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7134         struct rpc_cred *cred = NULL;
7135
7136         if (use_integrity) {
7137                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7138                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7139                 msg.rpc_cred = cred;
7140         }
7141
7142         dprintk("NFS call  secinfo %s\n", name->name);
7143
7144         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7145                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7146
7147         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7148                                 &res.seq_res, 0);
7149         dprintk("NFS reply  secinfo: %d\n", status);
7150
7151         if (cred)
7152                 put_rpccred(cred);
7153
7154         return status;
7155 }
7156
7157 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7158                       struct nfs4_secinfo_flavors *flavors)
7159 {
7160         struct nfs4_exception exception = { };
7161         int err;
7162         do {
7163                 err = -NFS4ERR_WRONGSEC;
7164
7165                 /* try to use integrity protection with machine cred */
7166                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7167                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7168
7169                 /*
7170                  * if unable to use integrity protection, or SECINFO with
7171                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7172                  * disallowed by spec, but exists in deployed servers) use
7173                  * the current filesystem's rpc_client and the user cred.
7174                  */
7175                 if (err == -NFS4ERR_WRONGSEC)
7176                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7177
7178                 trace_nfs4_secinfo(dir, name, err);
7179                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7180                                 &exception);
7181         } while (exception.retry);
7182         return err;
7183 }
7184
7185 #ifdef CONFIG_NFS_V4_1
7186 /*
7187  * Check the exchange flags returned by the server for invalid flags, having
7188  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7189  * DS flags set.
7190  */
7191 static int nfs4_check_cl_exchange_flags(u32 flags)
7192 {
7193         if (flags & ~EXCHGID4_FLAG_MASK_R)
7194                 goto out_inval;
7195         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7196             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7197                 goto out_inval;
7198         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7199                 goto out_inval;
7200         return NFS_OK;
7201 out_inval:
7202         return -NFS4ERR_INVAL;
7203 }
7204
7205 static bool
7206 nfs41_same_server_scope(struct nfs41_server_scope *a,
7207                         struct nfs41_server_scope *b)
7208 {
7209         if (a->server_scope_sz == b->server_scope_sz &&
7210             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7211                 return true;
7212
7213         return false;
7214 }
7215
7216 static void
7217 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7218 {
7219 }
7220
7221 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7222         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7223 };
7224
7225 /*
7226  * nfs4_proc_bind_one_conn_to_session()
7227  *
7228  * The 4.1 client currently uses the same TCP connection for the
7229  * fore and backchannel.
7230  */
7231 static
7232 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7233                 struct rpc_xprt *xprt,
7234                 struct nfs_client *clp,
7235                 struct rpc_cred *cred)
7236 {
7237         int status;
7238         struct nfs41_bind_conn_to_session_args args = {
7239                 .client = clp,
7240                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7241         };
7242         struct nfs41_bind_conn_to_session_res res;
7243         struct rpc_message msg = {
7244                 .rpc_proc =
7245                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7246                 .rpc_argp = &args,
7247                 .rpc_resp = &res,
7248                 .rpc_cred = cred,
7249         };
7250         struct rpc_task_setup task_setup_data = {
7251                 .rpc_client = clnt,
7252                 .rpc_xprt = xprt,
7253                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7254                 .rpc_message = &msg,
7255                 .flags = RPC_TASK_TIMEOUT,
7256         };
7257         struct rpc_task *task;
7258
7259         dprintk("--> %s\n", __func__);
7260
7261         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7262         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7263                 args.dir = NFS4_CDFC4_FORE;
7264
7265         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7266         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7267                 args.dir = NFS4_CDFC4_FORE;
7268
7269         task = rpc_run_task(&task_setup_data);
7270         if (!IS_ERR(task)) {
7271                 status = task->tk_status;
7272                 rpc_put_task(task);
7273         } else
7274                 status = PTR_ERR(task);
7275         trace_nfs4_bind_conn_to_session(clp, status);
7276         if (status == 0) {
7277                 if (memcmp(res.sessionid.data,
7278                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7279                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7280                         status = -EIO;
7281                         goto out;
7282                 }
7283                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7284                         dprintk("NFS: %s: Unexpected direction from server\n",
7285                                 __func__);
7286                         status = -EIO;
7287                         goto out;
7288                 }
7289                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7290                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7291                                 __func__);
7292                         status = -EIO;
7293                         goto out;
7294                 }
7295         }
7296 out:
7297         dprintk("<-- %s status= %d\n", __func__, status);
7298         return status;
7299 }
7300
7301 struct rpc_bind_conn_calldata {
7302         struct nfs_client *clp;
7303         struct rpc_cred *cred;
7304 };
7305
7306 static int
7307 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7308                 struct rpc_xprt *xprt,
7309                 void *calldata)
7310 {
7311         struct rpc_bind_conn_calldata *p = calldata;
7312
7313         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7314 }
7315
7316 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7317 {
7318         struct rpc_bind_conn_calldata data = {
7319                 .clp = clp,
7320                 .cred = cred,
7321         };
7322         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7323                         nfs4_proc_bind_conn_to_session_callback, &data);
7324 }
7325
7326 /*
7327  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7328  * and operations we'd like to see to enable certain features in the allow map
7329  */
7330 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7331         .how = SP4_MACH_CRED,
7332         .enforce.u.words = {
7333                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7334                       1 << (OP_EXCHANGE_ID - 32) |
7335                       1 << (OP_CREATE_SESSION - 32) |
7336                       1 << (OP_DESTROY_SESSION - 32) |
7337                       1 << (OP_DESTROY_CLIENTID - 32)
7338         },
7339         .allow.u.words = {
7340                 [0] = 1 << (OP_CLOSE) |
7341                       1 << (OP_OPEN_DOWNGRADE) |
7342                       1 << (OP_LOCKU) |
7343                       1 << (OP_DELEGRETURN) |
7344                       1 << (OP_COMMIT),
7345                 [1] = 1 << (OP_SECINFO - 32) |
7346                       1 << (OP_SECINFO_NO_NAME - 32) |
7347                       1 << (OP_LAYOUTRETURN - 32) |
7348                       1 << (OP_TEST_STATEID - 32) |
7349                       1 << (OP_FREE_STATEID - 32) |
7350                       1 << (OP_WRITE - 32)
7351         }
7352 };
7353
7354 /*
7355  * Select the state protection mode for client `clp' given the server results
7356  * from exchange_id in `sp'.
7357  *
7358  * Returns 0 on success, negative errno otherwise.
7359  */
7360 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7361                                  struct nfs41_state_protection *sp)
7362 {
7363         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7364                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7365                       1 << (OP_EXCHANGE_ID - 32) |
7366                       1 << (OP_CREATE_SESSION - 32) |
7367                       1 << (OP_DESTROY_SESSION - 32) |
7368                       1 << (OP_DESTROY_CLIENTID - 32)
7369         };
7370         unsigned int i;
7371
7372         if (sp->how == SP4_MACH_CRED) {
7373                 /* Print state protect result */
7374                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7375                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7376                         if (test_bit(i, sp->enforce.u.longs))
7377                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7378                         if (test_bit(i, sp->allow.u.longs))
7379                                 dfprintk(MOUNT, "  allow op %d\n", i);
7380                 }
7381
7382                 /* make sure nothing is on enforce list that isn't supported */
7383                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7384                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7385                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7386                                 return -EINVAL;
7387                         }
7388                 }
7389
7390                 /*
7391                  * Minimal mode - state operations are allowed to use machine
7392                  * credential.  Note this already happens by default, so the
7393                  * client doesn't have to do anything more than the negotiation.
7394                  *
7395                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7396                  *       we're already using the machine cred for exchange_id
7397                  *       and will never use a different cred.
