fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57
  58 #include "nfs4_fs.h"
  59 #include "delegation.h"
  60 #include "internal.h"
  61 #include "iostat.h"
  62 #include "callback.h"
  63 #include "pnfs.h"
  64 #include "netns.h"
  65 #include "nfs4idmap.h"
  66 #include "nfs4session.h"
  67 #include "fscache.h"
  68
  69 #include "nfs4trace.h"
  70
  71 #define NFSDBG_FACILITY         NFSDBG_PROC
  72
  73 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  74 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  75
  76 /* file attributes which can be mapped to nfs attributes */
  77 #define NFS4_VALID_ATTRS (ATTR_MODE \
  78         | ATTR_UID \
  79         | ATTR_GID \
  80         | ATTR_SIZE \
  81         | ATTR_ATIME \
  82         | ATTR_MTIME \
  83         | ATTR_CTIME \
  84         | ATTR_ATIME_SET \
  85         | ATTR_MTIME_SET)
  86
  87 struct nfs4_opendata;
  88 static int _nfs4_proc_open(struct nfs4_opendata *data);
  89 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  90 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  91 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  92 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
  93 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
  94 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
  95                             struct nfs_fattr *fattr, struct iattr *sattr,
  96                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
  97                             struct nfs4_label *olabel);
  98 #ifdef CONFIG_NFS_V4_1
  99 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 100                 struct rpc_cred *);
 101 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 102                 struct rpc_cred *, bool);
 103 #endif
 104
 105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 106 static inline struct nfs4_label *
 107 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 108         struct iattr *sattr, struct nfs4_label *label)
 109 {
 110         int err;
 111
 112         if (label == NULL)
 113                 return NULL;
 114
 115         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 116                 return NULL;
 117
 118         err = security_dentry_init_security(dentry, sattr->ia_mode,
 119                                 &dentry->d_name, (void **)&label->label, &label->len);
 120         if (err == 0)
 121                 return label;
 122
 123         return NULL;
 124 }
 125 static inline void
 126 nfs4_label_release_security(struct nfs4_label *label)
 127 {
 128         if (label)
 129                 security_release_secctx(label->label, label->len);
 130 }
 131 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 132 {
 133         if (label)
 134                 return server->attr_bitmask;
 135
 136         return server->attr_bitmask_nl;
 137 }
 138 #else
 139 static inline struct nfs4_label *
 140 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 141         struct iattr *sattr, struct nfs4_label *l)
 142 { return NULL; }
 143 static inline void
 144 nfs4_label_release_security(struct nfs4_label *label)
 145 { return; }
 146 static inline u32 *
 147 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 148 { return server->attr_bitmask; }
 149 #endif
 150
 151 /* Prevent leaks of NFSv4 errors into userland */
 152 static int nfs4_map_errors(int err)
 153 {
 154         if (err >= -1000)
 155                 return err;
 156         switch (err) {
 157         case -NFS4ERR_RESOURCE:
 158         case -NFS4ERR_LAYOUTTRYLATER:
 159         case -NFS4ERR_RECALLCONFLICT:
 160                 return -EREMOTEIO;
 161         case -NFS4ERR_WRONGSEC:
 162         case -NFS4ERR_WRONG_CRED:
 163                 return -EPERM;
 164         case -NFS4ERR_BADOWNER:
 165         case -NFS4ERR_BADNAME:
 166                 return -EINVAL;
 167         case -NFS4ERR_SHARE_DENIED:
 168                 return -EACCES;
 169         case -NFS4ERR_MINOR_VERS_MISMATCH:
 170                 return -EPROTONOSUPPORT;
 171         case -NFS4ERR_FILE_OPEN:
 172                 return -EBUSY;
 173         default:
 174                 dprintk("%s could not handle NFSv4 error %d\n",
 175                                 __func__, -err);
 176                 break;
 177         }
 178         return -EIO;
 179 }
 180
 181 /*
 182  * This is our standard bitmap for GETATTR requests.
 183  */
 184 const u32 nfs4_fattr_bitmap[3] = {
 185         FATTR4_WORD0_TYPE
 186         | FATTR4_WORD0_CHANGE
 187         | FATTR4_WORD0_SIZE
 188         | FATTR4_WORD0_FSID
 189         | FATTR4_WORD0_FILEID,
 190         FATTR4_WORD1_MODE
 191         | FATTR4_WORD1_NUMLINKS
 192         | FATTR4_WORD1_OWNER
 193         | FATTR4_WORD1_OWNER_GROUP
 194         | FATTR4_WORD1_RAWDEV
 195         | FATTR4_WORD1_SPACE_USED
 196         | FATTR4_WORD1_TIME_ACCESS
 197         | FATTR4_WORD1_TIME_METADATA
 198         | FATTR4_WORD1_TIME_MODIFY
 199         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 200 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 201         FATTR4_WORD2_SECURITY_LABEL
 202 #endif
 203 };
 204
 205 static const u32 nfs4_pnfs_open_bitmap[3] = {
 206         FATTR4_WORD0_TYPE
 207         | FATTR4_WORD0_CHANGE
 208         | FATTR4_WORD0_SIZE
 209         | FATTR4_WORD0_FSID
 210         | FATTR4_WORD0_FILEID,
 211         FATTR4_WORD1_MODE
 212         | FATTR4_WORD1_NUMLINKS
 213         | FATTR4_WORD1_OWNER
 214         | FATTR4_WORD1_OWNER_GROUP
 215         | FATTR4_WORD1_RAWDEV
 216         | FATTR4_WORD1_SPACE_USED
 217         | FATTR4_WORD1_TIME_ACCESS
 218         | FATTR4_WORD1_TIME_METADATA
 219         | FATTR4_WORD1_TIME_MODIFY,
 220         FATTR4_WORD2_MDSTHRESHOLD
 221 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 222         | FATTR4_WORD2_SECURITY_LABEL
 223 #endif
 224 };
 225
 226 static const u32 nfs4_open_noattr_bitmap[3] = {
 227         FATTR4_WORD0_TYPE
 228         | FATTR4_WORD0_FILEID,
 229 };
 230
 231 const u32 nfs4_statfs_bitmap[3] = {
 232         FATTR4_WORD0_FILES_AVAIL
 233         | FATTR4_WORD0_FILES_FREE
 234         | FATTR4_WORD0_FILES_TOTAL,
 235         FATTR4_WORD1_SPACE_AVAIL
 236         | FATTR4_WORD1_SPACE_FREE
 237         | FATTR4_WORD1_SPACE_TOTAL
 238 };
 239
 240 const u32 nfs4_pathconf_bitmap[3] = {
 241         FATTR4_WORD0_MAXLINK
 242         | FATTR4_WORD0_MAXNAME,
 243         0
 244 };
 245
 246 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 247                         | FATTR4_WORD0_MAXREAD
 248                         | FATTR4_WORD0_MAXWRITE
 249                         | FATTR4_WORD0_LEASE_TIME,
 250                         FATTR4_WORD1_TIME_DELTA
 251                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 252                         FATTR4_WORD2_LAYOUT_BLKSIZE
 253                         | FATTR4_WORD2_CLONE_BLKSIZE
 254 };
 255
 256 const u32 nfs4_fs_locations_bitmap[3] = {
 257         FATTR4_WORD0_TYPE
 258         | FATTR4_WORD0_CHANGE
 259         | FATTR4_WORD0_SIZE
 260         | FATTR4_WORD0_FSID
 261         | FATTR4_WORD0_FILEID
 262         | FATTR4_WORD0_FS_LOCATIONS,
 263         FATTR4_WORD1_MODE
 264         | FATTR4_WORD1_NUMLINKS
 265         | FATTR4_WORD1_OWNER
 266         | FATTR4_WORD1_OWNER_GROUP
 267         | FATTR4_WORD1_RAWDEV
 268         | FATTR4_WORD1_SPACE_USED
 269         | FATTR4_WORD1_TIME_ACCESS
 270         | FATTR4_WORD1_TIME_METADATA
 271         | FATTR4_WORD1_TIME_MODIFY
 272         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 273 };
 274
 275 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 276                 struct nfs4_readdir_arg *readdir)
 277 {
 278         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 279         __be32 *start, *p;
 280
 281         if (cookie > 2) {
 282                 readdir->cookie = cookie;
 283                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 284                 return;
 285         }
 286
 287         readdir->cookie = 0;
 288         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 289         if (cookie == 2)
 290                 return;
 291
 292         /*
 293          * NFSv4 servers do not return entries for '.' and '..'
 294          * Therefore, we fake these entries here.  We let '.'
 295          * have cookie 0 and '..' have cookie 1.  Note that
 296          * when talking to the server, we always send cookie 0
 297          * instead of 1 or 2.
 298          */
 299         start = p = kmap_atomic(*readdir->pages);
 300
 301         if (cookie == 0) {
 302                 *p++ = xdr_one;                                  /* next */
 303                 *p++ = xdr_zero;                   /* cookie, first word */
 304                 *p++ = xdr_one;                   /* cookie, second word */
 305                 *p++ = xdr_one;                             /* entry len */
 306                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 307                 p++;
 308                 *p++ = xdr_one;                         /* bitmap length */
 309                 *p++ = htonl(attrs);                           /* bitmap */
 310                 *p++ = htonl(12);             /* attribute buffer length */
 311                 *p++ = htonl(NF4DIR);
 312                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 313         }
 314
 315         *p++ = xdr_one;                                  /* next */
 316         *p++ = xdr_zero;                   /* cookie, first word */
 317         *p++ = xdr_two;                   /* cookie, second word */
 318         *p++ = xdr_two;                             /* entry len */
 319         memcpy(p, "..\0\0", 4);                         /* entry */
 320         p++;
 321         *p++ = xdr_one;                         /* bitmap length */
 322         *p++ = htonl(attrs);                           /* bitmap */
 323         *p++ = htonl(12);             /* attribute buffer length */
 324         *p++ = htonl(NF4DIR);
 325         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 326
 327         readdir->pgbase = (char *)p - (char *)start;
 328         readdir->count -= readdir->pgbase;
 329         kunmap_atomic(start);
 330 }
 331
 332 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 333                 nfs4_stateid *stateid,
 334                 struct rpc_cred *cred)
 335 {
 336         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 337
 338         ops->test_and_free_expired(server, stateid, cred);
 339 }
 340
 341 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 342                 nfs4_stateid *stateid,
 343                 struct rpc_cred *cred)
 344 {
 345         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 346         nfs4_test_and_free_stateid(server, stateid, cred);
 347 }
 348
 349 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 350                 const nfs4_stateid *stateid,
 351                 struct rpc_cred *cred)
 352 {
 353         nfs4_stateid tmp;
 354
 355         nfs4_stateid_copy(&tmp, stateid);
 356         __nfs4_free_revoked_stateid(server, &tmp, cred);
 357 }
 358
 359 static long nfs4_update_delay(long *timeout)
 360 {
 361         long ret;
 362         if (!timeout)
 363                 return NFS4_POLL_RETRY_MAX;
 364         if (*timeout <= 0)
 365                 *timeout = NFS4_POLL_RETRY_MIN;
 366         if (*timeout > NFS4_POLL_RETRY_MAX)
 367                 *timeout = NFS4_POLL_RETRY_MAX;
 368         ret = *timeout;
 369         *timeout <<= 1;
 370         return ret;
 371 }
 372
 373 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 374 {
 375         int res = 0;
 376
 377         might_sleep();
 378
 379         freezable_schedule_timeout_killable_unsafe(
 380                 nfs4_update_delay(timeout));
 381         if (fatal_signal_pending(current))
 382                 res = -ERESTARTSYS;
 383         return res;
 384 }
 385
 386 /* This is the error handling routine for processes that are allowed
 387  * to sleep.
 388  */
 389 static int nfs4_do_handle_exception(struct nfs_server *server,
 390                 int errorcode, struct nfs4_exception *exception)
 391 {
 392         struct nfs_client *clp = server->nfs_client;
 393         struct nfs4_state *state = exception->state;
 394         const nfs4_stateid *stateid = exception->stateid;
 395         struct inode *inode = exception->inode;
 396         int ret = errorcode;
 397
 398         exception->delay = 0;
 399         exception->recovering = 0;
 400         exception->retry = 0;
 401
 402         if (stateid == NULL && state != NULL)
 403                 stateid = &state->stateid;
 404
 405         switch(errorcode) {
 406                 case 0:
 407                         return 0;
 408                 case -NFS4ERR_DELEG_REVOKED:
 409                 case -NFS4ERR_ADMIN_REVOKED:
 410                 case -NFS4ERR_EXPIRED:
 411                 case -NFS4ERR_BAD_STATEID:
 412                         if (inode != NULL && stateid != NULL) {
 413                                 nfs_inode_find_state_and_recover(inode,
 414                                                 stateid);
 415                                 goto wait_on_recovery;
 416                         }
 417                 case -NFS4ERR_OPENMODE:
 418                         if (inode) {
 419                                 int err;
 420
 421                                 err = nfs_async_inode_return_delegation(inode,
 422                                                 stateid);
 423                                 if (err == 0)
 424                                         goto wait_on_recovery;
 425                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 426                                         exception->retry = 1;
 427                                         break;
 428                                 }
 429                         }
 430                         if (state == NULL)
 431                                 break;
 432                         ret = nfs4_schedule_stateid_recovery(server, state);
 433                         if (ret < 0)
 434                                 break;
 435                         goto wait_on_recovery;
 436                 case -NFS4ERR_STALE_STATEID:
 437                 case -NFS4ERR_STALE_CLIENTID:
 438                         nfs4_schedule_lease_recovery(clp);
 439                         goto wait_on_recovery;
 440                 case -NFS4ERR_MOVED:
 441                         ret = nfs4_schedule_migration_recovery(server);
 442                         if (ret < 0)
 443                                 break;
 444                         goto wait_on_recovery;
 445                 case -NFS4ERR_LEASE_MOVED:
 446                         nfs4_schedule_lease_moved_recovery(clp);
 447                         goto wait_on_recovery;
 448 #if defined(CONFIG_NFS_V4_1)
 449                 case -NFS4ERR_BADSESSION:
 450                 case -NFS4ERR_BADSLOT:
 451                 case -NFS4ERR_BAD_HIGH_SLOT:
 452                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 453                 case -NFS4ERR_DEADSESSION:
 454                 case -NFS4ERR_SEQ_FALSE_RETRY:
 455                 case -NFS4ERR_SEQ_MISORDERED:
 456                         dprintk("%s ERROR: %d Reset session\n", __func__,
 457                                 errorcode);
 458                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 459                         goto wait_on_recovery;
 460 #endif /* defined(CONFIG_NFS_V4_1) */
 461                 case -NFS4ERR_FILE_OPEN:
 462                         if (exception->timeout > HZ) {
 463                                 /* We have retried a decent amount, time to
 464                                  * fail
 465                                  */
 466                                 ret = -EBUSY;
 467                                 break;
 468                         }
 469                 case -NFS4ERR_DELAY:
 470                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 471                 case -NFS4ERR_GRACE:
 472                 case -NFS4ERR_LAYOUTTRYLATER:
 473                 case -NFS4ERR_RECALLCONFLICT:
 474                         exception->delay = 1;
 475                         return 0;
 476
 477                 case -NFS4ERR_RETRY_UNCACHED_REP:
 478                 case -NFS4ERR_OLD_STATEID:
 479                         exception->retry = 1;
 480                         break;
 481                 case -NFS4ERR_BADOWNER:
 482                         /* The following works around a Linux server bug! */
 483                 case -NFS4ERR_BADNAME:
 484                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 485                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 486                                 exception->retry = 1;
 487                                 printk(KERN_WARNING "NFS: v4 server %s "
 488                                                 "does not accept raw "
 489                                                 "uid/gids. "
 490                                                 "Reenabling the idmapper.\n",
 491                                                 server->nfs_client->cl_hostname);
 492                         }
 493         }
 494         /* We failed to handle the error */
 495         return nfs4_map_errors(ret);
 496 wait_on_recovery:
 497         exception->recovering = 1;
 498         return 0;
 499 }
 500
 501 /* This is the error handling routine for processes that are allowed
 502  * to sleep.
 503  */
 504 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 505 {
 506         struct nfs_client *clp = server->nfs_client;
 507         int ret;
 508
 509         ret = nfs4_do_handle_exception(server, errorcode, exception);
 510         if (exception->delay) {
 511                 ret = nfs4_delay(server->client, &exception->timeout);
 512                 goto out_retry;
 513         }
 514         if (exception->recovering) {
 515                 ret = nfs4_wait_clnt_recover(clp);
 516                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 517                         return -EIO;
 518                 goto out_retry;
 519         }
 520         return ret;
 521 out_retry:
 522         if (ret == 0)
 523                 exception->retry = 1;
 524         return ret;
 525 }
 526
 527 static int
 528 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 529                 int errorcode, struct nfs4_exception *exception)
 530 {
 531         struct nfs_client *clp = server->nfs_client;
 532         int ret;
 533
 534         ret = nfs4_do_handle_exception(server, errorcode, exception);
 535         if (exception->delay) {
 536                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 537                 goto out_retry;
 538         }
 539         if (exception->recovering) {
 540                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 541                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 542                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 543                 goto out_retry;
 544         }
 545         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 546                 ret = -EIO;
 547         return ret;
 548 out_retry:
 549         if (ret == 0)
 550                 exception->retry = 1;
 551         return ret;
 552 }
 553
 554 static int
 555 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 556                         struct nfs4_state *state, long *timeout)
 557 {
 558         struct nfs4_exception exception = {
 559                 .state = state,
 560         };
 561
 562         if (task->tk_status >= 0)
 563                 return 0;
 564         if (timeout)
 565                 exception.timeout = *timeout;
 566         task->tk_status = nfs4_async_handle_exception(task, server,
 567                         task->tk_status,
 568                         &exception);
 569         if (exception.delay && timeout)
 570                 *timeout = exception.timeout;
 571         if (exception.retry)
 572                 return -EAGAIN;
 573         return 0;
 574 }
 575
 576 /*
 577  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 578  * or 'false' otherwise.
 579  */
 580 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 581 {
 582         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 583         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 584 }
 585
 586 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 587 {
 588         spin_lock(&clp->cl_lock);
 589         if (time_before(clp->cl_last_renewal,timestamp))
 590                 clp->cl_last_renewal = timestamp;
 591         spin_unlock(&clp->cl_lock);
 592 }
 593
 594 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 595 {
 596         struct nfs_client *clp = server->nfs_client;
 597
 598         if (!nfs4_has_session(clp))
 599                 do_renew_lease(clp, timestamp);
 600 }
 601
 602 struct nfs4_call_sync_data {
 603         const struct nfs_server *seq_server;
 604         struct nfs4_sequence_args *seq_args;
 605         struct nfs4_sequence_res *seq_res;
 606 };
 607
 608 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 609                         struct nfs4_sequence_res *res, int cache_reply)
 610 {
 611         args->sa_slot = NULL;
 612         args->sa_cache_this = cache_reply;
 613         args->sa_privileged = 0;
 614
 615         res->sr_slot = NULL;
 616 }
 617
 618 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 619 {
 620         args->sa_privileged = 1;
 621 }
 622
 623 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 624 {
 625         struct nfs4_slot *slot = res->sr_slot;
 626         struct nfs4_slot_table *tbl;
 627
 628         tbl = slot->table;
 629         spin_lock(&tbl->slot_tbl_lock);
 630         if (!nfs41_wake_and_assign_slot(tbl, slot))
 631                 nfs4_free_slot(tbl, slot);
 632         spin_unlock(&tbl->slot_tbl_lock);
 633
 634         res->sr_slot = NULL;
 635 }
 636
 637 static int nfs40_sequence_done(struct rpc_task *task,
 638                                struct nfs4_sequence_res *res)
 639 {
 640         if (res->sr_slot != NULL)
 641                 nfs40_sequence_free_slot(res);
 642         return 1;
 643 }
 644
 645 #if defined(CONFIG_NFS_V4_1)
 646
 647 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 648 {
 649         struct nfs4_session *session;
 650         struct nfs4_slot_table *tbl;
 651         struct nfs4_slot *slot = res->sr_slot;
 652         bool send_new_highest_used_slotid = false;
 653
 654         tbl = slot->table;
 655         session = tbl->session;
 656
 657         /* Bump the slot sequence number */
 658         if (slot->seq_done)
 659                 slot->seq_nr++;
 660         slot->seq_done = 0;
 661
 662         spin_lock(&tbl->slot_tbl_lock);
 663         /* Be nice to the server: try to ensure that the last transmitted
 664          * value for highest_user_slotid <= target_highest_slotid
 665          */
 666         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 667                 send_new_highest_used_slotid = true;
 668
 669         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 670                 send_new_highest_used_slotid = false;
 671                 goto out_unlock;
 672         }
 673         nfs4_free_slot(tbl, slot);
 674
 675         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 676                 send_new_highest_used_slotid = false;
 677 out_unlock:
 678         spin_unlock(&tbl->slot_tbl_lock);
 679         res->sr_slot = NULL;
 680         if (send_new_highest_used_slotid)
 681                 nfs41_notify_server(session->clp);
 682         if (waitqueue_active(&tbl->slot_waitq))
 683                 wake_up_all(&tbl->slot_waitq);
 684 }
 685
 686 static int nfs41_sequence_process(struct rpc_task *task,
 687                 struct nfs4_sequence_res *res)
 688 {
 689         struct nfs4_session *session;
 690         struct nfs4_slot *slot = res->sr_slot;
 691         struct nfs_client *clp;
 692         bool interrupted = false;
 693         int ret = 1;
 694
 695         if (slot == NULL)
 696                 goto out_noaction;
 697         /* don't increment the sequence number if the task wasn't sent */
 698         if (!RPC_WAS_SENT(task))
 699                 goto out;
 700
 701         session = slot->table->session;
 702
 703         if (slot->interrupted) {
 704                 if (res->sr_status != -NFS4ERR_DELAY)
 705                         slot->interrupted = 0;
 706                 interrupted = true;
 707         }
 708
 709         trace_nfs4_sequence_done(session, res);
 710         /* Check the SEQUENCE operation status */
 711         switch (res->sr_status) {
 712         case 0:
 713                 /* If previous op on slot was interrupted and we reused
 714                  * the seq# and got a reply from the cache, then retry
 715                  */
 716                 if (task->tk_status == -EREMOTEIO && interrupted) {
 717                         ++slot->seq_nr;
 718                         goto retry_nowait;
 719                 }
 720                 /* Update the slot's sequence and clientid lease timer */
 721                 slot->seq_done = 1;
 722                 clp = session->clp;
 723                 do_renew_lease(clp, res->sr_timestamp);
 724                 /* Check sequence flags */
 725                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 726                                 !!slot->privileged);
 727                 nfs41_update_target_slotid(slot->table, slot, res);
 728                 break;
 729         case 1:
 730                 /*
 731                  * sr_status remains 1 if an RPC level error occurred.
 732                  * The server may or may not have processed the sequence
 733                  * operation..
 734                  * Mark the slot as having hosted an interrupted RPC call.
 735                  */
 736                 slot->interrupted = 1;
 737                 goto out;
 738         case -NFS4ERR_DELAY:
 739                 /* The server detected a resend of the RPC call and
 740                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 741                  * of RFC5661.
 742                  */
 743                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 744                         __func__,
 745                         slot->slot_nr,
 746                         slot->seq_nr);
 747                 goto out_retry;
 748         case -NFS4ERR_BADSLOT:
 749                 /*
 750                  * The slot id we used was probably retired. Try again
 751                  * using a different slot id.
 752                  */
 753                 goto retry_nowait;
 754         case -NFS4ERR_SEQ_MISORDERED:
 755                 /*
 756                  * Was the last operation on this sequence interrupted?
 757                  * If so, retry after bumping the sequence number.
 758                  */
 759                 if (interrupted) {
 760                         ++slot->seq_nr;
 761                         goto retry_nowait;
 762                 }
 763                 /*
 764                  * Could this slot have been previously retired?
 765                  * If so, then the server may be expecting seq_nr = 1!
 766                  */
 767                 if (slot->seq_nr != 1) {
 768                         slot->seq_nr = 1;
 769                         goto retry_nowait;
 770                 }
 771                 break;
 772         case -NFS4ERR_SEQ_FALSE_RETRY:
 773                 ++slot->seq_nr;
 774                 goto retry_nowait;
 775         default:
 776                 /* Just update the slot sequence no. */
 777                 slot->seq_done = 1;
 778         }
 779 out:
 780         /* The session may be reset by one of the error handlers. */
 781         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 782 out_noaction:
 783         return ret;
 784 retry_nowait:
 785         if (rpc_restart_call_prepare(task)) {
 786                 nfs41_sequence_free_slot(res);
 787                 task->tk_status = 0;
 788                 ret = 0;
 789         }
 790         goto out;
 791 out_retry:
 792         if (!rpc_restart_call(task))
 793                 goto out;
 794         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 795         return 0;
 796 }
 797
 798 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 799 {
 800         if (!nfs41_sequence_process(task, res))
 801                 return 0;
 802         if (res->sr_slot != NULL)
 803                 nfs41_sequence_free_slot(res);
 804         return 1;
 805
 806 }
 807 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 808
 809 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 810 {
 811         if (res->sr_slot == NULL)
 812                 return 1;
 813         if (res->sr_slot->table->session != NULL)
 814                 return nfs41_sequence_process(task, res);
 815         return nfs40_sequence_done(task, res);
 816 }
 817
 818 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 819 {
 820         if (res->sr_slot != NULL) {
 821                 if (res->sr_slot->table->session != NULL)
 822                         nfs41_sequence_free_slot(res);
 823                 else
 824                         nfs40_sequence_free_slot(res);
 825         }
 826 }
 827
 828 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 829 {
 830         if (res->sr_slot == NULL)
 831                 return 1;
 832         if (!res->sr_slot->table->session)
 833                 return nfs40_sequence_done(task, res);
 834         return nfs41_sequence_done(task, res);
 835 }
 836 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 837
 838 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 839 {
 840         struct nfs4_call_sync_data *data = calldata;
 841
 842         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 843
 844         nfs4_setup_sequence(data->seq_server->nfs_client,
 845                             data->seq_args, data->seq_res, task);
 846 }
 847
 848 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 849 {
 850         struct nfs4_call_sync_data *data = calldata;
 851
 852         nfs41_sequence_done(task, data->seq_res);
 853 }
 854
 855 static const struct rpc_call_ops nfs41_call_sync_ops = {
 856         .rpc_call_prepare = nfs41_call_sync_prepare,
 857         .rpc_call_done = nfs41_call_sync_done,
 858 };
 859
 860 #else   /* !CONFIG_NFS_V4_1 */
 861
 862 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 863 {
 864         return nfs40_sequence_done(task, res);
 865 }
 866
 867 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 868 {
 869         if (res->sr_slot != NULL)
 870                 nfs40_sequence_free_slot(res);
 871 }
 872
 873 int nfs4_sequence_done(struct rpc_task *task,
 874                        struct nfs4_sequence_res *res)
 875 {
 876         return nfs40_sequence_done(task, res);
 877 }
 878 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 879
 880 #endif  /* !CONFIG_NFS_V4_1 */
 881
 882 int nfs4_setup_sequence(const struct nfs_client *client,
 883                         struct nfs4_sequence_args *args,
 884                         struct nfs4_sequence_res *res,
 885                         struct rpc_task *task)
 886 {
 887         struct nfs4_session *session = nfs4_get_session(client);
 888         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
 889         struct nfs4_slot *slot;
 890
 891         /* slot already allocated? */
 892         if (res->sr_slot != NULL)
 893                 goto out_start;
 894
 895         if (session) {
 896                 tbl = &session->fc_slot_table;
 897                 task->tk_timeout = 0;
 898         }
 899
 900         spin_lock(&tbl->slot_tbl_lock);
 901         /* The state manager will wait until the slot table is empty */
 902         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 903                 goto out_sleep;
 904
 905         slot = nfs4_alloc_slot(tbl);
 906         if (IS_ERR(slot)) {
 907                 /* Try again in 1/4 second */
 908                 if (slot == ERR_PTR(-ENOMEM))
 909                         task->tk_timeout = HZ >> 2;
 910                 goto out_sleep;
 911         }
 912         spin_unlock(&tbl->slot_tbl_lock);
 913
 914         slot->privileged = args->sa_privileged ? 1 : 0;
 915         args->sa_slot = slot;
 916
 917         res->sr_slot = slot;
 918         if (session) {
 919                 res->sr_timestamp = jiffies;
 920                 res->sr_status_flags = 0;
 921                 res->sr_status = 1;
 922         }
 923
 924         trace_nfs4_setup_sequence(session, args);
 925 out_start:
 926         rpc_call_start(task);
 927         return 0;
 928
 929 out_sleep:
 930         if (args->sa_privileged)
 931                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 932                                 NULL, RPC_PRIORITY_PRIVILEGED);
 933         else
 934                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 935         spin_unlock(&tbl->slot_tbl_lock);
 936         return -EAGAIN;
 937 }
 938 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
 939
 940 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
 941 {
 942         struct nfs4_call_sync_data *data = calldata;
 943         nfs4_setup_sequence(data->seq_server->nfs_client,
 944                                 data->seq_args, data->seq_res, task);
 945 }
 946
 947 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
 948 {
 949         struct nfs4_call_sync_data *data = calldata;
 950         nfs4_sequence_done(task, data->seq_res);
 951 }
 952
 953 static const struct rpc_call_ops nfs40_call_sync_ops = {
 954         .rpc_call_prepare = nfs40_call_sync_prepare,
 955         .rpc_call_done = nfs40_call_sync_done,
 956 };
 957
 958 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
 959                                    struct nfs_server *server,
 960                                    struct rpc_message *msg,
 961                                    struct nfs4_sequence_args *args,
 962                                    struct nfs4_sequence_res *res)
 963 {
 964         int ret;
 965         struct rpc_task *task;
 966         struct nfs_client *clp = server->nfs_client;
 967         struct nfs4_call_sync_data data = {
 968                 .seq_server = server,
 969                 .seq_args = args,
 970                 .seq_res = res,
 971         };
 972         struct rpc_task_setup task_setup = {
 973                 .rpc_client = clnt,
 974                 .rpc_message = msg,
 975                 .callback_ops = clp->cl_mvops->call_sync_ops,
 976                 .callback_data = &data
 977         };
 978
 979         task = rpc_run_task(&task_setup);
 980         if (IS_ERR(task))
 981                 ret = PTR_ERR(task);
 982         else {
 983                 ret = task->tk_status;
 984                 rpc_put_task(task);
 985         }
 986         return ret;
 987 }
 988
 989 int nfs4_call_sync(struct rpc_clnt *clnt,
 990                    struct nfs_server *server,
 991                    struct rpc_message *msg,
 992                    struct nfs4_sequence_args *args,
 993                    struct nfs4_sequence_res *res,
 994                    int cache_reply)
 995 {
 996         nfs4_init_sequence(args, res, cache_reply);
 997         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
 998 }
 999
1000 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1001                 unsigned long timestamp)
1002 {
1003         struct nfs_inode *nfsi = NFS_I(dir);
1004
1005         spin_lock(&dir->i_lock);
1006         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1007         if (cinfo->atomic && cinfo->before == dir->i_version) {
1008                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1009                 nfsi->attrtimeo_timestamp = jiffies;
1010         } else {
1011                 nfs_force_lookup_revalidate(dir);
1012                 if (cinfo->before != dir->i_version)
1013                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1014                                 NFS_INO_INVALID_ACL;
1015         }
1016         dir->i_version = cinfo->after;
1017         nfsi->read_cache_jiffies = timestamp;
1018         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1019         nfs_fscache_invalidate(dir);
1020         spin_unlock(&dir->i_lock);
1021 }
1022
1023 struct nfs4_opendata {
1024         struct kref kref;
1025         struct nfs_openargs o_arg;
1026         struct nfs_openres o_res;
1027         struct nfs_open_confirmargs c_arg;
1028         struct nfs_open_confirmres c_res;
1029         struct nfs4_string owner_name;
1030         struct nfs4_string group_name;
1031         struct nfs4_label *a_label;
1032         struct nfs_fattr f_attr;
1033         struct nfs4_label *f_label;
1034         struct dentry *dir;
1035         struct dentry *dentry;
1036         struct nfs4_state_owner *owner;
1037         struct nfs4_state *state;
1038         struct iattr attrs;
1039         unsigned long timestamp;
1040         bool rpc_done;
1041         bool file_created;
1042         bool is_recover;
1043         bool cancelled;
1044         int rpc_status;
1045 };
1046
1047 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1048                 int err, struct nfs4_exception *exception)
1049 {
1050         if (err != -EINVAL)
1051                 return false;
1052         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1053                 return false;
1054         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1055         exception->retry = 1;
1056         return true;
1057 }
1058
1059 static u32
1060 nfs4_map_atomic_open_share(struct nfs_server *server,
1061                 fmode_t fmode, int openflags)
1062 {
1063         u32 res = 0;
1064
1065         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1066         case FMODE_READ:
1067                 res = NFS4_SHARE_ACCESS_READ;
1068                 break;
1069         case FMODE_WRITE:
1070                 res = NFS4_SHARE_ACCESS_WRITE;
1071                 break;
1072         case FMODE_READ|FMODE_WRITE:
1073                 res = NFS4_SHARE_ACCESS_BOTH;
1074         }
1075         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1076                 goto out;
1077         /* Want no delegation if we're using O_DIRECT */
1078         if (openflags & O_DIRECT)
1079                 res |= NFS4_SHARE_WANT_NO_DELEG;
1080 out:
1081         return res;
1082 }
1083
1084 static enum open_claim_type4
1085 nfs4_map_atomic_open_claim(struct nfs_server *server,
1086                 enum open_claim_type4 claim)
1087 {
1088         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1089                 return claim;
1090         switch (claim) {
1091         default:
1092                 return claim;
1093         case NFS4_OPEN_CLAIM_FH:
1094                 return NFS4_OPEN_CLAIM_NULL;
1095         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1096                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1097         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1098                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1099         }
1100 }
1101
1102 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1103 {
1104         p->o_res.f_attr = &p->f_attr;
1105         p->o_res.f_label = p->f_label;
1106         p->o_res.seqid = p->o_arg.seqid;
1107         p->c_res.seqid = p->c_arg.seqid;
1108         p->o_res.server = p->o_arg.server;
1109         p->o_res.access_request = p->o_arg.access;
1110         nfs_fattr_init(&p->f_attr);
1111         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1112 }
1113
1114 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1115                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1116                 const struct iattr *attrs,
1117                 struct nfs4_label *label,
1118                 enum open_claim_type4 claim,
1119                 gfp_t gfp_mask)
1120 {
1121         struct dentry *parent = dget_parent(dentry);
1122         struct inode *dir = d_inode(parent);
1123         struct nfs_server *server = NFS_SERVER(dir);
1124         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1125         struct nfs4_opendata *p;
1126
1127         p = kzalloc(sizeof(*p), gfp_mask);
1128         if (p == NULL)
1129                 goto err;
1130
1131         p->f_label = nfs4_label_alloc(server, gfp_mask);
1132         if (IS_ERR(p->f_label))
1133                 goto err_free_p;
1134
1135         p->a_label = nfs4_label_alloc(server, gfp_mask);
1136         if (IS_ERR(p->a_label))
1137                 goto err_free_f;
1138
1139         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1140         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1141         if (IS_ERR(p->o_arg.seqid))
1142                 goto err_free_label;
1143         nfs_sb_active(dentry->d_sb);
1144         p->dentry = dget(dentry);
1145         p->dir = parent;
1146         p->owner = sp;
1147         atomic_inc(&sp->so_count);
1148         p->o_arg.open_flags = flags;
1149         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1150         p->o_arg.umask = current_umask();
1151         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1152         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1153                         fmode, flags);
1154         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1155          * will return permission denied for all bits until close */
1156         if (!(flags & O_EXCL)) {
1157                 /* ask server to check for all possible rights as results
1158                  * are cached */
1159                 switch (p->o_arg.claim) {
1160                 default:
1161                         break;
1162                 case NFS4_OPEN_CLAIM_NULL:
1163                 case NFS4_OPEN_CLAIM_FH:
1164                         p->o_arg.access = NFS4_ACCESS_READ |
1165                                 NFS4_ACCESS_MODIFY |
1166                                 NFS4_ACCESS_EXTEND |
1167                                 NFS4_ACCESS_EXECUTE;
1168                 }
1169         }
1170         p->o_arg.clientid = server->nfs_client->cl_clientid;
1171         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1172         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1173         p->o_arg.name = &dentry->d_name;
1174         p->o_arg.server = server;
1175         p->o_arg.bitmask = nfs4_bitmask(server, label);
1176         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1177         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1178         switch (p->o_arg.claim) {
1179         case NFS4_OPEN_CLAIM_NULL:
1180         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1181         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1182                 p->o_arg.fh = NFS_FH(dir);
1183                 break;
1184         case NFS4_OPEN_CLAIM_PREVIOUS:
1185         case NFS4_OPEN_CLAIM_FH:
1186         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1187         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1188                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1189         }
1190         if (attrs != NULL && attrs->ia_valid != 0) {
1191                 __u32 verf[2];
1192
1193                 p->o_arg.u.attrs = &p->attrs;
1194                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1195
1196                 verf[0] = jiffies;
1197                 verf[1] = current->pid;
1198                 memcpy(p->o_arg.u.verifier.data, verf,
1199                                 sizeof(p->o_arg.u.verifier.data));
1200         }
1201         p->c_arg.fh = &p->o_res.fh;
1202         p->c_arg.stateid = &p->o_res.stateid;
1203         p->c_arg.seqid = p->o_arg.seqid;
1204         nfs4_init_opendata_res(p);
1205         kref_init(&p->kref);
1206         return p;
1207
1208 err_free_label:
1209         nfs4_label_free(p->a_label);
1210 err_free_f:
1211         nfs4_label_free(p->f_label);
