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 nfs4_state *state, 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_CHANGE
 229         | FATTR4_WORD0_FILEID,
 230 };
 231
 232 const u32 nfs4_statfs_bitmap[3] = {
 233         FATTR4_WORD0_FILES_AVAIL
 234         | FATTR4_WORD0_FILES_FREE
 235         | FATTR4_WORD0_FILES_TOTAL,
 236         FATTR4_WORD1_SPACE_AVAIL
 237         | FATTR4_WORD1_SPACE_FREE
 238         | FATTR4_WORD1_SPACE_TOTAL
 239 };
 240
 241 const u32 nfs4_pathconf_bitmap[3] = {
 242         FATTR4_WORD0_MAXLINK
 243         | FATTR4_WORD0_MAXNAME,
 244         0
 245 };
 246
 247 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 248                         | FATTR4_WORD0_MAXREAD
 249                         | FATTR4_WORD0_MAXWRITE
 250                         | FATTR4_WORD0_LEASE_TIME,
 251                         FATTR4_WORD1_TIME_DELTA
 252                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 253                         FATTR4_WORD2_LAYOUT_BLKSIZE
 254                         | FATTR4_WORD2_CLONE_BLKSIZE
 255 };
 256
 257 const u32 nfs4_fs_locations_bitmap[3] = {
 258         FATTR4_WORD0_CHANGE
 259         | FATTR4_WORD0_SIZE
 260         | FATTR4_WORD0_FSID
 261         | FATTR4_WORD0_FILEID
 262         | FATTR4_WORD0_FS_LOCATIONS,
 263         FATTR4_WORD1_OWNER
 264         | FATTR4_WORD1_OWNER_GROUP
 265         | FATTR4_WORD1_RAWDEV
 266         | FATTR4_WORD1_SPACE_USED
 267         | FATTR4_WORD1_TIME_ACCESS
 268         | FATTR4_WORD1_TIME_METADATA
 269         | FATTR4_WORD1_TIME_MODIFY
 270         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 271 };
 272
 273 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 274                 struct nfs4_readdir_arg *readdir)
 275 {
 276         __be32 *start, *p;
 277
 278         if (cookie > 2) {
 279                 readdir->cookie = cookie;
 280                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 281                 return;
 282         }
 283
 284         readdir->cookie = 0;
 285         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 286         if (cookie == 2)
 287                 return;
 288
 289         /*
 290          * NFSv4 servers do not return entries for '.' and '..'
 291          * Therefore, we fake these entries here.  We let '.'
 292          * have cookie 0 and '..' have cookie 1.  Note that
 293          * when talking to the server, we always send cookie 0
 294          * instead of 1 or 2.
 295          */
 296         start = p = kmap_atomic(*readdir->pages);
 297
 298         if (cookie == 0) {
 299                 *p++ = xdr_one;                                  /* next */
 300                 *p++ = xdr_zero;                   /* cookie, first word */
 301                 *p++ = xdr_one;                   /* cookie, second word */
 302                 *p++ = xdr_one;                             /* entry len */
 303                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 304                 p++;
 305                 *p++ = xdr_one;                         /* bitmap length */
 306                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 307                 *p++ = htonl(8);              /* attribute buffer length */
 308                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 309         }
 310
 311         *p++ = xdr_one;                                  /* next */
 312         *p++ = xdr_zero;                   /* cookie, first word */
 313         *p++ = xdr_two;                   /* cookie, second word */
 314         *p++ = xdr_two;                             /* entry len */
 315         memcpy(p, "..\0\0", 4);                         /* entry */
 316         p++;
 317         *p++ = xdr_one;                         /* bitmap length */
 318         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
 319         *p++ = htonl(8);              /* attribute buffer length */
 320         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 321
 322         readdir->pgbase = (char *)p - (char *)start;
 323         readdir->count -= readdir->pgbase;
 324         kunmap_atomic(start);
 325 }
 326
 327 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 328                 nfs4_stateid *stateid,
 329                 struct rpc_cred *cred)
 330 {
 331         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 332
 333         ops->test_and_free_expired(server, stateid, cred);
 334 }
 335