7398                  */
7399                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7400                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7401                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7402                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7403                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7404                         dfprintk(MOUNT, "  minimal mode enabled\n");
7405                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7406                 } else {
7407                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7408                         return -EINVAL;
7409                 }
7410
7411                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7412                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7413                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7414                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7415                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7416                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7417                 }
7418
7419                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7420                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7421                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7422                                 &clp->cl_sp4_flags);
7423                 }
7424
7425                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7426                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7427                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7428                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7429                 }
7430
7431                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7432                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7433                         dfprintk(MOUNT, "  stateid mode enabled\n");
7434                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7435                 }
7436
7437                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7438                         dfprintk(MOUNT, "  write mode enabled\n");
7439                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7440                 }
7441
7442                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7443                         dfprintk(MOUNT, "  commit mode enabled\n");
7444                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7445                 }
7446         }
7447
7448         return 0;
7449 }
7450
7451 struct nfs41_exchange_id_data {
7452         struct nfs41_exchange_id_res res;
7453         struct nfs41_exchange_id_args args;
7454         struct rpc_xprt *xprt;
7455         int rpc_status;
7456 };
7457
7458 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7459 {
7460         struct nfs41_exchange_id_data *cdata =
7461                                         (struct nfs41_exchange_id_data *)data;
7462         struct nfs_client *clp = cdata->args.client;
7463         int status = task->tk_status;
7464
7465         trace_nfs4_exchange_id(clp, status);
7466
7467         if (status == 0)
7468                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7469
7470         if (cdata->xprt && status == 0) {
7471                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7472                                                       cdata->xprt);
7473                 goto out;
7474         }
7475
7476         if (status  == 0)
7477                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7478
7479         if (status == 0) {
7480                 clp->cl_clientid = cdata->res.clientid;
7481                 clp->cl_exchange_flags = cdata->res.flags;
7482                 /* Client ID is not confirmed */
7483                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
7484                         clear_bit(NFS4_SESSION_ESTABLISHED,
7485                         &clp->cl_session->session_state);
7486                         clp->cl_seqid = cdata->res.seqid;
7487                 }
7488
7489                 kfree(clp->cl_serverowner);
7490                 clp->cl_serverowner = cdata->res.server_owner;
7491                 cdata->res.server_owner = NULL;
7492
7493                 /* use the most recent implementation id */
7494                 kfree(clp->cl_implid);
7495                 clp->cl_implid = cdata->res.impl_id;
7496                 cdata->res.impl_id = NULL;
7497
7498                 if (clp->cl_serverscope != NULL &&
7499                     !nfs41_same_server_scope(clp->cl_serverscope,
7500                                         cdata->res.server_scope)) {
7501                         dprintk("%s: server_scope mismatch detected\n",
7502                                 __func__);
7503                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7504                         kfree(clp->cl_serverscope);
7505                         clp->cl_serverscope = NULL;
7506                 }
7507
7508                 if (clp->cl_serverscope == NULL) {
7509                         clp->cl_serverscope = cdata->res.server_scope;
7510                         cdata->res.server_scope = NULL;
7511                 }
7512                 /* Save the EXCHANGE_ID verifier session trunk tests */
7513                 memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
7514                        sizeof(clp->cl_confirm.data));
7515         }
7516 out:
7517         cdata->rpc_status = status;
7518         return;
7519 }
7520
7521 static void nfs4_exchange_id_release(void *data)
7522 {
7523         struct nfs41_exchange_id_data *cdata =
7524                                         (struct nfs41_exchange_id_data *)data;
7525
7526         nfs_put_client(cdata->args.client);
7527         if (cdata->xprt) {
7528                 xprt_put(cdata->xprt);
7529                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7530         }
7531         kfree(cdata->res.impl_id);
7532         kfree(cdata->res.server_scope);
7533         kfree(cdata->res.server_owner);
7534         kfree(cdata);
7535 }
7536
7537 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7538         .rpc_call_done = nfs4_exchange_id_done,
7539         .rpc_release = nfs4_exchange_id_release,
7540 };
7541
7542 /*
7543  * _nfs4_proc_exchange_id()
7544  *
7545  * Wrapper for EXCHANGE_ID operation.
7546  */
7547 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7548                         u32 sp4_how, struct rpc_xprt *xprt)
7549 {
7550         nfs4_verifier verifier;
7551         struct rpc_message msg = {
7552                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7553                 .rpc_cred = cred,
7554         };
7555         struct rpc_task_setup task_setup_data = {
7556                 .rpc_client = clp->cl_rpcclient,
7557                 .callback_ops = &nfs4_exchange_id_call_ops,
7558                 .rpc_message = &msg,
7559                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7560         };
7561         struct nfs41_exchange_id_data *calldata;
7562         struct rpc_task *task;
7563         int status = -EIO;
7564
7565         if (!atomic_inc_not_zero(&clp->cl_count))
7566                 goto out;
7567
7568         status = -ENOMEM;
7569         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7570         if (!calldata)
7571                 goto out;
7572
7573         if (!xprt)
7574                 nfs4_init_boot_verifier(clp, &verifier);
7575
7576         status = nfs4_init_uniform_client_string(clp);
7577         if (status)
7578                 goto out_calldata;
7579
7580         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7581                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7582                 clp->cl_owner_id);
7583
7584         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7585                                                 GFP_NOFS);
7586         status = -ENOMEM;
7587         if (unlikely(calldata->res.server_owner == NULL))
7588                 goto out_calldata;
7589
7590         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7591                                         GFP_NOFS);
7592         if (unlikely(calldata->res.server_scope == NULL))
7593                 goto out_server_owner;
7594
7595         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7596         if (unlikely(calldata->res.impl_id == NULL))
7597                 goto out_server_scope;
7598
7599         switch (sp4_how) {
7600         case SP4_NONE:
7601                 calldata->args.state_protect.how = SP4_NONE;
7602                 break;
7603
7604         case SP4_MACH_CRED:
7605                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7606                 break;
7607
7608         default:
7609                 /* unsupported! */
7610                 WARN_ON_ONCE(1);
7611                 status = -EINVAL;
7612                 goto out_impl_id;
7613         }
7614         if (xprt) {
7615                 calldata->xprt = xprt;
7616                 task_setup_data.rpc_xprt = xprt;
7617                 task_setup_data.flags =
7618                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7619                 calldata->args.verifier = &clp->cl_confirm;
7620         } else {
7621                 calldata->args.verifier = &verifier;
7622         }
7623         calldata->args.client = clp;
7624 #ifdef CONFIG_NFS_V4_1_MIGRATION
7625         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7626         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7627         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7628 #else
7629         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7630         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7631 #endif
7632         msg.rpc_argp = &calldata->args;
7633         msg.rpc_resp = &calldata->res;
7634         task_setup_data.callback_data = calldata;
7635
7636         task = rpc_run_task(&task_setup_data);
7637         if (IS_ERR(task)) {
7638         status = PTR_ERR(task);
7639                 goto out_impl_id;
7640         }
7641
7642         if (!xprt) {
7643                 status = rpc_wait_for_completion_task(task);
7644                 if (!status)
7645                         status = calldata->rpc_status;
7646         } else  /* session trunking test */
7647                 status = calldata->rpc_status;
7648
7649         rpc_put_task(task);
7650 out:
7651         if (clp->cl_implid != NULL)
7652                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7653                         "domain: %s, name: %s, date: %llu,%u\n",
7654                         clp->cl_implid->domain, clp->cl_implid->name,
7655                         clp->cl_implid->date.seconds,
7656                         clp->cl_implid->date.nseconds);
7657         dprintk("NFS reply exchange_id: %d\n", status);
7658         return status;
7659
7660 out_impl_id:
7661         kfree(calldata->res.impl_id);
7662 out_server_scope:
7663         kfree(calldata->res.server_scope);
7664 out_server_owner:
7665         kfree(calldata->res.server_owner);
7666 out_calldata:
7667         kfree(calldata);
7668         goto out;
7669 }
7670
7671 /*
7672  * nfs4_proc_exchange_id()
7673  *
7674  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7675  *
7676  * Since the clientid has expired, all compounds using sessions
7677  * associated with the stale clientid will be returning
7678  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7679  * be in some phase of session reset.
7680  *
7681  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7682  */
7683 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7684 {
7685         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7686         int status;
7687
7688         /* try SP4_MACH_CRED if krb5i/p */
7689         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7690             authflavor == RPC_AUTH_GSS_KRB5P) {
7691                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7692                 if (!status)
7693                         return 0;
7694         }
7695
7696         /* try SP4_NONE */
7697         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7698 }
7699
7700 /**
7701  * nfs4_test_session_trunk
7702  *
7703  * This is an add_xprt_test() test function called from
7704  * rpc_clnt_setup_test_and_add_xprt.
7705  *
7706  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7707  * and is dereferrenced in nfs4_exchange_id_release
7708  *
7709  * Upon success, add the new transport to the rpc_clnt
7710  *
7711  * @clnt: struct rpc_clnt to get new transport
7712  * @xprt: the rpc_xprt to test
7713  * @data: call data for _nfs4_proc_exchange_id.