1212 err_free_p:
1213         kfree(p);
1214 err:
1215         dput(parent);
1216         return NULL;
1217 }
1218
1219 static void nfs4_opendata_free(struct kref *kref)
1220 {
1221         struct nfs4_opendata *p = container_of(kref,
1222                         struct nfs4_opendata, kref);
1223         struct super_block *sb = p->dentry->d_sb;
1224
1225         nfs_free_seqid(p->o_arg.seqid);
1226         nfs4_sequence_free_slot(&p->o_res.seq_res);
1227         if (p->state != NULL)
1228                 nfs4_put_open_state(p->state);
1229         nfs4_put_state_owner(p->owner);
1230
1231         nfs4_label_free(p->a_label);
1232         nfs4_label_free(p->f_label);
1233
1234         dput(p->dir);
1235         dput(p->dentry);
1236         nfs_sb_deactive(sb);
1237         nfs_fattr_free_names(&p->f_attr);
1238         kfree(p->f_attr.mdsthreshold);
1239         kfree(p);
1240 }
1241
1242 static void nfs4_opendata_put(struct nfs4_opendata *p)
1243 {
1244         if (p != NULL)
1245                 kref_put(&p->kref, nfs4_opendata_free);
1246 }
1247
1248 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1249                 fmode_t fmode)
1250 {
1251         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1252         case FMODE_READ|FMODE_WRITE:
1253                 return state->n_rdwr != 0;
1254         case FMODE_WRITE:
1255                 return state->n_wronly != 0;
1256         case FMODE_READ:
1257                 return state->n_rdonly != 0;
1258         }
1259         WARN_ON_ONCE(1);
1260         return false;
1261 }
1262
1263 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1264 {
1265         int ret = 0;
1266
1267         if (open_mode & (O_EXCL|O_TRUNC))
1268                 goto out;
1269         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1270                 case FMODE_READ:
1271                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1272                                 && state->n_rdonly != 0;
1273                         break;
1274                 case FMODE_WRITE:
1275                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1276                                 && state->n_wronly != 0;
1277                         break;
1278                 case FMODE_READ|FMODE_WRITE:
1279                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1280                                 && state->n_rdwr != 0;
1281         }
1282 out:
1283         return ret;
1284 }
1285
1286 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1287                 enum open_claim_type4 claim)
1288 {
1289         if (delegation == NULL)
1290                 return 0;
1291         if ((delegation->type & fmode) != fmode)
1292                 return 0;
1293         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1294                 return 0;
1295         switch (claim) {
1296         case NFS4_OPEN_CLAIM_NULL:
1297         case NFS4_OPEN_CLAIM_FH:
1298                 break;
1299         case NFS4_OPEN_CLAIM_PREVIOUS:
1300                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1301                         break;
1302         default:
1303                 return 0;
1304         }
1305         nfs_mark_delegation_referenced(delegation);
1306         return 1;
1307 }
1308
1309 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1310 {
1311         switch (fmode) {
1312                 case FMODE_WRITE:
1313                         state->n_wronly++;
1314                         break;
1315                 case FMODE_READ:
1316                         state->n_rdonly++;
1317                         break;
1318                 case FMODE_READ|FMODE_WRITE:
1319                         state->n_rdwr++;
1320         }
1321         nfs4_state_set_mode_locked(state, state->state | fmode);
1322 }
1323
1324 #ifdef CONFIG_NFS_V4_1
1325 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1326 {
1327         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1328                 return true;
1329         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1330                 return true;
1331         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1332                 return true;
1333         return false;
1334 }
1335 #endif /* CONFIG_NFS_V4_1 */
1336
1337 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1338 {
1339         struct nfs_client *clp = state->owner->so_server->nfs_client;
1340         bool need_recover = false;
1341
1342         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1343                 need_recover = true;
1344         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1345                 need_recover = true;
1346         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1347                 need_recover = true;
1348         if (need_recover)
1349                 nfs4_state_mark_reclaim_nograce(clp, state);
1350 }
1351
1352 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1353                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1354 {
1355         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1356                 return true;
1357         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1358                 nfs4_stateid_copy(freeme, &state->open_stateid);
1359                 nfs_test_and_clear_all_open_stateid(state);
1360                 return true;
1361         }
1362         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1363                 return true;
1364         return false;
1365 }
1366
1367 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1368 {
1369         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1370                 return;
1371         if (state->n_wronly)
1372                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1373         if (state->n_rdonly)
1374                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1375         if (state->n_rdwr)
1376                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1377         set_bit(NFS_OPEN_STATE, &state->flags);
1378 }
1379
1380 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1381                 nfs4_stateid *stateid, fmode_t fmode)
1382 {
1383         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1384         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1385         case FMODE_WRITE:
1386                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1387                 break;
1388         case FMODE_READ:
1389                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1390                 break;
1391         case 0:
1392                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1393                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1394                 clear_bit(NFS_OPEN_STATE, &state->flags);
1395         }
1396         if (stateid == NULL)
1397                 return;
1398         /* Handle OPEN+OPEN_DOWNGRADE races */
1399         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1400             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1401                 nfs_resync_open_stateid_locked(state);
1402                 return;
1403         }
1404         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1405                 nfs4_stateid_copy(&state->stateid, stateid);
1406         nfs4_stateid_copy(&state->open_stateid, stateid);
1407 }
1408
1409 static void nfs_clear_open_stateid(struct nfs4_state *state,
1410         nfs4_stateid *arg_stateid,
1411         nfs4_stateid *stateid, fmode_t fmode)
1412 {
1413         write_seqlock(&state->seqlock);
1414         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1415         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1416                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1417         write_sequnlock(&state->seqlock);
1418         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1419                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1420 }
1421
1422 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1423                 const nfs4_stateid *stateid, fmode_t fmode,
1424                 nfs4_stateid *freeme)
1425 {
1426         switch (fmode) {
1427                 case FMODE_READ:
1428                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1429                         break;
1430                 case FMODE_WRITE:
1431                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1432                         break;
1433                 case FMODE_READ|FMODE_WRITE:
1434                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1435         }
1436         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1437                 return;
1438         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1439                 nfs4_stateid_copy(&state->stateid, stateid);
1440         nfs4_stateid_copy(&state->open_stateid, stateid);
1441 }
1442
1443 static void __update_open_stateid(struct nfs4_state *state,
1444                 const nfs4_stateid *open_stateid,
1445                 const nfs4_stateid *deleg_stateid,
1446                 fmode_t fmode,
1447                 nfs4_stateid *freeme)
1448 {
1449         /*
1450          * Protect the call to nfs4_state_set_mode_locked and
1451          * serialise the stateid update
1452          */
1453         spin_lock(&state->owner->so_lock);
1454         write_seqlock(&state->seqlock);
1455         if (deleg_stateid != NULL) {
1456                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1457                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1458         }
1459         if (open_stateid != NULL)
1460                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1461         write_sequnlock(&state->seqlock);
1462         update_open_stateflags(state, fmode);
1463         spin_unlock(&state->owner->so_lock);
1464 }
1465
1466 static int update_open_stateid(struct nfs4_state *state,
1467                 const nfs4_stateid *open_stateid,
1468                 const nfs4_stateid *delegation,
1469                 fmode_t fmode)
1470 {
1471         struct nfs_server *server = NFS_SERVER(state->inode);
1472         struct nfs_client *clp = server->nfs_client;
1473         struct nfs_inode *nfsi = NFS_I(state->inode);
1474         struct nfs_delegation *deleg_cur;
1475         nfs4_stateid freeme = { };
1476         int ret = 0;
1477
1478         fmode &= (FMODE_READ|FMODE_WRITE);
1479
1480         rcu_read_lock();
1481         deleg_cur = rcu_dereference(nfsi->delegation);
1482         if (deleg_cur == NULL)
1483                 goto no_delegation;
1484
1485         spin_lock(&deleg_cur->lock);
1486         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1487            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1488             (deleg_cur->type & fmode) != fmode)
1489                 goto no_delegation_unlock;
1490
1491         if (delegation == NULL)
1492                 delegation = &deleg_cur->stateid;
1493         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1494                 goto no_delegation_unlock;
1495
1496         nfs_mark_delegation_referenced(deleg_cur);
1497         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1498                         fmode, &freeme);
1499         ret = 1;
1500 no_delegation_unlock:
1501         spin_unlock(&deleg_cur->lock);
1502 no_delegation:
1503         rcu_read_unlock();
1504
1505         if (!ret && open_stateid != NULL) {
1506                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1507                 ret = 1;
1508         }
1509         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1510                 nfs4_schedule_state_manager(clp);
1511         if (freeme.type != 0)
1512                 nfs4_test_and_free_stateid(server, &freeme,
1513                                 state->owner->so_cred);
1514
1515         return ret;
1516 }
1517
1518 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1519                 const nfs4_stateid *stateid)
1520 {
1521         struct nfs4_state *state = lsp->ls_state;
1522         bool ret = false;
1523
1524         spin_lock(&state->state_lock);
1525         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1526                 goto out_noupdate;
1527         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1528                 goto out_noupdate;
1529         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1530         ret = true;
1531 out_noupdate:
1532         spin_unlock(&state->state_lock);
1533         return ret;
1534 }
1535
1536 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1537 {
1538         struct nfs_delegation *delegation;
1539
1540         rcu_read_lock();
1541         delegation = rcu_dereference(NFS_I(inode)->delegation);
1542         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1543                 rcu_read_unlock();
1544                 return;
1545         }
1546         rcu_read_unlock();
1547         nfs4_inode_return_delegation(inode);
1548 }
1549
1550 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1551 {
1552         struct nfs4_state *state = opendata->state;
1553         struct nfs_inode *nfsi = NFS_I(state->inode);
1554         struct nfs_delegation *delegation;
1555         int open_mode = opendata->o_arg.open_flags;
1556         fmode_t fmode = opendata->o_arg.fmode;
1557         enum open_claim_type4 claim = opendata->o_arg.claim;
1558         nfs4_stateid stateid;
1559         int ret = -EAGAIN;
1560
1561         for (;;) {
1562                 spin_lock(&state->owner->so_lock);
1563                 if (can_open_cached(state, fmode, open_mode)) {
1564                         update_open_stateflags(state, fmode);
1565                         spin_unlock(&state->owner->so_lock);
1566                         goto out_return_state;
1567                 }
1568                 spin_unlock(&state->owner->so_lock);
1569                 rcu_read_lock();
1570                 delegation = rcu_dereference(nfsi->delegation);
1571                 if (!can_open_delegated(delegation, fmode, claim)) {
1572                         rcu_read_unlock();
1573                         break;
1574                 }
1575                 /* Save the delegation */
1576                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1577                 rcu_read_unlock();
1578                 nfs_release_seqid(opendata->o_arg.seqid);
1579                 if (!opendata->is_recover) {
1580                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1581                         if (ret != 0)
1582                                 goto out;
1583                 }
1584                 ret = -EAGAIN;
1585
1586                 /* Try to update the stateid using the delegation */
1587                 if (update_open_stateid(state, NULL, &stateid, fmode))
1588                         goto out_return_state;
1589         }
1590 out:
1591         return ERR_PTR(ret);
1592 out_return_state:
1593         atomic_inc(&state->count);
1594         return state;
1595 }
1596
1597 static void
1598 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1599 {
1600         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1601         struct nfs_delegation *delegation;
1602         int delegation_flags = 0;
1603
1604         rcu_read_lock();
1605         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1606         if (delegation)
1607                 delegation_flags = delegation->flags;
1608         rcu_read_unlock();
1609         switch (data->o_arg.claim) {
1610         default:
1611                 break;
1612         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1613         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1614                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1615                                    "returning a delegation for "
1616                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1617                                    clp->cl_hostname);
1618                 return;
1619         }
1620         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1621                 nfs_inode_set_delegation(state->inode,
1622                                          data->owner->so_cred,
1623                                          &data->o_res);
1624         else
1625                 nfs_inode_reclaim_delegation(state->inode,
1626                                              data->owner->so_cred,
1627                                              &data->o_res);
1628 }
1629
1630 /*
1631  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1632  * and update the nfs4_state.
1633  */
1634 static struct nfs4_state *
1635 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1636 {
1637         struct inode *inode = data->state->inode;
1638         struct nfs4_state *state = data->state;
1639         int ret;
1640
1641         if (!data->rpc_done) {
1642                 if (data->rpc_status)
1643                         return ERR_PTR(data->rpc_status);
1644                 /* cached opens have already been processed */
1645                 goto update;
1646         }
1647
1648         ret = nfs_refresh_inode(inode, &data->f_attr);
1649         if (ret)
1650                 return ERR_PTR(ret);
1651
1652         if (data->o_res.delegation_type != 0)
1653                 nfs4_opendata_check_deleg(data, state);
1654 update:
1655         update_open_stateid(state, &data->o_res.stateid, NULL,
1656                             data->o_arg.fmode);
1657         atomic_inc(&state->count);
1658
1659         return state;
1660 }
1661
1662 static struct nfs4_state *
1663 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1664 {
1665         struct inode *inode;
1666         struct nfs4_state *state = NULL;
1667         int ret;
1668
1669         if (!data->rpc_done) {
1670                 state = nfs4_try_open_cached(data);
1671                 trace_nfs4_cached_open(data->state);
1672                 goto out;
1673         }
1674
1675         ret = -EAGAIN;
1676         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1677                 goto err;
1678         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1679         ret = PTR_ERR(inode);
1680         if (IS_ERR(inode))
1681                 goto err;
1682         ret = -ENOMEM;
1683         state = nfs4_get_open_state(inode, data->owner);
1684         if (state == NULL)
1685                 goto err_put_inode;
1686         if (data->o_res.delegation_type != 0)
1687                 nfs4_opendata_check_deleg(data, state);
1688         update_open_stateid(state, &data->o_res.stateid, NULL,
1689                         data->o_arg.fmode);
1690         iput(inode);
1691 out:
1692         nfs_release_seqid(data->o_arg.seqid);
1693         return state;
1694 err_put_inode:
1695         iput(inode);
1696 err:
1697         return ERR_PTR(ret);
1698 }
1699
1700 static struct nfs4_state *
1701 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1702 {
1703         struct nfs4_state *ret;
1704
1705         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1706                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1707         else
1708                 ret = _nfs4_opendata_to_nfs4_state(data);
1709         nfs4_sequence_free_slot(&data->o_res.seq_res);
1710         return ret;
1711 }
1712
1713 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1714 {
1715         struct nfs_inode *nfsi = NFS_I(state->inode);
1716         struct nfs_open_context *ctx;
1717
1718         spin_lock(&state->inode->i_lock);
1719         list_for_each_entry(ctx, &nfsi->open_files, list) {
1720                 if (ctx->state != state)
1721                         continue;
1722                 get_nfs_open_context(ctx);
1723                 spin_unlock(&state->inode->i_lock);
1724                 return ctx;
1725         }
1726         spin_unlock(&state->inode->i_lock);
1727         return ERR_PTR(-ENOENT);
1728 }
1729
1730 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1731                 struct nfs4_state *state, enum open_claim_type4 claim)
1732 {
1733         struct nfs4_opendata *opendata;
1734
1735         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1736                         NULL, NULL, claim, GFP_NOFS);
1737         if (opendata == NULL)
1738                 return ERR_PTR(-ENOMEM);
1739         opendata->state = state;
1740         atomic_inc(&state->count);
1741         return opendata;
1742 }
1743
1744 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1745                 fmode_t fmode)
1746 {
1747         struct nfs4_state *newstate;
1748         int ret;
1749
1750         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1751                 return 0;
1752         opendata->o_arg.open_flags = 0;
1753         opendata->o_arg.fmode = fmode;
1754         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1755                         NFS_SB(opendata->dentry->d_sb),
1756                         fmode, 0);
1757         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1758         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1759         nfs4_init_opendata_res(opendata);
1760         ret = _nfs4_recover_proc_open(opendata);
1761         if (ret != 0)
1762                 return ret;
1763         newstate = nfs4_opendata_to_nfs4_state(opendata);
1764         if (IS_ERR(newstate))
1765                 return PTR_ERR(newstate);
1766         if (newstate != opendata->state)
1767                 ret = -ESTALE;
1768         nfs4_close_state(newstate, fmode);
1769         return ret;
1770 }
1771
1772 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1773 {
1774         int ret;
1775
1776         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1777         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1778         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1779         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1780         /* memory barrier prior to reading state->n_* */
1781         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1782         clear_bit(NFS_OPEN_STATE, &state->flags);
1783         smp_rmb();
1784         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1785         if (ret != 0)
1786                 return ret;
1787         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1788         if (ret != 0)
1789                 return ret;
1790         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1791         if (ret != 0)
1792                 return ret;
1793         /*
1794          * We may have performed cached opens for all three recoveries.
1795          * Check if we need to update the current stateid.
1796          */
1797         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1798             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1799                 write_seqlock(&state->seqlock);
1800                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1801                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1802                 write_sequnlock(&state->seqlock);
1803         }
1804         return 0;
1805 }
1806
1807 /*
1808  * OPEN_RECLAIM:
1809  *      reclaim state on the server after a reboot.
1810  */
1811 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1812 {
1813         struct nfs_delegation *delegation;
1814         struct nfs4_opendata *opendata;
1815         fmode_t delegation_type = 0;
1816         int status;
1817
1818         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1819                         NFS4_OPEN_CLAIM_PREVIOUS);
1820         if (IS_ERR(opendata))
1821                 return PTR_ERR(opendata);
1822         rcu_read_lock();
1823         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1824         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1825                 delegation_type = delegation->type;
1826         rcu_read_unlock();
1827         opendata->o_arg.u.delegation_type = delegation_type;
1828         status = nfs4_open_recover(opendata, state);
1829         nfs4_opendata_put(opendata);
1830         return status;
1831 }
1832
1833 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1834 {
1835         struct nfs_server *server = NFS_SERVER(state->inode);
1836         struct nfs4_exception exception = { };
1837         int err;
1838         do {
1839                 err = _nfs4_do_open_reclaim(ctx, state);
1840                 trace_nfs4_open_reclaim(ctx, 0, err);
1841                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1842                         continue;
1843                 if (err != -NFS4ERR_DELAY)
1844                         break;
1845                 nfs4_handle_exception(server, err, &exception);
1846         } while (exception.retry);
1847         return err;
1848 }
1849
1850 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1851 {
1852         struct nfs_open_context *ctx;
1853         int ret;
1854
1855         ctx = nfs4_state_find_open_context(state);
1856         if (IS_ERR(ctx))
1857                 return -EAGAIN;
1858         ret = nfs4_do_open_reclaim(ctx, state);
1859         put_nfs_open_context(ctx);
1860         return ret;
1861 }
1862
1863 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1864 {
1865         switch (err) {
1866                 default:
1867                         printk(KERN_ERR "NFS: %s: unhandled error "
1868                                         "%d.\n", __func__, err);
1869                 case 0:
1870                 case -ENOENT:
1871                 case -EAGAIN:
1872                 case -ESTALE:
1873                         break;
1874                 case -NFS4ERR_BADSESSION:
1875                 case -NFS4ERR_BADSLOT:
1876                 case -NFS4ERR_BAD_HIGH_SLOT:
1877                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1878                 case -NFS4ERR_DEADSESSION:
1879                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1880                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1881                         return -EAGAIN;
1882                 case -NFS4ERR_STALE_CLIENTID:
1883                 case -NFS4ERR_STALE_STATEID:
1884                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1885                         /* Don't recall a delegation if it was lost */
1886                         nfs4_schedule_lease_recovery(server->nfs_client);
1887                         return -EAGAIN;
1888                 case -NFS4ERR_MOVED:
1889                         nfs4_schedule_migration_recovery(server);
1890                         return -EAGAIN;
1891                 case -NFS4ERR_LEASE_MOVED:
1892                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1893                         return -EAGAIN;
1894                 case -NFS4ERR_DELEG_REVOKED:
1895                 case -NFS4ERR_ADMIN_REVOKED:
1896                 case -NFS4ERR_EXPIRED:
1897                 case -NFS4ERR_BAD_STATEID:
1898                 case -NFS4ERR_OPENMODE:
1899                         nfs_inode_find_state_and_recover(state->inode,
1900                                         stateid);
1901                         nfs4_schedule_stateid_recovery(server, state);
1902                         return -EAGAIN;
1903                 case -NFS4ERR_DELAY:
1904                 case -NFS4ERR_GRACE:
1905                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1906                         ssleep(1);
1907                         return -EAGAIN;
1908                 case -ENOMEM:
1909                 case -NFS4ERR_DENIED:
1910                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1911                         return 0;
1912         }
1913         return err;
1914 }
1915
1916 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1917                 struct nfs4_state *state, const nfs4_stateid *stateid,
1918                 fmode_t type)
1919 {
1920         struct nfs_server *server = NFS_SERVER(state->inode);
1921         struct nfs4_opendata *opendata;
1922         int err = 0;
1923
1924         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1925                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1926         if (IS_ERR(opendata))
1927                 return PTR_ERR(opendata);
1928         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1929         write_seqlock(&state->seqlock);
1930         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1931         write_sequnlock(&state->seqlock);
1932         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1933         switch (type & (FMODE_READ|FMODE_WRITE)) {
1934         case FMODE_READ|FMODE_WRITE:
1935         case FMODE_WRITE:
1936                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1937                 if (err)
1938                         break;
1939                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1940                 if (err)
1941                         break;
1942         case FMODE_READ:
1943                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1944         }
1945         nfs4_opendata_put(opendata);
1946         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1947 }
1948
1949 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1950 {
1951         struct nfs4_opendata *data = calldata;
1952
1953         nfs4_setup_sequence(data->o_arg.server->nfs_client,
1954                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
1955 }
1956
1957 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1958 {
1959         struct nfs4_opendata *data = calldata;
1960
1961         nfs40_sequence_done(task, &data->c_res.seq_res);
1962
1963         data->rpc_status = task->tk_status;
1964         if (data->rpc_status == 0) {
1965                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1966                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1967                 renew_lease(data->o_res.server, data->timestamp);
1968                 data->rpc_done = true;
1969         }
1970 }
1971
1972 static void nfs4_open_confirm_release(void *calldata)
1973 {
1974         struct nfs4_opendata *data = calldata;
1975         struct nfs4_state *state = NULL;
1976
1977         /* If this request hasn't been cancelled, do nothing */
1978         if (!data->cancelled)
1979                 goto out_free;
1980         /* In case of error, no cleanup! */
1981         if (!data->rpc_done)
1982                 goto out_free;
1983         state = nfs4_opendata_to_nfs4_state(data);
1984         if (!IS_ERR(state))
1985                 nfs4_close_state(state, data->o_arg.fmode);
1986 out_free:
1987         nfs4_opendata_put(data);
1988 }
1989
1990 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1991         .rpc_call_prepare = nfs4_open_confirm_prepare,
1992         .rpc_call_done = nfs4_open_confirm_done,
1993         .rpc_release = nfs4_open_confirm_release,
1994 };
1995
1996 /*
1997  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1998  */
1999 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2000 {
2001         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2002         struct rpc_task *task;
2003         struct  rpc_message msg = {
2004                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2005                 .rpc_argp = &data->c_arg,
2006                 .rpc_resp = &data->c_res,
2007                 .rpc_cred = data->owner->so_cred,
2008         };
2009         struct rpc_task_setup task_setup_data = {
2010                 .rpc_client = server->client,
2011                 .rpc_message = &msg,
2012                 .callback_ops = &nfs4_open_confirm_ops,
2013                 .callback_data = data,
2014                 .workqueue = nfsiod_workqueue,
2015                 .flags = RPC_TASK_ASYNC,
2016         };
2017         int status;
2018
2019         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2020         kref_get(&data->kref);
2021         data->rpc_done = false;
2022         data->rpc_status = 0;
2023         data->timestamp = jiffies;
2024         if (data->is_recover)
2025                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2026         task = rpc_run_task(&task_setup_data);
2027         if (IS_ERR(task))
2028                 return PTR_ERR(task);
2029         status = rpc_wait_for_completion_task(task);
2030         if (status != 0) {
2031                 data->cancelled = true;
2032                 smp_wmb();
2033         } else
2034                 status = data->rpc_status;
2035         rpc_put_task(task);
2036         return status;
2037 }
2038
2039 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2040 {
2041         struct nfs4_opendata *data = calldata;
2042         struct nfs4_state_owner *sp = data->owner;
2043         struct nfs_client *clp = sp->so_server->nfs_client;
2044         enum open_claim_type4 claim = data->o_arg.claim;
2045
2046         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2047                 goto out_wait;
2048         /*
2049          * Check if we still need to send an OPEN call, or if we can use
2050          * a delegation instead.
2051          */
2052         if (data->state != NULL) {
2053                 struct nfs_delegation *delegation;
2054
2055                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2056                         goto out_no_action;
2057                 rcu_read_lock();
2058                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2059                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2060                         goto unlock_no_action;
2061                 rcu_read_unlock();
2062         }
2063         /* Update client id. */
2064         data->o_arg.clientid = clp->cl_clientid;
2065         switch (claim) {
2066         default:
2067                 break;
2068         case NFS4_OPEN_CLAIM_PREVIOUS:
2069         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2070         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2071                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2072         case NFS4_OPEN_CLAIM_FH:
2073                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2074                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2075         }
2076         data->timestamp = jiffies;
2077         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2078                                 &data->o_arg.seq_args,
2079                                 &data->o_res.seq_res,
2080                                 task) != 0)
2081                 nfs_release_seqid(data->o_arg.seqid);
2082
2083         /* Set the create mode (note dependency on the session type) */
2084         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2085         if (data->o_arg.open_flags & O_EXCL) {
2086                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2087                 if (nfs4_has_persistent_session(clp))
2088                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2089                 else if (clp->cl_mvops->minor_version > 0)
2090                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2091         }
2092         return;
2093 unlock_no_action:
2094         trace_nfs4_cached_open(data->state);
2095         rcu_read_unlock();
2096 out_no_action:
2097         task->tk_action = NULL;
2098 out_wait:
2099         nfs4_sequence_done(task, &data->o_res.seq_res);
2100 }
2101
2102 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2103 {
2104         struct nfs4_opendata *data = calldata;
2105
2106         data->rpc_status = task->tk_status;
2107
2108         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2109                 return;
2110
2111         if (task->tk_status == 0) {
2112                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2113                         switch (data->o_res.f_attr->mode & S_IFMT) {
2114                         case S_IFREG:
2115                                 break;
2116                         case S_IFLNK:
2117                                 data->rpc_status = -ELOOP;
2118                                 break;
2119                         case S_IFDIR:
2120                                 data->rpc_status = -EISDIR;
2121                                 break;
2122                         default:
2123                                 data->rpc_status = -ENOTDIR;
2124                         }
2125                 }
2126                 renew_lease(data->o_res.server, data->timestamp);
2127                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2128                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2129         }
2130         data->rpc_done = true;
2131 }
2132
2133 static void nfs4_open_release(void *calldata)
2134 {
2135         struct nfs4_opendata *data = calldata;
2136         struct nfs4_state *state = NULL;
2137
2138         /* If this request hasn't been cancelled, do nothing */
2139         if (!data->cancelled)
2140                 goto out_free;
2141         /* In case of error, no cleanup! */
2142         if (data->rpc_status != 0 || !data->rpc_done)
2143                 goto out_free;
2144         /* In case we need an open_confirm, no cleanup! */
2145         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2146                 goto out_free;
2147         state = nfs4_opendata_to_nfs4_state(data);
2148         if (!IS_ERR(state))
2149                 nfs4_close_state(state, data->o_arg.fmode);
2150 out_free:
2151         nfs4_opendata_put(data);
2152 }
2153
2154 static const struct rpc_call_ops nfs4_open_ops = {
2155         .rpc_call_prepare = nfs4_open_prepare,
2156         .rpc_call_done = nfs4_open_done,
2157         .rpc_release = nfs4_open_release,
2158 };
2159
2160 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2161 {
2162         struct inode *dir = d_inode(data->dir);
2163         struct nfs_server *server = NFS_SERVER(dir);
2164         struct nfs_openargs *o_arg = &data->o_arg;
2165         struct nfs_openres *o_res = &data->o_res;
2166         struct rpc_task *task;
2167         struct rpc_message msg = {
2168                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2169                 .rpc_argp = o_arg,
2170                 .rpc_resp = o_res,
2171                 .rpc_cred = data->owner->so_cred,
2172         };
2173         struct rpc_task_setup task_setup_data = {
2174                 .rpc_client = server->client,
2175                 .rpc_message = &msg,
2176                 .callback_ops = &nfs4_open_ops,
2177                 .callback_data = data,
2178                 .workqueue = nfsiod_workqueue,
2179                 .flags = RPC_TASK_ASYNC,
2180         };
2181         int status;
2182
2183         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2184         kref_get(&data->kref);
2185         data->rpc_done = false;
2186         data->rpc_status = 0;
2187         data->cancelled = false;
2188         data->is_recover = false;
2189         if (isrecover) {
2190                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2191                 data->is_recover = true;
2192         }
2193         task = rpc_run_task(&task_setup_data);
2194         if (IS_ERR(task))
2195                 return PTR_ERR(task);
2196         status = rpc_wait_for_completion_task(task);
2197         if (status != 0) {
2198                 data->cancelled = true;
2199                 smp_wmb();
2200         } else
2201                 status = data->rpc_status;
2202         rpc_put_task(task);
2203
2204         return status;
2205 }
2206
2207 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2208 {
2209         struct inode *dir = d_inode(data->dir);
2210         struct nfs_openres *o_res = &data->o_res;
2211         int status;
2212
2213         status = nfs4_run_open_task(data, 1);
2214         if (status != 0 || !data->rpc_done)
2215                 return status;
2216
2217         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2218
2219         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2220                 status = _nfs4_proc_open_confirm(data);
2221
2222         return status;
2223 }
2224
2225 /*
2226  * Additional permission checks in order to distinguish between an
2227  * open for read, and an open for execute. This works around the
2228  * fact that NFSv4 OPEN treats read and execute permissions as being
2229  * the same.
2230  * Note that in the non-execute case, we want to turn off permission
2231  * checking if we just created a new file (POSIX open() semantics).
2232  */
2233 static int nfs4_opendata_access(struct rpc_cred *cred,
2234                                 struct nfs4_opendata *opendata,
2235                                 struct nfs4_state *state, fmode_t fmode,
2236                                 int openflags)
2237 {
2238         struct nfs_access_entry cache;
2239         u32 mask, flags;
2240
2241         /* access call failed or for some reason the server doesn't
2242          * support any access modes -- defer access call until later */
2243         if (opendata->o_res.access_supported == 0)
2244                 return 0;
2245
2246         mask = 0;
2247         /*
2248          * Use openflags to check for exec, because fmode won't
2249          * always have FMODE_EXEC set when file open for exec.
2250          */
2251         if (openflags & __FMODE_EXEC) {
2252                 /* ONLY check for exec rights */
2253                 if (S_ISDIR(state->inode->i_mode))
2254                         mask = NFS4_ACCESS_LOOKUP;
2255                 else
2256                         mask = NFS4_ACCESS_EXECUTE;
2257         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2258                 mask = NFS4_ACCESS_READ;
2259
2260         cache.cred = cred;
2261         cache.jiffies = jiffies;
2262         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2263         nfs_access_add_cache(state->inode, &cache);
2264
2265         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2266         if ((mask & ~cache.mask & flags) == 0)
2267                 return 0;
2268
2269         return -EACCES;
2270 }
2271
2272 /*
2273  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2274  */
2275 static int _nfs4_proc_open(struct nfs4_opendata *data)
2276 {
2277         struct inode *dir = d_inode(data->dir);
2278         struct nfs_server *server = NFS_SERVER(dir);
2279         struct nfs_openargs *o_arg = &data->o_arg;
2280         struct nfs_openres *o_res = &data->o_res;
2281         int status;
2282
2283         status = nfs4_run_open_task(data, 0);
2284         if (!data->rpc_done)
2285                 return status;
2286         if (status != 0) {
2287                 if (status == -NFS4ERR_BADNAME &&
2288                                 !(o_arg->open_flags & O_CREAT))
2289                         return -ENOENT;
2290                 return status;
2291         }
2292
2293         nfs_fattr_map_and_free_names(server, &data->f_attr);
2294
2295         if (o_arg->open_flags & O_CREAT) {
2296                 if (o_arg->open_flags & O_EXCL)
2297                         data->file_created = true;
2298                 else if (o_res->cinfo.before != o_res->cinfo.after)
2299                         data->file_created = true;
2300                 if (data->file_created || dir->i_version != o_res->cinfo.after)
2301                         update_changeattr(dir, &o_res->cinfo,
2302                                         o_res->f_attr->time_start);
2303         }
2304         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2305                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2306         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2307                 status = _nfs4_proc_open_confirm(data);
2308                 if (status != 0)
2309                         return status;
2310         }
2311         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2312                 nfs4_sequence_free_slot(&o_res->seq_res);
2313                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2314         }
2315         return 0;
2316 }
2317
2318 /*
2319  * OPEN_EXPIRED:
2320  *      reclaim state on the server after a network partition.