 336 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 337                 nfs4_stateid *stateid,
 338                 struct rpc_cred *cred)
 339 {
 340         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 341         nfs4_test_and_free_stateid(server, stateid, cred);
 342 }
 343
 344 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 345                 const nfs4_stateid *stateid,
 346                 struct rpc_cred *cred)
 347 {
 348         nfs4_stateid tmp;
 349
 350         nfs4_stateid_copy(&tmp, stateid);
 351         __nfs4_free_revoked_stateid(server, &tmp, cred);
 352 }
 353
 354 static long nfs4_update_delay(long *timeout)
 355 {
 356         long ret;
 357         if (!timeout)
 358                 return NFS4_POLL_RETRY_MAX;
 359         if (*timeout <= 0)
 360                 *timeout = NFS4_POLL_RETRY_MIN;
 361         if (*timeout > NFS4_POLL_RETRY_MAX)
 362                 *timeout = NFS4_POLL_RETRY_MAX;
 363         ret = *timeout;
 364         *timeout <<= 1;
 365         return ret;
 366 }
 367
 368 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
 369 {
 370         int res = 0;
 371
 372         might_sleep();
 373
 374         freezable_schedule_timeout_killable_unsafe(
 375                 nfs4_update_delay(timeout));
 376         if (fatal_signal_pending(current))
 377                 res = -ERESTARTSYS;
 378         return res;
 379 }
 380
 381 /* This is the error handling routine for processes that are allowed
 382  * to sleep.
 383  */
 384 static int nfs4_do_handle_exception(struct nfs_server *server,
 385                 int errorcode, struct nfs4_exception *exception)
 386 {
 387         struct nfs_client *clp = server->nfs_client;
 388         struct nfs4_state *state = exception->state;
 389         const nfs4_stateid *stateid = exception->stateid;
 390         struct inode *inode = exception->inode;
 391         int ret = errorcode;
 392
 393         exception->delay = 0;
 394         exception->recovering = 0;
 395         exception->retry = 0;
 396
 397         if (stateid == NULL && state != NULL)
 398                 stateid = &state->stateid;
 399
 400         switch(errorcode) {
 401                 case 0:
 402                         return 0;
 403                 case -NFS4ERR_DELEG_REVOKED:
 404                 case -NFS4ERR_ADMIN_REVOKED:
 405                 case -NFS4ERR_EXPIRED:
 406                 case -NFS4ERR_BAD_STATEID:
 407                         if (inode != NULL && stateid != NULL) {
 408                                 nfs_inode_find_state_and_recover(inode,
 409                                                 stateid);
 410                                 goto wait_on_recovery;
 411                         }
 412                 case -NFS4ERR_OPENMODE:
 413                         if (inode) {
 414                                 int err;
 415
 416                                 err = nfs_async_inode_return_delegation(inode,
 417                                                 stateid);
 418                                 if (err == 0)
 419                                         goto wait_on_recovery;
 420                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 421                                         exception->retry = 1;
 422                                         break;
 423                                 }
 424                         }
 425                         if (state == NULL)
 426                                 break;
 427                         ret = nfs4_schedule_stateid_recovery(server, state);
 428                         if (ret < 0)
 429                                 break;
 430                         goto wait_on_recovery;
 431                 case -NFS4ERR_STALE_STATEID:
 432                 case -NFS4ERR_STALE_CLIENTID:
 433                         nfs4_schedule_lease_recovery(clp);
 434                         goto wait_on_recovery;
 435                 case -NFS4ERR_MOVED:
 436                         ret = nfs4_schedule_migration_recovery(server);
 437                         if (ret < 0)
 438                                 break;
 439                         goto wait_on_recovery;
 440                 case -NFS4ERR_LEASE_MOVED:
 441                         nfs4_schedule_lease_moved_recovery(clp);
 442                         goto wait_on_recovery;
 443 #if defined(CONFIG_NFS_V4_1)
 444                 case -NFS4ERR_BADSESSION:
 445                 case -NFS4ERR_BADSLOT:
 446                 case -NFS4ERR_BAD_HIGH_SLOT:
 447                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 448                 case -NFS4ERR_DEADSESSION:
 449                 case -NFS4ERR_SEQ_FALSE_RETRY:
 450                 case -NFS4ERR_SEQ_MISORDERED:
 451                         dprintk("%s ERROR: %d Reset session\n", __func__,
 452                                 errorcode);
 453                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
 454                         goto wait_on_recovery;
 455 #endif /* defined(CONFIG_NFS_V4_1) */
 456                 case -NFS4ERR_FILE_OPEN:
 457                         if (exception->timeout > HZ) {
 458                                 /* We have retried a decent amount, time to
 459                                  * fail
 460                                  */
 461                                 ret = -EBUSY;
 462                                 break;
 463                         }
 464                 case -NFS4ERR_DELAY:
 465                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 466                 case -NFS4ERR_GRACE:
 467                 case -NFS4ERR_LAYOUTTRYLATER:
 468                 case -NFS4ERR_RECALLCONFLICT:
 469                         exception->delay = 1;
 470                         return 0;
 471
 472                 case -NFS4ERR_RETRY_UNCACHED_REP:
 473                 case -NFS4ERR_OLD_STATEID:
 474                         exception->retry = 1;
 475                         break;
 476                 case -NFS4ERR_BADOWNER:
 477                         /* The following works around a Linux server bug! */
 478                 case -NFS4ERR_BADNAME:
 479                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 480                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 481                                 exception->retry = 1;
 482                                 printk(KERN_WARNING "NFS: v4 server %s "
 483                                                 "does not accept raw "
 484                                                 "uid/gids. "
 485                                                 "Reenabling the idmapper.\n",
 486                                                 server->nfs_client->cl_hostname);
 487                         }
 488         }
 489         /* We failed to handle the error */
 490         return nfs4_map_errors(ret);
 491 wait_on_recovery:
 492         exception->recovering = 1;
 493         return 0;
 494 }
 495
 496 /* This is the error handling routine for processes that are allowed
 497  * to sleep.
 498  */
 499 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 500 {
 501         struct nfs_client *clp = server->nfs_client;
 502         int ret;
 503
 504         ret = nfs4_do_handle_exception(server, errorcode, exception);
 505         if (exception->delay) {
 506                 ret = nfs4_delay(server->client, &exception->timeout);
 507                 goto out_retry;
 508         }
 509         if (exception->recovering) {
 510                 ret = nfs4_wait_clnt_recover(clp);
 511                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 512                         return -EIO;
 513                 goto out_retry;
 514         }
 515         return ret;
 516 out_retry:
 517         if (ret == 0)
 518                 exception->retry = 1;
 519         return ret;
 520 }
 521
 522 static int
 523 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 524                 int errorcode, struct nfs4_exception *exception)
 525 {
 526         struct nfs_client *clp = server->nfs_client;
 527         int ret;
 528
 529         ret = nfs4_do_handle_exception(server, errorcode, exception);
 530         if (exception->delay) {
 531                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 532                 goto out_retry;
 533         }
 534         if (exception->recovering) {
 535                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 536                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 537                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 538                 goto out_retry;
 539         }
 540         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 541                 ret = -EIO;
 542         return ret;
 543 out_retry:
 544         if (ret == 0) {
 545                 exception->retry = 1;
 546                 /*
 547                  * For NFS4ERR_MOVED, the client transport will need to
 548                  * be recomputed after migration recovery has completed.
