fs/nfs/nfs4proc.c

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