2321  *      Assumes caller holds the appropriate lock
2322  */
2323 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2324 {
2325         struct nfs4_opendata *opendata;
2326         int ret;
2327
2328         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2329                         NFS4_OPEN_CLAIM_FH);
2330         if (IS_ERR(opendata))
2331                 return PTR_ERR(opendata);
2332         ret = nfs4_open_recover(opendata, state);
2333         if (ret == -ESTALE)
2334                 d_drop(ctx->dentry);
2335         nfs4_opendata_put(opendata);
2336         return ret;
2337 }
2338
2339 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2340 {
2341         struct nfs_server *server = NFS_SERVER(state->inode);
2342         struct nfs4_exception exception = { };
2343         int err;
2344
2345         do {
2346                 err = _nfs4_open_expired(ctx, state);
2347                 trace_nfs4_open_expired(ctx, 0, err);
2348                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2349                         continue;
2350                 switch (err) {
2351                 default:
2352                         goto out;
2353                 case -NFS4ERR_GRACE:
2354                 case -NFS4ERR_DELAY:
2355                         nfs4_handle_exception(server, err, &exception);
2356                         err = 0;
2357                 }
2358         } while (exception.retry);
2359 out:
2360         return err;
2361 }
2362
2363 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2364 {
2365         struct nfs_open_context *ctx;
2366         int ret;
2367
2368         ctx = nfs4_state_find_open_context(state);
2369         if (IS_ERR(ctx))
2370                 return -EAGAIN;
2371         ret = nfs4_do_open_expired(ctx, state);
2372         put_nfs_open_context(ctx);
2373         return ret;
2374 }
2375
2376 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2377                 const nfs4_stateid *stateid)
2378 {
2379         nfs_remove_bad_delegation(state->inode, stateid);
2380         write_seqlock(&state->seqlock);
2381         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2382         write_sequnlock(&state->seqlock);
2383         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2384 }
2385
2386 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2387 {
2388         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2389                 nfs_finish_clear_delegation_stateid(state, NULL);
2390 }
2391
2392 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2393 {
2394         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2395         nfs40_clear_delegation_stateid(state);
2396         return nfs4_open_expired(sp, state);
2397 }
2398
2399 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2400                 nfs4_stateid *stateid,
2401                 struct rpc_cred *cred)
2402 {
2403         return -NFS4ERR_BAD_STATEID;
2404 }
2405
2406 #if defined(CONFIG_NFS_V4_1)
2407 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2408                 nfs4_stateid *stateid,
2409                 struct rpc_cred *cred)
2410 {
2411         int status;
2412
2413         switch (stateid->type) {
2414         default:
2415                 break;
2416         case NFS4_INVALID_STATEID_TYPE:
2417         case NFS4_SPECIAL_STATEID_TYPE:
2418                 return -NFS4ERR_BAD_STATEID;
2419         case NFS4_REVOKED_STATEID_TYPE:
2420                 goto out_free;
2421         }
2422
2423         status = nfs41_test_stateid(server, stateid, cred);
2424         switch (status) {
2425         case -NFS4ERR_EXPIRED:
2426         case -NFS4ERR_ADMIN_REVOKED:
2427         case -NFS4ERR_DELEG_REVOKED:
2428                 break;
2429         default:
2430                 return status;
2431         }
2432 out_free:
2433         /* Ack the revoked state to the server */
2434         nfs41_free_stateid(server, stateid, cred, true);
2435         return -NFS4ERR_EXPIRED;
2436 }
2437
2438 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2439 {
2440         struct nfs_server *server = NFS_SERVER(state->inode);
2441         nfs4_stateid stateid;
2442         struct nfs_delegation *delegation;
2443         struct rpc_cred *cred;
2444         int status;
2445
2446         /* Get the delegation credential for use by test/free_stateid */
2447         rcu_read_lock();
2448         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2449         if (delegation == NULL) {
2450                 rcu_read_unlock();
2451                 return;
2452         }
2453
2454         nfs4_stateid_copy(&stateid, &delegation->stateid);
2455         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2456                 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2457                         &delegation->flags)) {
2458                 rcu_read_unlock();
2459                 nfs_finish_clear_delegation_stateid(state, &stateid);
2460                 return;
2461         }
2462
2463         cred = get_rpccred(delegation->cred);
2464         rcu_read_unlock();
2465         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2466         trace_nfs4_test_delegation_stateid(state, NULL, status);
2467         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2468                 nfs_finish_clear_delegation_stateid(state, &stateid);
2469
2470         put_rpccred(cred);
2471 }
2472
2473 /**
2474  * nfs41_check_expired_locks - possibly free a lock stateid
2475  *
2476  * @state: NFSv4 state for an inode
2477  *
2478  * Returns NFS_OK if recovery for this stateid is now finished.
2479  * Otherwise a negative NFS4ERR value is returned.
2480  */
2481 static int nfs41_check_expired_locks(struct nfs4_state *state)
2482 {
2483         int status, ret = NFS_OK;
2484         struct nfs4_lock_state *lsp, *prev = NULL;
2485         struct nfs_server *server = NFS_SERVER(state->inode);
2486
2487         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2488                 goto out;
2489
2490         spin_lock(&state->state_lock);
2491         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2492                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2493                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2494
2495                         atomic_inc(&lsp->ls_count);
2496                         spin_unlock(&state->state_lock);
2497
2498                         nfs4_put_lock_state(prev);
2499                         prev = lsp;
2500
2501                         status = nfs41_test_and_free_expired_stateid(server,
2502                                         &lsp->ls_stateid,
2503                                         cred);
2504                         trace_nfs4_test_lock_stateid(state, lsp, status);
2505                         if (status == -NFS4ERR_EXPIRED ||
2506                             status == -NFS4ERR_BAD_STATEID) {
2507                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2508                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2509                                 if (!recover_lost_locks)
2510                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2511                         } else if (status != NFS_OK) {
2512                                 ret = status;
2513                                 nfs4_put_lock_state(prev);
2514                                 goto out;
2515                         }
2516                         spin_lock(&state->state_lock);
2517                 }
2518         }
2519         spin_unlock(&state->state_lock);
2520         nfs4_put_lock_state(prev);
2521 out:
2522         return ret;
2523 }
2524
2525 /**
2526  * nfs41_check_open_stateid - possibly free an open stateid
2527  *
2528  * @state: NFSv4 state for an inode
2529  *
2530  * Returns NFS_OK if recovery for this stateid is now finished.
2531  * Otherwise a negative NFS4ERR value is returned.
2532  */
2533 static int nfs41_check_open_stateid(struct nfs4_state *state)
2534 {
2535         struct nfs_server *server = NFS_SERVER(state->inode);
2536         nfs4_stateid *stateid = &state->open_stateid;
2537         struct rpc_cred *cred = state->owner->so_cred;
2538         int status;
2539
2540         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2541                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2542                         if (nfs4_have_delegation(state->inode, state->state))
2543                                 return NFS_OK;
2544                         return -NFS4ERR_OPENMODE;
2545                 }
2546                 return -NFS4ERR_BAD_STATEID;
2547         }
2548         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2549         trace_nfs4_test_open_stateid(state, NULL, status);
2550         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2551                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2552                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2553                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2554                 clear_bit(NFS_OPEN_STATE, &state->flags);
2555                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2556                 return status;
2557         }
2558         if (nfs_open_stateid_recover_openmode(state))
2559                 return -NFS4ERR_OPENMODE;
2560         return NFS_OK;
2561 }
2562
2563 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2564 {
2565         int status;
2566
2567         nfs41_check_delegation_stateid(state);
2568         status = nfs41_check_expired_locks(state);
2569         if (status != NFS_OK)
2570                 return status;
2571         status = nfs41_check_open_stateid(state);
2572         if (status != NFS_OK)
2573                 status = nfs4_open_expired(sp, state);
2574         return status;
2575 }
2576 #endif
2577
2578 /*
2579  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2580  * fields corresponding to attributes that were used to store the verifier.
2581  * Make sure we clobber those fields in the later setattr call
2582  */
2583 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2584                                 struct iattr *sattr, struct nfs4_label **label)
2585 {
2586         const u32 *attrset = opendata->o_res.attrset;
2587
2588         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2589             !(sattr->ia_valid & ATTR_ATIME_SET))
2590                 sattr->ia_valid |= ATTR_ATIME;
2591
2592         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2593             !(sattr->ia_valid & ATTR_MTIME_SET))
2594                 sattr->ia_valid |= ATTR_MTIME;
2595
2596         /* Except MODE, it seems harmless of setting twice. */
2597         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2598                 (attrset[1] & FATTR4_WORD1_MODE ||
2599                  attrset[2] & FATTR4_WORD2_MODE_UMASK))
2600                 sattr->ia_valid &= ~ATTR_MODE;
2601
2602         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2603                 *label = NULL;
2604 }
2605
2606 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2607                 fmode_t fmode,
2608                 int flags,
2609                 struct nfs_open_context *ctx)
2610 {
2611         struct nfs4_state_owner *sp = opendata->owner;
2612         struct nfs_server *server = sp->so_server;
2613         struct dentry *dentry;
2614         struct nfs4_state *state;
2615         unsigned int seq;
2616         int ret;
2617
2618         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2619
2620         ret = _nfs4_proc_open(opendata);
2621         if (ret != 0)
2622                 goto out;
2623
2624         state = nfs4_opendata_to_nfs4_state(opendata);
2625         ret = PTR_ERR(state);
2626         if (IS_ERR(state))
2627                 goto out;
2628         ctx->state = state;
2629         if (server->caps & NFS_CAP_POSIX_LOCK)
2630                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2631         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2632                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2633
2634         dentry = opendata->dentry;
2635         if (d_really_is_negative(dentry)) {
2636                 struct dentry *alias;
2637                 d_drop(dentry);
2638                 alias = d_exact_alias(dentry, state->inode);
2639                 if (!alias)
2640                         alias = d_splice_alias(igrab(state->inode), dentry);
2641                 /* d_splice_alias() can't fail here - it's a non-directory */
2642                 if (alias) {
2643                         dput(ctx->dentry);
2644                         ctx->dentry = dentry = alias;
2645                 }
2646                 nfs_set_verifier(dentry,
2647                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2648         }
2649
2650         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2651         if (ret != 0)
2652                 goto out;
2653
2654         if (d_inode(dentry) == state->inode) {
2655                 nfs_inode_attach_open_context(ctx);
2656                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2657                         nfs4_schedule_stateid_recovery(server, state);
2658         }
2659 out:
2660         return ret;
2661 }
2662
2663 /*
2664  * Returns a referenced nfs4_state
2665  */
2666 static int _nfs4_do_open(struct inode *dir,
2667                         struct nfs_open_context *ctx,
2668                         int flags,
2669                         struct iattr *sattr,
2670                         struct nfs4_label *label,
2671                         int *opened)
2672 {
2673         struct nfs4_state_owner  *sp;
2674         struct nfs4_state     *state = NULL;
2675         struct nfs_server       *server = NFS_SERVER(dir);
2676         struct nfs4_opendata *opendata;
2677         struct dentry *dentry = ctx->dentry;
2678         struct rpc_cred *cred = ctx->cred;
2679         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2680         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2681         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2682         struct nfs4_label *olabel = NULL;
2683         int status;
2684
2685         /* Protect against reboot recovery conflicts */
2686         status = -ENOMEM;
2687         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2688         if (sp == NULL) {
2689                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2690                 goto out_err;
2691         }
2692         status = nfs4_client_recover_expired_lease(server->nfs_client);
2693         if (status != 0)
2694                 goto err_put_state_owner;
2695         if (d_really_is_positive(dentry))
2696                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2697         status = -ENOMEM;
2698         if (d_really_is_positive(dentry))
2699                 claim = NFS4_OPEN_CLAIM_FH;
2700         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2701                         label, claim, GFP_KERNEL);
2702         if (opendata == NULL)
2703                 goto err_put_state_owner;
2704
2705         if (label) {
2706                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2707                 if (IS_ERR(olabel)) {
2708                         status = PTR_ERR(olabel);
2709                         goto err_opendata_put;
2710                 }
2711         }
2712
2713         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2714                 if (!opendata->f_attr.mdsthreshold) {
2715                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2716                         if (!opendata->f_attr.mdsthreshold)
2717                                 goto err_free_label;
2718                 }
2719                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2720         }
2721         if (d_really_is_positive(dentry))
2722                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2723
2724         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2725         if (status != 0)
2726                 goto err_free_label;
2727         state = ctx->state;
2728
2729         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2730             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2731                 nfs4_exclusive_attrset(opendata, sattr, &label);
2732                 /*
2733                  * send create attributes which was not set by open
2734                  * with an extra setattr.
2735                  */
2736                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2737                         nfs_fattr_init(opendata->o_res.f_attr);
2738                         status = nfs4_do_setattr(state->inode, cred,
2739                                         opendata->o_res.f_attr, sattr,
2740                                         ctx, label, olabel);
2741                         if (status == 0) {
2742                                 nfs_setattr_update_inode(state->inode, sattr,
2743                                                 opendata->o_res.f_attr);
2744                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2745                         }
2746                 }
2747         }
2748         if (opened && opendata->file_created)
2749                 *opened |= FILE_CREATED;
2750
2751         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2752                 *ctx_th = opendata->f_attr.mdsthreshold;
2753                 opendata->f_attr.mdsthreshold = NULL;
2754         }
2755
2756         nfs4_label_free(olabel);
2757
2758         nfs4_opendata_put(opendata);
2759         nfs4_put_state_owner(sp);
2760         return 0;
2761 err_free_label:
2762         nfs4_label_free(olabel);
2763 err_opendata_put:
2764         nfs4_opendata_put(opendata);
2765 err_put_state_owner:
2766         nfs4_put_state_owner(sp);
2767 out_err:
2768         return status;
2769 }
2770
2771
2772 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2773                                         struct nfs_open_context *ctx,
2774                                         int flags,
2775                                         struct iattr *sattr,
2776                                         struct nfs4_label *label,
2777                                         int *opened)
2778 {
2779         struct nfs_server *server = NFS_SERVER(dir);
2780         struct nfs4_exception exception = { };
2781         struct nfs4_state *res;
2782         int status;
2783
2784         do {
2785                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2786                 res = ctx->state;
2787                 trace_nfs4_open_file(ctx, flags, status);
2788                 if (status == 0)
2789                         break;
2790                 /* NOTE: BAD_SEQID means the server and client disagree about the
2791                  * book-keeping w.r.t. state-changing operations
2792                  * (OPEN/CLOSE/LOCK/LOCKU...)
2793                  * It is actually a sign of a bug on the client or on the server.
2794                  *
2795                  * If we receive a BAD_SEQID error in the particular case of
2796                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2797                  * have unhashed the old state_owner for us, and that we can
2798                  * therefore safely retry using a new one. We should still warn
2799                  * the user though...
2800                  */
2801                 if (status == -NFS4ERR_BAD_SEQID) {
2802                         pr_warn_ratelimited("NFS: v4 server %s "
2803                                         " returned a bad sequence-id error!\n",
2804                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2805                         exception.retry = 1;
2806                         continue;
2807                 }
2808                 /*
2809                  * BAD_STATEID on OPEN means that the server cancelled our
2810                  * state before it received the OPEN_CONFIRM.
2811                  * Recover by retrying the request as per the discussion
2812                  * on Page 181 of RFC3530.
2813                  */
2814                 if (status == -NFS4ERR_BAD_STATEID) {
2815                         exception.retry = 1;
2816                         continue;
2817                 }
2818                 if (status == -EAGAIN) {
2819                         /* We must have found a delegation */
2820                         exception.retry = 1;
2821                         continue;
2822                 }
2823                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2824                         continue;
2825                 res = ERR_PTR(nfs4_handle_exception(server,
2826                                         status, &exception));
2827         } while (exception.retry);
2828         return res;
2829 }
2830
2831 static int _nfs4_do_setattr(struct inode *inode,
2832                             struct nfs_setattrargs *arg,
2833                             struct nfs_setattrres *res,
2834                             struct rpc_cred *cred,
2835                             struct nfs_open_context *ctx)
2836 {
2837         struct nfs_server *server = NFS_SERVER(inode);
2838         struct rpc_message msg = {
2839                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2840                 .rpc_argp       = arg,
2841                 .rpc_resp       = res,
2842                 .rpc_cred       = cred,
2843         };
2844         struct rpc_cred *delegation_cred = NULL;
2845         unsigned long timestamp = jiffies;
2846         fmode_t fmode;
2847         bool truncate;
2848         int status;
2849
2850         nfs_fattr_init(res->fattr);
2851
2852         /* Servers should only apply open mode checks for file size changes */
2853         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2854         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2855
2856         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2857                 /* Use that stateid */
2858         } else if (truncate && ctx != NULL) {
2859                 struct nfs_lock_context *l_ctx;
2860                 if (!nfs4_valid_open_stateid(ctx->state))
2861                         return -EBADF;
2862                 l_ctx = nfs_get_lock_context(ctx);
2863                 if (IS_ERR(l_ctx))
2864                         return PTR_ERR(l_ctx);
2865                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2866                                                 &arg->stateid, &delegation_cred);
2867                 nfs_put_lock_context(l_ctx);
2868                 if (status == -EIO)
2869                         return -EBADF;
2870         } else
2871                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2872         if (delegation_cred)
2873                 msg.rpc_cred = delegation_cred;
2874
2875         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2876
2877         put_rpccred(delegation_cred);
2878         if (status == 0 && ctx != NULL)
2879                 renew_lease(server, timestamp);
2880         trace_nfs4_setattr(inode, &arg->stateid, status);
2881         return status;
2882 }
2883
2884 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2885                            struct nfs_fattr *fattr, struct iattr *sattr,
2886                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2887                            struct nfs4_label *olabel)
2888 {
2889         struct nfs_server *server = NFS_SERVER(inode);
2890         struct nfs4_state *state = ctx ? ctx->state : NULL;
2891         struct nfs_setattrargs  arg = {
2892                 .fh             = NFS_FH(inode),
2893                 .iap            = sattr,
2894                 .server         = server,
2895                 .bitmask = server->attr_bitmask,
2896                 .label          = ilabel,
2897         };
2898         struct nfs_setattrres  res = {
2899                 .fattr          = fattr,
2900                 .label          = olabel,
2901                 .server         = server,
2902         };
2903         struct nfs4_exception exception = {
2904                 .state = state,
2905                 .inode = inode,
2906                 .stateid = &arg.stateid,
2907         };
2908         int err;
2909
2910         arg.bitmask = nfs4_bitmask(server, ilabel);
2911         if (ilabel)
2912                 arg.bitmask = nfs4_bitmask(server, olabel);
2913
2914         do {
2915                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
2916                 switch (err) {
2917                 case -NFS4ERR_OPENMODE:
2918                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2919                                 pr_warn_once("NFSv4: server %s is incorrectly "
2920                                                 "applying open mode checks to "
2921                                                 "a SETATTR that is not "
2922                                                 "changing file size.\n",
2923                                                 server->nfs_client->cl_hostname);
2924                         }
2925                         if (state && !(state->state & FMODE_WRITE)) {
2926                                 err = -EBADF;
2927                                 if (sattr->ia_valid & ATTR_OPEN)
2928                                         err = -EACCES;
2929                                 goto out;
2930                         }
2931                 }
2932                 err = nfs4_handle_exception(server, err, &exception);
2933         } while (exception.retry);
2934 out:
2935         return err;
2936 }
2937
2938 static bool
2939 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2940 {
2941         if (inode == NULL || !nfs_have_layout(inode))
2942                 return false;
2943
2944         return pnfs_wait_on_layoutreturn(inode, task);
2945 }
2946
2947 struct nfs4_closedata {
2948         struct inode *inode;
2949         struct nfs4_state *state;
2950         struct nfs_closeargs arg;
2951         struct nfs_closeres res;
2952         struct {
2953                 struct nfs4_layoutreturn_args arg;
2954                 struct nfs4_layoutreturn_res res;
2955                 struct nfs4_xdr_opaque_data ld_private;
2956                 u32 roc_barrier;
2957                 bool roc;
2958         } lr;
2959         struct nfs_fattr fattr;
2960         unsigned long timestamp;
2961 };
2962
2963 static void nfs4_free_closedata(void *data)
2964 {
2965         struct nfs4_closedata *calldata = data;
2966         struct nfs4_state_owner *sp = calldata->state->owner;
2967         struct super_block *sb = calldata->state->inode->i_sb;
2968
2969         if (calldata->lr.roc)
2970                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
2971                                 calldata->res.lr_ret);
2972         nfs4_put_open_state(calldata->state);
2973         nfs_free_seqid(calldata->arg.seqid);
2974         nfs4_put_state_owner(sp);
2975         nfs_sb_deactive(sb);
2976         kfree(calldata);
2977 }
2978
2979 static void nfs4_close_done(struct rpc_task *task, void *data)
2980 {
2981         struct nfs4_closedata *calldata = data;
2982         struct nfs4_state *state = calldata->state;
2983         struct nfs_server *server = NFS_SERVER(calldata->inode);
2984         nfs4_stateid *res_stateid = NULL;
2985
2986         dprintk("%s: begin!\n", __func__);
2987         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2988                 return;
2989         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2990
2991         /* Handle Layoutreturn errors */
2992         if (calldata->arg.lr_args && task->tk_status != 0) {
2993                 switch (calldata->res.lr_ret) {
2994                 default:
2995                         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
2996                         break;
2997                 case 0:
2998                         calldata->arg.lr_args = NULL;
2999                         calldata->res.lr_res = NULL;
3000                         break;
3001                 case -NFS4ERR_ADMIN_REVOKED:
3002                 case -NFS4ERR_DELEG_REVOKED:
3003                 case -NFS4ERR_EXPIRED:
3004                 case -NFS4ERR_BAD_STATEID:
3005                 case -NFS4ERR_OLD_STATEID:
3006                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3007                 case -NFS4ERR_WRONG_CRED:
3008                         calldata->arg.lr_args = NULL;
3009                         calldata->res.lr_res = NULL;
3010                         calldata->res.lr_ret = 0;
3011                         rpc_restart_call_prepare(task);
3012                         return;
3013                 }
3014         }
3015
3016         /* hmm. we are done with the inode, and in the process of freeing
3017          * the state_owner. we keep this around to process errors
3018          */
3019         switch (task->tk_status) {
3020                 case 0:
3021                         res_stateid = &calldata->res.stateid;
3022                         renew_lease(server, calldata->timestamp);
3023                         break;
3024                 case -NFS4ERR_ACCESS:
3025                         if (calldata->arg.bitmask != NULL) {
3026                                 calldata->arg.bitmask = NULL;
3027                                 calldata->res.fattr = NULL;
3028                                 task->tk_status = 0;
3029                                 rpc_restart_call_prepare(task);
3030                                 goto out_release;
3031
3032                         }
3033                         break;
3034                 case -NFS4ERR_ADMIN_REVOKED:
3035                 case -NFS4ERR_STALE_STATEID:
3036                 case -NFS4ERR_EXPIRED:
3037                         nfs4_free_revoked_stateid(server,
3038                                         &calldata->arg.stateid,
3039                                         task->tk_msg.rpc_cred);
3040                 case -NFS4ERR_OLD_STATEID:
3041                 case -NFS4ERR_BAD_STATEID:
3042                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3043                                                 &state->open_stateid)) {
3044                                 rpc_restart_call_prepare(task);
3045                                 goto out_release;
3046                         }
3047                         if (calldata->arg.fmode == 0)
3048                                 break;
3049                 default:
3050                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3051                                 rpc_restart_call_prepare(task);
3052                                 goto out_release;
3053                         }
3054         }
3055         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3056                         res_stateid, calldata->arg.fmode);
3057 out_release:
3058         nfs_release_seqid(calldata->arg.seqid);
3059         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3060         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3061 }
3062
3063 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3064 {
3065         struct nfs4_closedata *calldata = data;
3066         struct nfs4_state *state = calldata->state;
3067         struct inode *inode = calldata->inode;
3068         bool is_rdonly, is_wronly, is_rdwr;
3069         int call_close = 0;
3070
3071         dprintk("%s: begin!\n", __func__);
3072         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3073                 goto out_wait;
3074
3075         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3076         spin_lock(&state->owner->so_lock);
3077         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3078         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3079         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3080         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3081         /* Calculate the change in open mode */
3082         calldata->arg.fmode = 0;
3083         if (state->n_rdwr == 0) {
3084                 if (state->n_rdonly == 0)
3085                         call_close |= is_rdonly;
3086                 else if (is_rdonly)
3087                         calldata->arg.fmode |= FMODE_READ;
3088                 if (state->n_wronly == 0)
3089                         call_close |= is_wronly;
3090                 else if (is_wronly)
3091                         calldata->arg.fmode |= FMODE_WRITE;
3092                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3093                         call_close |= is_rdwr;
3094         } else if (is_rdwr)
3095                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3096
3097         if (!nfs4_valid_open_stateid(state) ||
3098             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3099                 call_close = 0;
3100         spin_unlock(&state->owner->so_lock);
3101
3102         if (!call_close) {
3103                 /* Note: exit _without_ calling nfs4_close_done */
3104                 goto out_no_action;
3105         }
3106
3107         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3108                 nfs_release_seqid(calldata->arg.seqid);
3109                 goto out_wait;
3110         }
3111
3112         if (calldata->arg.fmode == 0)
3113                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3114
3115         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3116                 /* Close-to-open cache consistency revalidation */
3117                 if (!nfs4_have_delegation(inode, FMODE_READ))
3118                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3119                 else
3120                         calldata->arg.bitmask = NULL;
3121         }
3122
3123         calldata->arg.share_access =
3124                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3125                                 calldata->arg.fmode, 0);
3126
3127         if (calldata->res.fattr == NULL)
3128                 calldata->arg.bitmask = NULL;
3129         else if (calldata->arg.bitmask == NULL)
3130                 calldata->res.fattr = NULL;
3131         calldata->timestamp = jiffies;
3132         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3133                                 &calldata->arg.seq_args,
3134                                 &calldata->res.seq_res,
3135                                 task) != 0)
3136                 nfs_release_seqid(calldata->arg.seqid);
3137         dprintk("%s: done!\n", __func__);
3138         return;
3139 out_no_action:
3140         task->tk_action = NULL;
3141 out_wait:
3142         nfs4_sequence_done(task, &calldata->res.seq_res);
3143 }
3144
3145 static const struct rpc_call_ops nfs4_close_ops = {
3146         .rpc_call_prepare = nfs4_close_prepare,
3147         .rpc_call_done = nfs4_close_done,
3148         .rpc_release = nfs4_free_closedata,
3149 };
3150
3151 /*
3152  * It is possible for data to be read/written from a mem-mapped file
3153  * after the sys_close call (which hits the vfs layer as a flush).
3154  * This means that we can't safely call nfsv4 close on a file until
3155  * the inode is cleared. This in turn means that we are not good
3156  * NFSv4 citizens - we do not indicate to the server to update the file's
3157  * share state even when we are done with one of the three share
3158  * stateid's in the inode.
3159  *
3160  * NOTE: Caller must be holding the sp->so_owner semaphore!
3161  */
3162 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3163 {
3164         struct nfs_server *server = NFS_SERVER(state->inode);
3165         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3166         struct nfs4_closedata *calldata;
3167         struct nfs4_state_owner *sp = state->owner;
3168         struct rpc_task *task;
3169         struct rpc_message msg = {
3170                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3171                 .rpc_cred = state->owner->so_cred,
3172         };
3173         struct rpc_task_setup task_setup_data = {
3174                 .rpc_client = server->client,
3175                 .rpc_message = &msg,
3176                 .callback_ops = &nfs4_close_ops,
3177                 .workqueue = nfsiod_workqueue,
3178                 .flags = RPC_TASK_ASYNC,
3179         };
3180         int status = -ENOMEM;
3181
3182         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3183                 &task_setup_data.rpc_client, &msg);
3184
3185         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3186         if (calldata == NULL)
3187                 goto out;
3188         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3189         calldata->inode = state->inode;
3190         calldata->state = state;
3191         calldata->arg.fh = NFS_FH(state->inode);
3192         /* Serialization for the sequence id */
3193         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3194         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3195         if (IS_ERR(calldata->arg.seqid))
3196                 goto out_free_calldata;
3197         nfs_fattr_init(&calldata->fattr);
3198         calldata->arg.fmode = 0;
3199         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3200         calldata->res.fattr = &calldata->fattr;
3201         calldata->res.seqid = calldata->arg.seqid;
3202         calldata->res.server = server;
3203         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3204         calldata->lr.roc = pnfs_roc(state->inode,
3205                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3206         if (calldata->lr.roc) {
3207                 calldata->arg.lr_args = &calldata->lr.arg;
3208                 calldata->res.lr_res = &calldata->lr.res;
3209         }
3210         nfs_sb_active(calldata->inode->i_sb);
3211
3212         msg.rpc_argp = &calldata->arg;
3213         msg.rpc_resp = &calldata->res;
3214         task_setup_data.callback_data = calldata;
3215         task = rpc_run_task(&task_setup_data);
3216         if (IS_ERR(task))
3217                 return PTR_ERR(task);
3218         status = 0;
3219         if (wait)
3220                 status = rpc_wait_for_completion_task(task);
3221         rpc_put_task(task);
3222         return status;
3223 out_free_calldata:
3224         kfree(calldata);
3225 out:
3226         nfs4_put_open_state(state);
3227         nfs4_put_state_owner(sp);
3228         return status;
3229 }
3230
3231 static struct inode *
3232 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3233                 int open_flags, struct iattr *attr, int *opened)
3234 {
3235         struct nfs4_state *state;
3236         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3237
3238         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3239
3240         /* Protect against concurrent sillydeletes */
3241         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3242
3243         nfs4_label_release_security(label);
3244
3245         if (IS_ERR(state))
3246                 return ERR_CAST(state);
3247         return state->inode;
3248 }
3249
3250 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3251 {
3252         if (ctx->state == NULL)
3253                 return;
3254         if (is_sync)
3255                 nfs4_close_sync(ctx->state, ctx->mode);
3256         else
3257                 nfs4_close_state(ctx->state, ctx->mode);
3258 }
3259
3260 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3261 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3262 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3263
3264 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3265 {
3266         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3267         struct nfs4_server_caps_arg args = {
3268                 .fhandle = fhandle,
3269                 .bitmask = bitmask,
3270         };
3271         struct nfs4_server_caps_res res = {};
3272         struct rpc_message msg = {
3273                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3274                 .rpc_argp = &args,
3275                 .rpc_resp = &res,
3276         };
3277         int status;
3278         int i;
3279
3280         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3281                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3282                      FATTR4_WORD0_LINK_SUPPORT |
3283                      FATTR4_WORD0_SYMLINK_SUPPORT |
3284                      FATTR4_WORD0_ACLSUPPORT;
3285         if (minorversion)
3286                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3287
3288         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3289         if (status == 0) {
3290                 /* Sanity check the server answers */
3291                 switch (minorversion) {
3292                 case 0:
3293                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3294                         res.attr_bitmask[2] = 0;
3295                         break;
3296                 case 1:
3297                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3298                         break;
3299                 case 2:
3300                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3301                 }
3302                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3303                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3304                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3305                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3306                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3307                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3308                                 NFS_CAP_SECURITY_LABEL);
3309                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3310                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3311                         server->caps |= NFS_CAP_ACLS;
3312                 if (res.has_links != 0)
3313                         server->caps |= NFS_CAP_HARDLINKS;
3314                 if (res.has_symlinks != 0)
3315                         server->caps |= NFS_CAP_SYMLINKS;
3316                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3317                         server->caps |= NFS_CAP_FILEID;
3318                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3319                         server->caps |= NFS_CAP_MODE;
3320                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3321                         server->caps |= NFS_CAP_NLINK;
3322                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3323                         server->caps |= NFS_CAP_OWNER;
3324                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3325                         server->caps |= NFS_CAP_OWNER_GROUP;
3326                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3327                         server->caps |= NFS_CAP_ATIME;
3328                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3329                         server->caps |= NFS_CAP_CTIME;
3330                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3331                         server->caps |= NFS_CAP_MTIME;
3332 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3333                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3334                         server->caps |= NFS_CAP_SECURITY_LABEL;
3335 #endif
3336                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3337                                 sizeof(server->attr_bitmask));
3338                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3339
3340                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3341                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3342                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3343                 server->cache_consistency_bitmask[2] = 0;
3344
3345                 /* Avoid a regression due to buggy server */
3346                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3347                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3348                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3349                         sizeof(server->exclcreat_bitmask));
3350
3351                 server->acl_bitmask = res.acl_bitmask;
3352                 server->fh_expire_type = res.fh_expire_type;
3353         }
3354
3355         return status;
3356 }
3357
3358 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3359 {
3360         struct nfs4_exception exception = { };
3361         int err;
3362         do {
3363                 err = nfs4_handle_exception(server,
3364                                 _nfs4_server_capabilities(server, fhandle),
3365                                 &exception);
3366         } while (exception.retry);
3367         return err;
3368 }
3369
3370 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3371                 struct nfs_fsinfo *info)
3372 {
3373         u32 bitmask[3];
3374         struct nfs4_lookup_root_arg args = {
3375                 .bitmask = bitmask,
3376         };
3377         struct nfs4_lookup_res res = {
3378                 .server = server,
3379                 .fattr = info->fattr,
3380                 .fh = fhandle,
3381         };
3382         struct rpc_message msg = {
3383                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3384                 .rpc_argp = &args,
3385                 .rpc_resp = &res,
3386         };
3387
3388         bitmask[0] = nfs4_fattr_bitmap[0];
3389         bitmask[1] = nfs4_fattr_bitmap[1];
3390         /*
3391          * Process the label in the upcoming getfattr
3392          */
3393         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3394
3395         nfs_fattr_init(info->fattr);
3396         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3397 }
3398
3399 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3400                 struct nfs_fsinfo *info)
3401 {
3402         struct nfs4_exception exception = { };
3403         int err;
3404         do {
3405                 err = _nfs4_lookup_root(server, fhandle, info);
3406                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3407                 switch (err) {
3408                 case 0:
3409                 case -NFS4ERR_WRONGSEC:
3410                         goto out;
3411                 default:
3412                         err = nfs4_handle_exception(server, err, &exception);
3413                 }
3414         } while (exception.retry);
3415 out:
3416         return err;
3417 }
3418
3419 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3420                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3421 {
3422         struct rpc_auth_create_args auth_args = {
3423                 .pseudoflavor = flavor,
3424         };
3425         struct rpc_auth *auth;
3426
3427         auth = rpcauth_create(&auth_args, server->client);
3428         if (IS_ERR(auth))
3429                 return -EACCES;
3430         return nfs4_lookup_root(server, fhandle, info);
3431 }
3432
3433 /*
3434  * Retry pseudoroot lookup with various security flavors.  We do this when:
3435  *
3436  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3437  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3438  *
3439  * Returns zero on success, or a negative NFS4ERR value, or a
3440  * negative errno value.