7714  */
7715 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7716                             void *data)
7717 {
7718         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7719         u32 sp4_how;
7720
7721         dprintk("--> %s try %s\n", __func__,
7722                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7723
7724         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7725
7726         /* Test connection for session trunking. Async exchange_id call */
7727         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7728 }
7729 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7730
7731 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7732                 struct rpc_cred *cred)
7733 {
7734         struct rpc_message msg = {
7735                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7736                 .rpc_argp = clp,
7737                 .rpc_cred = cred,
7738         };
7739         int status;
7740
7741         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7742         trace_nfs4_destroy_clientid(clp, status);
7743         if (status)
7744                 dprintk("NFS: Got error %d from the server %s on "
7745                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7746         return status;
7747 }
7748
7749 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7750                 struct rpc_cred *cred)
7751 {
7752         unsigned int loop;
7753         int ret;
7754
7755         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7756                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7757                 switch (ret) {
7758                 case -NFS4ERR_DELAY:
7759                 case -NFS4ERR_CLIENTID_BUSY:
7760                         ssleep(1);
7761                         break;
7762                 default:
7763                         return ret;
7764                 }
7765         }
7766         return 0;
7767 }
7768
7769 int nfs4_destroy_clientid(struct nfs_client *clp)
7770 {
7771         struct rpc_cred *cred;
7772         int ret = 0;
7773
7774         if (clp->cl_mvops->minor_version < 1)
7775                 goto out;
7776         if (clp->cl_exchange_flags == 0)
7777                 goto out;
7778         if (clp->cl_preserve_clid)
7779                 goto out;
7780         cred = nfs4_get_clid_cred(clp);
7781         ret = nfs4_proc_destroy_clientid(clp, cred);
7782         if (cred)
7783                 put_rpccred(cred);
7784         switch (ret) {
7785         case 0:
7786         case -NFS4ERR_STALE_CLIENTID:
7787                 clp->cl_exchange_flags = 0;
7788         }
7789 out:
7790         return ret;
7791 }
7792
7793 struct nfs4_get_lease_time_data {
7794         struct nfs4_get_lease_time_args *args;
7795         struct nfs4_get_lease_time_res *res;
7796         struct nfs_client *clp;
7797 };
7798
7799 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7800                                         void *calldata)
7801 {
7802         struct nfs4_get_lease_time_data *data =
7803                         (struct nfs4_get_lease_time_data *)calldata;
7804
7805         dprintk("--> %s\n", __func__);
7806         /* just setup sequence, do not trigger session recovery
7807            since we're invoked within one */
7808         nfs41_setup_sequence(data->clp->cl_session,
7809                         &data->args->la_seq_args,
7810                         &data->res->lr_seq_res,
7811                         task);
7812         dprintk("<-- %s\n", __func__);
7813 }
7814
7815 /*
7816  * Called from nfs4_state_manager thread for session setup, so don't recover
7817  * from sequence operation or clientid errors.
7818  */
7819 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7820 {
7821         struct nfs4_get_lease_time_data *data =
7822                         (struct nfs4_get_lease_time_data *)calldata;
7823
7824         dprintk("--> %s\n", __func__);
7825         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7826                 return;
7827         switch (task->tk_status) {
7828         case -NFS4ERR_DELAY:
7829         case -NFS4ERR_GRACE:
7830                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7831                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7832                 task->tk_status = 0;
7833                 /* fall through */
7834         case -NFS4ERR_RETRY_UNCACHED_REP:
7835                 rpc_restart_call_prepare(task);
7836                 return;
7837         }
7838         dprintk("<-- %s\n", __func__);
7839 }
7840
7841 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7842         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7843         .rpc_call_done = nfs4_get_lease_time_done,
7844 };
7845
7846 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7847 {
7848         struct rpc_task *task;
7849         struct nfs4_get_lease_time_args args;
7850         struct nfs4_get_lease_time_res res = {
7851                 .lr_fsinfo = fsinfo,
7852         };
7853         struct nfs4_get_lease_time_data data = {
7854                 .args = &args,
7855                 .res = &res,
7856                 .clp = clp,
7857         };
7858         struct rpc_message msg = {
7859                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7860                 .rpc_argp = &args,
7861                 .rpc_resp = &res,
7862         };
7863         struct rpc_task_setup task_setup = {
7864                 .rpc_client = clp->cl_rpcclient,
7865                 .rpc_message = &msg,
7866                 .callback_ops = &nfs4_get_lease_time_ops,
7867                 .callback_data = &data,
7868                 .flags = RPC_TASK_TIMEOUT,
7869         };
7870         int status;
7871
7872         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7873         nfs4_set_sequence_privileged(&args.la_seq_args);
7874         dprintk("--> %s\n", __func__);
7875         task = rpc_run_task(&task_setup);
7876
7877         if (IS_ERR(task))
7878                 status = PTR_ERR(task);
7879         else {
7880                 status = task->tk_status;
7881                 rpc_put_task(task);
7882         }
7883         dprintk("<-- %s return %d\n", __func__, status);
7884
7885         return status;
7886 }
7887
7888 /*
7889  * Initialize the values to be used by the client in CREATE_SESSION
7890  * If nfs4_init_session set the fore channel request and response sizes,
7891  * use them.
7892  *
7893  * Set the back channel max_resp_sz_cached to zero to force the client to
7894  * always set csa_cachethis to FALSE because the current implementation
7895  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7896  */
7897 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7898                                     struct rpc_clnt *clnt)
7899 {
7900         unsigned int max_rqst_sz, max_resp_sz;
7901         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7902
7903         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7904         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7905
7906         /* Fore channel attributes */
7907         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7908         args->fc_attrs.max_resp_sz = max_resp_sz;
7909         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7910         args->fc_attrs.max_reqs = max_session_slots;
7911
7912         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7913                 "max_ops=%u max_reqs=%u\n",
7914                 __func__,
7915                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7916                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7917
7918         /* Back channel attributes */
7919         args->bc_attrs.max_rqst_sz = max_bc_payload;
7920         args->bc_attrs.max_resp_sz = max_bc_payload;
7921         args->bc_attrs.max_resp_sz_cached = 0;
7922         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7923         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7924
7925         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7926                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7927                 __func__,
7928                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7929                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7930                 args->bc_attrs.max_reqs);
7931 }
7932
7933 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7934                 struct nfs41_create_session_res *res)
7935 {
7936         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7937         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7938
7939         if (rcvd->max_resp_sz > sent->max_resp_sz)
7940                 return -EINVAL;
7941         /*
7942          * Our requested max_ops is the minimum we need; we're not
7943          * prepared to break up compounds into smaller pieces than that.
7944          * So, no point even trying to continue if the server won't
7945          * cooperate:
7946          */
7947         if (rcvd->max_ops < sent->max_ops)
7948                 return -EINVAL;
7949         if (rcvd->max_reqs == 0)
7950                 return -EINVAL;
7951         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7952                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7953         return 0;
7954 }
7955
7956 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7957                 struct nfs41_create_session_res *res)
7958 {
7959         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7960         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7961
7962         if (!(res->flags & SESSION4_BACK_CHAN))
7963                 goto out;
7964         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7965                 return -EINVAL;
7966         if (rcvd->max_resp_sz < sent->max_resp_sz)
7967                 return -EINVAL;
7968         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7969                 return -EINVAL;
7970         if (rcvd->max_ops > sent->max_ops)
7971                 return -EINVAL;
7972         if (rcvd->max_reqs > sent->max_reqs)
7973                 return -EINVAL;
7974 out:
7975         return 0;
7976 }
7977
7978 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7979                                      struct nfs41_create_session_res *res)
7980 {
7981         int ret;
7982
7983         ret = nfs4_verify_fore_channel_attrs(args, res);
7984         if (ret)
7985                 return ret;
7986         return nfs4_verify_back_channel_attrs(args, res);
7987 }
7988
7989 static void nfs4_update_session(struct nfs4_session *session,
7990                 struct nfs41_create_session_res *res)
7991 {
7992         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7993         /* Mark client id and session as being confirmed */
7994         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7995         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7996         session->flags = res->flags;
7997         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7998         if (res->flags & SESSION4_BACK_CHAN)
7999                 memcpy(&session->bc_attrs, &res->bc_attrs,
8000                                 sizeof(session->bc_attrs));
8001 }
8002
8003 static int _nfs4_proc_create_session(struct nfs_client *clp,
8004                 struct rpc_cred *cred)
8005 {
8006         struct nfs4_session *session = clp->cl_session;
8007         struct nfs41_create_session_args args = {
8008                 .client = clp,
8009                 .clientid = clp->cl_clientid,
8010                 .seqid = clp->cl_seqid,
8011                 .cb_program = NFS4_CALLBACK,
8012         };
8013         struct nfs41_create_session_res res;
8014
8015         struct rpc_message msg = {
8016                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8017                 .rpc_argp = &args,
8018                 .rpc_resp = &res,
8019                 .rpc_cred = cred,
8020         };
8021         int status;
8022
8023         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8024         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8025
8026         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8027         trace_nfs4_create_session(clp, status);
8028
8029         switch (status) {
8030         case -NFS4ERR_STALE_CLIENTID:
8031         case -NFS4ERR_DELAY:
8032         case -ETIMEDOUT:
8033         case -EACCES:
8034         case -EAGAIN:
8035                 goto out;
8036         };
8037
8038         clp->cl_seqid++;
8039         if (!status) {
8040                 /* Verify the session's negotiated channel_attrs values */
8041                 status = nfs4_verify_channel_attrs(&args, &res);
8042                 /* Increment the clientid slot sequence id */
8043                 if (status)
8044                         goto out;
8045                 nfs4_update_session(session, &res);
8046         }
8047 out:
8048         return status;
8049 }
8050
8051 /*
8052  * Issues a CREATE_SESSION operation to the server.
8053  * It is the responsibility of the caller to verify the session is
8054  * expired before calling this routine.
8055  */
8056 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8057 {
8058         int status;
8059         unsigned *ptr;
8060         struct nfs4_session *session = clp->cl_session;
8061
8062         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8063
8064         status = _nfs4_proc_create_session(clp, cred);
8065         if (status)
8066                 goto out;
8067
8068         /* Init or reset the session slot tables */
8069         status = nfs4_setup_session_slot_tables(session);
8070         dprintk("slot table setup returned %d\n", status);
8071         if (status)
8072                 goto out;
8073
8074         ptr = (unsigned *)&session->sess_id.data[0];
8075         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8076                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8077 out:
8078         dprintk("<-- %s\n", __func__);
8079         return status;
8080 }
8081
8082 /*
8083  * Issue the over-the-wire RPC DESTROY_SESSION.