 549                  */
 550                 if (errorcode == -NFS4ERR_MOVED)
 551                         rpc_task_release_transport(task);
 552         }
 553         return ret;
 554 }
 555
 556 static int
 557 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 558                         struct nfs4_state *state, long *timeout)
 559 {
 560         struct nfs4_exception exception = {
 561                 .state = state,
 562         };
 563
 564         if (task->tk_status >= 0)
 565                 return 0;
 566         if (timeout)
 567                 exception.timeout = *timeout;
 568         task->tk_status = nfs4_async_handle_exception(task, server,
 569                         task->tk_status,
 570                         &exception);
 571         if (exception.delay && timeout)
 572                 *timeout = exception.timeout;
 573         if (exception.retry)
 574                 return -EAGAIN;
 575         return 0;
 576 }
 577
 578 /*
 579  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 580  * or 'false' otherwise.
 581  */
 582 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 583 {
 584         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 585
 586         if (flavor == RPC_AUTH_GSS_KRB5I ||
 587             flavor == RPC_AUTH_GSS_KRB5P)
 588                 return true;
 589
 590         return false;
 591 }
 592
 593 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 594 {
 595         spin_lock(&clp->cl_lock);
 596         if (time_before(clp->cl_last_renewal,timestamp))
 597                 clp->cl_last_renewal = timestamp;
 598         spin_unlock(&clp->cl_lock);
 599 }
 600
 601 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 602 {
 603         struct nfs_client *clp = server->nfs_client;
 604
 605         if (!nfs4_has_session(clp))
 606                 do_renew_lease(clp, timestamp);
 607 }
 608
 609 struct nfs4_call_sync_data {
 610         const struct nfs_server *seq_server;
 611         struct nfs4_sequence_args *seq_args;
 612         struct nfs4_sequence_res *seq_res;
 613 };
 614
 615 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 616                         struct nfs4_sequence_res *res, int cache_reply)
 617 {
 618         args->sa_slot = NULL;
 619         args->sa_cache_this = cache_reply;
 620         args->sa_privileged = 0;
 621
 622         res->sr_slot = NULL;
 623 }
 624
 625 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
 626 {
 627         args->sa_privileged = 1;
 628 }
 629
 630 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
 631                          struct nfs4_sequence_args *args,
 632                          struct nfs4_sequence_res *res,
 633                          struct rpc_task *task)
 634 {
 635         struct nfs4_slot *slot;
 636
 637         /* slot already allocated? */
 638         if (res->sr_slot != NULL)
 639                 goto out_start;
 640
 641         spin_lock(&tbl->slot_tbl_lock);
 642         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
 643                 goto out_sleep;
 644
 645         slot = nfs4_alloc_slot(tbl);
 646         if (IS_ERR(slot)) {
 647                 if (slot == ERR_PTR(-ENOMEM))
 648                         task->tk_timeout = HZ >> 2;
 649                 goto out_sleep;
 650         }
 651         spin_unlock(&tbl->slot_tbl_lock);
 652
 653         slot->privileged = args->sa_privileged ? 1 : 0;
 654         args->sa_slot = slot;
 655         res->sr_slot = slot;
 656
 657 out_start:
 658         rpc_call_start(task);
 659         return 0;
 660
 661 out_sleep:
 662         if (args->sa_privileged)
 663                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 664                                 NULL, RPC_PRIORITY_PRIVILEGED);
 665         else
 666                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 667         spin_unlock(&tbl->slot_tbl_lock);
 668         return -EAGAIN;
 669 }
 670 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
 671
 672 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 673 {
 674         struct nfs4_slot *slot = res->sr_slot;
 675         struct nfs4_slot_table *tbl;
 676
 677         tbl = slot->table;
 678         spin_lock(&tbl->slot_tbl_lock);
 679         if (!nfs41_wake_and_assign_slot(tbl, slot))