3441  */
3442 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3443                               struct nfs_fsinfo *info)
3444 {
3445         /* Per 3530bis 15.33.5 */
3446         static const rpc_authflavor_t flav_array[] = {
3447                 RPC_AUTH_GSS_KRB5P,
3448                 RPC_AUTH_GSS_KRB5I,
3449                 RPC_AUTH_GSS_KRB5,
3450                 RPC_AUTH_UNIX,                  /* courtesy */
3451                 RPC_AUTH_NULL,
3452         };
3453         int status = -EPERM;
3454         size_t i;
3455
3456         if (server->auth_info.flavor_len > 0) {
3457                 /* try each flavor specified by user */
3458                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3459                         status = nfs4_lookup_root_sec(server, fhandle, info,
3460                                                 server->auth_info.flavors[i]);
3461                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3462                                 continue;
3463                         break;
3464                 }
3465         } else {
3466                 /* no flavors specified by user, try default list */
3467                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3468                         status = nfs4_lookup_root_sec(server, fhandle, info,
3469                                                       flav_array[i]);
3470                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3471                                 continue;
3472                         break;
3473                 }
3474         }
3475
3476         /*
3477          * -EACCESS could mean that the user doesn't have correct permissions
3478          * to access the mount.  It could also mean that we tried to mount
3479          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3480          * existing mount programs don't handle -EACCES very well so it should
3481          * be mapped to -EPERM instead.
3482          */
3483         if (status == -EACCES)
3484                 status = -EPERM;
3485         return status;
3486 }
3487
3488 /**
3489  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3490  * @server: initialized nfs_server handle
3491  * @fhandle: we fill in the pseudo-fs root file handle
3492  * @info: we fill in an FSINFO struct
3493  * @auth_probe: probe the auth flavours
3494  *
3495  * Returns zero on success, or a negative errno.
3496  */
3497 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3498                          struct nfs_fsinfo *info,
3499                          bool auth_probe)
3500 {
3501         int status = 0;
3502
3503         if (!auth_probe)
3504                 status = nfs4_lookup_root(server, fhandle, info);
3505
3506         if (auth_probe || status == NFS4ERR_WRONGSEC)
3507                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3508                                 fhandle, info);
3509
3510         if (status == 0)
3511                 status = nfs4_server_capabilities(server, fhandle);
3512         if (status == 0)
3513                 status = nfs4_do_fsinfo(server, fhandle, info);
3514
3515         return nfs4_map_errors(status);
3516 }
3517
3518 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3519                               struct nfs_fsinfo *info)
3520 {
3521         int error;
3522         struct nfs_fattr *fattr = info->fattr;
3523         struct nfs4_label *label = NULL;
3524
3525         error = nfs4_server_capabilities(server, mntfh);
3526         if (error < 0) {
3527                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3528                 return error;
3529         }
3530
3531         label = nfs4_label_alloc(server, GFP_KERNEL);
3532         if (IS_ERR(label))
3533                 return PTR_ERR(label);
3534
3535         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3536         if (error < 0) {
3537                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3538                 goto err_free_label;
3539         }
3540
3541         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3542             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3543                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3544
3545 err_free_label:
3546         nfs4_label_free(label);
3547
3548         return error;
3549 }
3550
3551 /*
3552  * Get locations and (maybe) other attributes of a referral.
3553  * Note that we'll actually follow the referral later when
3554  * we detect fsid mismatch in inode revalidation
3555  */
3556 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3557                              const struct qstr *name, struct nfs_fattr *fattr,
3558                              struct nfs_fh *fhandle)
3559 {
3560         int status = -ENOMEM;
3561         struct page *page = NULL;
3562         struct nfs4_fs_locations *locations = NULL;
3563
3564         page = alloc_page(GFP_KERNEL);
3565         if (page == NULL)
3566                 goto out;
3567         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3568         if (locations == NULL)
3569                 goto out;
3570
3571         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3572         if (status != 0)
3573                 goto out;
3574
3575         /*
3576          * If the fsid didn't change, this is a migration event, not a
3577          * referral.  Cause us to drop into the exception handler, which
3578          * will kick off migration recovery.
3579          */
3580         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3581                 dprintk("%s: server did not return a different fsid for"
3582                         " a referral at %s\n", __func__, name->name);
3583                 status = -NFS4ERR_MOVED;
3584                 goto out;
3585         }
3586         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3587         nfs_fixup_referral_attributes(&locations->fattr);
3588
3589         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3590         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3591         memset(fhandle, 0, sizeof(struct nfs_fh));
3592 out:
3593         if (page)
3594                 __free_page(page);
3595         kfree(locations);
3596         return status;
3597 }
3598
3599 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3600                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3601 {
3602         struct nfs4_getattr_arg args = {
3603                 .fh = fhandle,
3604                 .bitmask = server->attr_bitmask,
3605         };
3606         struct nfs4_getattr_res res = {
3607                 .fattr = fattr,
3608                 .label = label,
3609                 .server = server,
3610         };
3611         struct rpc_message msg = {
3612                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3613                 .rpc_argp = &args,
3614                 .rpc_resp = &res,
3615         };
3616
3617         args.bitmask = nfs4_bitmask(server, label);
3618
3619         nfs_fattr_init(fattr);
3620         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3621 }
3622
3623 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3624                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3625 {
3626         struct nfs4_exception exception = { };
3627         int err;
3628         do {
3629                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3630                 trace_nfs4_getattr(server, fhandle, fattr, err);
3631                 err = nfs4_handle_exception(server, err,
3632                                 &exception);
3633         } while (exception.retry);
3634         return err;
3635 }
3636
3637 /*
3638  * The file is not closed if it is opened due to the a request to change
3639  * the size of the file. The open call will not be needed once the
3640  * VFS layer lookup-intents are implemented.
3641  *
3642  * Close is called when the inode is destroyed.
3643  * If we haven't opened the file for O_WRONLY, we
3644  * need to in the size_change case to obtain a stateid.
3645  *
3646  * Got race?
3647  * Because OPEN is always done by name in nfsv4, it is
3648  * possible that we opened a different file by the same
3649  * name.  We can recognize this race condition, but we
3650  * can't do anything about it besides returning an error.
3651  *
3652  * This will be fixed with VFS changes (lookup-intent).
3653  */
3654 static int
3655 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3656                   struct iattr *sattr)
3657 {
3658         struct inode *inode = d_inode(dentry);
3659         struct rpc_cred *cred = NULL;
3660         struct nfs_open_context *ctx = NULL;
3661         struct nfs4_label *label = NULL;
3662         int status;
3663
3664         if (pnfs_ld_layoutret_on_setattr(inode) &&
3665             sattr->ia_valid & ATTR_SIZE &&
3666             sattr->ia_size < i_size_read(inode))
3667                 pnfs_commit_and_return_layout(inode);
3668
3669         nfs_fattr_init(fattr);
3670
3671         /* Deal with open(O_TRUNC) */
3672         if (sattr->ia_valid & ATTR_OPEN)
3673                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3674
3675         /* Optimization: if the end result is no change, don't RPC */
3676         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3677                 return 0;
3678
3679         /* Search for an existing open(O_WRITE) file */
3680         if (sattr->ia_valid & ATTR_FILE) {
3681
3682                 ctx = nfs_file_open_context(sattr->ia_file);
3683                 if (ctx)
3684                         cred = ctx->cred;
3685         }
3686
3687         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3688         if (IS_ERR(label))
3689                 return PTR_ERR(label);
3690
3691         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3692         if (status == 0) {
3693                 nfs_setattr_update_inode(inode, sattr, fattr);
3694                 nfs_setsecurity(inode, fattr, label);
3695         }
3696         nfs4_label_free(label);
3697         return status;
3698 }
3699
3700 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3701                 const struct qstr *name, struct nfs_fh *fhandle,
3702                 struct nfs_fattr *fattr, struct nfs4_label *label)
3703 {
3704         struct nfs_server *server = NFS_SERVER(dir);
3705         int                    status;
3706         struct nfs4_lookup_arg args = {
3707                 .bitmask = server->attr_bitmask,
3708                 .dir_fh = NFS_FH(dir),
3709                 .name = name,
3710         };
3711         struct nfs4_lookup_res res = {
3712                 .server = server,
3713                 .fattr = fattr,
3714                 .label = label,
3715                 .fh = fhandle,
3716         };
3717         struct rpc_message msg = {
3718                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3719                 .rpc_argp = &args,
3720                 .rpc_resp = &res,
3721         };
3722
3723         args.bitmask = nfs4_bitmask(server, label);
3724
3725         nfs_fattr_init(fattr);
3726
3727         dprintk("NFS call  lookup %s\n", name->name);
3728         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3729         dprintk("NFS reply lookup: %d\n", status);
3730         return status;
3731 }
3732
3733 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3734 {
3735         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3736                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3737         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3738         fattr->nlink = 2;
3739 }
3740
3741 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3742                                    const struct qstr *name, struct nfs_fh *fhandle,
3743                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3744 {
3745         struct nfs4_exception exception = { };
3746         struct rpc_clnt *client = *clnt;
3747         int err;
3748         do {
3749                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3750                 trace_nfs4_lookup(dir, name, err);
3751                 switch (err) {
3752                 case -NFS4ERR_BADNAME:
3753                         err = -ENOENT;
3754                         goto out;
3755                 case -NFS4ERR_MOVED:
3756                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3757                         if (err == -NFS4ERR_MOVED)
3758                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3759                         goto out;
3760                 case -NFS4ERR_WRONGSEC:
3761                         err = -EPERM;
3762                         if (client != *clnt)
3763                                 goto out;
3764                         client = nfs4_negotiate_security(client, dir, name);
3765                         if (IS_ERR(client))
3766                                 return PTR_ERR(client);
3767
3768                         exception.retry = 1;
3769                         break;
3770                 default:
3771                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3772                 }
3773         } while (exception.retry);
3774
3775 out:
3776         if (err == 0)
3777                 *clnt = client;
3778         else if (client != *clnt)
3779                 rpc_shutdown_client(client);
3780
3781         return err;
3782 }
3783
3784 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3785                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3786                             struct nfs4_label *label)
3787 {
3788         int status;
3789         struct rpc_clnt *client = NFS_CLIENT(dir);
3790
3791         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3792         if (client != NFS_CLIENT(dir)) {
3793                 rpc_shutdown_client(client);
3794                 nfs_fixup_secinfo_attributes(fattr);
3795         }
3796         return status;
3797 }
3798
3799 struct rpc_clnt *
3800 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3801                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3802 {
3803         struct rpc_clnt *client = NFS_CLIENT(dir);
3804         int status;
3805
3806         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3807         if (status < 0)
3808                 return ERR_PTR(status);
3809         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3810 }
3811
3812 static int _nfs4_proc_lookupp(struct inode *inode,
3813                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3814                 struct nfs4_label *label)
3815 {
3816         struct rpc_clnt *clnt = NFS_CLIENT(inode);
3817         struct nfs_server *server = NFS_SERVER(inode);
3818         int                    status;
3819         struct nfs4_lookupp_arg args = {
3820                 .bitmask = server->attr_bitmask,
3821                 .fh = NFS_FH(inode),
3822         };
3823         struct nfs4_lookupp_res res = {
3824                 .server = server,
3825                 .fattr = fattr,
3826                 .label = label,
3827                 .fh = fhandle,
3828         };
3829         struct rpc_message msg = {
3830                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
3831                 .rpc_argp = &args,
3832                 .rpc_resp = &res,
3833         };
3834
3835         args.bitmask = nfs4_bitmask(server, label);
3836
3837         nfs_fattr_init(fattr);
3838
3839         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
3840         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
3841                                 &res.seq_res, 0);
3842         dprintk("NFS reply lookupp: %d\n", status);
3843         return status;
3844 }
3845
3846 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
3847                              struct nfs_fattr *fattr, struct nfs4_label *label)
3848 {
3849         struct nfs4_exception exception = { };
3850         int err;
3851         do {
3852                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
3853                 trace_nfs4_lookupp(inode, err);
3854                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3855                                 &exception);
3856         } while (exception.retry);
3857         return err;
3858 }
3859
3860 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3861 {
3862         struct nfs_server *server = NFS_SERVER(inode);
3863         struct nfs4_accessargs args = {
3864                 .fh = NFS_FH(inode),
3865                 .bitmask = server->cache_consistency_bitmask,
3866         };
3867         struct nfs4_accessres res = {
3868                 .server = server,
3869         };
3870         struct rpc_message msg = {
3871                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3872                 .rpc_argp = &args,
3873                 .rpc_resp = &res,
3874                 .rpc_cred = entry->cred,
3875         };
3876         int mode = entry->mask;
3877         int status = 0;
3878
3879         /*
3880          * Determine which access bits we want to ask for...
3881          */
3882         if (mode & MAY_READ)
3883                 args.access |= NFS4_ACCESS_READ;
3884         if (S_ISDIR(inode->i_mode)) {
3885                 if (mode & MAY_WRITE)
3886                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3887                 if (mode & MAY_EXEC)
3888                         args.access |= NFS4_ACCESS_LOOKUP;
3889         } else {
3890                 if (mode & MAY_WRITE)
3891                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3892                 if (mode & MAY_EXEC)
3893                         args.access |= NFS4_ACCESS_EXECUTE;
3894         }
3895
3896         res.fattr = nfs_alloc_fattr();
3897         if (res.fattr == NULL)
3898                 return -ENOMEM;
3899
3900         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3901         if (!status) {
3902                 nfs_access_set_mask(entry, res.access);
3903                 nfs_refresh_inode(inode, res.fattr);
3904         }
3905         nfs_free_fattr(res.fattr);
3906         return status;
3907 }
3908
3909 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3910 {
3911         struct nfs4_exception exception = { };
3912         int err;
3913         do {
3914                 err = _nfs4_proc_access(inode, entry);
3915                 trace_nfs4_access(inode, err);
3916                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3917                                 &exception);
3918         } while (exception.retry);
3919         return err;
3920 }
3921
3922 /*
3923  * TODO: For the time being, we don't try to get any attributes
3924  * along with any of the zero-copy operations READ, READDIR,
3925  * READLINK, WRITE.
3926  *
3927  * In the case of the first three, we want to put the GETATTR
3928  * after the read-type operation -- this is because it is hard
3929  * to predict the length of a GETATTR response in v4, and thus
3930  * align the READ data correctly.  This means that the GETATTR
3931  * may end up partially falling into the page cache, and we should
3932  * shift it into the 'tail' of the xdr_buf before processing.
3933  * To do this efficiently, we need to know the total length
3934  * of data received, which doesn't seem to be available outside
3935  * of the RPC layer.
3936  *
3937  * In the case of WRITE, we also want to put the GETATTR after
3938  * the operation -- in this case because we want to make sure
3939  * we get the post-operation mtime and size.
3940  *
3941  * Both of these changes to the XDR layer would in fact be quite
3942  * minor, but I decided to leave them for a subsequent patch.
3943  */
3944 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3945                 unsigned int pgbase, unsigned int pglen)
3946 {
3947         struct nfs4_readlink args = {
3948                 .fh       = NFS_FH(inode),
3949                 .pgbase   = pgbase,
3950                 .pglen    = pglen,
3951                 .pages    = &page,
3952         };
3953         struct nfs4_readlink_res res;
3954         struct rpc_message msg = {
3955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3956                 .rpc_argp = &args,
3957                 .rpc_resp = &res,
3958         };
3959
3960         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3961 }
3962
3963 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3964                 unsigned int pgbase, unsigned int pglen)
3965 {
3966         struct nfs4_exception exception = { };
3967         int err;
3968         do {
3969                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3970                 trace_nfs4_readlink(inode, err);
3971                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3972                                 &exception);
3973         } while (exception.retry);
3974         return err;
3975 }
3976
3977 /*
3978  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3979  */
3980 static int
3981 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3982                  int flags)
3983 {
3984         struct nfs_server *server = NFS_SERVER(dir);
3985         struct nfs4_label l, *ilabel = NULL;
3986         struct nfs_open_context *ctx;
3987         struct nfs4_state *state;
3988         int status = 0;
3989
3990         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
3991         if (IS_ERR(ctx))
3992                 return PTR_ERR(ctx);
3993
3994         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3995
3996         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
3997                 sattr->ia_mode &= ~current_umask();
3998         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3999         if (IS_ERR(state)) {
4000                 status = PTR_ERR(state);
4001                 goto out;
4002         }
4003 out:
4004         nfs4_label_release_security(ilabel);
4005         put_nfs_open_context(ctx);
4006         return status;
4007 }
4008
4009 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4010 {
4011         struct nfs_server *server = NFS_SERVER(dir);
4012         struct nfs_removeargs args = {
4013                 .fh = NFS_FH(dir),
4014                 .name = *name,
4015         };
4016         struct nfs_removeres res = {
4017                 .server = server,
4018         };
4019         struct rpc_message msg = {
4020                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4021                 .rpc_argp = &args,
4022                 .rpc_resp = &res,
4023         };
4024         unsigned long timestamp = jiffies;
4025         int status;
4026
4027         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4028         if (status == 0)
4029                 update_changeattr(dir, &res.cinfo, timestamp);
4030         return status;
4031 }
4032
4033 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4034 {
4035         struct nfs4_exception exception = { };
4036         int err;
4037         do {
4038                 err = _nfs4_proc_remove(dir, name);
4039                 trace_nfs4_remove(dir, name, err);
4040                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4041                                 &exception);
4042         } while (exception.retry);
4043         return err;
4044 }
4045
4046 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4047 {
4048         struct nfs_server *server = NFS_SERVER(dir);
4049         struct nfs_removeargs *args = msg->rpc_argp;
4050         struct nfs_removeres *res = msg->rpc_resp;
4051
4052         res->server = server;
4053         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4054         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4055
4056         nfs_fattr_init(res->dir_attr);
4057 }
4058
4059 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4060 {
4061         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4062                         &data->args.seq_args,
4063                         &data->res.seq_res,
4064                         task);
4065 }
4066
4067 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4068 {
4069         struct nfs_unlinkdata *data = task->tk_calldata;
4070         struct nfs_removeres *res = &data->res;
4071
4072         if (!nfs4_sequence_done(task, &res->seq_res))
4073                 return 0;
4074         if (nfs4_async_handle_error(task, res->server, NULL,
4075                                     &data->timeout) == -EAGAIN)
4076                 return 0;
4077         if (task->tk_status == 0)
4078                 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4079         return 1;
4080 }
4081
4082 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4083 {
4084         struct nfs_server *server = NFS_SERVER(dir);
4085         struct nfs_renameargs *arg = msg->rpc_argp;
4086         struct nfs_renameres *res = msg->rpc_resp;
4087
4088         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4089         res->server = server;
4090         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4091 }
4092
4093 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4094 {
4095         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4096                         &data->args.seq_args,
4097                         &data->res.seq_res,
4098                         task);
4099 }
4100
4101 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4102                                  struct inode *new_dir)
4103 {
4104         struct nfs_renamedata *data = task->tk_calldata;
4105         struct nfs_renameres *res = &data->res;
4106
4107         if (!nfs4_sequence_done(task, &res->seq_res))
4108                 return 0;
4109         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4110                 return 0;
4111
4112         if (task->tk_status == 0) {
4113                 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4114                 if (new_dir != old_dir)
4115                         update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4116         }
4117         return 1;
4118 }
4119
4120 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4121 {
4122         struct nfs_server *server = NFS_SERVER(inode);
4123         struct nfs4_link_arg arg = {
4124                 .fh     = NFS_FH(inode),
4125                 .dir_fh = NFS_FH(dir),
4126                 .name   = name,
4127                 .bitmask = server->attr_bitmask,
4128         };
4129         struct nfs4_link_res res = {
4130                 .server = server,
4131                 .label = NULL,
4132         };
4133         struct rpc_message msg = {
4134                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4135                 .rpc_argp = &arg,
4136                 .rpc_resp = &res,
4137         };
4138         int status = -ENOMEM;
4139
4140         res.fattr = nfs_alloc_fattr();
4141         if (res.fattr == NULL)
4142                 goto out;
4143
4144         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4145         if (IS_ERR(res.label)) {
4146                 status = PTR_ERR(res.label);
4147                 goto out;
4148         }
4149         arg.bitmask = nfs4_bitmask(server, res.label);
4150
4151         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4152         if (!status) {
4153                 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4154                 status = nfs_post_op_update_inode(inode, res.fattr);
4155                 if (!status)
4156                         nfs_setsecurity(inode, res.fattr, res.label);
4157         }
4158
4159
4160         nfs4_label_free(res.label);
4161
4162 out:
4163         nfs_free_fattr(res.fattr);
4164         return status;
4165 }
4166
4167 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4168 {
4169         struct nfs4_exception exception = { };
4170         int err;
4171         do {
4172                 err = nfs4_handle_exception(NFS_SERVER(inode),
4173                                 _nfs4_proc_link(inode, dir, name),
4174                                 &exception);
4175         } while (exception.retry);
4176         return err;
4177 }
4178
4179 struct nfs4_createdata {
4180         struct rpc_message msg;
4181         struct nfs4_create_arg arg;
4182         struct nfs4_create_res res;
4183         struct nfs_fh fh;
4184         struct nfs_fattr fattr;
4185         struct nfs4_label *label;
4186 };
4187
4188 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4189                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4190 {
4191         struct nfs4_createdata *data;
4192
4193         data = kzalloc(sizeof(*data), GFP_KERNEL);
4194         if (data != NULL) {
4195                 struct nfs_server *server = NFS_SERVER(dir);
4196
4197                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4198                 if (IS_ERR(data->label))
4199                         goto out_free;
4200
4201                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4202                 data->msg.rpc_argp = &data->arg;
4203                 data->msg.rpc_resp = &data->res;
4204                 data->arg.dir_fh = NFS_FH(dir);
4205                 data->arg.server = server;
4206                 data->arg.name = name;
4207                 data->arg.attrs = sattr;
4208                 data->arg.ftype = ftype;
4209                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4210                 data->arg.umask = current_umask();
4211                 data->res.server = server;
4212                 data->res.fh = &data->fh;
4213                 data->res.fattr = &data->fattr;
4214                 data->res.label = data->label;
4215                 nfs_fattr_init(data->res.fattr);
4216         }
4217         return data;
4218 out_free:
4219         kfree(data);
4220         return NULL;
4221 }
4222
4223 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4224 {
4225         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4226                                     &data->arg.seq_args, &data->res.seq_res, 1);
4227         if (status == 0) {
4228                 update_changeattr(dir, &data->res.dir_cinfo,
4229                                 data->res.fattr->time_start);
4230                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4231         }
4232         return status;
4233 }
4234
4235 static void nfs4_free_createdata(struct nfs4_createdata *data)
4236 {
4237         nfs4_label_free(data->label);
4238         kfree(data);
4239 }
4240
4241 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4242                 struct page *page, unsigned int len, struct iattr *sattr,
4243                 struct nfs4_label *label)
4244 {
4245         struct nfs4_createdata *data;
4246         int status = -ENAMETOOLONG;
4247
4248         if (len > NFS4_MAXPATHLEN)
4249                 goto out;
4250
4251         status = -ENOMEM;
4252         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4253         if (data == NULL)
4254                 goto out;
4255
4256         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4257         data->arg.u.symlink.pages = &page;
4258         data->arg.u.symlink.len = len;
4259         data->arg.label = label;
4260
4261         status = nfs4_do_create(dir, dentry, data);
4262
4263         nfs4_free_createdata(data);
4264 out:
4265         return status;
4266 }
4267
4268 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4269                 struct page *page, unsigned int len, struct iattr *sattr)
4270 {
4271         struct nfs4_exception exception = { };
4272         struct nfs4_label l, *label = NULL;
4273         int err;
4274
4275         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4276
4277         do {
4278                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4279                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4280                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4281                                 &exception);
4282         } while (exception.retry);
4283
4284         nfs4_label_release_security(label);
4285         return err;
4286 }
4287
4288 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4289                 struct iattr *sattr, struct nfs4_label *label)
4290 {
4291         struct nfs4_createdata *data;
4292         int status = -ENOMEM;
4293
4294         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4295         if (data == NULL)
4296                 goto out;
4297
4298         data->arg.label = label;
4299         status = nfs4_do_create(dir, dentry, data);
4300
4301         nfs4_free_createdata(data);
4302 out:
4303         return status;
4304 }
4305
4306 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4307                 struct iattr *sattr)
4308 {
4309         struct nfs_server *server = NFS_SERVER(dir);
4310         struct nfs4_exception exception = { };
4311         struct nfs4_label l, *label = NULL;
4312         int err;
4313
4314         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4315
4316         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4317                 sattr->ia_mode &= ~current_umask();
4318         do {
4319                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4320                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4321                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4322                                 &exception);
4323         } while (exception.retry);
4324         nfs4_label_release_security(label);
4325
4326         return err;
4327 }
4328
4329 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4330                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4331 {
4332         struct inode            *dir = d_inode(dentry);
4333         struct nfs4_readdir_arg args = {
4334                 .fh = NFS_FH(dir),
4335                 .pages = pages,
4336                 .pgbase = 0,
4337                 .count = count,
4338                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4339                 .plus = plus,
4340         };
4341         struct nfs4_readdir_res res;
4342         struct rpc_message msg = {
4343                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4344                 .rpc_argp = &args,
4345                 .rpc_resp = &res,
4346                 .rpc_cred = cred,
4347         };
4348         int                     status;
4349
4350         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4351                         dentry,
4352                         (unsigned long long)cookie);
4353         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4354         res.pgbase = args.pgbase;
4355         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4356         if (status >= 0) {
4357                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4358                 status += args.pgbase;
4359         }
4360
4361         nfs_invalidate_atime(dir);
4362
4363         dprintk("%s: returns %d\n", __func__, status);
4364         return status;
4365 }
4366
4367 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4368                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4369 {
4370         struct nfs4_exception exception = { };
4371         int err;
4372         do {
4373                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4374                                 pages, count, plus);
4375                 trace_nfs4_readdir(d_inode(dentry), err);
4376                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4377                                 &exception);
4378         } while (exception.retry);
4379         return err;
4380 }
4381
4382 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4383                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4384 {
4385         struct nfs4_createdata *data;
4386         int mode = sattr->ia_mode;
4387         int status = -ENOMEM;
4388
4389         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4390         if (data == NULL)
4391                 goto out;
4392
4393         if (S_ISFIFO(mode))
4394                 data->arg.ftype = NF4FIFO;
4395         else if (S_ISBLK(mode)) {
4396                 data->arg.ftype = NF4BLK;
4397                 data->arg.u.device.specdata1 = MAJOR(rdev);
4398                 data->arg.u.device.specdata2 = MINOR(rdev);
4399         }
4400         else if (S_ISCHR(mode)) {
4401                 data->arg.ftype = NF4CHR;
4402                 data->arg.u.device.specdata1 = MAJOR(rdev);
4403                 data->arg.u.device.specdata2 = MINOR(rdev);
4404         } else if (!S_ISSOCK(mode)) {
4405                 status = -EINVAL;
4406                 goto out_free;
4407         }
4408
4409         data->arg.label = label;
4410         status = nfs4_do_create(dir, dentry, data);
4411 out_free:
4412         nfs4_free_createdata(data);
4413 out:
4414         return status;
4415 }
4416
4417 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4418                 struct iattr *sattr, dev_t rdev)
4419 {
4420         struct nfs_server *server = NFS_SERVER(dir);
4421         struct nfs4_exception exception = { };
4422         struct nfs4_label l, *label = NULL;
4423         int err;
4424
4425         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4426
4427         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4428                 sattr->ia_mode &= ~current_umask();
4429         do {
4430                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4431                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4432                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4433                                 &exception);
4434         } while (exception.retry);
4435
4436         nfs4_label_release_security(label);
4437
4438         return err;
4439 }
4440
4441 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4442                  struct nfs_fsstat *fsstat)
4443 {
4444         struct nfs4_statfs_arg args = {
4445                 .fh = fhandle,
4446                 .bitmask = server->attr_bitmask,
4447         };
4448         struct nfs4_statfs_res res = {
4449                 .fsstat = fsstat,
4450         };
4451         struct rpc_message msg = {
4452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4453                 .rpc_argp = &args,
4454                 .rpc_resp = &res,
4455         };
4456
4457         nfs_fattr_init(fsstat->fattr);
4458         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4459 }
4460
4461 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4462 {
4463         struct nfs4_exception exception = { };
4464         int err;
4465         do {
4466                 err = nfs4_handle_exception(server,
4467                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4468                                 &exception);
4469         } while (exception.retry);
4470         return err;
4471 }
4472
4473 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4474                 struct nfs_fsinfo *fsinfo)
4475 {
4476         struct nfs4_fsinfo_arg args = {
4477                 .fh = fhandle,
4478                 .bitmask = server->attr_bitmask,
4479         };
4480         struct nfs4_fsinfo_res res = {
4481                 .fsinfo = fsinfo,
4482         };
4483         struct rpc_message msg = {
4484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4485                 .rpc_argp = &args,
4486                 .rpc_resp = &res,
4487         };
4488
4489         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4490 }
4491
4492 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4493 {
4494         struct nfs4_exception exception = { };
4495         unsigned long now = jiffies;
4496         int err;
4497
4498         do {
4499                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4500                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4501                 if (err == 0) {
4502                         nfs4_set_lease_period(server->nfs_client,
4503                                         fsinfo->lease_time * HZ,
4504                                         now);
4505                         break;
4506                 }
4507                 err = nfs4_handle_exception(server, err, &exception);
4508         } while (exception.retry);
4509         return err;
4510 }
4511
4512 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4513 {
4514         int error;
4515
4516         nfs_fattr_init(fsinfo->fattr);
4517         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4518         if (error == 0) {
4519                 /* block layout checks this! */
4520                 server->pnfs_blksize = fsinfo->blksize;
4521                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4522         }
4523
4524         return error;
4525 }
4526
4527 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4528                 struct nfs_pathconf *pathconf)
4529 {
4530         struct nfs4_pathconf_arg args = {
4531                 .fh = fhandle,
4532                 .bitmask = server->attr_bitmask,
4533         };
4534         struct nfs4_pathconf_res res = {
4535                 .pathconf = pathconf,
4536         };
4537         struct rpc_message msg = {
4538                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4539                 .rpc_argp = &args,
4540                 .rpc_resp = &res,
4541         };
4542
4543         /* None of the pathconf attributes are mandatory to implement */
4544         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4545                 memset(pathconf, 0, sizeof(*pathconf));
4546                 return 0;
4547         }
4548
4549         nfs_fattr_init(pathconf->fattr);
4550         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4551 }
4552
4553 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4554                 struct nfs_pathconf *pathconf)
4555 {
4556         struct nfs4_exception exception = { };
4557         int err;
4558
4559         do {
4560                 err = nfs4_handle_exception(server,
4561                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4562                                 &exception);
4563         } while (exception.retry);
4564         return err;
4565 }
4566
4567 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4568                 const struct nfs_open_context *ctx,
4569                 const struct nfs_lock_context *l_ctx,
4570                 fmode_t fmode)
4571 {
4572         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4573 }
4574 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4575
4576 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4577                 const struct nfs_open_context *ctx,
4578                 const struct nfs_lock_context *l_ctx,
4579                 fmode_t fmode)
4580 {
4581         nfs4_stateid current_stateid;
4582
4583         /* If the current stateid represents a lost lock, then exit */
4584         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4585                 return true;
4586         return nfs4_stateid_match(stateid, &current_stateid);
4587 }
4588
4589 static bool nfs4_error_stateid_expired(int err)
4590 {
4591         switch (err) {
4592         case -NFS4ERR_DELEG_REVOKED:
4593         case -NFS4ERR_ADMIN_REVOKED:
4594         case -NFS4ERR_BAD_STATEID:
4595         case -NFS4ERR_STALE_STATEID:
4596         case -NFS4ERR_OLD_STATEID:
4597         case -NFS4ERR_OPENMODE:
4598         case -NFS4ERR_EXPIRED:
4599                 return true;
4600         }
4601         return false;
4602 }
4603
4604 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4605 {
4606         struct nfs_server *server = NFS_SERVER(hdr->inode);
4607
4608         trace_nfs4_read(hdr, task->tk_status);
4609         if (task->tk_status < 0) {
4610                 struct nfs4_exception exception = {
4611                         .inode = hdr->inode,
4612                         .state = hdr->args.context->state,
4613                         .stateid = &hdr->args.stateid,
4614                 };
4615                 task->tk_status = nfs4_async_handle_exception(task,
4616                                 server, task->tk_status, &exception);
4617                 if (exception.retry) {
4618                         rpc_restart_call_prepare(task);
4619                         return -EAGAIN;
4620                 }
4621         }
4622
4623         if (task->tk_status > 0)
4624                 renew_lease(server, hdr->timestamp);
4625         return 0;
4626 }
4627
4628 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4629                 struct nfs_pgio_args *args)
4630 {
4631
4632         if (!nfs4_error_stateid_expired(task->tk_status) ||
4633                 nfs4_stateid_is_current(&args->stateid,
4634                                 args->context,
4635                                 args->lock_context,
4636                                 FMODE_READ))
4637                 return false;
4638         rpc_restart_call_prepare(task);
4639         return true;
4640 }
4641
4642 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4643 {
4644
4645         dprintk("--> %s\n", __func__);
4646
4647         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4648                 return -EAGAIN;
4649         if (nfs4_read_stateid_changed(task, &hdr->args))
4650                 return -EAGAIN;
4651         if (task->tk_status > 0)
4652                 nfs_invalidate_atime(hdr->inode);
4653         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4654                                     nfs4_read_done_cb(task, hdr);
4655 }
4656
4657 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4658                                  struct rpc_message *msg)
4659 {
4660         hdr->timestamp   = jiffies;
4661         if (!hdr->pgio_done_cb)
4662                 hdr->pgio_done_cb = nfs4_read_done_cb;
4663         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4664         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4665 }
4666
4667 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4668                                       struct nfs_pgio_header *hdr)
4669 {
4670         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4671                         &hdr->args.seq_args,
4672                         &hdr->res.seq_res,
4673                         task))
4674                 return 0;
4675         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4676                                 hdr->args.lock_context,
4677                                 hdr->rw_mode) == -EIO)
4678                 return -EIO;
4679         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4680                 return -EIO;
4681         return 0;
4682 }
4683
4684 static int nfs4_write_done_cb(struct rpc_task *task,
4685                               struct nfs_pgio_header *hdr)
4686 {
4687         struct inode *inode = hdr->inode;
4688
4689         trace_nfs4_write(hdr, task->tk_status);
4690         if (task->tk_status < 0) {
4691                 struct nfs4_exception exception = {
4692                         .inode = hdr->inode,
4693                         .state = hdr->args.context->state,
4694                         .stateid = &hdr->args.stateid,
4695                 };
4696                 task->tk_status = nfs4_async_handle_exception(task,
4697                                 NFS_SERVER(inode), task->tk_status,
4698                                 &exception);
4699                 if (exception.retry) {
4700                         rpc_restart_call_prepare(task);
4701                         return -EAGAIN;
4702                 }
4703         }
4704         if (task->tk_status >= 0) {
4705                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4706                 nfs_writeback_update_inode(hdr);
4707         }
4708         return 0;
4709 }
4710
4711 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4712                 struct nfs_pgio_args *args)
4713 {
4714
4715         if (!nfs4_error_stateid_expired(task->tk_status) ||
4716                 nfs4_stateid_is_current(&args->stateid,
4717                                 args->context,
4718                                 args->lock_context,
4719                                 FMODE_WRITE))
4720                 return false;
4721         rpc_restart_call_prepare(task);
4722         return true;
4723 }
4724
4725 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4726 {
4727         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4728                 return -EAGAIN;
4729         if (nfs4_write_stateid_changed(task, &hdr->args))
4730                 return -EAGAIN;
4731         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4732                 nfs4_write_done_cb(task, hdr);
4733 }
4734
4735 static
4736 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4737 {
4738         /* Don't request attributes for pNFS or O_DIRECT writes */
4739         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4740                 return false;
4741         /* Otherwise, request attributes if and only if we don't hold
4742          * a delegation
4743          */
4744         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4745 }
4746
4747 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4748                                   struct rpc_message *msg)
4749 {
4750         struct nfs_server *server = NFS_SERVER(hdr->inode);
4751
4752         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4753                 hdr->args.bitmask = NULL;
4754                 hdr->res.fattr = NULL;
4755         } else
4756                 hdr->args.bitmask = server->cache_consistency_bitmask;
4757
4758         if (!hdr->pgio_done_cb)
4759                 hdr->pgio_done_cb = nfs4_write_done_cb;
4760         hdr->res.server = server;
4761         hdr->timestamp   = jiffies;
4762
4763         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4764         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4765 }
4766
4767 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4768 {
4769         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
4770                         &data->args.seq_args,
4771                         &data->res.seq_res,
4772                         task);
4773 }
4774
4775 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4776 {
4777         struct inode *inode = data->inode;
4778
4779         trace_nfs4_commit(data, task->tk_status);
4780         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4781                                     NULL, NULL) == -EAGAIN) {
4782                 rpc_restart_call_prepare(task);
4783                 return -EAGAIN;
4784         }
4785         return 0;
4786 }
4787
4788 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4789 {
4790         if (!nfs4_sequence_done(task, &data->res.seq_res))
4791                 return -EAGAIN;
4792         return data->commit_done_cb(task, data);
4793 }
4794
4795 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4796 {
4797         struct nfs_server *server = NFS_SERVER(data->inode);
4798
4799         if (data->commit_done_cb == NULL)
4800                 data->commit_done_cb = nfs4_commit_done_cb;
4801         data->res.server = server;
4802         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4803         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4804 }
4805
4806 struct nfs4_renewdata {
4807         struct nfs_client       *client;
4808         unsigned long           timestamp;
4809 };
4810
4811 /*
4812  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4813  * standalone procedure for queueing an asynchronous RENEW.