8084  * The caller must serialize access to this routine.
8085  */
8086 int nfs4_proc_destroy_session(struct nfs4_session *session,
8087                 struct rpc_cred *cred)
8088 {
8089         struct rpc_message msg = {
8090                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8091                 .rpc_argp = session,
8092                 .rpc_cred = cred,
8093         };
8094         int status = 0;
8095
8096         dprintk("--> nfs4_proc_destroy_session\n");
8097
8098         /* session is still being setup */
8099         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8100                 return 0;
8101
8102         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8103         trace_nfs4_destroy_session(session->clp, status);
8104
8105         if (status)
8106                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8107                         "Session has been destroyed regardless...\n", status);
8108
8109         dprintk("<-- nfs4_proc_destroy_session\n");
8110         return status;
8111 }
8112
8113 /*
8114  * Renew the cl_session lease.
8115  */
8116 struct nfs4_sequence_data {
8117         struct nfs_client *clp;
8118         struct nfs4_sequence_args args;
8119         struct nfs4_sequence_res res;
8120 };
8121
8122 static void nfs41_sequence_release(void *data)
8123 {
8124         struct nfs4_sequence_data *calldata = data;
8125         struct nfs_client *clp = calldata->clp;
8126
8127         if (atomic_read(&clp->cl_count) > 1)
8128                 nfs4_schedule_state_renewal(clp);
8129         nfs_put_client(clp);
8130         kfree(calldata);
8131 }
8132
8133 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8134 {
8135         switch(task->tk_status) {
8136         case -NFS4ERR_DELAY:
8137                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8138                 return -EAGAIN;
8139         default:
8140                 nfs4_schedule_lease_recovery(clp);
8141         }
8142         return 0;
8143 }
8144
8145 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8146 {
8147         struct nfs4_sequence_data *calldata = data;
8148         struct nfs_client *clp = calldata->clp;
8149
8150         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8151                 return;
8152
8153         trace_nfs4_sequence(clp, task->tk_status);
8154         if (task->tk_status < 0) {
8155                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8156                 if (atomic_read(&clp->cl_count) == 1)
8157                         goto out;
8158
8159                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8160                         rpc_restart_call_prepare(task);
8161                         return;
8162                 }
8163         }
8164         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8165 out:
8166         dprintk("<-- %s\n", __func__);
8167 }
8168
8169 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8170 {
8171         struct nfs4_sequence_data *calldata = data;
8172         struct nfs_client *clp = calldata->clp;
8173         struct nfs4_sequence_args *args;
8174         struct nfs4_sequence_res *res;
8175
8176         args = task->tk_msg.rpc_argp;
8177         res = task->tk_msg.rpc_resp;
8178
8179         nfs41_setup_sequence(clp->cl_session, args, res, task);
8180 }
8181
8182 static const struct rpc_call_ops nfs41_sequence_ops = {
8183         .rpc_call_done = nfs41_sequence_call_done,
8184         .rpc_call_prepare = nfs41_sequence_prepare,
8185         .rpc_release = nfs41_sequence_release,
8186 };
8187
8188 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8189                 struct rpc_cred *cred,
8190                 bool is_privileged)
8191 {
8192         struct nfs4_sequence_data *calldata;
8193         struct rpc_message msg = {
8194                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8195                 .rpc_cred = cred,
8196         };
8197         struct rpc_task_setup task_setup_data = {
8198                 .rpc_client = clp->cl_rpcclient,
8199                 .rpc_message = &msg,
8200                 .callback_ops = &nfs41_sequence_ops,
8201                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8202         };
8203
8204         if (!atomic_inc_not_zero(&clp->cl_count))
8205                 return ERR_PTR(-EIO);
8206         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8207         if (calldata == NULL) {
8208                 nfs_put_client(clp);
8209                 return ERR_PTR(-ENOMEM);
8210         }
8211         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8212         if (is_privileged)
8213                 nfs4_set_sequence_privileged(&calldata->args);
8214         msg.rpc_argp = &calldata->args;
8215         msg.rpc_resp = &calldata->res;
8216         calldata->clp = clp;
8217         task_setup_data.callback_data = calldata;
8218
8219         return rpc_run_task(&task_setup_data);
8220 }
8221
8222 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8223 {
8224         struct rpc_task *task;
8225         int ret = 0;
8226
8227         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8228                 return -EAGAIN;
8229         task = _nfs41_proc_sequence(clp, cred, false);
8230         if (IS_ERR(task))
8231                 ret = PTR_ERR(task);
8232         else
8233                 rpc_put_task_async(task);
8234         dprintk("<-- %s status=%d\n", __func__, ret);
8235         return ret;
8236 }
8237
8238 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8239 {
8240         struct rpc_task *task;
8241         int ret;
8242
8243         task = _nfs41_proc_sequence(clp, cred, true);
8244         if (IS_ERR(task)) {
8245                 ret = PTR_ERR(task);
8246                 goto out;
8247         }
8248         ret = rpc_wait_for_completion_task(task);
8249         if (!ret)
8250                 ret = task->tk_status;
8251         rpc_put_task(task);
8252 out:
8253         dprintk("<-- %s status=%d\n", __func__, ret);
8254         return ret;
8255 }
8256
8257 struct nfs4_reclaim_complete_data {
8258         struct nfs_client *clp;
8259         struct nfs41_reclaim_complete_args arg;
8260         struct nfs41_reclaim_complete_res res;
8261 };
8262
8263 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8264 {
8265         struct nfs4_reclaim_complete_data *calldata = data;
8266
8267         nfs41_setup_sequence(calldata->clp->cl_session,
8268                         &calldata->arg.seq_args,
8269                         &calldata->res.seq_res,
8270                         task);
8271 }
8272
8273 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8274 {
8275         switch(task->tk_status) {
8276         case 0:
8277         case -NFS4ERR_COMPLETE_ALREADY:
8278         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8279                 break;
8280         case -NFS4ERR_DELAY:
8281                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8282                 /* fall through */
8283         case -NFS4ERR_RETRY_UNCACHED_REP:
8284                 return -EAGAIN;
8285         default:
8286                 nfs4_schedule_lease_recovery(clp);
8287         }
8288         return 0;
8289 }
8290
8291 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8292 {
8293         struct nfs4_reclaim_complete_data *calldata = data;
8294         struct nfs_client *clp = calldata->clp;
8295         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8296
8297         dprintk("--> %s\n", __func__);
8298         if (!nfs41_sequence_done(task, res))
8299                 return;
8300
8301         trace_nfs4_reclaim_complete(clp, task->tk_status);
8302         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8303                 rpc_restart_call_prepare(task);
8304                 return;
8305         }
8306         dprintk("<-- %s\n", __func__);
8307 }
8308
8309 static void nfs4_free_reclaim_complete_data(void *data)
8310 {
8311         struct nfs4_reclaim_complete_data *calldata = data;
8312
8313         kfree(calldata);
8314 }
8315
8316 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8317         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8318         .rpc_call_done = nfs4_reclaim_complete_done,
8319         .rpc_release = nfs4_free_reclaim_complete_data,
8320 };
8321
8322 /*
8323  * Issue a global reclaim complete.
8324  */
8325 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8326                 struct rpc_cred *cred)
8327 {
8328         struct nfs4_reclaim_complete_data *calldata;
8329         struct rpc_task *task;
8330         struct rpc_message msg = {
8331                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8332                 .rpc_cred = cred,
8333         };
8334         struct rpc_task_setup task_setup_data = {
8335                 .rpc_client = clp->cl_rpcclient,
8336                 .rpc_message = &msg,
8337                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8338                 .flags = RPC_TASK_ASYNC,
8339         };
8340         int status = -ENOMEM;
8341
8342         dprintk("--> %s\n", __func__);
8343         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8344         if (calldata == NULL)
8345                 goto out;
8346         calldata->clp = clp;
8347         calldata->arg.one_fs = 0;
8348
8349         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8350         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8351         msg.rpc_argp = &calldata->arg;
8352         msg.rpc_resp = &calldata->res;
8353         task_setup_data.callback_data = calldata;
8354         task = rpc_run_task(&task_setup_data);
8355         if (IS_ERR(task)) {
8356                 status = PTR_ERR(task);
8357                 goto out;
8358         }
8359         status = nfs4_wait_for_completion_rpc_task(task);
8360         if (status == 0)
8361                 status = task->tk_status;
8362         rpc_put_task(task);
8363         return 0;
8364 out:
8365         dprintk("<-- %s status=%d\n", __func__, status);
8366         return status;
8367 }
8368
8369 static void
8370 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8371 {
8372         struct nfs4_layoutget *lgp = calldata;
8373         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8374         struct nfs4_session *session = nfs4_get_session(server);
8375
8376         dprintk("--> %s\n", __func__);
8377         nfs41_setup_sequence(session, &lgp->args.seq_args,
8378                                 &lgp->res.seq_res, task);
8379         dprintk("<-- %s\n", __func__);
8380 }
8381
8382 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8383 {
8384         struct nfs4_layoutget *lgp = calldata;
8385
8386         dprintk("--> %s\n", __func__);
8387         nfs41_sequence_process(task, &lgp->res.seq_res);
8388         dprintk("<-- %s\n", __func__);
8389 }
8390
8391 static int
8392 nfs4_layoutget_handle_exception(struct rpc_task *task,
8393                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8394 {
8395         struct inode *inode = lgp->args.inode;
8396         struct nfs_server *server = NFS_SERVER(inode);
8397         struct pnfs_layout_hdr *lo;
8398         int nfs4err = task->tk_status;
8399         int err, status = 0;
8400         LIST_HEAD(head);
8401
8402         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8403
8404         switch (nfs4err) {
8405         case 0:
8406                 goto out;
8407
8408         /*
8409          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8410          * on the file. set tk_status to -ENODATA to tell upper layer to
8411          * retry go inband.