 680                 nfs4_free_slot(tbl, slot);
 681         spin_unlock(&tbl->slot_tbl_lock);
 682
 683         res->sr_slot = NULL;
 684 }
 685
 686 static int nfs40_sequence_done(struct rpc_task *task,
 687                                struct nfs4_sequence_res *res)
 688 {
 689         if (res->sr_slot != NULL)
 690                 nfs40_sequence_free_slot(res);
 691         return 1;
 692 }
 693
 694 #if defined(CONFIG_NFS_V4_1)
 695
 696 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 697 {
 698         struct nfs4_session *session;
 699         struct nfs4_slot_table *tbl;
 700         struct nfs4_slot *slot = res->sr_slot;
 701         bool send_new_highest_used_slotid = false;
 702
 703         tbl = slot->table;
 704         session = tbl->session;
 705
 706         /* Bump the slot sequence number */
 707         if (slot->seq_done)
 708                 slot->seq_nr++;
 709         slot->seq_done = 0;
 710
 711         spin_lock(&tbl->slot_tbl_lock);
 712         /* Be nice to the server: try to ensure that the last transmitted
 713          * value for highest_user_slotid <= target_highest_slotid
 714          */
 715         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 716                 send_new_highest_used_slotid = true;
 717
 718         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 719                 send_new_highest_used_slotid = false;
 720                 goto out_unlock;
 721         }
 722         nfs4_free_slot(tbl, slot);
 723
 724         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 725                 send_new_highest_used_slotid = false;
 726 out_unlock:
 727         spin_unlock(&tbl->slot_tbl_lock);
 728         res->sr_slot = NULL;
 729         if (send_new_highest_used_slotid)
 730                 nfs41_notify_server(session->clp);
 731         if (waitqueue_active(&tbl->slot_waitq))
 732                 wake_up_all(&tbl->slot_waitq);
 733 }
 734
 735 static int nfs41_sequence_process(struct rpc_task *task,
 736                 struct nfs4_sequence_res *res)
 737 {
 738         struct nfs4_session *session;
 739         struct nfs4_slot *slot = res->sr_slot;
 740         struct nfs_client *clp;
 741         bool interrupted = false;
 742         int ret = 1;
 743
 744         if (slot == NULL)
 745                 goto out_noaction;
 746         /* don't increment the sequence number if the task wasn't sent */
 747         if (!RPC_WAS_SENT(task))
 748                 goto out;
 749
 750         session = slot->table->session;
 751
 752         if (slot->interrupted) {
 753                 slot->interrupted = 0;
 754                 interrupted = true;
 755         }
 756
 757         trace_nfs4_sequence_done(session, res);
 758         /* Check the SEQUENCE operation status */
 759         switch (res->sr_status) {
 760         case 0:
 761                 /* If previous op on slot was interrupted and we reused
 762                  * the seq# and got a reply from the cache, then retry
 763                  */
 764                 if (task->tk_status == -EREMOTEIO && interrupted) {
 765                         ++slot->seq_nr;
 766                         goto retry_nowait;
 767                 }
 768                 /* Update the slot's sequence and clientid lease timer */
 769                 slot->seq_done = 1;
 770                 clp = session->clp;
 771                 do_renew_lease(clp, res->sr_timestamp);
 772                 /* Check sequence flags */
 773                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 774                                 !!slot->privileged);
 775                 nfs41_update_target_slotid(slot->table, slot, res);
 776                 break;
 777         case 1:
 778                 /*
 779                  * sr_status remains 1 if an RPC level error occurred.
 780                  * The server may or may not have processed the sequence
 781                  * operation..
 782                  * Mark the slot as having hosted an interrupted RPC call.
 783                  */
 784                 slot->interrupted = 1;
 785                 goto out;
 786         case -NFS4ERR_DELAY:
 787                 /* The server detected a resend of the RPC call and
 788                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 789                  * of RFC5661.