4814  */
4815 static void nfs4_renew_release(void *calldata)
4816 {
4817         struct nfs4_renewdata *data = calldata;
4818         struct nfs_client *clp = data->client;
4819
4820         if (atomic_read(&clp->cl_count) > 1)
4821                 nfs4_schedule_state_renewal(clp);
4822         nfs_put_client(clp);
4823         kfree(data);
4824 }
4825
4826 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4827 {
4828         struct nfs4_renewdata *data = calldata;
4829         struct nfs_client *clp = data->client;
4830         unsigned long timestamp = data->timestamp;
4831
4832         trace_nfs4_renew_async(clp, task->tk_status);
4833         switch (task->tk_status) {
4834         case 0:
4835                 break;
4836         case -NFS4ERR_LEASE_MOVED:
4837                 nfs4_schedule_lease_moved_recovery(clp);
4838                 break;
4839         default:
4840                 /* Unless we're shutting down, schedule state recovery! */
4841                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4842                         return;
4843                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4844                         nfs4_schedule_lease_recovery(clp);
4845                         return;
4846                 }
4847                 nfs4_schedule_path_down_recovery(clp);
4848         }
4849         do_renew_lease(clp, timestamp);
4850 }
4851
4852 static const struct rpc_call_ops nfs4_renew_ops = {
4853         .rpc_call_done = nfs4_renew_done,
4854         .rpc_release = nfs4_renew_release,
4855 };
4856
4857 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4858 {
4859         struct rpc_message msg = {
4860                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4861                 .rpc_argp       = clp,
4862                 .rpc_cred       = cred,
4863         };
4864         struct nfs4_renewdata *data;
4865
4866         if (renew_flags == 0)
4867                 return 0;
4868         if (!atomic_inc_not_zero(&clp->cl_count))
4869                 return -EIO;
4870         data = kmalloc(sizeof(*data), GFP_NOFS);
4871         if (data == NULL) {
4872                 nfs_put_client(clp);
4873                 return -ENOMEM;
4874         }
4875         data->client = clp;
4876         data->timestamp = jiffies;
4877         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4878                         &nfs4_renew_ops, data);
4879 }
4880
4881 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4882 {
4883         struct rpc_message msg = {
4884                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4885                 .rpc_argp       = clp,
4886                 .rpc_cred       = cred,
4887         };
4888         unsigned long now = jiffies;
4889         int status;
4890
4891         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4892         if (status < 0)
4893                 return status;
4894         do_renew_lease(clp, now);
4895         return 0;
4896 }
4897
4898 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4899 {
4900         return server->caps & NFS_CAP_ACLS;
4901 }
4902
4903 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4904  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4905  * the stack.
4906  */
4907 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4908
4909 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4910                 struct page **pages)
4911 {
4912         struct page *newpage, **spages;
4913         int rc = 0;
4914         size_t len;
4915         spages = pages;
4916
4917         do {
4918                 len = min_t(size_t, PAGE_SIZE, buflen);
4919                 newpage = alloc_page(GFP_KERNEL);
4920
4921                 if (newpage == NULL)
4922                         goto unwind;
4923                 memcpy(page_address(newpage), buf, len);
4924                 buf += len;
4925                 buflen -= len;
4926                 *pages++ = newpage;
4927                 rc++;
4928         } while (buflen != 0);
4929
4930         return rc;
4931
4932 unwind:
4933         for(; rc > 0; rc--)
4934                 __free_page(spages[rc-1]);
4935         return -ENOMEM;
4936 }
4937
4938 struct nfs4_cached_acl {
4939         int cached;
4940         size_t len;
4941         char data[0];
4942 };
4943
4944 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4945 {
4946         struct nfs_inode *nfsi = NFS_I(inode);
4947
4948         spin_lock(&inode->i_lock);
4949         kfree(nfsi->nfs4_acl);
4950         nfsi->nfs4_acl = acl;
4951         spin_unlock(&inode->i_lock);
4952 }
4953
4954 static void nfs4_zap_acl_attr(struct inode *inode)
4955 {
4956         nfs4_set_cached_acl(inode, NULL);
4957 }
4958
4959 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4960 {
4961         struct nfs_inode *nfsi = NFS_I(inode);
4962         struct nfs4_cached_acl *acl;
4963         int ret = -ENOENT;
4964
4965         spin_lock(&inode->i_lock);
4966         acl = nfsi->nfs4_acl;
4967         if (acl == NULL)
4968                 goto out;
4969         if (buf == NULL) /* user is just asking for length */
4970                 goto out_len;
4971         if (acl->cached == 0)
4972                 goto out;
4973         ret = -ERANGE; /* see getxattr(2) man page */
4974         if (acl->len > buflen)
4975                 goto out;
4976         memcpy(buf, acl->data, acl->len);
4977 out_len:
4978         ret = acl->len;
4979 out:
4980         spin_unlock(&inode->i_lock);
4981         return ret;
4982 }
4983
4984 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4985 {
4986         struct nfs4_cached_acl *acl;
4987         size_t buflen = sizeof(*acl) + acl_len;
4988
4989         if (buflen <= PAGE_SIZE) {
4990                 acl = kmalloc(buflen, GFP_KERNEL);
4991                 if (acl == NULL)
4992                         goto out;
4993                 acl->cached = 1;
4994                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4995         } else {
4996                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4997                 if (acl == NULL)
4998                         goto out;
4999                 acl->cached = 0;
5000         }
5001         acl->len = acl_len;
5002 out:
5003         nfs4_set_cached_acl(inode, acl);
5004 }
5005
5006 /*
5007  * The getxattr API returns the required buffer length when called with a
5008  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5009  * the required buf.  On a NULL buf, we send a page of data to the server
5010  * guessing that the ACL request can be serviced by a page. If so, we cache
5011  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5012  * the cache. If not so, we throw away the page, and cache the required
5013  * length. The next getxattr call will then produce another round trip to
5014  * the server, this time with the input buf of the required size.
5015  */
5016 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5017 {
5018         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5019         struct nfs_getaclargs args = {
5020                 .fh = NFS_FH(inode),
5021                 .acl_pages = pages,
5022                 .acl_len = buflen,
5023         };
5024         struct nfs_getaclres res = {
5025                 .acl_len = buflen,
5026         };
5027         struct rpc_message msg = {
5028                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5029                 .rpc_argp = &args,
5030                 .rpc_resp = &res,
5031         };
5032         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5033         int ret = -ENOMEM, i;
5034
5035         if (npages > ARRAY_SIZE(pages))
5036                 return -ERANGE;
5037
5038         for (i = 0; i < npages; i++) {
5039                 pages[i] = alloc_page(GFP_KERNEL);
5040                 if (!pages[i])
5041                         goto out_free;
5042         }
5043
5044         /* for decoding across pages */
5045         res.acl_scratch = alloc_page(GFP_KERNEL);
5046         if (!res.acl_scratch)
5047                 goto out_free;
5048
5049         args.acl_len = npages * PAGE_SIZE;
5050
5051         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5052                 __func__, buf, buflen, npages, args.acl_len);
5053         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5054                              &msg, &args.seq_args, &res.seq_res, 0);
5055         if (ret)
5056                 goto out_free;
5057
5058         /* Handle the case where the passed-in buffer is too short */
5059         if (res.acl_flags & NFS4_ACL_TRUNC) {
5060                 /* Did the user only issue a request for the acl length? */
5061                 if (buf == NULL)
5062                         goto out_ok;
5063                 ret = -ERANGE;
5064                 goto out_free;
5065         }
5066         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5067         if (buf) {
5068                 if (res.acl_len > buflen) {
5069                         ret = -ERANGE;
5070                         goto out_free;
5071                 }
5072                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5073         }
5074 out_ok:
5075         ret = res.acl_len;
5076 out_free:
5077         for (i = 0; i < npages; i++)
5078                 if (pages[i])
5079                         __free_page(pages[i]);
5080         if (res.acl_scratch)
5081                 __free_page(res.acl_scratch);
5082         return ret;
5083 }
5084
5085 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5086 {
5087         struct nfs4_exception exception = { };
5088         ssize_t ret;
5089         do {
5090                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5091                 trace_nfs4_get_acl(inode, ret);
5092                 if (ret >= 0)
5093                         break;
5094                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5095         } while (exception.retry);
5096         return ret;
5097 }
5098
5099 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5100 {
5101         struct nfs_server *server = NFS_SERVER(inode);
5102         int ret;
5103
5104         if (!nfs4_server_supports_acls(server))
5105                 return -EOPNOTSUPP;
5106         ret = nfs_revalidate_inode(server, inode);
5107         if (ret < 0)
5108                 return ret;
5109         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5110                 nfs_zap_acl_cache(inode);
5111         ret = nfs4_read_cached_acl(inode, buf, buflen);
5112         if (ret != -ENOENT)
5113                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5114                  * but no cached acl data, just the acl length */
5115                 return ret;
5116         return nfs4_get_acl_uncached(inode, buf, buflen);
5117 }
5118
5119 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5120 {
5121         struct nfs_server *server = NFS_SERVER(inode);
5122         struct page *pages[NFS4ACL_MAXPAGES];
5123         struct nfs_setaclargs arg = {
5124                 .fh             = NFS_FH(inode),
5125                 .acl_pages      = pages,
5126                 .acl_len        = buflen,
5127         };
5128         struct nfs_setaclres res;
5129         struct rpc_message msg = {
5130                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5131                 .rpc_argp       = &arg,
5132                 .rpc_resp       = &res,
5133         };
5134         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5135         int ret, i;
5136
5137         if (!nfs4_server_supports_acls(server))
5138                 return -EOPNOTSUPP;
5139         if (npages > ARRAY_SIZE(pages))
5140                 return -ERANGE;
5141         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5142         if (i < 0)
5143                 return i;
5144         nfs4_inode_return_delegation(inode);
5145         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5146
5147         /*
5148          * Free each page after tx, so the only ref left is
5149          * held by the network stack
5150          */
5151         for (; i > 0; i--)
5152                 put_page(pages[i-1]);
5153
5154         /*
5155          * Acl update can result in inode attribute update.
5156          * so mark the attribute cache invalid.
5157          */
5158         spin_lock(&inode->i_lock);
5159         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5160         spin_unlock(&inode->i_lock);
5161         nfs_access_zap_cache(inode);
5162         nfs_zap_acl_cache(inode);
5163         return ret;
5164 }
5165
5166 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5167 {
5168         struct nfs4_exception exception = { };
5169         int err;
5170         do {
5171                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5172                 trace_nfs4_set_acl(inode, err);
5173                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5174                                 &exception);
5175         } while (exception.retry);
5176         return err;
5177 }
5178
5179 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5180 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5181                                         size_t buflen)
5182 {
5183         struct nfs_server *server = NFS_SERVER(inode);
5184         struct nfs_fattr fattr;
5185         struct nfs4_label label = {0, 0, buflen, buf};
5186
5187         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5188         struct nfs4_getattr_arg arg = {
5189                 .fh             = NFS_FH(inode),
5190                 .bitmask        = bitmask,
5191         };
5192         struct nfs4_getattr_res res = {
5193                 .fattr          = &fattr,
5194                 .label          = &label,
5195                 .server         = server,
5196         };
5197         struct rpc_message msg = {
5198                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5199                 .rpc_argp       = &arg,
5200                 .rpc_resp       = &res,
5201         };
5202         int ret;
5203
5204         nfs_fattr_init(&fattr);
5205
5206         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5207         if (ret)
5208                 return ret;
5209         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5210                 return -ENOENT;
5211         if (buflen < label.len)
5212                 return -ERANGE;
5213         return 0;
5214 }
5215
5216 static int nfs4_get_security_label(struct inode *inode, void *buf,
5217                                         size_t buflen)
5218 {
5219         struct nfs4_exception exception = { };
5220         int err;
5221
5222         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5223                 return -EOPNOTSUPP;
5224
5225         do {
5226                 err = _nfs4_get_security_label(inode, buf, buflen);
5227                 trace_nfs4_get_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 _nfs4_do_set_security_label(struct inode *inode,
5235                 struct nfs4_label *ilabel,
5236                 struct nfs_fattr *fattr,
5237                 struct nfs4_label *olabel)
5238 {
5239
5240         struct iattr sattr = {0};
5241         struct nfs_server *server = NFS_SERVER(inode);
5242         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5243         struct nfs_setattrargs arg = {
5244                 .fh             = NFS_FH(inode),
5245                 .iap            = &sattr,
5246                 .server         = server,
5247                 .bitmask        = bitmask,
5248                 .label          = ilabel,
5249         };
5250         struct nfs_setattrres res = {
5251                 .fattr          = fattr,
5252                 .label          = olabel,
5253                 .server         = server,
5254         };
5255         struct rpc_message msg = {
5256                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5257                 .rpc_argp       = &arg,
5258                 .rpc_resp       = &res,
5259         };
5260         int status;
5261
5262         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5263
5264         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5265         if (status)
5266                 dprintk("%s failed: %d\n", __func__, status);
5267
5268         return status;
5269 }
5270
5271 static int nfs4_do_set_security_label(struct inode *inode,
5272                 struct nfs4_label *ilabel,
5273                 struct nfs_fattr *fattr,
5274                 struct nfs4_label *olabel)
5275 {
5276         struct nfs4_exception exception = { };
5277         int err;
5278
5279         do {
5280                 err = _nfs4_do_set_security_label(inode, ilabel,
5281                                 fattr, olabel);
5282                 trace_nfs4_set_security_label(inode, err);
5283                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5284                                 &exception);
5285         } while (exception.retry);
5286         return err;
5287 }
5288
5289 static int
5290 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5291 {
5292         struct nfs4_label ilabel, *olabel = NULL;
5293         struct nfs_fattr fattr;
5294         struct rpc_cred *cred;
5295         int status;
5296
5297         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5298                 return -EOPNOTSUPP;
5299
5300         nfs_fattr_init(&fattr);
5301
5302         ilabel.pi = 0;
5303         ilabel.lfs = 0;
5304         ilabel.label = (char *)buf;
5305         ilabel.len = buflen;
5306
5307         cred = rpc_lookup_cred();
5308         if (IS_ERR(cred))
5309                 return PTR_ERR(cred);
5310
5311         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5312         if (IS_ERR(olabel)) {
5313                 status = -PTR_ERR(olabel);
5314                 goto out;
5315         }
5316
5317         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5318         if (status == 0)
5319                 nfs_setsecurity(inode, &fattr, olabel);
5320
5321         nfs4_label_free(olabel);
5322 out:
5323         put_rpccred(cred);
5324         return status;
5325 }
5326 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5327
5328
5329 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5330                                     nfs4_verifier *bootverf)
5331 {
5332         __be32 verf[2];
5333
5334         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5335                 /* An impossible timestamp guarantees this value
5336                  * will never match a generated boot time. */
5337                 verf[0] = cpu_to_be32(U32_MAX);
5338                 verf[1] = cpu_to_be32(U32_MAX);
5339         } else {
5340                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5341                 u64 ns = ktime_to_ns(nn->boot_time);
5342
5343                 verf[0] = cpu_to_be32(ns >> 32);
5344                 verf[1] = cpu_to_be32(ns);
5345         }
5346         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5347 }
5348
5349 static int
5350 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5351 {
5352         size_t len;
5353         char *str;
5354
5355         if (clp->cl_owner_id != NULL)
5356                 return 0;
5357
5358         rcu_read_lock();
5359         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5360                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5361                 1 +
5362                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5363                 1;
5364         rcu_read_unlock();
5365
5366         if (len > NFS4_OPAQUE_LIMIT + 1)
5367                 return -EINVAL;
5368
5369         /*
5370          * Since this string is allocated at mount time, and held until the
5371          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5372          * about a memory-reclaim deadlock.
5373          */
5374         str = kmalloc(len, GFP_KERNEL);
5375         if (!str)
5376                 return -ENOMEM;
5377
5378         rcu_read_lock();
5379         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5380                         clp->cl_ipaddr,
5381                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5382                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5383         rcu_read_unlock();
5384
5385         clp->cl_owner_id = str;
5386         return 0;
5387 }
5388
5389 static int
5390 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5391 {
5392         size_t len;
5393         char *str;
5394
5395         len = 10 + 10 + 1 + 10 + 1 +
5396                 strlen(nfs4_client_id_uniquifier) + 1 +
5397                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5398
5399         if (len > NFS4_OPAQUE_LIMIT + 1)
5400                 return -EINVAL;
5401
5402         /*
5403          * Since this string is allocated at mount time, and held until the
5404          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5405          * about a memory-reclaim deadlock.
5406          */
5407         str = kmalloc(len, GFP_KERNEL);
5408         if (!str)
5409                 return -ENOMEM;
5410
5411         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5412                         clp->rpc_ops->version, clp->cl_minorversion,
5413                         nfs4_client_id_uniquifier,
5414                         clp->cl_rpcclient->cl_nodename);
5415         clp->cl_owner_id = str;
5416         return 0;
5417 }
5418
5419 static int
5420 nfs4_init_uniform_client_string(struct nfs_client *clp)
5421 {
5422         size_t len;
5423         char *str;
5424
5425         if (clp->cl_owner_id != NULL)
5426                 return 0;
5427
5428         if (nfs4_client_id_uniquifier[0] != '\0')
5429                 return nfs4_init_uniquifier_client_string(clp);
5430
5431         len = 10 + 10 + 1 + 10 + 1 +
5432                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5433
5434         if (len > NFS4_OPAQUE_LIMIT + 1)
5435                 return -EINVAL;
5436
5437         /*
5438          * Since this string is allocated at mount time, and held until the
5439          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5440          * about a memory-reclaim deadlock.
5441          */
5442         str = kmalloc(len, GFP_KERNEL);
5443         if (!str)
5444                 return -ENOMEM;
5445
5446         scnprintf(str, len, "Linux NFSv%u.%u %s",
5447                         clp->rpc_ops->version, clp->cl_minorversion,
5448                         clp->cl_rpcclient->cl_nodename);
5449         clp->cl_owner_id = str;
5450         return 0;
5451 }
5452
5453 /*
5454  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5455  * services.  Advertise one based on the address family of the
5456  * clientaddr.
5457  */
5458 static unsigned int
5459 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5460 {
5461         if (strchr(clp->cl_ipaddr, ':') != NULL)
5462                 return scnprintf(buf, len, "tcp6");
5463         else
5464                 return scnprintf(buf, len, "tcp");
5465 }
5466
5467 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5468 {
5469         struct nfs4_setclientid *sc = calldata;
5470
5471         if (task->tk_status == 0)
5472                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5473 }
5474
5475 static const struct rpc_call_ops nfs4_setclientid_ops = {
5476         .rpc_call_done = nfs4_setclientid_done,
5477 };
5478
5479 /**
5480  * nfs4_proc_setclientid - Negotiate client ID
5481  * @clp: state data structure
5482  * @program: RPC program for NFSv4 callback service
5483  * @port: IP port number for NFS4 callback service
5484  * @cred: RPC credential to use for this call
5485  * @res: where to place the result
5486  *
5487  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5488  */
5489 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5490                 unsigned short port, struct rpc_cred *cred,
5491                 struct nfs4_setclientid_res *res)
5492 {
5493         nfs4_verifier sc_verifier;
5494         struct nfs4_setclientid setclientid = {
5495                 .sc_verifier = &sc_verifier,
5496                 .sc_prog = program,
5497                 .sc_clnt = clp,
5498         };
5499         struct rpc_message msg = {
5500                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5501                 .rpc_argp = &setclientid,
5502                 .rpc_resp = res,
5503                 .rpc_cred = cred,
5504         };
5505         struct rpc_task *task;
5506         struct rpc_task_setup task_setup_data = {
5507                 .rpc_client = clp->cl_rpcclient,
5508                 .rpc_message = &msg,
5509                 .callback_ops = &nfs4_setclientid_ops,
5510                 .callback_data = &setclientid,
5511                 .flags = RPC_TASK_TIMEOUT,
5512         };
5513         int status;
5514
5515         /* nfs_client_id4 */
5516         nfs4_init_boot_verifier(clp, &sc_verifier);
5517
5518         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5519                 status = nfs4_init_uniform_client_string(clp);
5520         else
5521                 status = nfs4_init_nonuniform_client_string(clp);
5522
5523         if (status)
5524                 goto out;
5525
5526         /* cb_client4 */
5527         setclientid.sc_netid_len =
5528                                 nfs4_init_callback_netid(clp,
5529                                                 setclientid.sc_netid,
5530                                                 sizeof(setclientid.sc_netid));
5531         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5532                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5533                                 clp->cl_ipaddr, port >> 8, port & 255);
5534
5535         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5536                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5537                 clp->cl_owner_id);
5538         task = rpc_run_task(&task_setup_data);
5539         if (IS_ERR(task)) {
5540                 status = PTR_ERR(task);
5541                 goto out;
5542         }
5543         status = task->tk_status;
5544         if (setclientid.sc_cred) {
5545                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5546                 put_rpccred(setclientid.sc_cred);
5547         }
5548         rpc_put_task(task);
5549 out:
5550         trace_nfs4_setclientid(clp, status);
5551         dprintk("NFS reply setclientid: %d\n", status);
5552         return status;
5553 }
5554
5555 /**
5556  * nfs4_proc_setclientid_confirm - Confirm client ID
5557  * @clp: state data structure
5558  * @res: result of a previous SETCLIENTID
5559  * @cred: RPC credential to use for this call
5560  *
5561  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5562  */
5563 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5564                 struct nfs4_setclientid_res *arg,
5565                 struct rpc_cred *cred)
5566 {
5567         struct rpc_message msg = {
5568                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5569                 .rpc_argp = arg,
5570                 .rpc_cred = cred,
5571         };
5572         int status;
5573
5574         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5575                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5576                 clp->cl_clientid);
5577         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5578         trace_nfs4_setclientid_confirm(clp, status);
5579         dprintk("NFS reply setclientid_confirm: %d\n", status);
5580         return status;
5581 }
5582
5583 struct nfs4_delegreturndata {
5584         struct nfs4_delegreturnargs args;
5585         struct nfs4_delegreturnres res;
5586         struct nfs_fh fh;
5587         nfs4_stateid stateid;
5588         unsigned long timestamp;
5589         struct {
5590                 struct nfs4_layoutreturn_args arg;
5591                 struct nfs4_layoutreturn_res res;
5592                 struct nfs4_xdr_opaque_data ld_private;
5593                 u32 roc_barrier;
5594                 bool roc;
5595         } lr;
5596         struct nfs_fattr fattr;
5597         int rpc_status;
5598         struct inode *inode;
5599 };
5600
5601 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5602 {
5603         struct nfs4_delegreturndata *data = calldata;
5604
5605         if (!nfs4_sequence_done(task, &data->res.seq_res))
5606                 return;
5607
5608         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5609
5610         /* Handle Layoutreturn errors */
5611         if (data->args.lr_args && task->tk_status != 0) {
5612                 switch(data->res.lr_ret) {
5613                 default:
5614                         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5615                         break;
5616                 case 0:
5617                         data->args.lr_args = NULL;
5618                         data->res.lr_res = NULL;
5619                         break;
5620                 case -NFS4ERR_ADMIN_REVOKED:
5621                 case -NFS4ERR_DELEG_REVOKED:
5622                 case -NFS4ERR_EXPIRED:
5623                 case -NFS4ERR_BAD_STATEID:
5624                 case -NFS4ERR_OLD_STATEID:
5625                 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5626                 case -NFS4ERR_WRONG_CRED:
5627                         data->args.lr_args = NULL;
5628                         data->res.lr_res = NULL;
5629                         data->res.lr_ret = 0;
5630                         rpc_restart_call_prepare(task);
5631                         return;
5632                 }
5633         }
5634
5635         switch (task->tk_status) {
5636         case 0:
5637                 renew_lease(data->res.server, data->timestamp);
5638                 break;
5639         case -NFS4ERR_ADMIN_REVOKED:
5640         case -NFS4ERR_DELEG_REVOKED:
5641         case -NFS4ERR_EXPIRED:
5642                 nfs4_free_revoked_stateid(data->res.server,
5643                                 data->args.stateid,
5644                                 task->tk_msg.rpc_cred);
5645         case -NFS4ERR_BAD_STATEID:
5646         case -NFS4ERR_OLD_STATEID:
5647         case -NFS4ERR_STALE_STATEID:
5648                 task->tk_status = 0;
5649                 break;
5650         case -NFS4ERR_ACCESS:
5651                 if (data->args.bitmask) {
5652                         data->args.bitmask = NULL;
5653                         data->res.fattr = NULL;
5654                         task->tk_status = 0;
5655                         rpc_restart_call_prepare(task);
5656                         return;
5657                 }
5658         default:
5659                 if (nfs4_async_handle_error(task, data->res.server,
5660                                             NULL, NULL) == -EAGAIN) {
5661                         rpc_restart_call_prepare(task);
5662                         return;
5663                 }
5664         }
5665         data->rpc_status = task->tk_status;
5666 }
5667
5668 static void nfs4_delegreturn_release(void *calldata)
5669 {
5670         struct nfs4_delegreturndata *data = calldata;
5671         struct inode *inode = data->inode;
5672
5673         if (inode) {
5674                 if (data->lr.roc)
5675                         pnfs_roc_release(&data->lr.arg, &data->lr.res,
5676                                         data->res.lr_ret);
5677                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5678                 nfs_iput_and_deactive(inode);
5679         }
5680         kfree(calldata);
5681 }
5682
5683 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5684 {
5685         struct nfs4_delegreturndata *d_data;
5686
5687         d_data = (struct nfs4_delegreturndata *)data;
5688
5689         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5690                 return;
5691
5692         nfs4_setup_sequence(d_data->res.server->nfs_client,
5693                         &d_data->args.seq_args,
5694                         &d_data->res.seq_res,
5695                         task);
5696 }
5697
5698 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5699         .rpc_call_prepare = nfs4_delegreturn_prepare,
5700         .rpc_call_done = nfs4_delegreturn_done,
5701         .rpc_release = nfs4_delegreturn_release,
5702 };
5703
5704 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5705 {
5706         struct nfs4_delegreturndata *data;
5707         struct nfs_server *server = NFS_SERVER(inode);
5708         struct rpc_task *task;
5709         struct rpc_message msg = {
5710                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5711                 .rpc_cred = cred,
5712         };
5713         struct rpc_task_setup task_setup_data = {
5714                 .rpc_client = server->client,
5715                 .rpc_message = &msg,
5716                 .callback_ops = &nfs4_delegreturn_ops,
5717                 .flags = RPC_TASK_ASYNC,
5718         };
5719         int status = 0;
5720
5721         data = kzalloc(sizeof(*data), GFP_NOFS);
5722         if (data == NULL)
5723                 return -ENOMEM;
5724         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5725
5726         nfs4_state_protect(server->nfs_client,
5727                         NFS_SP4_MACH_CRED_CLEANUP,
5728                         &task_setup_data.rpc_client, &msg);
5729
5730         data->args.fhandle = &data->fh;
5731         data->args.stateid = &data->stateid;
5732         data->args.bitmask = server->cache_consistency_bitmask;
5733         nfs_copy_fh(&data->fh, NFS_FH(inode));
5734         nfs4_stateid_copy(&data->stateid, stateid);
5735         data->res.fattr = &data->fattr;
5736         data->res.server = server;
5737         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5738         data->lr.arg.ld_private = &data->lr.ld_private;
5739         nfs_fattr_init(data->res.fattr);
5740         data->timestamp = jiffies;
5741         data->rpc_status = 0;
5742         data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5743         data->inode = nfs_igrab_and_active(inode);
5744         if (data->inode) {
5745                 if (data->lr.roc) {
5746                         data->args.lr_args = &data->lr.arg;
5747                         data->res.lr_res = &data->lr.res;
5748                 }
5749         } else if (data->lr.roc) {
5750                 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5751                 data->lr.roc = false;
5752         }
5753
5754         task_setup_data.callback_data = data;
5755         msg.rpc_argp = &data->args;
5756         msg.rpc_resp = &data->res;
5757         task = rpc_run_task(&task_setup_data);
5758         if (IS_ERR(task))
5759                 return PTR_ERR(task);
5760         if (!issync)
5761                 goto out;
5762         status = rpc_wait_for_completion_task(task);
5763         if (status != 0)
5764                 goto out;
5765         status = data->rpc_status;
5766 out:
5767         rpc_put_task(task);
5768         return status;
5769 }
5770
5771 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5772 {
5773         struct nfs_server *server = NFS_SERVER(inode);
5774         struct nfs4_exception exception = { };
5775         int err;
5776         do {
5777                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5778                 trace_nfs4_delegreturn(inode, stateid, err);
5779                 switch (err) {
5780                         case -NFS4ERR_STALE_STATEID:
5781                         case -NFS4ERR_EXPIRED:
5782                         case 0:
5783                                 return 0;
5784                 }
5785                 err = nfs4_handle_exception(server, err, &exception);
5786         } while (exception.retry);
5787         return err;
5788 }
5789
5790 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5791 {
5792         struct inode *inode = state->inode;
5793         struct nfs_server *server = NFS_SERVER(inode);
5794         struct nfs_client *clp = server->nfs_client;
5795         struct nfs_lockt_args arg = {
5796                 .fh = NFS_FH(inode),
5797                 .fl = request,
5798         };
5799         struct nfs_lockt_res res = {
5800                 .denied = request,
5801         };
5802         struct rpc_message msg = {
5803                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5804                 .rpc_argp       = &arg,
5805                 .rpc_resp       = &res,
5806                 .rpc_cred       = state->owner->so_cred,
5807         };
5808         struct nfs4_lock_state *lsp;
5809         int status;
5810
5811         arg.lock_owner.clientid = clp->cl_clientid;
5812         status = nfs4_set_lock_state(state, request);
5813         if (status != 0)
5814                 goto out;
5815         lsp = request->fl_u.nfs4_fl.owner;
5816         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5817         arg.lock_owner.s_dev = server->s_dev;
5818         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5819         switch (status) {
5820                 case 0:
5821                         request->fl_type = F_UNLCK;
5822                         break;
5823                 case -NFS4ERR_DENIED:
5824                         status = 0;
5825         }
5826         request->fl_ops->fl_release_private(request);
5827         request->fl_ops = NULL;
5828 out:
5829         return status;
5830 }
5831
5832 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5833 {
5834         struct nfs4_exception exception = { };
5835         int err;
5836
5837         do {
5838                 err = _nfs4_proc_getlk(state, cmd, request);
5839                 trace_nfs4_get_lock(request, state, cmd, err);
5840                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5841                                 &exception);
5842         } while (exception.retry);
5843         return err;
5844 }
5845
5846 struct nfs4_unlockdata {
5847         struct nfs_locku_args arg;
5848         struct nfs_locku_res res;
5849         struct nfs4_lock_state *lsp;
5850         struct nfs_open_context *ctx;
5851         struct nfs_lock_context *l_ctx;
5852         struct file_lock fl;
5853         struct nfs_server *server;
5854         unsigned long timestamp;
5855 };
5856
5857 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5858                 struct nfs_open_context *ctx,
5859                 struct nfs4_lock_state *lsp,
5860                 struct nfs_seqid *seqid)
5861 {
5862         struct nfs4_unlockdata *p;
5863         struct inode *inode = lsp->ls_state->inode;
5864
5865         p = kzalloc(sizeof(*p), GFP_NOFS);
5866         if (p == NULL)
5867                 return NULL;
5868         p->arg.fh = NFS_FH(inode);
5869         p->arg.fl = &p->fl;
5870         p->arg.seqid = seqid;
5871         p->res.seqid = seqid;
5872         p->lsp = lsp;
5873         atomic_inc(&lsp->ls_count);
5874         /* Ensure we don't close file until we're done freeing locks! */
5875         p->ctx = get_nfs_open_context(ctx);
5876         p->l_ctx = nfs_get_lock_context(ctx);
5877         memcpy(&p->fl, fl, sizeof(p->fl));
5878         p->server = NFS_SERVER(inode);
5879         return p;
5880 }
5881
5882 static void nfs4_locku_release_calldata(void *data)
5883 {
5884         struct nfs4_unlockdata *calldata = data;
5885         nfs_free_seqid(calldata->arg.seqid);
5886         nfs4_put_lock_state(calldata->lsp);
5887         nfs_put_lock_context(calldata->l_ctx);
5888         put_nfs_open_context(calldata->ctx);
5889         kfree(calldata);
5890 }
5891
5892 static void nfs4_locku_done(struct rpc_task *task, void *data)
5893 {
5894         struct nfs4_unlockdata *calldata = data;
5895
5896         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5897                 return;
5898         switch (task->tk_status) {
5899                 case 0:
5900                         renew_lease(calldata->server, calldata->timestamp);
5901                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5902                         if (nfs4_update_lock_stateid(calldata->lsp,
5903                                         &calldata->res.stateid))
5904                                 break;
5905                 case -NFS4ERR_ADMIN_REVOKED:
5906                 case -NFS4ERR_EXPIRED:
5907                         nfs4_free_revoked_stateid(calldata->server,
5908                                         &calldata->arg.stateid,
5909                                         task->tk_msg.rpc_cred);
5910                 case -NFS4ERR_BAD_STATEID:
5911                 case -NFS4ERR_OLD_STATEID:
5912                 case -NFS4ERR_STALE_STATEID:
5913                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5914                                                 &calldata->lsp->ls_stateid))
5915                                 rpc_restart_call_prepare(task);
5916                         break;
5917                 default:
5918                         if (nfs4_async_handle_error(task, calldata->server,
5919                                                     NULL, NULL) == -EAGAIN)
5920                                 rpc_restart_call_prepare(task);
5921         }
5922         nfs_release_seqid(calldata->arg.seqid);
5923 }
5924
5925 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5926 {
5927         struct nfs4_unlockdata *calldata = data;
5928
5929         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
5930                 nfs_async_iocounter_wait(task, calldata->l_ctx))
5931                 return;
5932
5933         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5934                 goto out_wait;
5935         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5936         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5937                 /* Note: exit _without_ running nfs4_locku_done */
5938                 goto out_no_action;
5939         }
5940         calldata->timestamp = jiffies;
5941         if (nfs4_setup_sequence(calldata->server->nfs_client,
5942                                 &calldata->arg.seq_args,
5943                                 &calldata->res.seq_res,
5944                                 task) != 0)
5945                 nfs_release_seqid(calldata->arg.seqid);
5946         return;
5947 out_no_action:
5948         task->tk_action = NULL;
5949 out_wait:
5950         nfs4_sequence_done(task, &calldata->res.seq_res);
5951 }
5952
5953 static const struct rpc_call_ops nfs4_locku_ops = {
5954         .rpc_call_prepare = nfs4_locku_prepare,
5955         .rpc_call_done = nfs4_locku_done,
5956         .rpc_release = nfs4_locku_release_calldata,
5957 };
5958
5959 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5960                 struct nfs_open_context *ctx,
5961                 struct nfs4_lock_state *lsp,
5962                 struct nfs_seqid *seqid)
5963 {
5964         struct nfs4_unlockdata *data;
5965         struct rpc_message msg = {
5966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5967                 .rpc_cred = ctx->cred,
5968         };
5969         struct rpc_task_setup task_setup_data = {
5970                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5971                 .rpc_message = &msg,
5972                 .callback_ops = &nfs4_locku_ops,
5973                 .workqueue = nfsiod_workqueue,
5974                 .flags = RPC_TASK_ASYNC,
5975         };
5976
5977         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5978                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5979
5980         /* Ensure this is an unlock - when canceling a lock, the
5981          * canceled lock is passed in, and it won't be an unlock.