8412          */
8413         case -NFS4ERR_LAYOUTUNAVAILABLE:
8414                 status = -ENODATA;
8415                 goto out;
8416         /*
8417          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8418          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8419          */
8420         case -NFS4ERR_BADLAYOUT:
8421                 status = -EOVERFLOW;
8422                 goto out;
8423         /*
8424          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8425          * (or clients) writing to the same RAID stripe except when
8426          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8427          *
8428          * Treat it like we would RECALLCONFLICT -- we retry for a little
8429          * while, and then eventually give up.
8430          */
8431         case -NFS4ERR_LAYOUTTRYLATER:
8432                 if (lgp->args.minlength == 0) {
8433                         status = -EOVERFLOW;
8434                         goto out;
8435                 }
8436                 status = -EBUSY;
8437                 break;
8438         case -NFS4ERR_RECALLCONFLICT:
8439                 status = -ERECALLCONFLICT;
8440                 break;
8441         case -NFS4ERR_DELEG_REVOKED:
8442         case -NFS4ERR_ADMIN_REVOKED:
8443         case -NFS4ERR_EXPIRED:
8444         case -NFS4ERR_BAD_STATEID:
8445                 exception->timeout = 0;
8446                 spin_lock(&inode->i_lock);
8447                 lo = NFS_I(inode)->layout;
8448                 /* If the open stateid was bad, then recover it. */
8449                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8450                     nfs4_stateid_match_other(&lgp->args.stateid,
8451                                         &lgp->args.ctx->state->stateid)) {
8452                         spin_unlock(&inode->i_lock);
8453                         exception->state = lgp->args.ctx->state;
8454                         exception->stateid = &lgp->args.stateid;
8455                         break;
8456                 }
8457
8458                 /*
8459                  * Mark the bad layout state as invalid, then retry
8460                  */
8461                 pnfs_mark_layout_stateid_invalid(lo, &head);
8462                 spin_unlock(&inode->i_lock);
8463                 pnfs_free_lseg_list(&head);
8464                 status = -EAGAIN;
8465                 goto out;
8466         }
8467
8468         err = nfs4_handle_exception(server, nfs4err, exception);
8469         if (!status) {
8470                 if (exception->retry)
8471                         status = -EAGAIN;
8472                 else
8473                         status = err;
8474         }
8475 out:
8476         dprintk("<-- %s\n", __func__);
8477         return status;
8478 }
8479
8480 static size_t max_response_pages(struct nfs_server *server)
8481 {
8482         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8483         return nfs_page_array_len(0, max_resp_sz);
8484 }
8485
8486 static void nfs4_free_pages(struct page **pages, size_t size)
8487 {
8488         int i;
8489
8490         if (!pages)
8491                 return;
8492
8493         for (i = 0; i < size; i++) {
8494                 if (!pages[i])
8495                         break;
8496                 __free_page(pages[i]);
8497         }
8498         kfree(pages);
8499 }
8500
8501 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8502 {
8503         struct page **pages;
8504         int i;
8505
8506         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8507         if (!pages) {
8508                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8509                 return NULL;
8510         }
8511
8512         for (i = 0; i < size; i++) {
8513                 pages[i] = alloc_page(gfp_flags);
8514                 if (!pages[i]) {
8515                         dprintk("%s: failed to allocate page\n", __func__);
8516                         nfs4_free_pages(pages, size);
8517                         return NULL;
8518                 }
8519         }
8520
8521         return pages;
8522 }
8523
8524 static void nfs4_layoutget_release(void *calldata)
8525 {
8526         struct nfs4_layoutget *lgp = calldata;
8527         struct inode *inode = lgp->args.inode;
8528         struct nfs_server *server = NFS_SERVER(inode);
8529         size_t max_pages = max_response_pages(server);
8530
8531         dprintk("--> %s\n", __func__);
8532         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8533         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8534         put_nfs_open_context(lgp->args.ctx);
8535         kfree(calldata);
8536         dprintk("<-- %s\n", __func__);
8537 }
8538
8539 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8540         .rpc_call_prepare = nfs4_layoutget_prepare,
8541         .rpc_call_done = nfs4_layoutget_done,
8542         .rpc_release = nfs4_layoutget_release,
8543 };
8544
8545 struct pnfs_layout_segment *
8546 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8547 {
8548         struct inode *inode = lgp->args.inode;
8549         struct nfs_server *server = NFS_SERVER(inode);
8550         size_t max_pages = max_response_pages(server);
8551         struct rpc_task *task;
8552         struct rpc_message msg = {
8553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8554                 .rpc_argp = &lgp->args,
8555                 .rpc_resp = &lgp->res,
8556                 .rpc_cred = lgp->cred,
8557         };
8558         struct rpc_task_setup task_setup_data = {
8559                 .rpc_client = server->client,
8560                 .rpc_message = &msg,
8561                 .callback_ops = &nfs4_layoutget_call_ops,
8562                 .callback_data = lgp,
8563                 .flags = RPC_TASK_ASYNC,
8564         };
8565         struct pnfs_layout_segment *lseg = NULL;
8566         struct nfs4_exception exception = {
8567                 .inode = inode,
8568                 .timeout = *timeout,
8569         };
8570         int status = 0;
8571
8572         dprintk("--> %s\n", __func__);
8573
8574         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8575         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8576
8577         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8578         if (!lgp->args.layout.pages) {
8579                 nfs4_layoutget_release(lgp);
8580                 return ERR_PTR(-ENOMEM);
8581         }
8582         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8583
8584         lgp->res.layoutp = &lgp->args.layout;
8585         lgp->res.seq_res.sr_slot = NULL;
8586         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8587
8588         task = rpc_run_task(&task_setup_data);
8589         if (IS_ERR(task))
8590                 return ERR_CAST(task);
8591         status = nfs4_wait_for_completion_rpc_task(task);
8592         if (status == 0) {
8593                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8594                 *timeout = exception.timeout;
8595         }
8596
8597         trace_nfs4_layoutget(lgp->args.ctx,
8598                         &lgp->args.range,
8599                         &lgp->res.range,
8600                         &lgp->res.stateid,
8601                         status);
8602
8603         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8604         if (status == 0 && lgp->res.layoutp->len)
8605                 lseg = pnfs_layout_process(lgp);
8606         nfs4_sequence_free_slot(&lgp->res.seq_res);
8607         rpc_put_task(task);
8608         dprintk("<-- %s status=%d\n", __func__, status);
8609         if (status)
8610                 return ERR_PTR(status);
8611         return lseg;
8612 }
8613
8614 static void
8615 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8616 {
8617         struct nfs4_layoutreturn *lrp = calldata;
8618
8619         dprintk("--> %s\n", __func__);
8620         nfs41_setup_sequence(lrp->clp->cl_session,
8621                         &lrp->args.seq_args,
8622                         &lrp->res.seq_res,
8623                         task);
8624 }
8625
8626 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8627 {
8628         struct nfs4_layoutreturn *lrp = calldata;
8629         struct nfs_server *server;
8630
8631         dprintk("--> %s\n", __func__);
8632
8633         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8634                 return;
8635
8636         server = NFS_SERVER(lrp->args.inode);
8637         switch (task->tk_status) {
8638         default:
8639                 task->tk_status = 0;
8640         case 0:
8641                 break;
8642         case -NFS4ERR_DELAY:
8643                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8644                         break;
8645                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8646                 rpc_restart_call_prepare(task);
8647                 return;
8648         }
8649         dprintk("<-- %s\n", __func__);
8650 }
8651
8652 static void nfs4_layoutreturn_release(void *calldata)
8653 {
8654         struct nfs4_layoutreturn *lrp = calldata;
8655         struct pnfs_layout_hdr *lo = lrp->args.layout;
8656
8657         dprintk("--> %s\n", __func__);
8658         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8659                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8660         nfs4_sequence_free_slot(&lrp->res.seq_res);
8661         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8662                 lrp->ld_private.ops->free(&lrp->ld_private);
8663         pnfs_put_layout_hdr(lrp->args.layout);
8664         nfs_iput_and_deactive(lrp->inode);
8665         kfree(calldata);
8666         dprintk("<-- %s\n", __func__);
8667 }
8668
8669 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8670         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8671         .rpc_call_done = nfs4_layoutreturn_done,
8672         .rpc_release = nfs4_layoutreturn_release,