 790                  */
 791                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 792                         __func__,
 793                         slot->slot_nr,
 794                         slot->seq_nr);
 795                 goto out_retry;
 796         case -NFS4ERR_BADSLOT:
 797                 /*
 798                  * The slot id we used was probably retired. Try again
 799                  * using a different slot id.
 800                  */
 801                 goto retry_nowait;
 802         case -NFS4ERR_SEQ_MISORDERED:
 803                 /*
 804                  * Was the last operation on this sequence interrupted?
 805                  * If so, retry after bumping the sequence number.
 806                  */
 807                 if (interrupted) {
 808                         ++slot->seq_nr;
 809                         goto retry_nowait;
 810                 }
 811                 /*
 812                  * Could this slot have been previously retired?
 813                  * If so, then the server may be expecting seq_nr = 1!
 814                  */
 815                 if (slot->seq_nr != 1) {
 816                         slot->seq_nr = 1;
 817                         goto retry_nowait;
 818                 }
 819                 break;
 820         case -NFS4ERR_SEQ_FALSE_RETRY:
 821                 ++slot->seq_nr;
 822                 goto retry_nowait;
 823         default:
 824                 /* Just update the slot sequence no. */
 825                 slot->seq_done = 1;
 826         }
 827 out:
 828         /* The session may be reset by one of the error handlers. */
 829         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 830 out_noaction:
 831         return ret;
 832 retry_nowait:
 833         if (rpc_restart_call_prepare(task)) {
 834                 nfs41_sequence_free_slot(res);
 835                 task->tk_status = 0;
 836                 ret = 0;
 837         }
 838         goto out;
 839 out_retry:
 840         if (!rpc_restart_call(task))
 841                 goto out;
 842         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 843         return 0;
 844 }
 845
 846 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 847 {
 848         if (!nfs41_sequence_process(task, res))
 849                 return 0;
 850         if (res->sr_slot != NULL)
 851                 nfs41_sequence_free_slot(res);
 852         return 1;
 853
 854 }
 855 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 856
 857 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 858 {
 859         if (res->sr_slot == NULL)
 860                 return 1;
 861         if (res->sr_slot->table->session != NULL)
 862                 return nfs41_sequence_process(task, res);
 863         return nfs40_sequence_done(task, res);
 864 }
 865
 866 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 867 {
 868         if (res->sr_slot != NULL) {
 869                 if (res->sr_slot->table->session != NULL)
 870                         nfs41_sequence_free_slot(res);
 871                 else
 872                         nfs40_sequence_free_slot(res);
 873         }
 874 }
 875
 876 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 877 {
 878         if (res->sr_slot == NULL)
 879                 return 1;
 880         if (!res->sr_slot->table->session)
 881                 return nfs40_sequence_done(task, res);
 882         return nfs41_sequence_done(task, res);
 883 }
 884 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 885
 886 int nfs41_setup_sequence(struct nfs4_session *session,
 887                                 struct nfs4_sequence_args *args,
 888                                 struct nfs4_sequence_res *res,
 889                                 struct rpc_task *task)
 890 {
 891         struct nfs4_slot *slot;
 892         struct nfs4_slot_table *tbl;
 893
 894         dprintk("--> %s\n", __func__);
 895         /* slot already allocated? */
 896         if (res->sr_slot != NULL)
 897                 goto out_success;
 898
 899         tbl = &session->fc_slot_table;
 900
 901         task->tk_timeout = 0;
 902
 903         spin_lock(&tbl->slot_tbl_lock);
 904         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
 905             !args->sa_privileged) {
 906                 /* The state manager will wait until the slot table is empty */
 907                 dprintk("%s session is draining\n", __func__);
 908                 goto out_sleep;
 909         }
 910
 911         slot = nfs4_alloc_slot(tbl);
 912         if (IS_ERR(slot)) {
 913                 /* If out of memory, try again in 1/4 second */
 914                 if (slot == ERR_PTR(-ENOMEM))
 915                         task->tk_timeout = HZ >> 2;
 916                 dprintk("<-- %s: no free slots\n", __func__);
 917                 goto out_sleep;
 918         }
 919         spin_unlock(&tbl->slot_tbl_lock);
 920
 921         slot->privileged = args->sa_privileged ? 1 : 0;
 922         args->sa_slot = slot;
 923
 924         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
 925                         slot->slot_nr, slot->seq_nr);
 926
 927         res->sr_slot = slot;
 928         res->sr_timestamp = jiffies;
 929         res->sr_status_flags = 0;
 930         /*
 931          * sr_status is only set in decode_sequence, and so will remain
 932          * set to 1 if an rpc level failure occurs.