5982          */
5983         fl->fl_type = F_UNLCK;
5984         if (fl->fl_flags & FL_CLOSE)
5985                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
5986
5987         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5988         if (data == NULL) {
5989                 nfs_free_seqid(seqid);
5990                 return ERR_PTR(-ENOMEM);
5991         }
5992
5993         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5994         msg.rpc_argp = &data->arg;
5995         msg.rpc_resp = &data->res;
5996         task_setup_data.callback_data = data;
5997         return rpc_run_task(&task_setup_data);
5998 }
5999
6000 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6001 {
6002         struct inode *inode = state->inode;
6003         struct nfs4_state_owner *sp = state->owner;
6004         struct nfs_inode *nfsi = NFS_I(inode);
6005         struct nfs_seqid *seqid;
6006         struct nfs4_lock_state *lsp;
6007         struct rpc_task *task;
6008         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6009         int status = 0;
6010         unsigned char fl_flags = request->fl_flags;
6011
6012         status = nfs4_set_lock_state(state, request);
6013         /* Unlock _before_ we do the RPC call */
6014         request->fl_flags |= FL_EXISTS;
6015         /* Exclude nfs_delegation_claim_locks() */
6016         mutex_lock(&sp->so_delegreturn_mutex);
6017         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6018         down_read(&nfsi->rwsem);
6019         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6020                 up_read(&nfsi->rwsem);
6021                 mutex_unlock(&sp->so_delegreturn_mutex);
6022                 goto out;
6023         }
6024         up_read(&nfsi->rwsem);
6025         mutex_unlock(&sp->so_delegreturn_mutex);
6026         if (status != 0)
6027                 goto out;
6028         /* Is this a delegated lock? */
6029         lsp = request->fl_u.nfs4_fl.owner;
6030         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6031                 goto out;
6032         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6033         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6034         status = -ENOMEM;
6035         if (IS_ERR(seqid))
6036                 goto out;
6037         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6038         status = PTR_ERR(task);
6039         if (IS_ERR(task))
6040                 goto out;
6041         status = rpc_wait_for_completion_task(task);
6042         rpc_put_task(task);
6043 out:
6044         request->fl_flags = fl_flags;
6045         trace_nfs4_unlock(request, state, F_SETLK, status);
6046         return status;
6047 }
6048
6049 struct nfs4_lockdata {
6050         struct nfs_lock_args arg;
6051         struct nfs_lock_res res;
6052         struct nfs4_lock_state *lsp;
6053         struct nfs_open_context *ctx;
6054         struct file_lock fl;
6055         unsigned long timestamp;
6056         int rpc_status;
6057         int cancelled;
6058         struct nfs_server *server;
6059 };
6060
6061 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6062                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6063                 gfp_t gfp_mask)
6064 {
6065         struct nfs4_lockdata *p;
6066         struct inode *inode = lsp->ls_state->inode;
6067         struct nfs_server *server = NFS_SERVER(inode);
6068         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6069
6070         p = kzalloc(sizeof(*p), gfp_mask);
6071         if (p == NULL)
6072                 return NULL;
6073
6074         p->arg.fh = NFS_FH(inode);
6075         p->arg.fl = &p->fl;
6076         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6077         if (IS_ERR(p->arg.open_seqid))
6078                 goto out_free;
6079         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6080         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6081         if (IS_ERR(p->arg.lock_seqid))
6082                 goto out_free_seqid;
6083         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6084         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6085         p->arg.lock_owner.s_dev = server->s_dev;
6086         p->res.lock_seqid = p->arg.lock_seqid;
6087         p->lsp = lsp;
6088         p->server = server;
6089         atomic_inc(&lsp->ls_count);
6090         p->ctx = get_nfs_open_context(ctx);
6091         memcpy(&p->fl, fl, sizeof(p->fl));
6092         return p;
6093 out_free_seqid:
6094         nfs_free_seqid(p->arg.open_seqid);
6095 out_free:
6096         kfree(p);
6097         return NULL;
6098 }
6099
6100 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6101 {
6102         struct nfs4_lockdata *data = calldata;
6103         struct nfs4_state *state = data->lsp->ls_state;
6104
6105         dprintk("%s: begin!\n", __func__);
6106         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6107                 goto out_wait;
6108         /* Do we need to do an open_to_lock_owner? */
6109         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6110                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6111                         goto out_release_lock_seqid;
6112                 }
6113                 nfs4_stateid_copy(&data->arg.open_stateid,
6114                                 &state->open_stateid);
6115                 data->arg.new_lock_owner = 1;
6116                 data->res.open_seqid = data->arg.open_seqid;
6117         } else {
6118                 data->arg.new_lock_owner = 0;
6119                 nfs4_stateid_copy(&data->arg.lock_stateid,
6120                                 &data->lsp->ls_stateid);
6121         }
6122         if (!nfs4_valid_open_stateid(state)) {
6123                 data->rpc_status = -EBADF;
6124                 task->tk_action = NULL;
6125                 goto out_release_open_seqid;
6126         }
6127         data->timestamp = jiffies;
6128         if (nfs4_setup_sequence(data->server->nfs_client,
6129                                 &data->arg.seq_args,
6130                                 &data->res.seq_res,
6131                                 task) == 0)
6132                 return;
6133 out_release_open_seqid:
6134         nfs_release_seqid(data->arg.open_seqid);
6135 out_release_lock_seqid:
6136         nfs_release_seqid(data->arg.lock_seqid);
6137 out_wait:
6138         nfs4_sequence_done(task, &data->res.seq_res);
6139         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6140 }
6141
6142 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6143 {
6144         struct nfs4_lockdata *data = calldata;
6145         struct nfs4_lock_state *lsp = data->lsp;
6146
6147         dprintk("%s: begin!\n", __func__);
6148
6149         if (!nfs4_sequence_done(task, &data->res.seq_res))
6150                 return;
6151
6152         data->rpc_status = task->tk_status;
6153         switch (task->tk_status) {
6154         case 0:
6155                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6156                                 data->timestamp);
6157                 if (data->arg.new_lock) {
6158                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6159                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6160                                 rpc_restart_call_prepare(task);
6161                                 break;
6162                         }
6163                 }
6164                 if (data->arg.new_lock_owner != 0) {
6165                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6166                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6167                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6168                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6169                         rpc_restart_call_prepare(task);
6170                 break;
6171         case -NFS4ERR_BAD_STATEID:
6172         case -NFS4ERR_OLD_STATEID:
6173         case -NFS4ERR_STALE_STATEID:
6174         case -NFS4ERR_EXPIRED:
6175                 if (data->arg.new_lock_owner != 0) {
6176                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6177                                                 &lsp->ls_state->open_stateid))
6178                                 rpc_restart_call_prepare(task);
6179                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6180                                                 &lsp->ls_stateid))
6181                                 rpc_restart_call_prepare(task);
6182         }
6183         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6184 }
6185
6186 static void nfs4_lock_release(void *calldata)
6187 {
6188         struct nfs4_lockdata *data = calldata;
6189
6190         dprintk("%s: begin!\n", __func__);
6191         nfs_free_seqid(data->arg.open_seqid);
6192         if (data->cancelled) {
6193                 struct rpc_task *task;
6194                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6195                                 data->arg.lock_seqid);
6196                 if (!IS_ERR(task))
6197                         rpc_put_task_async(task);
6198                 dprintk("%s: cancelling lock!\n", __func__);
6199         } else
6200                 nfs_free_seqid(data->arg.lock_seqid);
6201         nfs4_put_lock_state(data->lsp);
6202         put_nfs_open_context(data->ctx);
6203         kfree(data);
6204         dprintk("%s: done!\n", __func__);
6205 }
6206
6207 static const struct rpc_call_ops nfs4_lock_ops = {
6208         .rpc_call_prepare = nfs4_lock_prepare,
6209         .rpc_call_done = nfs4_lock_done,
6210         .rpc_release = nfs4_lock_release,
6211 };
6212
6213 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6214 {
6215         switch (error) {
6216         case -NFS4ERR_ADMIN_REVOKED:
6217         case -NFS4ERR_EXPIRED:
6218         case -NFS4ERR_BAD_STATEID:
6219                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6220                 if (new_lock_owner != 0 ||
6221                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6222                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6223                 break;
6224         case -NFS4ERR_STALE_STATEID:
6225                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6226                 nfs4_schedule_lease_recovery(server->nfs_client);
6227         };
6228 }
6229
6230 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6231 {
6232         struct nfs4_lockdata *data;
6233         struct rpc_task *task;
6234         struct rpc_message msg = {
6235                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6236                 .rpc_cred = state->owner->so_cred,
6237         };
6238         struct rpc_task_setup task_setup_data = {
6239                 .rpc_client = NFS_CLIENT(state->inode),
6240                 .rpc_message = &msg,
6241                 .callback_ops = &nfs4_lock_ops,
6242                 .workqueue = nfsiod_workqueue,
6243                 .flags = RPC_TASK_ASYNC,
6244         };
6245         int ret;
6246
6247         dprintk("%s: begin!\n", __func__);
6248         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6249                         fl->fl_u.nfs4_fl.owner,
6250                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6251         if (data == NULL)
6252                 return -ENOMEM;
6253         if (IS_SETLKW(cmd))
6254                 data->arg.block = 1;
6255         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6256         msg.rpc_argp = &data->arg;
6257         msg.rpc_resp = &data->res;
6258         task_setup_data.callback_data = data;
6259         if (recovery_type > NFS_LOCK_NEW) {
6260                 if (recovery_type == NFS_LOCK_RECLAIM)
6261                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6262                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6263         } else
6264                 data->arg.new_lock = 1;
6265         task = rpc_run_task(&task_setup_data);
6266         if (IS_ERR(task))
6267                 return PTR_ERR(task);
6268         ret = rpc_wait_for_completion_task(task);
6269         if (ret == 0) {
6270                 ret = data->rpc_status;
6271                 if (ret)
6272                         nfs4_handle_setlk_error(data->server, data->lsp,
6273                                         data->arg.new_lock_owner, ret);
6274         } else
6275                 data->cancelled = true;
6276         rpc_put_task(task);
6277         dprintk("%s: done, ret = %d!\n", __func__, ret);
6278         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6279         return ret;
6280 }
6281
6282 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6283 {
6284         struct nfs_server *server = NFS_SERVER(state->inode);
6285         struct nfs4_exception exception = {
6286                 .inode = state->inode,
6287         };
6288         int err;
6289
6290         do {
6291                 /* Cache the lock if possible... */
6292                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6293                         return 0;
6294                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6295                 if (err != -NFS4ERR_DELAY)
6296                         break;
6297                 nfs4_handle_exception(server, err, &exception);
6298         } while (exception.retry);
6299         return err;
6300 }
6301
6302 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6303 {
6304         struct nfs_server *server = NFS_SERVER(state->inode);
6305         struct nfs4_exception exception = {
6306                 .inode = state->inode,
6307         };
6308         int err;
6309
6310         err = nfs4_set_lock_state(state, request);
6311         if (err != 0)
6312                 return err;
6313         if (!recover_lost_locks) {
6314                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6315                 return 0;
6316         }
6317         do {
6318                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6319                         return 0;
6320                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6321                 switch (err) {
6322                 default:
6323                         goto out;
6324                 case -NFS4ERR_GRACE:
6325                 case -NFS4ERR_DELAY:
6326                         nfs4_handle_exception(server, err, &exception);
6327                         err = 0;
6328                 }
6329         } while (exception.retry);
6330 out:
6331         return err;
6332 }
6333
6334 #if defined(CONFIG_NFS_V4_1)
6335 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6336 {
6337         struct nfs4_lock_state *lsp;
6338         int status;
6339
6340         status = nfs4_set_lock_state(state, request);
6341         if (status != 0)
6342                 return status;
6343         lsp = request->fl_u.nfs4_fl.owner;
6344         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6345             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6346                 return 0;
6347         return nfs4_lock_expired(state, request);
6348 }
6349 #endif
6350
6351 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6352 {
6353         struct nfs_inode *nfsi = NFS_I(state->inode);
6354         struct nfs4_state_owner *sp = state->owner;
6355         unsigned char fl_flags = request->fl_flags;
6356         int status;
6357
6358         request->fl_flags |= FL_ACCESS;
6359         status = locks_lock_inode_wait(state->inode, request);
6360         if (status < 0)
6361                 goto out;
6362         mutex_lock(&sp->so_delegreturn_mutex);
6363         down_read(&nfsi->rwsem);
6364         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6365                 /* Yes: cache locks! */
6366                 /* ...but avoid races with delegation recall... */
6367                 request->fl_flags = fl_flags & ~FL_SLEEP;
6368                 status = locks_lock_inode_wait(state->inode, request);
6369                 up_read(&nfsi->rwsem);
6370                 mutex_unlock(&sp->so_delegreturn_mutex);
6371                 goto out;
6372         }
6373         up_read(&nfsi->rwsem);
6374         mutex_unlock(&sp->so_delegreturn_mutex);
6375         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6376 out:
6377         request->fl_flags = fl_flags;
6378         return status;
6379 }
6380
6381 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6382 {
6383         struct nfs4_exception exception = {
6384                 .state = state,
6385                 .inode = state->inode,
6386         };
6387         int err;
6388
6389         do {
6390                 err = _nfs4_proc_setlk(state, cmd, request);
6391                 if (err == -NFS4ERR_DENIED)
6392                         err = -EAGAIN;
6393                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6394                                 err, &exception);
6395         } while (exception.retry);
6396         return err;
6397 }
6398
6399 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6400 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6401
6402 static int
6403 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6404                         struct file_lock *request)
6405 {
6406         int             status = -ERESTARTSYS;
6407         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6408
6409         while(!signalled()) {
6410                 status = nfs4_proc_setlk(state, cmd, request);
6411                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6412                         break;
6413                 freezable_schedule_timeout_interruptible(timeout);
6414                 timeout *= 2;
6415                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6416                 status = -ERESTARTSYS;
6417         }
6418         return status;
6419 }
6420
6421 #ifdef CONFIG_NFS_V4_1
6422 struct nfs4_lock_waiter {
6423         struct task_struct      *task;
6424         struct inode            *inode;
6425         struct nfs_lowner       *owner;
6426         bool                    notified;
6427 };
6428
6429 static int
6430 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
6431 {
6432         int ret;
6433         struct cb_notify_lock_args *cbnl = key;
6434         struct nfs4_lock_waiter *waiter = wait->private;
6435         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6436                                 *wowner = waiter->owner;
6437
6438         /* Only wake if the callback was for the same owner */
6439         if (lowner->clientid != wowner->clientid ||
6440             lowner->id != wowner->id             ||
6441             lowner->s_dev != wowner->s_dev)
6442                 return 0;
6443
6444         /* Make sure it's for the right inode */
6445         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6446                 return 0;
6447
6448         waiter->notified = true;
6449
6450         /* override "private" so we can use default_wake_function */
6451         wait->private = waiter->task;
6452         ret = autoremove_wake_function(wait, mode, flags, key);
6453         wait->private = waiter;
6454         return ret;
6455 }
6456
6457 static int
6458 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6459 {
6460         int status = -ERESTARTSYS;
6461         unsigned long flags;
6462         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6463         struct nfs_server *server = NFS_SERVER(state->inode);
6464         struct nfs_client *clp = server->nfs_client;
6465         wait_queue_head_t *q = &clp->cl_lock_waitq;
6466         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6467                                     .id = lsp->ls_seqid.owner_id,
6468                                     .s_dev = server->s_dev };
6469         struct nfs4_lock_waiter waiter = { .task  = current,
6470                                            .inode = state->inode,
6471                                            .owner = &owner,
6472                                            .notified = false };
6473         wait_queue_entry_t wait;
6474
6475         /* Don't bother with waitqueue if we don't expect a callback */
6476         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6477                 return nfs4_retry_setlk_simple(state, cmd, request);
6478
6479         init_wait(&wait);
6480         wait.private = &waiter;
6481         wait.func = nfs4_wake_lock_waiter;
6482         add_wait_queue(q, &wait);
6483
6484         while(!signalled()) {
6485                 status = nfs4_proc_setlk(state, cmd, request);
6486                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6487                         break;
6488
6489                 status = -ERESTARTSYS;
6490                 spin_lock_irqsave(&q->lock, flags);
6491                 if (waiter.notified) {
6492                         spin_unlock_irqrestore(&q->lock, flags);
6493                         continue;
6494                 }
6495                 set_current_state(TASK_INTERRUPTIBLE);
6496                 spin_unlock_irqrestore(&q->lock, flags);
6497
6498                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6499         }
6500
6501         finish_wait(q, &wait);
6502         return status;
6503 }
6504 #else /* !CONFIG_NFS_V4_1 */
6505 static inline int
6506 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6507 {
6508         return nfs4_retry_setlk_simple(state, cmd, request);
6509 }
6510 #endif
6511
6512 static int
6513 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6514 {
6515         struct nfs_open_context *ctx;
6516         struct nfs4_state *state;
6517         int status;
6518
6519         /* verify open state */
6520         ctx = nfs_file_open_context(filp);
6521         state = ctx->state;
6522
6523         if (IS_GETLK(cmd)) {
6524                 if (state != NULL)
6525                         return nfs4_proc_getlk(state, F_GETLK, request);
6526                 return 0;
6527         }
6528
6529         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6530                 return -EINVAL;
6531
6532         if (request->fl_type == F_UNLCK) {
6533                 if (state != NULL)
6534                         return nfs4_proc_unlck(state, cmd, request);
6535                 return 0;
6536         }
6537
6538         if (state == NULL)
6539                 return -ENOLCK;
6540
6541         if ((request->fl_flags & FL_POSIX) &&
6542             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6543                 return -ENOLCK;
6544
6545         status = nfs4_set_lock_state(state, request);
6546         if (status != 0)
6547                 return status;
6548
6549         return nfs4_retry_setlk(state, cmd, request);
6550 }
6551
6552 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6553 {
6554         struct nfs_server *server = NFS_SERVER(state->inode);
6555         int err;
6556
6557         err = nfs4_set_lock_state(state, fl);
6558         if (err != 0)
6559                 return err;
6560         err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6561         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
6562 }
6563
6564 struct nfs_release_lockowner_data {
6565         struct nfs4_lock_state *lsp;
6566         struct nfs_server *server;
6567         struct nfs_release_lockowner_args args;
6568         struct nfs_release_lockowner_res res;
6569         unsigned long timestamp;
6570 };
6571
6572 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6573 {
6574         struct nfs_release_lockowner_data *data = calldata;
6575         struct nfs_server *server = data->server;
6576         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
6577                            &data->res.seq_res, task);
6578         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6579         data->timestamp = jiffies;
6580 }
6581
6582 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6583 {
6584         struct nfs_release_lockowner_data *data = calldata;
6585         struct nfs_server *server = data->server;
6586
6587         nfs40_sequence_done(task, &data->res.seq_res);
6588
6589         switch (task->tk_status) {
6590         case 0:
6591                 renew_lease(server, data->timestamp);
6592                 break;
6593         case -NFS4ERR_STALE_CLIENTID:
6594         case -NFS4ERR_EXPIRED:
6595                 nfs4_schedule_lease_recovery(server->nfs_client);
6596                 break;
6597         case -NFS4ERR_LEASE_MOVED:
6598         case -NFS4ERR_DELAY:
6599                 if (nfs4_async_handle_error(task, server,
6600                                             NULL, NULL) == -EAGAIN)
6601                         rpc_restart_call_prepare(task);
6602         }
6603 }
6604
6605 static void nfs4_release_lockowner_release(void *calldata)
6606 {
6607         struct nfs_release_lockowner_data *data = calldata;
6608         nfs4_free_lock_state(data->server, data->lsp);
6609         kfree(calldata);
6610 }
6611
6612 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6613         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6614         .rpc_call_done = nfs4_release_lockowner_done,
6615         .rpc_release = nfs4_release_lockowner_release,
6616 };
6617
6618 static void
6619 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6620 {
6621         struct nfs_release_lockowner_data *data;
6622         struct rpc_message msg = {
6623                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6624         };
6625
6626         if (server->nfs_client->cl_mvops->minor_version != 0)
6627                 return;
6628
6629         data = kmalloc(sizeof(*data), GFP_NOFS);
6630         if (!data)
6631                 return;
6632         data->lsp = lsp;
6633         data->server = server;
6634         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6635         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6636         data->args.lock_owner.s_dev = server->s_dev;
6637
6638         msg.rpc_argp = &data->args;
6639         msg.rpc_resp = &data->res;
6640         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6641         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6642 }
6643
6644 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6645
6646 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6647                                    struct dentry *unused, struct inode *inode,
6648                                    const char *key, const void *buf,
6649                                    size_t buflen, int flags)
6650 {
6651         return nfs4_proc_set_acl(inode, buf, buflen);
6652 }
6653
6654 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6655                                    struct dentry *unused, struct inode *inode,
6656                                    const char *key, void *buf, size_t buflen)
6657 {
6658         return nfs4_proc_get_acl(inode, buf, buflen);
6659 }
6660
6661 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6662 {
6663         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6664 }
6665
6666 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6667
6668 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6669                                      struct dentry *unused, struct inode *inode,
6670                                      const char *key, const void *buf,
6671                                      size_t buflen, int flags)
6672 {
6673         if (security_ismaclabel(key))
6674                 return nfs4_set_security_label(inode, buf, buflen);
6675
6676         return -EOPNOTSUPP;
6677 }
6678
6679 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6680                                      struct dentry *unused, struct inode *inode,
6681                                      const char *key, void *buf, size_t buflen)
6682 {
6683         if (security_ismaclabel(key))
6684                 return nfs4_get_security_label(inode, buf, buflen);
6685         return -EOPNOTSUPP;
6686 }
6687
6688 static ssize_t
6689 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6690 {
6691         int len = 0;
6692
6693         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6694                 len = security_inode_listsecurity(inode, list, list_len);
6695                 if (list_len && len > list_len)
6696                         return -ERANGE;
6697         }
6698         return len;
6699 }
6700
6701 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6702         .prefix = XATTR_SECURITY_PREFIX,
6703         .get    = nfs4_xattr_get_nfs4_label,
6704         .set    = nfs4_xattr_set_nfs4_label,
6705 };
6706
6707 #else
6708
6709 static ssize_t
6710 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6711 {
6712         return 0;
6713 }
6714
6715 #endif
6716
6717 /*
6718  * nfs_fhget will use either the mounted_on_fileid or the fileid
6719  */
6720 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6721 {
6722         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6723                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6724               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6725               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6726                 return;
6727
6728         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6729                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6730         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6731         fattr->nlink = 2;
6732 }
6733
6734 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6735                                    const struct qstr *name,
6736                                    struct nfs4_fs_locations *fs_locations,
6737                                    struct page *page)
6738 {
6739         struct nfs_server *server = NFS_SERVER(dir);
6740         u32 bitmask[3] = {
6741                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6742         };
6743         struct nfs4_fs_locations_arg args = {
6744                 .dir_fh = NFS_FH(dir),
6745                 .name = name,
6746                 .page = page,
6747                 .bitmask = bitmask,
6748         };
6749         struct nfs4_fs_locations_res res = {
6750                 .fs_locations = fs_locations,
6751         };
6752         struct rpc_message msg = {
6753                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6754                 .rpc_argp = &args,
6755                 .rpc_resp = &res,
6756         };
6757         int status;
6758
6759         dprintk("%s: start\n", __func__);
6760
6761         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6762          * is not supported */
6763         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6764                 bitmask[1] |= FATTR4_WORD1_MOUNTED_ON_FILEID;
6765         else
6766                 bitmask[0] |= FATTR4_WORD0_FILEID;
6767
6768         nfs_fattr_init(&fs_locations->fattr);
6769         fs_locations->server = server;
6770         fs_locations->nlocations = 0;
6771         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6772         dprintk("%s: returned status = %d\n", __func__, status);
6773         return status;
6774 }
6775
6776 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6777                            const struct qstr *name,
6778                            struct nfs4_fs_locations *fs_locations,
6779                            struct page *page)
6780 {
6781         struct nfs4_exception exception = { };
6782         int err;
6783         do {
6784                 err = _nfs4_proc_fs_locations(client, dir, name,
6785                                 fs_locations, page);
6786                 trace_nfs4_get_fs_locations(dir, name, err);
6787                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6788                                 &exception);
6789         } while (exception.retry);
6790         return err;
6791 }
6792
6793 /*
6794  * This operation also signals the server that this client is
6795  * performing migration recovery.  The server can stop returning
6796  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6797  * appended to this compound to identify the client ID which is
6798  * performing recovery.
6799  */
6800 static int _nfs40_proc_get_locations(struct inode *inode,
6801                                      struct nfs4_fs_locations *locations,
6802                                      struct page *page, struct rpc_cred *cred)
6803 {
6804         struct nfs_server *server = NFS_SERVER(inode);
6805         struct rpc_clnt *clnt = server->client;
6806         u32 bitmask[2] = {
6807                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6808         };
6809         struct nfs4_fs_locations_arg args = {
6810                 .clientid       = server->nfs_client->cl_clientid,
6811                 .fh             = NFS_FH(inode),
6812                 .page           = page,
6813                 .bitmask        = bitmask,
6814                 .migration      = 1,            /* skip LOOKUP */
6815                 .renew          = 1,            /* append RENEW */
6816         };
6817         struct nfs4_fs_locations_res res = {
6818                 .fs_locations   = locations,
6819                 .migration      = 1,
6820                 .renew          = 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         unsigned long now = jiffies;
6829         int status;
6830
6831         nfs_fattr_init(&locations->fattr);
6832         locations->server = server;
6833         locations->nlocations = 0;
6834
6835         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6836         nfs4_set_sequence_privileged(&args.seq_args);
6837         status = nfs4_call_sync_sequence(clnt, server, &msg,
6838                                         &args.seq_args, &res.seq_res);
6839         if (status)
6840                 return status;
6841
6842         renew_lease(server, now);
6843         return 0;
6844 }
6845
6846 #ifdef CONFIG_NFS_V4_1
6847
6848 /*
6849  * This operation also signals the server that this client is
6850  * performing migration recovery.  The server can stop asserting
6851  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6852  * performing this operation is identified in the SEQUENCE
6853  * operation in this compound.
6854  *
6855  * When the client supports GETATTR(fs_locations_info), it can
6856  * be plumbed in here.
6857  */
6858 static int _nfs41_proc_get_locations(struct inode *inode,
6859                                      struct nfs4_fs_locations *locations,
6860                                      struct page *page, struct rpc_cred *cred)
6861 {
6862         struct nfs_server *server = NFS_SERVER(inode);
6863         struct rpc_clnt *clnt = server->client;
6864         u32 bitmask[2] = {
6865                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6866         };
6867         struct nfs4_fs_locations_arg args = {
6868                 .fh             = NFS_FH(inode),
6869                 .page           = page,
6870                 .bitmask        = bitmask,
6871                 .migration      = 1,            /* skip LOOKUP */
6872         };
6873         struct nfs4_fs_locations_res res = {
6874                 .fs_locations   = locations,
6875                 .migration      = 1,
6876         };
6877         struct rpc_message msg = {
6878                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6879                 .rpc_argp       = &args,
6880                 .rpc_resp       = &res,
6881                 .rpc_cred       = cred,
6882         };
6883         int status;
6884
6885         nfs_fattr_init(&locations->fattr);
6886         locations->server = server;
6887         locations->nlocations = 0;
6888
6889         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6890         nfs4_set_sequence_privileged(&args.seq_args);
6891         status = nfs4_call_sync_sequence(clnt, server, &msg,
6892                                         &args.seq_args, &res.seq_res);
6893         if (status == NFS4_OK &&
6894             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6895                 status = -NFS4ERR_LEASE_MOVED;
6896         return status;
6897 }
6898
6899 #endif  /* CONFIG_NFS_V4_1 */
6900
6901 /**
6902  * nfs4_proc_get_locations - discover locations for a migrated FSID
6903  * @inode: inode on FSID that is migrating
6904  * @locations: result of query
6905  * @page: buffer
6906  * @cred: credential to use for this operation
6907  *
6908  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6909  * operation failed, or a negative errno if a local error occurred.
6910  *
6911  * On success, "locations" is filled in, but if the server has
6912  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6913  * asserted.
6914  *
6915  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6916  * from this client that require migration recovery.
6917  */
6918 int nfs4_proc_get_locations(struct inode *inode,
6919                             struct nfs4_fs_locations *locations,
6920                             struct page *page, struct rpc_cred *cred)
6921 {
6922         struct nfs_server *server = NFS_SERVER(inode);
6923         struct nfs_client *clp = server->nfs_client;
6924         const struct nfs4_mig_recovery_ops *ops =
6925                                         clp->cl_mvops->mig_recovery_ops;
6926         struct nfs4_exception exception = { };
6927         int status;
6928
6929         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6930                 (unsigned long long)server->fsid.major,
6931                 (unsigned long long)server->fsid.minor,
6932                 clp->cl_hostname);
6933         nfs_display_fhandle(NFS_FH(inode), __func__);
6934
6935         do {
6936                 status = ops->get_locations(inode, locations, page, cred);
6937                 if (status != -NFS4ERR_DELAY)
6938                         break;
6939                 nfs4_handle_exception(server, status, &exception);
6940         } while (exception.retry);
6941         return status;
6942 }
6943
6944 /*
6945  * This operation also signals the server that this client is
6946  * performing "lease moved" recovery.  The server can stop
6947  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6948  * is appended to this compound to identify the client ID which is
6949  * performing recovery.
6950  */
6951 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6952 {
6953         struct nfs_server *server = NFS_SERVER(inode);
6954         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6955         struct rpc_clnt *clnt = server->client;
6956         struct nfs4_fsid_present_arg args = {
6957                 .fh             = NFS_FH(inode),
6958                 .clientid       = clp->cl_clientid,
6959                 .renew          = 1,            /* append RENEW */
6960         };
6961         struct nfs4_fsid_present_res res = {
6962                 .renew          = 1,
6963         };
6964         struct rpc_message msg = {
6965                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6966                 .rpc_argp       = &args,
6967                 .rpc_resp       = &res,
6968                 .rpc_cred       = cred,
6969         };
6970         unsigned long now = jiffies;
6971         int status;
6972
6973         res.fh = nfs_alloc_fhandle();
6974         if (res.fh == NULL)
6975                 return -ENOMEM;
6976
6977         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6978         nfs4_set_sequence_privileged(&args.seq_args);
6979         status = nfs4_call_sync_sequence(clnt, server, &msg,
6980                                                 &args.seq_args, &res.seq_res);
6981         nfs_free_fhandle(res.fh);
6982         if (status)
6983                 return status;
6984
6985         do_renew_lease(clp, now);
6986         return 0;
6987 }
6988
6989 #ifdef CONFIG_NFS_V4_1
6990
6991 /*
6992  * This operation also signals the server that this client is
6993  * performing "lease moved" recovery.  The server can stop asserting
6994  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6995  * this operation is identified in the SEQUENCE operation in this
6996  * compound.
6997  */
6998 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6999 {
7000         struct nfs_server *server = NFS_SERVER(inode);
7001         struct rpc_clnt *clnt = server->client;
7002         struct nfs4_fsid_present_arg args = {
7003                 .fh             = NFS_FH(inode),
7004         };
7005         struct nfs4_fsid_present_res res = {
7006         };
7007         struct rpc_message msg = {
7008                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7009                 .rpc_argp       = &args,
7010                 .rpc_resp       = &res,
7011                 .rpc_cred       = cred,
7012         };
7013         int status;
7014
7015         res.fh = nfs_alloc_fhandle();
7016         if (res.fh == NULL)
7017                 return -ENOMEM;
7018
7019         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7020         nfs4_set_sequence_privileged(&args.seq_args);
7021         status = nfs4_call_sync_sequence(clnt, server, &msg,
7022                                                 &args.seq_args, &res.seq_res);
7023         nfs_free_fhandle(res.fh);
7024         if (status == NFS4_OK &&
7025             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7026                 status = -NFS4ERR_LEASE_MOVED;
7027         return status;
7028 }
7029
7030 #endif  /* CONFIG_NFS_V4_1 */
7031
7032 /**
7033  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7034  * @inode: inode on FSID to check
7035  * @cred: credential to use for this operation
7036  *
7037  * Server indicates whether the FSID is present, moved, or not
7038  * recognized.  This operation is necessary to clear a LEASE_MOVED
7039  * condition for this client ID.
7040  *
7041  * Returns NFS4_OK if the FSID is present on this server,
7042  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7043  *  NFS4ERR code if some error occurred on the server, or a
7044  *  negative errno if a local failure occurred.
7045  */
7046 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7047 {
7048         struct nfs_server *server = NFS_SERVER(inode);
7049         struct nfs_client *clp = server->nfs_client;
7050         const struct nfs4_mig_recovery_ops *ops =
7051                                         clp->cl_mvops->mig_recovery_ops;
7052         struct nfs4_exception exception = { };
7053         int status;
7054
7055         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7056                 (unsigned long long)server->fsid.major,
7057                 (unsigned long long)server->fsid.minor,
7058                 clp->cl_hostname);
7059         nfs_display_fhandle(NFS_FH(inode), __func__);
7060
7061         do {
7062                 status = ops->fsid_present(inode, cred);
7063                 if (status != -NFS4ERR_DELAY)
7064                         break;
7065                 nfs4_handle_exception(server, status, &exception);
7066         } while (exception.retry);
7067         return status;
7068 }
7069
7070 /**
7071  * If 'use_integrity' is true and the state managment nfs_client
7072  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7073  * and the machine credential as per RFC3530bis and RFC5661 Security
7074  * Considerations sections. Otherwise, just use the user cred with the
7075  * filesystem's rpc_client.