8673 };
8674
8675 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8676 {
8677         struct rpc_task *task;
8678         struct rpc_message msg = {
8679                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8680                 .rpc_argp = &lrp->args,
8681                 .rpc_resp = &lrp->res,
8682                 .rpc_cred = lrp->cred,
8683         };
8684         struct rpc_task_setup task_setup_data = {
8685                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8686                 .rpc_message = &msg,
8687                 .callback_ops = &nfs4_layoutreturn_call_ops,
8688                 .callback_data = lrp,
8689         };
8690         int status = 0;
8691
8692         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8693                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8694                         &task_setup_data.rpc_client, &msg);
8695
8696         dprintk("--> %s\n", __func__);
8697         if (!sync) {
8698                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8699                 if (!lrp->inode) {
8700                         nfs4_layoutreturn_release(lrp);
8701                         return -EAGAIN;
8702                 }
8703                 task_setup_data.flags |= RPC_TASK_ASYNC;
8704         }
8705         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8706         task = rpc_run_task(&task_setup_data);
8707         if (IS_ERR(task))
8708                 return PTR_ERR(task);
8709         if (sync)
8710                 status = task->tk_status;
8711         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8712         dprintk("<-- %s status=%d\n", __func__, status);
8713         rpc_put_task(task);
8714         return status;
8715 }
8716
8717 static int
8718 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8719                 struct pnfs_device *pdev,
8720                 struct rpc_cred *cred)
8721 {
8722         struct nfs4_getdeviceinfo_args args = {
8723                 .pdev = pdev,
8724                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8725                         NOTIFY_DEVICEID4_DELETE,
8726         };
8727         struct nfs4_getdeviceinfo_res res = {
8728                 .pdev = pdev,
8729         };
8730         struct rpc_message msg = {
8731                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8732                 .rpc_argp = &args,
8733                 .rpc_resp = &res,
8734                 .rpc_cred = cred,
8735         };
8736         int status;
8737
8738         dprintk("--> %s\n", __func__);
8739         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8740         if (res.notification & ~args.notify_types)
8741                 dprintk("%s: unsupported notification\n", __func__);
8742         if (res.notification != args.notify_types)
8743                 pdev->nocache = 1;
8744
8745         dprintk("<-- %s status=%d\n", __func__, status);
8746
8747         return status;
8748 }
8749
8750 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8751                 struct pnfs_device *pdev,
8752                 struct rpc_cred *cred)
8753 {
8754         struct nfs4_exception exception = { };
8755         int err;
8756
8757         do {
8758                 err = nfs4_handle_exception(server,
8759                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8760                                         &exception);
8761         } while (exception.retry);
8762         return err;
8763 }
8764 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8765
8766 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8767 {
8768         struct nfs4_layoutcommit_data *data = calldata;
8769         struct nfs_server *server = NFS_SERVER(data->args.inode);
8770         struct nfs4_session *session = nfs4_get_session(server);
8771
8772         nfs41_setup_sequence(session,
8773                         &data->args.seq_args,
8774                         &data->res.seq_res,
8775                         task);
8776 }
8777
8778 static void
8779 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8780 {
8781         struct nfs4_layoutcommit_data *data = calldata;
8782         struct nfs_server *server = NFS_SERVER(data->args.inode);
8783
8784         if (!nfs41_sequence_done(task, &data->res.seq_res))
8785                 return;
8786
8787         switch (task->tk_status) { /* Just ignore these failures */
8788         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8789         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8790         case -NFS4ERR_BADLAYOUT:     /* no layout */
8791         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8792                 task->tk_status = 0;
8793         case 0:
8794                 break;
8795         default:
8796                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8797                         rpc_restart_call_prepare(task);
8798                         return;
8799                 }
8800         }
8801 }
8802
8803 static void nfs4_layoutcommit_release(void *calldata)
8804 {
8805         struct nfs4_layoutcommit_data *data = calldata;
8806
8807         pnfs_cleanup_layoutcommit(data);
8808         nfs_post_op_update_inode_force_wcc(data->args.inode,
8809                                            data->res.fattr);
8810         put_rpccred(data->cred);
8811         nfs_iput_and_deactive(data->inode);
8812         kfree(data);
8813 }
8814
8815 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8816         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8817         .rpc_call_done = nfs4_layoutcommit_done,
8818         .rpc_release = nfs4_layoutcommit_release,
8819 };
8820
8821 int
8822 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8823 {
8824         struct rpc_message msg = {
8825                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8826                 .rpc_argp = &data->args,
8827                 .rpc_resp = &data->res,
8828                 .rpc_cred = data->cred,
8829         };
8830         struct rpc_task_setup task_setup_data = {
8831                 .task = &data->task,
8832                 .rpc_client = NFS_CLIENT(data->args.inode),
8833                 .rpc_message = &msg,
8834                 .callback_ops = &nfs4_layoutcommit_ops,
8835                 .callback_data = data,
8836         };
8837         struct rpc_task *task;
8838         int status = 0;
8839
8840         dprintk("NFS: initiating layoutcommit call. sync %d "
8841                 "lbw: %llu inode %lu\n", sync,
8842                 data->args.lastbytewritten,
8843                 data->args.inode->i_ino);
8844
8845         if (!sync) {
8846                 data->inode = nfs_igrab_and_active(data->args.inode);
8847                 if (data->inode == NULL) {
8848                         nfs4_layoutcommit_release(data);
8849                         return -EAGAIN;
8850                 }
8851                 task_setup_data.flags = RPC_TASK_ASYNC;
8852         }
8853         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8854         task = rpc_run_task(&task_setup_data);
8855         if (IS_ERR(task))
8856                 return PTR_ERR(task);
8857         if (sync)
8858                 status = task->tk_status;
8859         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8860         dprintk("%s: status %d\n", __func__, status);
8861         rpc_put_task(task);
8862         return status;
8863 }
8864
8865 /**
8866  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8867  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8868  */
8869 static int
8870 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8871                     struct nfs_fsinfo *info,
8872                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8873 {
8874         struct nfs41_secinfo_no_name_args args = {
8875                 .style = SECINFO_STYLE_CURRENT_FH,
8876         };
8877         struct nfs4_secinfo_res res = {
8878                 .flavors = flavors,
8879         };
8880         struct rpc_message msg = {
8881                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8882                 .rpc_argp = &args,
8883                 .rpc_resp = &res,
8884         };
8885         struct rpc_clnt *clnt = server->client;
8886         struct rpc_cred *cred = NULL;
8887         int status;
8888
8889         if (use_integrity) {
8890                 clnt = server->nfs_client->cl_rpcclient;
8891                 cred = nfs4_get_clid_cred(server->nfs_client);
8892                 msg.rpc_cred = cred;
8893         }
8894
8895         dprintk("--> %s\n", __func__);
8896         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8897                                 &res.seq_res, 0);
8898         dprintk("<-- %s status=%d\n", __func__, status);
8899
8900         if (cred)
8901                 put_rpccred(cred);
8902
8903         return status;
8904 }
8905
8906 static int
8907 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8908                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8909 {
8910         struct nfs4_exception exception = { };
8911         int err;
8912         do {
8913                 /* first try using integrity protection */
8914                 err = -NFS4ERR_WRONGSEC;
8915
8916                 /* try to use integrity protection with machine cred */
8917                 if (_nfs4_is_integrity_protected(server->nfs_client))
8918                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8919                                                           flavors, true);
8920
8921                 /*
8922                  * if unable to use integrity protection, or SECINFO with
8923                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8924                  * disallowed by spec, but exists in deployed servers) use
8925                  * the current filesystem's rpc_client and the user cred.
8926                  */
8927                 if (err == -NFS4ERR_WRONGSEC)
8928                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8929                                                           flavors, false);
8930
8931                 switch (err) {
8932                 case 0:
8933                 case -NFS4ERR_WRONGSEC:
8934                 case -ENOTSUPP:
8935                         goto out;
8936                 default:
8937                         err = nfs4_handle_exception(server, err, &exception);
8938                 }
8939         } while (exception.retry);
8940 out:
8941         return err;
8942 }
8943
8944 static int
8945 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8946                     struct nfs_fsinfo *info)
8947 {
8948         int err;
8949         struct page *page;
8950         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8951         struct nfs4_secinfo_flavors *flavors;
8952         struct nfs4_secinfo4 *secinfo;
8953         int i;
8954
8955         page = alloc_page(GFP_KERNEL);
8956         if (!page) {
8957                 err = -ENOMEM;
8958                 goto out;
8959         }
8960
8961         flavors = page_address(page);
8962         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8963
8964         /*
8965          * Fall back on "guess and check" method if
8966          * the server doesn't support SECINFO_NO_NAME
8967          */
8968         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8969                 err = nfs4_find_root_sec(server, fhandle, info);
8970                 goto out_freepage;
8971         }
8972         if (err)
8973                 goto out_freepage;
8974
8975         for (i = 0; i < flavors->num_flavors; i++) {
8976                 secinfo = &flavors->flavors[i];
8977
8978                 switch (secinfo->flavor) {
8979                 case RPC_AUTH_NULL:
8980                 case RPC_AUTH_UNIX:
8981                 case RPC_AUTH_GSS:
8982                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8983                                         &secinfo->flavor_info);
8984                         break;
8985                 default:
8986                         flavor = RPC_AUTH_MAXFLAVOR;
8987                         break;
8988                 }
8989
8990                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8991                         flavor = RPC_AUTH_MAXFLAVOR;
8992
8993                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8994                         err = nfs4_lookup_root_sec(server, fhandle,
8995                                                    info, flavor);
8996                         if (!err)
8997                                 break;
8998                 }
8999         }
9000
9001         if (flavor == RPC_AUTH_MAXFLAVOR)
9002                 err = -EPERM;
9003
9004 out_freepage:
9005         put_page(page);
9006         if (err == -EACCES)
9007                 return -EPERM;
9008 out:
9009         return err;
9010 }
9011
9012 static int _nfs41_test_stateid(struct nfs_server *server,
9013                 nfs4_stateid *stateid,
9014                 struct rpc_cred *cred)
9015 {
9016         int status;
9017         struct nfs41_test_stateid_args args = {
9018                 .stateid = stateid,
9019         };
9020         struct nfs41_test_stateid_res res;
9021         struct rpc_message msg = {
9022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9023                 .rpc_argp = &args,
9024                 .rpc_resp = &res,
9025                 .rpc_cred = cred,
9026         };
9027         struct rpc_clnt *rpc_client = server->client;
9028
9029         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9030                 &rpc_client, &msg);
9031
9032         dprintk("NFS call  test_stateid %p\n", stateid);
9033         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
9034         nfs4_set_sequence_privileged(&args.seq_args);
9035         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9036                         &args.seq_args, &res.seq_res);
9037         if (status != NFS_OK) {
9038                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9039                 return status;
9040         }
9041         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9042         return -res.status;
9043 }
9044
9045 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9046                 int err, struct nfs4_exception *exception)
9047 {
9048         exception->retry = 0;
9049         switch(err) {
9050         case -NFS4ERR_DELAY:
9051         case -NFS4ERR_RETRY_UNCACHED_REP:
9052                 nfs4_handle_exception(server, err, exception);
9053                 break;
9054         case -NFS4ERR_BADSESSION:
9055         case -NFS4ERR_BADSLOT:
9056         case -NFS4ERR_BAD_HIGH_SLOT:
9057         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9058         case -NFS4ERR_DEADSESSION:
9059                 nfs4_do_handle_exception(server, err, exception);
9060         }
9061 }
9062
9063 /**
9064  * nfs41_test_stateid - perform a TEST_STATEID operation
9065  *
9066  * @server: server / transport on which to perform the operation
9067  * @stateid: state ID to test
9068  * @cred: credential
9069  *
9070  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9071  * Otherwise a negative NFS4ERR value is returned if the operation
9072  * failed or the state ID is not currently valid.
9073  */
9074 static int nfs41_test_stateid(struct nfs_server *server,
9075                 nfs4_stateid *stateid,
9076                 struct rpc_cred *cred)
9077 {
9078         struct nfs4_exception exception = { };
9079         int err;
9080         do {
9081                 err = _nfs41_test_stateid(server, stateid, cred);
9082                 nfs4_handle_delay_or_session_error(server, err, &exception);
9083         } while (exception.retry);
9084         return err;
9085 }
9086
9087 struct nfs_free_stateid_data {
9088         struct nfs_server *server;
9089         struct nfs41_free_stateid_args args;
9090         struct nfs41_free_stateid_res res;
9091 };
9092
9093 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9094 {
9095         struct nfs_free_stateid_data *data = calldata;
9096         nfs41_setup_sequence(nfs4_get_session(data->server),
9097                         &data->args.seq_args,
9098                         &data->res.seq_res,
9099                         task);
9100 }
9101
9102 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9103 {
9104         struct nfs_free_stateid_data *data = calldata;
9105
9106         nfs41_sequence_done(task, &data->res.seq_res);
9107
9108         switch (task->tk_status) {
9109         case -NFS4ERR_DELAY:
9110                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9111                         rpc_restart_call_prepare(task);
9112         }
9113 }
9114
9115 static void nfs41_free_stateid_release(void *calldata)
9116 {
9117         kfree(calldata);
9118 }
9119
9120 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9121         .rpc_call_prepare = nfs41_free_stateid_prepare,
9122         .rpc_call_done = nfs41_free_stateid_done,
9123         .rpc_release = nfs41_free_stateid_release,
9124 };
9125
9126 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9127                 const nfs4_stateid *stateid,
9128                 struct rpc_cred *cred,
9129                 bool privileged)
9130 {
9131         struct rpc_message msg = {
9132                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9133                 .rpc_cred = cred,
9134         };
9135         struct rpc_task_setup task_setup = {
9136                 .rpc_client = server->client,
9137                 .rpc_message = &msg,
9138                 .callback_ops = &nfs41_free_stateid_ops,
9139                 .flags = RPC_TASK_ASYNC,
9140         };
9141         struct nfs_free_stateid_data *data;
9142
9143         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9144                 &task_setup.rpc_client, &msg);
9145
9146         dprintk("NFS call  free_stateid %p\n", stateid);
9147         data = kmalloc(sizeof(*data), GFP_NOFS);
9148         if (!data)
9149                 return ERR_PTR(-ENOMEM);
9150         data->server = server;
9151         nfs4_stateid_copy(&data->args.stateid, stateid);
9152
9153         task_setup.callback_data = data;
9154
9155         msg.rpc_argp = &data->args;
9156         msg.rpc_resp = &data->res;
9157         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9158         if (privileged)
9159                 nfs4_set_sequence_privileged(&data->args.seq_args);
9160
9161         return rpc_run_task(&task_setup);
9162 }
9163
9164 /**
9165  * nfs41_free_stateid - perform a FREE_STATEID operation
9166  *
9167  * @server: server / transport on which to perform the operation
9168  * @stateid: state ID to release
9169  * @cred: credential
9170  * @is_recovery: set to true if this call needs to be privileged
9171  *
9172  * Note: this function is always asynchronous.