7076  */
7077 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7078 {
7079         int status;
7080         struct nfs4_secinfo_arg args = {
7081                 .dir_fh = NFS_FH(dir),
7082                 .name   = name,
7083         };
7084         struct nfs4_secinfo_res res = {
7085                 .flavors     = flavors,
7086         };
7087         struct rpc_message msg = {
7088                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7089                 .rpc_argp = &args,
7090                 .rpc_resp = &res,
7091         };
7092         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7093         struct rpc_cred *cred = NULL;
7094
7095         if (use_integrity) {
7096                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7097                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7098                 msg.rpc_cred = cred;
7099         }
7100
7101         dprintk("NFS call  secinfo %s\n", name->name);
7102
7103         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7104                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7105
7106         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7107                                 &res.seq_res, 0);
7108         dprintk("NFS reply  secinfo: %d\n", status);
7109
7110         if (cred)
7111                 put_rpccred(cred);
7112
7113         return status;
7114 }
7115
7116 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7117                       struct nfs4_secinfo_flavors *flavors)
7118 {
7119         struct nfs4_exception exception = { };
7120         int err;
7121         do {
7122                 err = -NFS4ERR_WRONGSEC;
7123
7124                 /* try to use integrity protection with machine cred */
7125                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7126                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7127
7128                 /*
7129                  * if unable to use integrity protection, or SECINFO with
7130                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7131                  * disallowed by spec, but exists in deployed servers) use
7132                  * the current filesystem's rpc_client and the user cred.
7133                  */
7134                 if (err == -NFS4ERR_WRONGSEC)
7135                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7136
7137                 trace_nfs4_secinfo(dir, name, err);
7138                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7139                                 &exception);
7140         } while (exception.retry);
7141         return err;
7142 }
7143
7144 #ifdef CONFIG_NFS_V4_1
7145 /*
7146  * Check the exchange flags returned by the server for invalid flags, having
7147  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7148  * DS flags set.
7149  */
7150 static int nfs4_check_cl_exchange_flags(u32 flags)
7151 {
7152         if (flags & ~EXCHGID4_FLAG_MASK_R)
7153                 goto out_inval;
7154         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7155             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7156                 goto out_inval;
7157         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7158                 goto out_inval;
7159         return NFS_OK;
7160 out_inval:
7161         return -NFS4ERR_INVAL;
7162 }
7163
7164 static bool
7165 nfs41_same_server_scope(struct nfs41_server_scope *a,
7166                         struct nfs41_server_scope *b)
7167 {
7168         if (a->server_scope_sz != b->server_scope_sz)
7169                 return false;
7170         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
7171 }
7172
7173 static void
7174 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7175 {
7176 }
7177
7178 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7179         .rpc_call_done =  &nfs4_bind_one_conn_to_session_done,
7180 };
7181
7182 /*
7183  * nfs4_proc_bind_one_conn_to_session()
7184  *
7185  * The 4.1 client currently uses the same TCP connection for the
7186  * fore and backchannel.
7187  */
7188 static
7189 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7190                 struct rpc_xprt *xprt,
7191                 struct nfs_client *clp,
7192                 struct rpc_cred *cred)
7193 {
7194         int status;
7195         struct nfs41_bind_conn_to_session_args args = {
7196                 .client = clp,
7197                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7198         };
7199         struct nfs41_bind_conn_to_session_res res;
7200         struct rpc_message msg = {
7201                 .rpc_proc =
7202                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7203                 .rpc_argp = &args,
7204                 .rpc_resp = &res,
7205                 .rpc_cred = cred,
7206         };
7207         struct rpc_task_setup task_setup_data = {
7208                 .rpc_client = clnt,
7209                 .rpc_xprt = xprt,
7210                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7211                 .rpc_message = &msg,
7212                 .flags = RPC_TASK_TIMEOUT,
7213         };
7214         struct rpc_task *task;
7215
7216         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7217         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7218                 args.dir = NFS4_CDFC4_FORE;
7219
7220         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7221         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7222                 args.dir = NFS4_CDFC4_FORE;
7223
7224         task = rpc_run_task(&task_setup_data);
7225         if (!IS_ERR(task)) {
7226                 status = task->tk_status;
7227                 rpc_put_task(task);
7228         } else
7229                 status = PTR_ERR(task);
7230         trace_nfs4_bind_conn_to_session(clp, status);
7231         if (status == 0) {
7232                 if (memcmp(res.sessionid.data,
7233                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7234                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7235                         return -EIO;
7236                 }
7237                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7238                         dprintk("NFS: %s: Unexpected direction from server\n",
7239                                 __func__);
7240                         return -EIO;
7241                 }
7242                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7243                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7244                                 __func__);
7245                         return -EIO;
7246                 }
7247         }
7248
7249         return status;
7250 }
7251
7252 struct rpc_bind_conn_calldata {
7253         struct nfs_client *clp;
7254         struct rpc_cred *cred;
7255 };
7256
7257 static int
7258 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7259                 struct rpc_xprt *xprt,
7260                 void *calldata)
7261 {
7262         struct rpc_bind_conn_calldata *p = calldata;
7263
7264         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7265 }
7266
7267 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7268 {
7269         struct rpc_bind_conn_calldata data = {
7270                 .clp = clp,
7271                 .cred = cred,
7272         };
7273         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7274                         nfs4_proc_bind_conn_to_session_callback, &data);
7275 }
7276
7277 /*
7278  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7279  * and operations we'd like to see to enable certain features in the allow map
7280  */
7281 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7282         .how = SP4_MACH_CRED,
7283         .enforce.u.words = {
7284                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7285                       1 << (OP_EXCHANGE_ID - 32) |
7286                       1 << (OP_CREATE_SESSION - 32) |
7287                       1 << (OP_DESTROY_SESSION - 32) |
7288                       1 << (OP_DESTROY_CLIENTID - 32)
7289         },
7290         .allow.u.words = {
7291                 [0] = 1 << (OP_CLOSE) |
7292                       1 << (OP_OPEN_DOWNGRADE) |
7293                       1 << (OP_LOCKU) |
7294                       1 << (OP_DELEGRETURN) |
7295                       1 << (OP_COMMIT),
7296                 [1] = 1 << (OP_SECINFO - 32) |
7297                       1 << (OP_SECINFO_NO_NAME - 32) |
7298                       1 << (OP_LAYOUTRETURN - 32) |
7299                       1 << (OP_TEST_STATEID - 32) |
7300                       1 << (OP_FREE_STATEID - 32) |
7301                       1 << (OP_WRITE - 32)
7302         }
7303 };
7304
7305 /*
7306  * Select the state protection mode for client `clp' given the server results
7307  * from exchange_id in `sp'.
7308  *
7309  * Returns 0 on success, negative errno otherwise.
7310  */
7311 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7312                                  struct nfs41_state_protection *sp)
7313 {
7314         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7315                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7316                       1 << (OP_EXCHANGE_ID - 32) |
7317                       1 << (OP_CREATE_SESSION - 32) |
7318                       1 << (OP_DESTROY_SESSION - 32) |
7319                       1 << (OP_DESTROY_CLIENTID - 32)
7320         };
7321         unsigned int i;
7322
7323         if (sp->how == SP4_MACH_CRED) {
7324                 /* Print state protect result */
7325                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7326                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7327                         if (test_bit(i, sp->enforce.u.longs))
7328                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7329                         if (test_bit(i, sp->allow.u.longs))
7330                                 dfprintk(MOUNT, "  allow op %d\n", i);
7331                 }
7332
7333                 /* make sure nothing is on enforce list that isn't supported */
7334                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7335                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7336                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7337                                 return -EINVAL;
7338                         }
7339                 }
7340
7341                 /*
7342                  * Minimal mode - state operations are allowed to use machine
7343                  * credential.  Note this already happens by default, so the
7344                  * client doesn't have to do anything more than the negotiation.
7345                  *
7346                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7347                  *       we're already using the machine cred for exchange_id
7348                  *       and will never use a different cred.
7349                  */
7350                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7351                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7352                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7353                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7354                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7355                         dfprintk(MOUNT, "  minimal mode enabled\n");
7356                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7357                 } else {
7358                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7359                         return -EINVAL;
7360                 }
7361
7362                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7363                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7364                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7365                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7366                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7367                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7368                 }
7369
7370                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7371                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7372                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7373                                 &clp->cl_sp4_flags);
7374                 }
7375
7376                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7377                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7378                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7379                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7380                 }
7381
7382                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7383                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7384                         dfprintk(MOUNT, "  stateid mode enabled\n");
7385                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7386                 }
7387
7388                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7389                         dfprintk(MOUNT, "  write mode enabled\n");
7390                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7391                 }
7392
7393                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7394                         dfprintk(MOUNT, "  commit mode enabled\n");
7395                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7396                 }
7397         }
7398
7399         return 0;
7400 }
7401
7402 struct nfs41_exchange_id_data {
7403         struct nfs41_exchange_id_res res;
7404         struct nfs41_exchange_id_args args;
7405         struct rpc_xprt *xprt;
7406         int rpc_status;
7407 };
7408
7409 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7410 {
7411         struct nfs41_exchange_id_data *cdata =
7412                                         (struct nfs41_exchange_id_data *)data;
7413         struct nfs_client *clp = cdata->args.client;
7414         int status = task->tk_status;
7415
7416         trace_nfs4_exchange_id(clp, status);
7417
7418         if (status == 0)
7419                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7420
7421         if (cdata->xprt && status == 0) {
7422                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7423                                                       cdata->xprt);
7424                 goto out;
7425         }
7426
7427         if (status  == 0)
7428                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7429
7430         if (status == 0) {
7431                 clp->cl_clientid = cdata->res.clientid;
7432                 clp->cl_exchange_flags = cdata->res.flags;
7433                 clp->cl_seqid = cdata->res.seqid;
7434                 /* Client ID is not confirmed */
7435                 if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R))
7436                         clear_bit(NFS4_SESSION_ESTABLISHED,
7437                                   &clp->cl_session->session_state);
7438
7439                 kfree(clp->cl_serverowner);
7440                 clp->cl_serverowner = cdata->res.server_owner;
7441                 cdata->res.server_owner = NULL;
7442
7443                 /* use the most recent implementation id */
7444                 kfree(clp->cl_implid);
7445                 clp->cl_implid = cdata->res.impl_id;
7446                 cdata->res.impl_id = NULL;
7447
7448                 if (clp->cl_serverscope != NULL &&
7449                     !nfs41_same_server_scope(clp->cl_serverscope,
7450                                         cdata->res.server_scope)) {
7451                         dprintk("%s: server_scope mismatch detected\n",
7452                                 __func__);
7453                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7454                         kfree(clp->cl_serverscope);
7455                         clp->cl_serverscope = NULL;
7456                 }
7457
7458                 if (clp->cl_serverscope == NULL) {
7459                         clp->cl_serverscope = cdata->res.server_scope;
7460                         cdata->res.server_scope = NULL;
7461                 }
7462                 /* Save the EXCHANGE_ID verifier session trunk tests */
7463                 memcpy(clp->cl_confirm.data, cdata->args.verifier.data,
7464                        sizeof(clp->cl_confirm.data));
7465         }
7466 out:
7467         cdata->rpc_status = status;
7468         return;
7469 }
7470
7471 static void nfs4_exchange_id_release(void *data)
7472 {
7473         struct nfs41_exchange_id_data *cdata =
7474                                         (struct nfs41_exchange_id_data *)data;
7475
7476         nfs_put_client(cdata->args.client);
7477         kfree(cdata->res.impl_id);
7478         kfree(cdata->res.server_scope);
7479         kfree(cdata->res.server_owner);
7480         kfree(cdata);
7481 }
7482
7483 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7484         .rpc_call_done = nfs4_exchange_id_done,
7485         .rpc_release = nfs4_exchange_id_release,
7486 };
7487
7488 /*
7489  * _nfs4_proc_exchange_id()
7490  *
7491  * Wrapper for EXCHANGE_ID operation.
7492  */
7493 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7494                         u32 sp4_how, struct rpc_xprt *xprt)
7495 {
7496         struct rpc_message msg = {
7497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7498                 .rpc_cred = cred,
7499         };
7500         struct rpc_task_setup task_setup_data = {
7501                 .rpc_client = clp->cl_rpcclient,
7502                 .callback_ops = &nfs4_exchange_id_call_ops,
7503                 .rpc_message = &msg,
7504                 .flags = RPC_TASK_TIMEOUT,
7505         };
7506         struct nfs41_exchange_id_data *calldata;
7507         struct rpc_task *task;
7508         int status;
7509
7510         if (!atomic_inc_not_zero(&clp->cl_count))
7511                 return -EIO;
7512
7513         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7514         if (!calldata) {
7515                 nfs_put_client(clp);
7516                 return -ENOMEM;
7517         }
7518
7519         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7520
7521         status = nfs4_init_uniform_client_string(clp);
7522         if (status)
7523                 goto out_calldata;
7524
7525         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7526                                                 GFP_NOFS);
7527         status = -ENOMEM;
7528         if (unlikely(calldata->res.server_owner == NULL))
7529                 goto out_calldata;
7530
7531         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7532                                         GFP_NOFS);
7533         if (unlikely(calldata->res.server_scope == NULL))
7534                 goto out_server_owner;
7535
7536         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7537         if (unlikely(calldata->res.impl_id == NULL))
7538                 goto out_server_scope;
7539
7540         switch (sp4_how) {
7541         case SP4_NONE:
7542                 calldata->args.state_protect.how = SP4_NONE;
7543                 break;
7544
7545         case SP4_MACH_CRED:
7546                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7547                 break;
7548
7549         default:
7550                 /* unsupported! */
7551                 WARN_ON_ONCE(1);
7552                 status = -EINVAL;
7553                 goto out_impl_id;
7554         }
7555         if (xprt) {
7556                 calldata->xprt = xprt;
7557                 task_setup_data.rpc_xprt = xprt;
7558                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
7559                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7560                                 sizeof(calldata->args.verifier.data));
7561         }
7562         calldata->args.client = clp;
7563 #ifdef CONFIG_NFS_V4_1_MIGRATION
7564         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7565         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7566         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7567 #else
7568         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7569         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7570 #endif
7571         msg.rpc_argp = &calldata->args;
7572         msg.rpc_resp = &calldata->res;
7573         task_setup_data.callback_data = calldata;
7574
7575         task = rpc_run_task(&task_setup_data);
7576         if (IS_ERR(task))
7577                 return PTR_ERR(task);
7578
7579         status = calldata->rpc_status;
7580
7581         rpc_put_task(task);
7582 out:
7583         return status;
7584
7585 out_impl_id:
7586         kfree(calldata->res.impl_id);
7587 out_server_scope:
7588         kfree(calldata->res.server_scope);
7589 out_server_owner:
7590         kfree(calldata->res.server_owner);
7591 out_calldata:
7592         kfree(calldata);
7593         nfs_put_client(clp);
7594         goto out;
7595 }
7596
7597 /*
7598  * nfs4_proc_exchange_id()
7599  *
7600  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7601  *
7602  * Since the clientid has expired, all compounds using sessions
7603  * associated with the stale clientid will be returning
7604  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7605  * be in some phase of session reset.
7606  *
7607  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7608  */
7609 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7610 {
7611         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7612         int status;
7613
7614         /* try SP4_MACH_CRED if krb5i/p */
7615         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7616             authflavor == RPC_AUTH_GSS_KRB5P) {
7617                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7618                 if (!status)
7619                         return 0;
7620         }
7621
7622         /* try SP4_NONE */
7623         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7624 }
7625
7626 /**
7627  * nfs4_test_session_trunk
7628  *
7629  * This is an add_xprt_test() test function called from
7630  * rpc_clnt_setup_test_and_add_xprt.
7631  *
7632  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7633  * and is dereferrenced in nfs4_exchange_id_release
7634  *
7635  * Upon success, add the new transport to the rpc_clnt
7636  *
7637  * @clnt: struct rpc_clnt to get new transport
7638  * @xprt: the rpc_xprt to test
7639  * @data: call data for _nfs4_proc_exchange_id.
7640  */
7641 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7642                             void *data)
7643 {
7644         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7645         u32 sp4_how;
7646
7647         dprintk("--> %s try %s\n", __func__,
7648                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7649
7650         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7651
7652         /* Test connection for session trunking. Async exchange_id call */
7653         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7654 }
7655 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7656
7657 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7658                 struct rpc_cred *cred)
7659 {
7660         struct rpc_message msg = {
7661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7662                 .rpc_argp = clp,
7663                 .rpc_cred = cred,
7664         };
7665         int status;
7666
7667         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7668         trace_nfs4_destroy_clientid(clp, status);
7669         if (status)
7670                 dprintk("NFS: Got error %d from the server %s on "
7671                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7672         return status;
7673 }
7674
7675 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7676                 struct rpc_cred *cred)
7677 {
7678         unsigned int loop;
7679         int ret;
7680
7681         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7682                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7683                 switch (ret) {
7684                 case -NFS4ERR_DELAY:
7685                 case -NFS4ERR_CLIENTID_BUSY:
7686                         ssleep(1);
7687                         break;
7688                 default:
7689                         return ret;
7690                 }
7691         }
7692         return 0;
7693 }
7694
7695 int nfs4_destroy_clientid(struct nfs_client *clp)
7696 {
7697         struct rpc_cred *cred;
7698         int ret = 0;
7699
7700         if (clp->cl_mvops->minor_version < 1)
7701                 goto out;
7702         if (clp->cl_exchange_flags == 0)
7703                 goto out;
7704         if (clp->cl_preserve_clid)
7705                 goto out;
7706         cred = nfs4_get_clid_cred(clp);
7707         ret = nfs4_proc_destroy_clientid(clp, cred);
7708         if (cred)
7709                 put_rpccred(cred);
7710         switch (ret) {
7711         case 0:
7712         case -NFS4ERR_STALE_CLIENTID:
7713                 clp->cl_exchange_flags = 0;
7714         }
7715 out:
7716         return ret;
7717 }
7718
7719 struct nfs4_get_lease_time_data {
7720         struct nfs4_get_lease_time_args *args;
7721         struct nfs4_get_lease_time_res *res;
7722         struct nfs_client *clp;
7723 };
7724
7725 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7726                                         void *calldata)
7727 {
7728         struct nfs4_get_lease_time_data *data =
7729                         (struct nfs4_get_lease_time_data *)calldata;
7730
7731         dprintk("--> %s\n", __func__);
7732         /* just setup sequence, do not trigger session recovery
7733            since we're invoked within one */
7734         nfs4_setup_sequence(data->clp,
7735                         &data->args->la_seq_args,
7736                         &data->res->lr_seq_res,
7737                         task);
7738         dprintk("<-- %s\n", __func__);
7739 }
7740
7741 /*
7742  * Called from nfs4_state_manager thread for session setup, so don't recover
7743  * from sequence operation or clientid errors.
7744  */
7745 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7746 {
7747         struct nfs4_get_lease_time_data *data =
7748                         (struct nfs4_get_lease_time_data *)calldata;
7749
7750         dprintk("--> %s\n", __func__);
7751         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7752                 return;
7753         switch (task->tk_status) {
7754         case -NFS4ERR_DELAY:
7755         case -NFS4ERR_GRACE:
7756                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7757                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7758                 task->tk_status = 0;
7759                 /* fall through */
7760         case -NFS4ERR_RETRY_UNCACHED_REP:
7761                 rpc_restart_call_prepare(task);
7762                 return;
7763         }
7764         dprintk("<-- %s\n", __func__);
7765 }
7766
7767 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7768         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7769         .rpc_call_done = nfs4_get_lease_time_done,
7770 };
7771
7772 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7773 {
7774         struct rpc_task *task;
7775         struct nfs4_get_lease_time_args args;
7776         struct nfs4_get_lease_time_res res = {
7777                 .lr_fsinfo = fsinfo,
7778         };
7779         struct nfs4_get_lease_time_data data = {
7780                 .args = &args,
7781                 .res = &res,
7782                 .clp = clp,
7783         };
7784         struct rpc_message msg = {
7785                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7786                 .rpc_argp = &args,
7787                 .rpc_resp = &res,
7788         };
7789         struct rpc_task_setup task_setup = {
7790                 .rpc_client = clp->cl_rpcclient,
7791                 .rpc_message = &msg,
7792                 .callback_ops = &nfs4_get_lease_time_ops,
7793                 .callback_data = &data,
7794                 .flags = RPC_TASK_TIMEOUT,
7795         };
7796         int status;
7797
7798         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7799         nfs4_set_sequence_privileged(&args.la_seq_args);
7800         task = rpc_run_task(&task_setup);
7801
7802         if (IS_ERR(task))
7803                 return PTR_ERR(task);
7804
7805         status = task->tk_status;
7806         rpc_put_task(task);
7807         return status;
7808 }
7809
7810 /*
7811  * Initialize the values to be used by the client in CREATE_SESSION
7812  * If nfs4_init_session set the fore channel request and response sizes,
7813  * use them.
7814  *
7815  * Set the back channel max_resp_sz_cached to zero to force the client to
7816  * always set csa_cachethis to FALSE because the current implementation
7817  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7818  */
7819 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7820                                     struct rpc_clnt *clnt)
7821 {
7822         unsigned int max_rqst_sz, max_resp_sz;
7823         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7824
7825         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7826         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7827
7828         /* Fore channel attributes */
7829         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7830         args->fc_attrs.max_resp_sz = max_resp_sz;
7831         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7832         args->fc_attrs.max_reqs = max_session_slots;
7833
7834         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7835                 "max_ops=%u max_reqs=%u\n",
7836                 __func__,
7837                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7838                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7839
7840         /* Back channel attributes */
7841         args->bc_attrs.max_rqst_sz = max_bc_payload;
7842         args->bc_attrs.max_resp_sz = max_bc_payload;
7843         args->bc_attrs.max_resp_sz_cached = 0;
7844         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7845         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7846
7847         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7848                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7849                 __func__,
7850                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7851                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7852                 args->bc_attrs.max_reqs);
7853 }
7854
7855 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7856                 struct nfs41_create_session_res *res)
7857 {
7858         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7859         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7860
7861         if (rcvd->max_resp_sz > sent->max_resp_sz)
7862                 return -EINVAL;
7863         /*
7864          * Our requested max_ops is the minimum we need; we're not
7865          * prepared to break up compounds into smaller pieces than that.
7866          * So, no point even trying to continue if the server won't
7867          * cooperate:
7868          */
7869         if (rcvd->max_ops < sent->max_ops)
7870                 return -EINVAL;
7871         if (rcvd->max_reqs == 0)
7872                 return -EINVAL;
7873         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7874                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7875         return 0;
7876 }
7877
7878 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7879                 struct nfs41_create_session_res *res)
7880 {
7881         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7882         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7883
7884         if (!(res->flags & SESSION4_BACK_CHAN))
7885                 goto out;
7886         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7887                 return -EINVAL;
7888         if (rcvd->max_resp_sz < sent->max_resp_sz)
7889                 return -EINVAL;
7890         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7891                 return -EINVAL;
7892         if (rcvd->max_ops > sent->max_ops)
7893                 return -EINVAL;
7894         if (rcvd->max_reqs > sent->max_reqs)
7895                 return -EINVAL;
7896 out:
7897         return 0;
7898 }
7899
7900 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7901                                      struct nfs41_create_session_res *res)
7902 {
7903         int ret;
7904
7905         ret = nfs4_verify_fore_channel_attrs(args, res);
7906         if (ret)
7907                 return ret;
7908         return nfs4_verify_back_channel_attrs(args, res);
7909 }
7910
7911 static void nfs4_update_session(struct nfs4_session *session,
7912                 struct nfs41_create_session_res *res)
7913 {
7914         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7915         /* Mark client id and session as being confirmed */
7916         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7917         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7918         session->flags = res->flags;
7919         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7920         if (res->flags & SESSION4_BACK_CHAN)
7921                 memcpy(&session->bc_attrs, &res->bc_attrs,
7922                                 sizeof(session->bc_attrs));
7923 }
7924
7925 static int _nfs4_proc_create_session(struct nfs_client *clp,
7926                 struct rpc_cred *cred)
7927 {
7928         struct nfs4_session *session = clp->cl_session;
7929         struct nfs41_create_session_args args = {
7930                 .client = clp,
7931                 .clientid = clp->cl_clientid,
7932                 .seqid = clp->cl_seqid,
7933                 .cb_program = NFS4_CALLBACK,
7934         };
7935         struct nfs41_create_session_res res;
7936
7937         struct rpc_message msg = {
7938                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7939                 .rpc_argp = &args,
7940                 .rpc_resp = &res,
7941                 .rpc_cred = cred,
7942         };
7943         int status;
7944
7945         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7946         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7947
7948         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7949         trace_nfs4_create_session(clp, status);
7950
7951         switch (status) {
7952         case -NFS4ERR_STALE_CLIENTID:
7953         case -NFS4ERR_DELAY:
7954         case -ETIMEDOUT:
7955         case -EACCES:
7956         case -EAGAIN:
7957                 goto out;
7958         };
7959
7960         clp->cl_seqid++;
7961         if (!status) {
7962                 /* Verify the session's negotiated channel_attrs values */
7963                 status = nfs4_verify_channel_attrs(&args, &res);
7964                 /* Increment the clientid slot sequence id */
7965                 if (status)
7966                         goto out;
7967                 nfs4_update_session(session, &res);
7968         }
7969 out:
7970         return status;
7971 }
7972
7973 /*
7974  * Issues a CREATE_SESSION operation to the server.
7975  * It is the responsibility of the caller to verify the session is
7976  * expired before calling this routine.
7977  */
7978 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
7979 {
7980         int status;
7981         unsigned *ptr;
7982         struct nfs4_session *session = clp->cl_session;
7983
7984         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
7985
7986         status = _nfs4_proc_create_session(clp, cred);
7987         if (status)
7988                 goto out;
7989
7990         /* Init or reset the session slot tables */
7991         status = nfs4_setup_session_slot_tables(session);
7992         dprintk("slot table setup returned %d\n", status);
7993         if (status)
7994                 goto out;
7995
7996         ptr = (unsigned *)&session->sess_id.data[0];
7997         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
7998                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
7999 out:
8000         dprintk("<-- %s\n", __func__);
8001         return status;
8002 }
8003
8004 /*
8005  * Issue the over-the-wire RPC DESTROY_SESSION.
8006  * The caller must serialize access to this routine.
8007  */
8008 int nfs4_proc_destroy_session(struct nfs4_session *session,
8009                 struct rpc_cred *cred)
8010 {
8011         struct rpc_message msg = {
8012                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8013                 .rpc_argp = session,
8014                 .rpc_cred = cred,
8015         };
8016         int status = 0;
8017
8018         dprintk("--> nfs4_proc_destroy_session\n");
8019
8020         /* session is still being setup */
8021         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8022                 return 0;
8023
8024         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8025         trace_nfs4_destroy_session(session->clp, status);
8026
8027         if (status)
8028                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8029                         "Session has been destroyed regardless...\n", status);
8030
8031         dprintk("<-- nfs4_proc_destroy_session\n");
8032         return status;
8033 }
8034
8035 /*
8036  * Renew the cl_session lease.
8037  */
8038 struct nfs4_sequence_data {
8039         struct nfs_client *clp;
8040         struct nfs4_sequence_args args;
8041         struct nfs4_sequence_res res;
8042 };
8043
8044 static void nfs41_sequence_release(void *data)
8045 {
8046         struct nfs4_sequence_data *calldata = data;
8047         struct nfs_client *clp = calldata->clp;
8048
8049         if (atomic_read(&clp->cl_count) > 1)
8050                 nfs4_schedule_state_renewal(clp);
8051         nfs_put_client(clp);
8052         kfree(calldata);
8053 }
8054
8055 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8056 {
8057         switch(task->tk_status) {
8058         case -NFS4ERR_DELAY:
8059                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8060                 return -EAGAIN;
8061         default:
8062                 nfs4_schedule_lease_recovery(clp);
8063         }
8064         return 0;
8065 }
8066
8067 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8068 {
8069         struct nfs4_sequence_data *calldata = data;
8070         struct nfs_client *clp = calldata->clp;
8071
8072         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8073                 return;
8074
8075         trace_nfs4_sequence(clp, task->tk_status);
8076         if (task->tk_status < 0) {
8077                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8078                 if (atomic_read(&clp->cl_count) == 1)
8079                         goto out;
8080
8081                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8082                         rpc_restart_call_prepare(task);
8083                         return;
8084                 }
8085         }
8086         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8087 out:
8088         dprintk("<-- %s\n", __func__);
8089 }
8090
8091 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8092 {
8093         struct nfs4_sequence_data *calldata = data;
8094         struct nfs_client *clp = calldata->clp;
8095         struct nfs4_sequence_args *args;
8096         struct nfs4_sequence_res *res;
8097
8098         args = task->tk_msg.rpc_argp;
8099         res = task->tk_msg.rpc_resp;
8100
8101         nfs4_setup_sequence(clp, args, res, task);
8102 }
8103
8104 static const struct rpc_call_ops nfs41_sequence_ops = {
8105         .rpc_call_done = nfs41_sequence_call_done,
8106         .rpc_call_prepare = nfs41_sequence_prepare,
8107         .rpc_release = nfs41_sequence_release,
8108 };
8109
8110 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8111                 struct rpc_cred *cred,
8112                 bool is_privileged)
8113 {
8114         struct nfs4_sequence_data *calldata;
8115         struct rpc_message msg = {
8116                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8117                 .rpc_cred = cred,
8118         };
8119         struct rpc_task_setup task_setup_data = {
8120                 .rpc_client = clp->cl_rpcclient,
8121                 .rpc_message = &msg,
8122                 .callback_ops = &nfs41_sequence_ops,
8123                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8124         };
8125
8126         if (!atomic_inc_not_zero(&clp->cl_count))
8127                 return ERR_PTR(-EIO);
8128         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8129         if (calldata == NULL) {
8130                 nfs_put_client(clp);
8131                 return ERR_PTR(-ENOMEM);
8132         }
8133         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8134         if (is_privileged)
8135                 nfs4_set_sequence_privileged(&calldata->args);
8136         msg.rpc_argp = &calldata->args;
8137         msg.rpc_resp = &calldata->res;
8138         calldata->clp = clp;
8139         task_setup_data.callback_data = calldata;
8140
8141         return rpc_run_task(&task_setup_data);
8142 }
8143
8144 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8145 {
8146         struct rpc_task *task;
8147         int ret = 0;
8148
8149         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8150                 return -EAGAIN;
8151         task = _nfs41_proc_sequence(clp, cred, false);
8152         if (IS_ERR(task))
8153                 ret = PTR_ERR(task);
8154         else
8155                 rpc_put_task_async(task);
8156         dprintk("<-- %s status=%d\n", __func__, ret);
8157         return ret;
8158 }
8159
8160 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8161 {
8162         struct rpc_task *task;
8163         int ret;
8164
8165         task = _nfs41_proc_sequence(clp, cred, true);
8166         if (IS_ERR(task)) {
8167                 ret = PTR_ERR(task);
8168                 goto out;
8169         }
8170         ret = rpc_wait_for_completion_task(task);
8171         if (!ret)
8172                 ret = task->tk_status;
8173         rpc_put_task(task);
8174 out:
8175         dprintk("<-- %s status=%d\n", __func__, ret);
8176         return ret;
8177 }
8178
8179 struct nfs4_reclaim_complete_data {
8180         struct nfs_client *clp;
8181         struct nfs41_reclaim_complete_args arg;
8182         struct nfs41_reclaim_complete_res res;
8183 };
8184
8185 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8186 {
8187         struct nfs4_reclaim_complete_data *calldata = data;
8188
8189         nfs4_setup_sequence(calldata->clp,
8190                         &calldata->arg.seq_args,
8191                         &calldata->res.seq_res,
8192                         task);
8193 }
8194
8195 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8196 {
8197         switch(task->tk_status) {
8198         case 0:
8199         case -NFS4ERR_COMPLETE_ALREADY:
8200         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8201                 break;
8202         case -NFS4ERR_DELAY:
8203                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8204                 /* fall through */
8205         case -NFS4ERR_RETRY_UNCACHED_REP:
8206                 return -EAGAIN;
8207         case -NFS4ERR_BADSESSION:
8208         case -NFS4ERR_DEADSESSION:
8209         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8210                 nfs4_schedule_session_recovery(clp->cl_session,
8211                                 task->tk_status);
8212                 break;
8213         default:
8214                 nfs4_schedule_lease_recovery(clp);
8215         }
8216         return 0;
8217 }
8218
8219 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8220 {
8221         struct nfs4_reclaim_complete_data *calldata = data;
8222         struct nfs_client *clp = calldata->clp;
8223         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8224
8225         dprintk("--> %s\n", __func__);
8226         if (!nfs41_sequence_done(task, res))
8227                 return;
8228
8229         trace_nfs4_reclaim_complete(clp, task->tk_status);
8230         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8231                 rpc_restart_call_prepare(task);
8232                 return;
8233         }
8234         dprintk("<-- %s\n", __func__);
8235 }
8236
8237 static void nfs4_free_reclaim_complete_data(void *data)
8238 {
8239         struct nfs4_reclaim_complete_data *calldata = data;
8240
8241         kfree(calldata);
8242 }
8243
8244 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8245         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8246         .rpc_call_done = nfs4_reclaim_complete_done,
8247         .rpc_release = nfs4_free_reclaim_complete_data,
8248 };
8249
8250 /*
8251  * Issue a global reclaim complete.
8252  */
8253 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8254                 struct rpc_cred *cred)
8255 {
8256         struct nfs4_reclaim_complete_data *calldata;
8257         struct rpc_task *task;
8258         struct rpc_message msg = {
8259                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8260                 .rpc_cred = cred,
8261         };
8262         struct rpc_task_setup task_setup_data = {
8263                 .rpc_client = clp->cl_rpcclient,
8264                 .rpc_message = &msg,
8265                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8266                 .flags = RPC_TASK_ASYNC,
8267         };
8268         int status = -ENOMEM;
8269
8270         dprintk("--> %s\n", __func__);
8271         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8272         if (calldata == NULL)
8273                 goto out;
8274         calldata->clp = clp;
8275         calldata->arg.one_fs = 0;
8276
8277         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8278         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8279         msg.rpc_argp = &calldata->arg;
8280         msg.rpc_resp = &calldata->res;
8281         task_setup_data.callback_data = calldata;
8282         task = rpc_run_task(&task_setup_data);
8283         if (IS_ERR(task)) {
8284                 status = PTR_ERR(task);
8285                 goto out;
8286         }
8287         status = rpc_wait_for_completion_task(task);
8288         if (status == 0)
8289                 status = task->tk_status;
8290         rpc_put_task(task);
8291 out:
8292         dprintk("<-- %s status=%d\n", __func__, status);
8293         return status;
8294 }
8295
8296 static void
8297 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8298 {
8299         struct nfs4_layoutget *lgp = calldata;
8300         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8301
8302         dprintk("--> %s\n", __func__);
8303         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
8304                                 &lgp->res.seq_res, task);
8305         dprintk("<-- %s\n", __func__);
8306 }
8307
8308 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8309 {
8310         struct nfs4_layoutget *lgp = calldata;
8311
8312         dprintk("--> %s\n", __func__);
8313         nfs41_sequence_process(task, &lgp->res.seq_res);
8314         dprintk("<-- %s\n", __func__);
8315 }
8316
8317 static int
8318 nfs4_layoutget_handle_exception(struct rpc_task *task,
8319                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8320 {
8321         struct inode *inode = lgp->args.inode;
8322         struct nfs_server *server = NFS_SERVER(inode);
8323         struct pnfs_layout_hdr *lo;
8324         int nfs4err = task->tk_status;
8325         int err, status = 0;
8326         LIST_HEAD(head);
8327
8328         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8329
8330         switch (nfs4err) {
8331         case 0:
8332                 goto out;
8333
8334         /*
8335          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8336          * on the file. set tk_status to -ENODATA to tell upper layer to
8337          * retry go inband.