9173  */
9174 static int nfs41_free_stateid(struct nfs_server *server,
9175                 const nfs4_stateid *stateid,
9176                 struct rpc_cred *cred,
9177                 bool is_recovery)
9178 {
9179         struct rpc_task *task;
9180
9181         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9182         if (IS_ERR(task))
9183                 return PTR_ERR(task);
9184         rpc_put_task(task);
9185         return 0;
9186 }
9187
9188 static void
9189 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9190 {
9191         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9192
9193         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9194         nfs4_free_lock_state(server, lsp);
9195 }
9196
9197 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9198                 const nfs4_stateid *s2)
9199 {
9200         if (s1->type != s2->type)
9201                 return false;
9202
9203         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9204                 return false;
9205
9206         if (s1->seqid == s2->seqid)
9207                 return true;
9208         if (s1->seqid == 0 || s2->seqid == 0)
9209                 return true;
9210
9211         return false;
9212 }
9213
9214 #endif /* CONFIG_NFS_V4_1 */
9215
9216 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9217                 const nfs4_stateid *s2)
9218 {
9219         return nfs4_stateid_match(s1, s2);
9220 }
9221
9222
9223 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9224         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9225         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9226         .recover_open   = nfs4_open_reclaim,
9227         .recover_lock   = nfs4_lock_reclaim,
9228         .establish_clid = nfs4_init_clientid,
9229         .detect_trunking = nfs40_discover_server_trunking,
9230 };
9231
9232 #if defined(CONFIG_NFS_V4_1)
9233 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9234         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9235         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9236         .recover_open   = nfs4_open_reclaim,
9237         .recover_lock   = nfs4_lock_reclaim,
9238         .establish_clid = nfs41_init_clientid,
9239         .reclaim_complete = nfs41_proc_reclaim_complete,
9240         .detect_trunking = nfs41_discover_server_trunking,
9241 };
9242 #endif /* CONFIG_NFS_V4_1 */
9243
9244 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9245         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9246         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9247         .recover_open   = nfs40_open_expired,
9248         .recover_lock   = nfs4_lock_expired,
9249         .establish_clid = nfs4_init_clientid,
9250 };
9251
9252 #if defined(CONFIG_NFS_V4_1)
9253 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9254         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9255         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9256         .recover_open   = nfs41_open_expired,
9257         .recover_lock   = nfs41_lock_expired,
9258         .establish_clid = nfs41_init_clientid,
9259 };
9260 #endif /* CONFIG_NFS_V4_1 */
9261
9262 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9263         .sched_state_renewal = nfs4_proc_async_renew,
9264         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9265         .renew_lease = nfs4_proc_renew,
9266 };
9267
9268 #if defined(CONFIG_NFS_V4_1)
9269 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9270         .sched_state_renewal = nfs41_proc_async_sequence,
9271         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9272         .renew_lease = nfs4_proc_sequence,
9273 };
9274 #endif
9275
9276 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9277         .get_locations = _nfs40_proc_get_locations,
9278         .fsid_present = _nfs40_proc_fsid_present,
9279 };
9280
9281 #if defined(CONFIG_NFS_V4_1)
9282 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9283         .get_locations = _nfs41_proc_get_locations,
9284         .fsid_present = _nfs41_proc_fsid_present,
9285 };
9286 #endif  /* CONFIG_NFS_V4_1 */
9287
9288 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9289         .minor_version = 0,
9290         .init_caps = NFS_CAP_READDIRPLUS
9291                 | NFS_CAP_ATOMIC_OPEN
9292                 | NFS_CAP_POSIX_LOCK,
9293         .init_client = nfs40_init_client,
9294         .shutdown_client = nfs40_shutdown_client,
9295         .match_stateid = nfs4_match_stateid,
9296         .find_root_sec = nfs4_find_root_sec,
9297         .free_lock_state = nfs4_release_lockowner,
9298         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9299         .alloc_seqid = nfs_alloc_seqid,
9300         .call_sync_ops = &nfs40_call_sync_ops,
9301         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9302         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9303         .state_renewal_ops = &nfs40_state_renewal_ops,
9304         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9305 };
9306
9307 #if defined(CONFIG_NFS_V4_1)
9308 static struct nfs_seqid *
9309 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9310 {
9311         return NULL;
9312 }
9313
9314 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9315         .minor_version = 1,
9316         .init_caps = NFS_CAP_READDIRPLUS
9317                 | NFS_CAP_ATOMIC_OPEN
9318                 | NFS_CAP_POSIX_LOCK
9319                 | NFS_CAP_STATEID_NFSV41
9320                 | NFS_CAP_ATOMIC_OPEN_V1,
9321         .init_client = nfs41_init_client,
9322         .shutdown_client = nfs41_shutdown_client,
9323         .match_stateid = nfs41_match_stateid,
9324         .find_root_sec = nfs41_find_root_sec,
9325         .free_lock_state = nfs41_free_lock_state,
9326         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9327         .alloc_seqid = nfs_alloc_no_seqid,
9328         .session_trunk = nfs4_test_session_trunk,
9329         .call_sync_ops = &nfs41_call_sync_ops,
9330         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9331         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9332         .state_renewal_ops = &nfs41_state_renewal_ops,
9333         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9334 };
9335 #endif
9336
9337 #if defined(CONFIG_NFS_V4_2)
9338 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9339         .minor_version = 2,
9340         .init_caps = NFS_CAP_READDIRPLUS
9341                 | NFS_CAP_ATOMIC_OPEN
9342                 | NFS_CAP_POSIX_LOCK
9343                 | NFS_CAP_STATEID_NFSV41
9344                 | NFS_CAP_ATOMIC_OPEN_V1
9345                 | NFS_CAP_ALLOCATE
9346                 | NFS_CAP_COPY
9347                 | NFS_CAP_DEALLOCATE
9348                 | NFS_CAP_SEEK
9349                 | NFS_CAP_LAYOUTSTATS
9350                 | NFS_CAP_CLONE,
9351         .init_client = nfs41_init_client,
9352         .shutdown_client = nfs41_shutdown_client,
9353         .match_stateid = nfs41_match_stateid,
9354         .find_root_sec = nfs41_find_root_sec,
9355         .free_lock_state = nfs41_free_lock_state,
9356         .call_sync_ops = &nfs41_call_sync_ops,
9357         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9358         .alloc_seqid = nfs_alloc_no_seqid,
9359         .session_trunk = nfs4_test_session_trunk,
9360         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9361         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9362         .state_renewal_ops = &nfs41_state_renewal_ops,
9363         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9364 };
9365 #endif
9366
9367 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9368         [0] = &nfs_v4_0_minor_ops,
9369 #if defined(CONFIG_NFS_V4_1)
9370         [1] = &nfs_v4_1_minor_ops,
9371 #endif
9372 #if defined(CONFIG_NFS_V4_2)
9373         [2] = &nfs_v4_2_minor_ops,
9374 #endif
9375 };
9376
9377 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9378 {
9379         ssize_t error, error2;
9380
9381         error = generic_listxattr(dentry, list, size);
9382         if (error < 0)
9383                 return error;
9384         if (list) {
9385                 list += error;
9386                 size -= error;
9387         }
9388
9389         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9390         if (error2 < 0)
9391                 return error2;
9392         return error + error2;
9393 }
9394
9395 static const struct inode_operations nfs4_dir_inode_operations = {
9396         .create         = nfs_create,
9397         .lookup         = nfs_lookup,
9398         .atomic_open    = nfs_atomic_open,
9399         .link           = nfs_link,
9400         .unlink         = nfs_unlink,
9401         .symlink        = nfs_symlink,
9402         .mkdir          = nfs_mkdir,
9403         .rmdir          = nfs_rmdir,
9404         .mknod          = nfs_mknod,
9405         .rename         = nfs_rename,
9406         .permission     = nfs_permission,
9407         .getattr        = nfs_getattr,
9408         .setattr        = nfs_setattr,
9409         .listxattr      = nfs4_listxattr,
9410 };
9411
9412 static const struct inode_operations nfs4_file_inode_operations = {
9413         .permission     = nfs_permission,
9414         .getattr        = nfs_getattr,
9415         .setattr        = nfs_setattr,
9416         .listxattr      = nfs4_listxattr,
9417 };
9418
9419 const struct nfs_rpc_ops nfs_v4_clientops = {
9420         .version        = 4,                    /* protocol version */
9421         .dentry_ops     = &nfs4_dentry_operations,
9422         .dir_inode_ops  = &nfs4_dir_inode_operations,
9423         .file_inode_ops = &nfs4_file_inode_operations,
9424         .file_ops       = &nfs4_file_operations,
9425         .getroot        = nfs4_proc_get_root,
9426         .submount       = nfs4_submount,
9427         .try_mount      = nfs4_try_mount,
9428         .getattr        = nfs4_proc_getattr,
9429         .setattr        = nfs4_proc_setattr,
9430         .lookup         = nfs4_proc_lookup,
9431         .access         = nfs4_proc_access,
9432         .readlink       = nfs4_proc_readlink,
9433         .create         = nfs4_proc_create,
9434         .remove         = nfs4_proc_remove,
9435         .unlink_setup   = nfs4_proc_unlink_setup,
9436         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9437         .unlink_done    = nfs4_proc_unlink_done,
9438         .rename_setup   = nfs4_proc_rename_setup,
9439         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9440         .rename_done    = nfs4_proc_rename_done,
9441         .link           = nfs4_proc_link,
9442         .symlink        = nfs4_proc_symlink,
9443         .mkdir          = nfs4_proc_mkdir,
9444         .rmdir          = nfs4_proc_remove,
9445         .readdir        = nfs4_proc_readdir,
9446         .mknod          = nfs4_proc_mknod,
9447         .statfs         = nfs4_proc_statfs,
9448         .fsinfo         = nfs4_proc_fsinfo,
9449         .pathconf       = nfs4_proc_pathconf,
9450         .set_capabilities = nfs4_server_capabilities,
9451         .decode_dirent  = nfs4_decode_dirent,
9452         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9453         .read_setup     = nfs4_proc_read_setup,
9454         .read_done      = nfs4_read_done,
9455         .write_setup    = nfs4_proc_write_setup,
9456         .write_done     = nfs4_write_done,
9457         .commit_setup   = nfs4_proc_commit_setup,
9458         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9459         .commit_done    = nfs4_commit_done,
9460         .lock           = nfs4_proc_lock,
9461         .clear_acl_cache = nfs4_zap_acl_attr,
9462         .close_context  = nfs4_close_context,
9463         .open_context   = nfs4_atomic_open,
9464         .have_delegation = nfs4_have_delegation,
9465         .return_delegation = nfs4_inode_return_delegation,
9466         .alloc_client   = nfs4_alloc_client,
9467         .init_client    = nfs4_init_client,
9468         .free_client    = nfs4_free_client,
9469         .create_server  = nfs4_create_server,
9470         .clone_server   = nfs_clone_server,
9471 };
9472
9473 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9474         .name   = XATTR_NAME_NFSV4_ACL,
9475         .list   = nfs4_xattr_list_nfs4_acl,
9476         .get    = nfs4_xattr_get_nfs4_acl,
9477         .set    = nfs4_xattr_set_nfs4_acl,
9478 };
9479
9480 const struct xattr_handler *nfs4_xattr_handlers[] = {
9481         &nfs4_xattr_nfs4_acl_handler,
9482 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9483         &nfs4_xattr_nfs4_label_handler,
9484 #endif
9485         NULL
9486 };
9487
9488 /*
9489  * Local variables:
9490  *  c-basic-offset: 8
9491  * End:
9492  */