8338          */
8339         case -NFS4ERR_LAYOUTUNAVAILABLE:
8340                 status = -ENODATA;
8341                 goto out;
8342         /*
8343          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8344          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8345          */
8346         case -NFS4ERR_BADLAYOUT:
8347                 status = -EOVERFLOW;
8348                 goto out;
8349         /*
8350          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8351          * (or clients) writing to the same RAID stripe except when
8352          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8353          *
8354          * Treat it like we would RECALLCONFLICT -- we retry for a little
8355          * while, and then eventually give up.
8356          */
8357         case -NFS4ERR_LAYOUTTRYLATER:
8358                 if (lgp->args.minlength == 0) {
8359                         status = -EOVERFLOW;
8360                         goto out;
8361                 }
8362                 status = -EBUSY;
8363                 break;
8364         case -NFS4ERR_RECALLCONFLICT:
8365                 status = -ERECALLCONFLICT;
8366                 break;
8367         case -NFS4ERR_DELEG_REVOKED:
8368         case -NFS4ERR_ADMIN_REVOKED:
8369         case -NFS4ERR_EXPIRED:
8370         case -NFS4ERR_BAD_STATEID:
8371                 exception->timeout = 0;
8372                 spin_lock(&inode->i_lock);
8373                 lo = NFS_I(inode)->layout;
8374                 /* If the open stateid was bad, then recover it. */
8375                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8376                     nfs4_stateid_match_other(&lgp->args.stateid,
8377                                         &lgp->args.ctx->state->stateid)) {
8378                         spin_unlock(&inode->i_lock);
8379                         exception->state = lgp->args.ctx->state;
8380                         exception->stateid = &lgp->args.stateid;
8381                         break;
8382                 }
8383
8384                 /*
8385                  * Mark the bad layout state as invalid, then retry
8386                  */
8387                 pnfs_mark_layout_stateid_invalid(lo, &head);
8388                 spin_unlock(&inode->i_lock);
8389                 nfs_commit_inode(inode, 0);
8390                 pnfs_free_lseg_list(&head);
8391                 status = -EAGAIN;
8392                 goto out;
8393         }
8394
8395         nfs4_sequence_free_slot(&lgp->res.seq_res);
8396         err = nfs4_handle_exception(server, nfs4err, exception);
8397         if (!status) {
8398                 if (exception->retry)
8399                         status = -EAGAIN;
8400                 else
8401                         status = err;
8402         }
8403 out:
8404         dprintk("<-- %s\n", __func__);
8405         return status;
8406 }
8407
8408 static size_t max_response_pages(struct nfs_server *server)
8409 {
8410         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8411         return nfs_page_array_len(0, max_resp_sz);
8412 }
8413
8414 static void nfs4_free_pages(struct page **pages, size_t size)
8415 {
8416         int i;
8417
8418         if (!pages)
8419                 return;
8420
8421         for (i = 0; i < size; i++) {
8422                 if (!pages[i])
8423                         break;
8424                 __free_page(pages[i]);
8425         }
8426         kfree(pages);
8427 }
8428
8429 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8430 {
8431         struct page **pages;
8432         int i;
8433
8434         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8435         if (!pages) {
8436                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8437                 return NULL;
8438         }
8439
8440         for (i = 0; i < size; i++) {
8441                 pages[i] = alloc_page(gfp_flags);
8442                 if (!pages[i]) {
8443                         dprintk("%s: failed to allocate page\n", __func__);
8444                         nfs4_free_pages(pages, size);
8445                         return NULL;
8446                 }
8447         }
8448
8449         return pages;
8450 }
8451
8452 static void nfs4_layoutget_release(void *calldata)
8453 {
8454         struct nfs4_layoutget *lgp = calldata;
8455         struct inode *inode = lgp->args.inode;
8456         struct nfs_server *server = NFS_SERVER(inode);
8457         size_t max_pages = max_response_pages(server);
8458
8459         dprintk("--> %s\n", __func__);
8460         nfs4_sequence_free_slot(&lgp->res.seq_res);
8461         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8462         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8463         put_nfs_open_context(lgp->args.ctx);
8464         kfree(calldata);
8465         dprintk("<-- %s\n", __func__);
8466 }
8467
8468 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8469         .rpc_call_prepare = nfs4_layoutget_prepare,
8470         .rpc_call_done = nfs4_layoutget_done,
8471         .rpc_release = nfs4_layoutget_release,
8472 };
8473
8474 struct pnfs_layout_segment *
8475 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8476 {
8477         struct inode *inode = lgp->args.inode;
8478         struct nfs_server *server = NFS_SERVER(inode);
8479         size_t max_pages = max_response_pages(server);
8480         struct rpc_task *task;
8481         struct rpc_message msg = {
8482                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8483                 .rpc_argp = &lgp->args,
8484                 .rpc_resp = &lgp->res,
8485                 .rpc_cred = lgp->cred,
8486         };
8487         struct rpc_task_setup task_setup_data = {
8488                 .rpc_client = server->client,
8489                 .rpc_message = &msg,
8490                 .callback_ops = &nfs4_layoutget_call_ops,
8491                 .callback_data = lgp,
8492                 .flags = RPC_TASK_ASYNC,
8493         };
8494         struct pnfs_layout_segment *lseg = NULL;
8495         struct nfs4_exception exception = {
8496                 .inode = inode,
8497                 .timeout = *timeout,
8498         };
8499         int status = 0;
8500
8501         dprintk("--> %s\n", __func__);
8502
8503         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8504         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8505
8506         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8507         if (!lgp->args.layout.pages) {
8508                 nfs4_layoutget_release(lgp);
8509                 return ERR_PTR(-ENOMEM);
8510         }
8511         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8512
8513         lgp->res.layoutp = &lgp->args.layout;
8514         lgp->res.seq_res.sr_slot = NULL;
8515         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8516
8517         task = rpc_run_task(&task_setup_data);
8518         if (IS_ERR(task))
8519                 return ERR_CAST(task);
8520         status = rpc_wait_for_completion_task(task);
8521         if (status == 0) {
8522                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8523                 *timeout = exception.timeout;
8524         }
8525
8526         trace_nfs4_layoutget(lgp->args.ctx,
8527                         &lgp->args.range,
8528                         &lgp->res.range,
8529                         &lgp->res.stateid,
8530                         status);
8531
8532         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8533         if (status == 0 && lgp->res.layoutp->len)
8534                 lseg = pnfs_layout_process(lgp);
8535         rpc_put_task(task);
8536         dprintk("<-- %s status=%d\n", __func__, status);
8537         if (status)
8538                 return ERR_PTR(status);
8539         return lseg;
8540 }
8541
8542 static void
8543 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8544 {
8545         struct nfs4_layoutreturn *lrp = calldata;
8546
8547         dprintk("--> %s\n", __func__);
8548         nfs4_setup_sequence(lrp->clp,
8549                         &lrp->args.seq_args,
8550                         &lrp->res.seq_res,
8551                         task);
8552 }
8553
8554 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8555 {
8556         struct nfs4_layoutreturn *lrp = calldata;
8557         struct nfs_server *server;
8558
8559         dprintk("--> %s\n", __func__);
8560
8561         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8562                 return;
8563
8564         server = NFS_SERVER(lrp->args.inode);
8565         switch (task->tk_status) {
8566         default:
8567                 task->tk_status = 0;
8568         case 0:
8569                 break;
8570         case -NFS4ERR_DELAY:
8571                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8572                         break;
8573                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8574                 rpc_restart_call_prepare(task);
8575                 return;
8576         }
8577         dprintk("<-- %s\n", __func__);
8578 }
8579
8580 static void nfs4_layoutreturn_release(void *calldata)
8581 {
8582         struct nfs4_layoutreturn *lrp = calldata;
8583         struct pnfs_layout_hdr *lo = lrp->args.layout;
8584
8585         dprintk("--> %s\n", __func__);
8586         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
8587                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
8588         nfs4_sequence_free_slot(&lrp->res.seq_res);
8589         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
8590                 lrp->ld_private.ops->free(&lrp->ld_private);
8591         pnfs_put_layout_hdr(lrp->args.layout);
8592         nfs_iput_and_deactive(lrp->inode);
8593         kfree(calldata);
8594         dprintk("<-- %s\n", __func__);
8595 }
8596
8597 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8598         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8599         .rpc_call_done = nfs4_layoutreturn_done,
8600         .rpc_release = nfs4_layoutreturn_release,
8601 };
8602
8603 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8604 {
8605         struct rpc_task *task;
8606         struct rpc_message msg = {
8607                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8608                 .rpc_argp = &lrp->args,
8609                 .rpc_resp = &lrp->res,
8610                 .rpc_cred = lrp->cred,
8611         };
8612         struct rpc_task_setup task_setup_data = {
8613                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8614                 .rpc_message = &msg,
8615                 .callback_ops = &nfs4_layoutreturn_call_ops,
8616                 .callback_data = lrp,
8617         };
8618         int status = 0;
8619
8620         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8621                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8622                         &task_setup_data.rpc_client, &msg);
8623
8624         dprintk("--> %s\n", __func__);
8625         if (!sync) {
8626                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8627                 if (!lrp->inode) {
8628                         nfs4_layoutreturn_release(lrp);
8629                         return -EAGAIN;
8630                 }
8631                 task_setup_data.flags |= RPC_TASK_ASYNC;
8632         }
8633         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8634         task = rpc_run_task(&task_setup_data);
8635         if (IS_ERR(task))
8636                 return PTR_ERR(task);
8637         if (sync)
8638                 status = task->tk_status;
8639         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8640         dprintk("<-- %s status=%d\n", __func__, status);
8641         rpc_put_task(task);
8642         return status;
8643 }
8644
8645 static int
8646 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8647                 struct pnfs_device *pdev,
8648                 struct rpc_cred *cred)
8649 {
8650         struct nfs4_getdeviceinfo_args args = {
8651                 .pdev = pdev,
8652                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8653                         NOTIFY_DEVICEID4_DELETE,
8654         };
8655         struct nfs4_getdeviceinfo_res res = {
8656                 .pdev = pdev,
8657         };
8658         struct rpc_message msg = {
8659                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8660                 .rpc_argp = &args,
8661                 .rpc_resp = &res,
8662                 .rpc_cred = cred,
8663         };
8664         int status;
8665
8666         dprintk("--> %s\n", __func__);
8667         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8668         if (res.notification & ~args.notify_types)
8669                 dprintk("%s: unsupported notification\n", __func__);
8670         if (res.notification != args.notify_types)
8671                 pdev->nocache = 1;
8672
8673         dprintk("<-- %s status=%d\n", __func__, status);
8674
8675         return status;
8676 }
8677
8678 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8679                 struct pnfs_device *pdev,
8680                 struct rpc_cred *cred)
8681 {
8682         struct nfs4_exception exception = { };
8683         int err;
8684
8685         do {
8686                 err = nfs4_handle_exception(server,
8687                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8688                                         &exception);
8689         } while (exception.retry);
8690         return err;
8691 }
8692 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8693
8694 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8695 {
8696         struct nfs4_layoutcommit_data *data = calldata;
8697         struct nfs_server *server = NFS_SERVER(data->args.inode);
8698
8699         nfs4_setup_sequence(server->nfs_client,
8700                         &data->args.seq_args,
8701                         &data->res.seq_res,
8702                         task);
8703 }
8704
8705 static void
8706 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8707 {
8708         struct nfs4_layoutcommit_data *data = calldata;
8709         struct nfs_server *server = NFS_SERVER(data->args.inode);
8710
8711         if (!nfs41_sequence_done(task, &data->res.seq_res))
8712                 return;
8713
8714         switch (task->tk_status) { /* Just ignore these failures */
8715         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8716         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8717         case -NFS4ERR_BADLAYOUT:     /* no layout */
8718         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8719                 task->tk_status = 0;
8720         case 0:
8721                 break;
8722         default:
8723                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8724                         rpc_restart_call_prepare(task);
8725                         return;
8726                 }
8727         }
8728 }
8729
8730 static void nfs4_layoutcommit_release(void *calldata)
8731 {
8732         struct nfs4_layoutcommit_data *data = calldata;
8733
8734         pnfs_cleanup_layoutcommit(data);
8735         nfs_post_op_update_inode_force_wcc(data->args.inode,
8736                                            data->res.fattr);
8737         put_rpccred(data->cred);
8738         nfs_iput_and_deactive(data->inode);
8739         kfree(data);
8740 }
8741
8742 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8743         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8744         .rpc_call_done = nfs4_layoutcommit_done,
8745         .rpc_release = nfs4_layoutcommit_release,
8746 };
8747
8748 int
8749 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8750 {
8751         struct rpc_message msg = {
8752                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8753                 .rpc_argp = &data->args,
8754                 .rpc_resp = &data->res,
8755                 .rpc_cred = data->cred,
8756         };
8757         struct rpc_task_setup task_setup_data = {
8758                 .task = &data->task,
8759                 .rpc_client = NFS_CLIENT(data->args.inode),
8760                 .rpc_message = &msg,
8761                 .callback_ops = &nfs4_layoutcommit_ops,
8762                 .callback_data = data,
8763         };
8764         struct rpc_task *task;
8765         int status = 0;
8766
8767         dprintk("NFS: initiating layoutcommit call. sync %d "
8768                 "lbw: %llu inode %lu\n", sync,
8769                 data->args.lastbytewritten,
8770                 data->args.inode->i_ino);
8771
8772         if (!sync) {
8773                 data->inode = nfs_igrab_and_active(data->args.inode);
8774                 if (data->inode == NULL) {
8775                         nfs4_layoutcommit_release(data);
8776                         return -EAGAIN;
8777                 }
8778                 task_setup_data.flags = RPC_TASK_ASYNC;
8779         }
8780         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8781         task = rpc_run_task(&task_setup_data);
8782         if (IS_ERR(task))
8783                 return PTR_ERR(task);
8784         if (sync)
8785                 status = task->tk_status;
8786         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8787         dprintk("%s: status %d\n", __func__, status);
8788         rpc_put_task(task);
8789         return status;
8790 }
8791
8792 /**
8793  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8794  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8795  */
8796 static int
8797 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8798                     struct nfs_fsinfo *info,
8799                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8800 {
8801         struct nfs41_secinfo_no_name_args args = {
8802                 .style = SECINFO_STYLE_CURRENT_FH,
8803         };
8804         struct nfs4_secinfo_res res = {
8805                 .flavors = flavors,
8806         };
8807         struct rpc_message msg = {
8808                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8809                 .rpc_argp = &args,
8810                 .rpc_resp = &res,
8811         };
8812         struct rpc_clnt *clnt = server->client;
8813         struct rpc_cred *cred = NULL;
8814         int status;
8815
8816         if (use_integrity) {
8817                 clnt = server->nfs_client->cl_rpcclient;
8818                 cred = nfs4_get_clid_cred(server->nfs_client);
8819                 msg.rpc_cred = cred;
8820         }
8821
8822         dprintk("--> %s\n", __func__);
8823         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8824                                 &res.seq_res, 0);
8825         dprintk("<-- %s status=%d\n", __func__, status);
8826
8827         if (cred)
8828                 put_rpccred(cred);
8829
8830         return status;
8831 }
8832
8833 static int
8834 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8835                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8836 {
8837         struct nfs4_exception exception = { };
8838         int err;
8839         do {
8840                 /* first try using integrity protection */
8841                 err = -NFS4ERR_WRONGSEC;
8842
8843                 /* try to use integrity protection with machine cred */
8844                 if (_nfs4_is_integrity_protected(server->nfs_client))
8845                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8846                                                           flavors, true);
8847
8848                 /*
8849                  * if unable to use integrity protection, or SECINFO with
8850                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8851                  * disallowed by spec, but exists in deployed servers) use
8852                  * the current filesystem's rpc_client and the user cred.
8853                  */
8854                 if (err == -NFS4ERR_WRONGSEC)
8855                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8856                                                           flavors, false);
8857
8858                 switch (err) {
8859                 case 0:
8860                 case -NFS4ERR_WRONGSEC:
8861                 case -ENOTSUPP:
8862                         goto out;
8863                 default:
8864                         err = nfs4_handle_exception(server, err, &exception);
8865                 }
8866         } while (exception.retry);
8867 out:
8868         return err;
8869 }
8870
8871 static int
8872 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8873                     struct nfs_fsinfo *info)
8874 {
8875         int err;
8876         struct page *page;
8877         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8878         struct nfs4_secinfo_flavors *flavors;
8879         struct nfs4_secinfo4 *secinfo;
8880         int i;
8881
8882         page = alloc_page(GFP_KERNEL);
8883         if (!page) {
8884                 err = -ENOMEM;
8885                 goto out;
8886         }
8887
8888         flavors = page_address(page);
8889         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8890
8891         /*
8892          * Fall back on "guess and check" method if
8893          * the server doesn't support SECINFO_NO_NAME
8894          */
8895         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8896                 err = nfs4_find_root_sec(server, fhandle, info);
8897                 goto out_freepage;
8898         }
8899         if (err)
8900                 goto out_freepage;
8901
8902         for (i = 0; i < flavors->num_flavors; i++) {
8903                 secinfo = &flavors->flavors[i];
8904
8905                 switch (secinfo->flavor) {
8906                 case RPC_AUTH_NULL:
8907                 case RPC_AUTH_UNIX:
8908                 case RPC_AUTH_GSS:
8909                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8910                                         &secinfo->flavor_info);
8911                         break;
8912                 default:
8913                         flavor = RPC_AUTH_MAXFLAVOR;
8914                         break;
8915                 }
8916
8917                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8918                         flavor = RPC_AUTH_MAXFLAVOR;
8919
8920                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8921                         err = nfs4_lookup_root_sec(server, fhandle,
8922                                                    info, flavor);
8923                         if (!err)
8924                                 break;
8925                 }
8926         }
8927
8928         if (flavor == RPC_AUTH_MAXFLAVOR)
8929                 err = -EPERM;
8930
8931 out_freepage:
8932         put_page(page);
8933         if (err == -EACCES)
8934                 return -EPERM;
8935 out:
8936         return err;
8937 }
8938
8939 static int _nfs41_test_stateid(struct nfs_server *server,
8940                 nfs4_stateid *stateid,
8941                 struct rpc_cred *cred)
8942 {
8943         int status;
8944         struct nfs41_test_stateid_args args = {
8945                 .stateid = stateid,
8946         };
8947         struct nfs41_test_stateid_res res;
8948         struct rpc_message msg = {
8949                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8950                 .rpc_argp = &args,
8951                 .rpc_resp = &res,
8952                 .rpc_cred = cred,
8953         };
8954         struct rpc_clnt *rpc_client = server->client;
8955
8956         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8957                 &rpc_client, &msg);
8958
8959         dprintk("NFS call  test_stateid %p\n", stateid);
8960         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8961         nfs4_set_sequence_privileged(&args.seq_args);
8962         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8963                         &args.seq_args, &res.seq_res);
8964         if (status != NFS_OK) {
8965                 dprintk("NFS reply test_stateid: failed, %d\n", status);
8966                 return status;
8967         }
8968         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
8969         return -res.status;
8970 }
8971
8972 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
8973                 int err, struct nfs4_exception *exception)
8974 {
8975         exception->retry = 0;
8976         switch(err) {
8977         case -NFS4ERR_DELAY:
8978         case -NFS4ERR_RETRY_UNCACHED_REP:
8979                 nfs4_handle_exception(server, err, exception);
8980                 break;
8981         case -NFS4ERR_BADSESSION:
8982         case -NFS4ERR_BADSLOT:
8983         case -NFS4ERR_BAD_HIGH_SLOT:
8984         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8985         case -NFS4ERR_DEADSESSION:
8986                 nfs4_do_handle_exception(server, err, exception);
8987         }
8988 }
8989
8990 /**
8991  * nfs41_test_stateid - perform a TEST_STATEID operation
8992  *
8993  * @server: server / transport on which to perform the operation
8994  * @stateid: state ID to test
8995  * @cred: credential
8996  *
8997  * Returns NFS_OK if the server recognizes that "stateid" is valid.
8998  * Otherwise a negative NFS4ERR value is returned if the operation
8999  * failed or the state ID is not currently valid.
9000  */
9001 static int nfs41_test_stateid(struct nfs_server *server,
9002                 nfs4_stateid *stateid,
9003                 struct rpc_cred *cred)
9004 {
9005         struct nfs4_exception exception = { };
9006         int err;
9007         do {
9008                 err = _nfs41_test_stateid(server, stateid, cred);
9009                 nfs4_handle_delay_or_session_error(server, err, &exception);
9010         } while (exception.retry);
9011         return err;
9012 }
9013
9014 struct nfs_free_stateid_data {
9015         struct nfs_server *server;
9016         struct nfs41_free_stateid_args args;
9017         struct nfs41_free_stateid_res res;
9018 };
9019
9020 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9021 {
9022         struct nfs_free_stateid_data *data = calldata;
9023         nfs4_setup_sequence(data->server->nfs_client,
9024                         &data->args.seq_args,
9025                         &data->res.seq_res,
9026                         task);
9027 }
9028
9029 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9030 {
9031         struct nfs_free_stateid_data *data = calldata;
9032
9033         nfs41_sequence_done(task, &data->res.seq_res);
9034
9035         switch (task->tk_status) {
9036         case -NFS4ERR_DELAY:
9037                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9038                         rpc_restart_call_prepare(task);
9039         }
9040 }
9041
9042 static void nfs41_free_stateid_release(void *calldata)
9043 {
9044         kfree(calldata);
9045 }
9046
9047 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9048         .rpc_call_prepare = nfs41_free_stateid_prepare,
9049         .rpc_call_done = nfs41_free_stateid_done,
9050         .rpc_release = nfs41_free_stateid_release,
9051 };
9052
9053 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9054                 const nfs4_stateid *stateid,
9055                 struct rpc_cred *cred,
9056                 bool privileged)
9057 {
9058         struct rpc_message msg = {
9059                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9060                 .rpc_cred = cred,
9061         };
9062         struct rpc_task_setup task_setup = {
9063                 .rpc_client = server->client,
9064                 .rpc_message = &msg,
9065                 .callback_ops = &nfs41_free_stateid_ops,
9066                 .flags = RPC_TASK_ASYNC,
9067         };
9068         struct nfs_free_stateid_data *data;
9069
9070         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9071                 &task_setup.rpc_client, &msg);
9072
9073         dprintk("NFS call  free_stateid %p\n", stateid);
9074         data = kmalloc(sizeof(*data), GFP_NOFS);
9075         if (!data)
9076                 return ERR_PTR(-ENOMEM);
9077         data->server = server;
9078         nfs4_stateid_copy(&data->args.stateid, stateid);
9079
9080         task_setup.callback_data = data;
9081
9082         msg.rpc_argp = &data->args;
9083         msg.rpc_resp = &data->res;
9084         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9085         if (privileged)
9086                 nfs4_set_sequence_privileged(&data->args.seq_args);
9087
9088         return rpc_run_task(&task_setup);
9089 }
9090
9091 /**
9092  * nfs41_free_stateid - perform a FREE_STATEID operation
9093  *
9094  * @server: server / transport on which to perform the operation
9095  * @stateid: state ID to release
9096  * @cred: credential
9097  * @is_recovery: set to true if this call needs to be privileged
9098  *
9099  * Note: this function is always asynchronous.
9100  */
9101 static int nfs41_free_stateid(struct nfs_server *server,
9102                 const nfs4_stateid *stateid,
9103                 struct rpc_cred *cred,
9104                 bool is_recovery)
9105 {
9106         struct rpc_task *task;
9107
9108         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9109         if (IS_ERR(task))
9110                 return PTR_ERR(task);
9111         rpc_put_task(task);
9112         return 0;
9113 }
9114
9115 static void
9116 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9117 {
9118         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9119
9120         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9121         nfs4_free_lock_state(server, lsp);
9122 }
9123
9124 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9125                 const nfs4_stateid *s2)
9126 {
9127         if (s1->type != s2->type)
9128                 return false;
9129
9130         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9131                 return false;
9132
9133         if (s1->seqid == s2->seqid)
9134                 return true;
9135
9136         return s1->seqid == 0 || s2->seqid == 0;
9137 }
9138
9139 #endif /* CONFIG_NFS_V4_1 */
9140
9141 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9142                 const nfs4_stateid *s2)
9143 {
9144         return nfs4_stateid_match(s1, s2);
9145 }
9146
9147
9148 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9149         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9150         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9151         .recover_open   = nfs4_open_reclaim,
9152         .recover_lock   = nfs4_lock_reclaim,
9153         .establish_clid = nfs4_init_clientid,
9154         .detect_trunking = nfs40_discover_server_trunking,
9155 };
9156
9157 #if defined(CONFIG_NFS_V4_1)
9158 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9159         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9160         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9161         .recover_open   = nfs4_open_reclaim,
9162         .recover_lock   = nfs4_lock_reclaim,
9163         .establish_clid = nfs41_init_clientid,
9164         .reclaim_complete = nfs41_proc_reclaim_complete,
9165         .detect_trunking = nfs41_discover_server_trunking,
9166 };
9167 #endif /* CONFIG_NFS_V4_1 */
9168
9169 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9170         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9171         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9172         .recover_open   = nfs40_open_expired,
9173         .recover_lock   = nfs4_lock_expired,
9174         .establish_clid = nfs4_init_clientid,
9175 };
9176
9177 #if defined(CONFIG_NFS_V4_1)
9178 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9179         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9180         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9181         .recover_open   = nfs41_open_expired,
9182         .recover_lock   = nfs41_lock_expired,
9183         .establish_clid = nfs41_init_clientid,
9184 };
9185 #endif /* CONFIG_NFS_V4_1 */
9186
9187 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9188         .sched_state_renewal = nfs4_proc_async_renew,
9189         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9190         .renew_lease = nfs4_proc_renew,
9191 };
9192
9193 #if defined(CONFIG_NFS_V4_1)
9194 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9195         .sched_state_renewal = nfs41_proc_async_sequence,
9196         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9197         .renew_lease = nfs4_proc_sequence,
9198 };
9199 #endif
9200
9201 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9202         .get_locations = _nfs40_proc_get_locations,
9203         .fsid_present = _nfs40_proc_fsid_present,
9204 };
9205
9206 #if defined(CONFIG_NFS_V4_1)
9207 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9208         .get_locations = _nfs41_proc_get_locations,
9209         .fsid_present = _nfs41_proc_fsid_present,
9210 };
9211 #endif  /* CONFIG_NFS_V4_1 */
9212
9213 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9214         .minor_version = 0,
9215         .init_caps = NFS_CAP_READDIRPLUS
9216                 | NFS_CAP_ATOMIC_OPEN
9217                 | NFS_CAP_POSIX_LOCK,
9218         .init_client = nfs40_init_client,
9219         .shutdown_client = nfs40_shutdown_client,
9220         .match_stateid = nfs4_match_stateid,
9221         .find_root_sec = nfs4_find_root_sec,
9222         .free_lock_state = nfs4_release_lockowner,
9223         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9224         .alloc_seqid = nfs_alloc_seqid,
9225         .call_sync_ops = &nfs40_call_sync_ops,
9226         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9227         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9228         .state_renewal_ops = &nfs40_state_renewal_ops,
9229         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9230 };
9231
9232 #if defined(CONFIG_NFS_V4_1)
9233 static struct nfs_seqid *
9234 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9235 {
9236         return NULL;
9237 }
9238
9239 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9240         .minor_version = 1,
9241         .init_caps = NFS_CAP_READDIRPLUS
9242                 | NFS_CAP_ATOMIC_OPEN
9243                 | NFS_CAP_POSIX_LOCK
9244                 | NFS_CAP_STATEID_NFSV41
9245                 | NFS_CAP_ATOMIC_OPEN_V1,
9246         .init_client = nfs41_init_client,
9247         .shutdown_client = nfs41_shutdown_client,
9248         .match_stateid = nfs41_match_stateid,
9249         .find_root_sec = nfs41_find_root_sec,
9250         .free_lock_state = nfs41_free_lock_state,
9251         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9252         .alloc_seqid = nfs_alloc_no_seqid,
9253         .session_trunk = nfs4_test_session_trunk,
9254         .call_sync_ops = &nfs41_call_sync_ops,
9255         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9256         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9257         .state_renewal_ops = &nfs41_state_renewal_ops,
9258         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9259 };
9260 #endif
9261
9262 #if defined(CONFIG_NFS_V4_2)
9263 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9264         .minor_version = 2,
9265         .init_caps = NFS_CAP_READDIRPLUS
9266                 | NFS_CAP_ATOMIC_OPEN
9267                 | NFS_CAP_POSIX_LOCK
9268                 | NFS_CAP_STATEID_NFSV41
9269                 | NFS_CAP_ATOMIC_OPEN_V1
9270                 | NFS_CAP_ALLOCATE
9271                 | NFS_CAP_COPY
9272                 | NFS_CAP_DEALLOCATE
9273                 | NFS_CAP_SEEK
9274                 | NFS_CAP_LAYOUTSTATS
9275                 | NFS_CAP_CLONE,
9276         .init_client = nfs41_init_client,
9277         .shutdown_client = nfs41_shutdown_client,
9278         .match_stateid = nfs41_match_stateid,
9279         .find_root_sec = nfs41_find_root_sec,
9280         .free_lock_state = nfs41_free_lock_state,
9281         .call_sync_ops = &nfs41_call_sync_ops,
9282         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9283         .alloc_seqid = nfs_alloc_no_seqid,
9284         .session_trunk = nfs4_test_session_trunk,
9285         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9286         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9287         .state_renewal_ops = &nfs41_state_renewal_ops,
9288         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9289 };
9290 #endif
9291
9292 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9293         [0] = &nfs_v4_0_minor_ops,
9294 #if defined(CONFIG_NFS_V4_1)
9295         [1] = &nfs_v4_1_minor_ops,
9296 #endif
9297 #if defined(CONFIG_NFS_V4_2)
9298         [2] = &nfs_v4_2_minor_ops,
9299 #endif
9300 };
9301
9302 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9303 {
9304         ssize_t error, error2;
9305
9306         error = generic_listxattr(dentry, list, size);
9307         if (error < 0)
9308                 return error;
9309         if (list) {
9310                 list += error;
9311                 size -= error;
9312         }
9313
9314         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9315         if (error2 < 0)
9316                 return error2;
9317         return error + error2;
9318 }
9319
9320 static const struct inode_operations nfs4_dir_inode_operations = {
9321         .create         = nfs_create,
9322         .lookup         = nfs_lookup,
9323         .atomic_open    = nfs_atomic_open,
9324         .link           = nfs_link,
9325         .unlink         = nfs_unlink,
9326         .symlink        = nfs_symlink,
9327         .mkdir          = nfs_mkdir,
9328         .rmdir          = nfs_rmdir,
9329         .mknod          = nfs_mknod,
9330         .rename         = nfs_rename,
9331         .permission     = nfs_permission,
9332         .getattr        = nfs_getattr,
9333         .setattr        = nfs_setattr,
9334         .listxattr      = nfs4_listxattr,
9335 };
9336
9337 static const struct inode_operations nfs4_file_inode_operations = {
9338         .permission     = nfs_permission,
9339         .getattr        = nfs_getattr,
9340         .setattr        = nfs_setattr,
9341         .listxattr      = nfs4_listxattr,
9342 };
9343
9344 const struct nfs_rpc_ops nfs_v4_clientops = {
9345         .version        = 4,                    /* protocol version */
9346         .dentry_ops     = &nfs4_dentry_operations,
9347         .dir_inode_ops  = &nfs4_dir_inode_operations,
9348         .file_inode_ops = &nfs4_file_inode_operations,
9349         .file_ops       = &nfs4_file_operations,
9350         .getroot        = nfs4_proc_get_root,
9351         .submount       = nfs4_submount,
9352         .try_mount      = nfs4_try_mount,
9353         .getattr        = nfs4_proc_getattr,
9354         .setattr        = nfs4_proc_setattr,
9355         .lookup         = nfs4_proc_lookup,
9356         .lookupp        = nfs4_proc_lookupp,
9357         .access         = nfs4_proc_access,
9358         .readlink       = nfs4_proc_readlink,
9359         .create         = nfs4_proc_create,
9360         .remove         = nfs4_proc_remove,
9361         .unlink_setup   = nfs4_proc_unlink_setup,
9362         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9363         .unlink_done    = nfs4_proc_unlink_done,
9364         .rename_setup   = nfs4_proc_rename_setup,
9365         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9366         .rename_done    = nfs4_proc_rename_done,
9367         .link           = nfs4_proc_link,
9368         .symlink        = nfs4_proc_symlink,
9369         .mkdir          = nfs4_proc_mkdir,
9370         .rmdir          = nfs4_proc_remove,
9371         .readdir        = nfs4_proc_readdir,
9372         .mknod          = nfs4_proc_mknod,
9373         .statfs         = nfs4_proc_statfs,
9374         .fsinfo         = nfs4_proc_fsinfo,
9375         .pathconf       = nfs4_proc_pathconf,
9376         .set_capabilities = nfs4_server_capabilities,
9377         .decode_dirent  = nfs4_decode_dirent,
9378         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9379         .read_setup     = nfs4_proc_read_setup,
9380         .read_done      = nfs4_read_done,
9381         .write_setup    = nfs4_proc_write_setup,
9382         .write_done     = nfs4_write_done,
9383         .commit_setup   = nfs4_proc_commit_setup,
9384         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9385         .commit_done    = nfs4_commit_done,
9386         .lock           = nfs4_proc_lock,
9387         .clear_acl_cache = nfs4_zap_acl_attr,
9388         .close_context  = nfs4_close_context,
9389         .open_context   = nfs4_atomic_open,
9390         .have_delegation = nfs4_have_delegation,
9391         .return_delegation = nfs4_inode_return_delegation,
9392         .alloc_client   = nfs4_alloc_client,
9393         .init_client    = nfs4_init_client,
9394         .free_client    = nfs4_free_client,
9395         .create_server  = nfs4_create_server,
9396         .clone_server   = nfs_clone_server,
9397 };
9398
9399 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9400         .name   = XATTR_NAME_NFSV4_ACL,
9401         .list   = nfs4_xattr_list_nfs4_acl,
9402         .get    = nfs4_xattr_get_nfs4_acl,
9403         .set    = nfs4_xattr_set_nfs4_acl,
9404 };
9405
9406 const struct xattr_handler *nfs4_xattr_handlers[] = {
9407         &nfs4_xattr_nfs4_acl_handler,
9408 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9409         &nfs4_xattr_nfs4_label_handler,
9410 #endif
9411         NULL
9412 };
9413
9414 /*
9415  * Local variables:
9416  *  c-basic-offset: 8
9417  * End:
9418  */