 933          */
 934         res->sr_status = 1;
 935         trace_nfs4_setup_sequence(session, args);
 936 out_success:
 937         rpc_call_start(task);
 938         return 0;
 939 out_sleep:
 940         /* Privileged tasks are queued with top priority */
 941         if (args->sa_privileged)
 942                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
 943                                 NULL, RPC_PRIORITY_PRIVILEGED);
 944         else
 945                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
 946         spin_unlock(&tbl->slot_tbl_lock);
 947         return -EAGAIN;
 948 }
 949 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
 950
 951 static int nfs4_setup_sequence(const struct nfs_server *server,
 952                                struct nfs4_sequence_args *args,
 953                                struct nfs4_sequence_res *res,
 954                                struct rpc_task *task)
 955 {
 956         struct nfs4_session *session = nfs4_get_session(server);
 957         int ret = 0;
 958
 959         if (!session)
 960                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
 961                                             args, res, task);
 962
 963         dprintk("--> %s clp %p session %p sr_slot %u\n",
 964                 __func__, session->clp, session, res->sr_slot ?
 965                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
 966
 967         ret = nfs41_setup_sequence(session, args, res, task);
 968
 969         dprintk("<-- %s status=%d\n", __func__, ret);
 970         return ret;
 971 }
 972
 973 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 974 {
 975         struct nfs4_call_sync_data *data = calldata;
 976         struct nfs4_session *session = nfs4_get_session(data->seq_server);
 977
 978         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 979
 980         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
 981 }
 982
 983 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 984 {
 985         struct nfs4_call_sync_data *data = calldata;
 986
 987         nfs41_sequence_done(task, data->seq_res);
 988 }
 989
 990 static const struct rpc_call_ops nfs41_call_sync_ops = {
 991         .rpc_call_prepare = nfs41_call_sync_prepare,
 992         .rpc_call_done = nfs41_call_sync_done,
 993 };
 994
 995 #else   /* !CONFIG_NFS_V4_1 */
 996
 997 static int nfs4_setup_sequence(const struct nfs_server *server,
 998                                struct nfs4_sequence_args *args,
 999                                struct nfs4_sequence_res *res,
1000                                struct rpc_task *task)
1001 {
1002         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
1003                                     args, res, task);
1004 }
1005
1006 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1007 {
1008         return nfs40_sequence_done(task, res);
1009 }
1010
1011 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1012 {
1013         if (res->sr_slot != NULL)
1014                 nfs40_sequence_free_slot(res);
1015 }
1016
1017 int nfs4_sequence_done(struct rpc_task *task,
1018                        struct nfs4_sequence_res *res)
1019 {
1020         return nfs40_sequence_done(task, res);
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1023
1024 #endif  /* !CONFIG_NFS_V4_1 */
1025
1026 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1027 {
1028         struct nfs4_call_sync_data *data = calldata;
1029         nfs4_setup_sequence(data->seq_server,
1030                                 data->seq_args, data->seq_res, task);
1031 }
1032
1033 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1034 {
1035         struct nfs4_call_sync_data *data = calldata;
1036         nfs4_sequence_done(task, data->seq_res);
1037 }
1038
1039 static const struct rpc_call_ops nfs40_call_sync_ops = {
1040         .rpc_call_prepare = nfs40_call_sync_prepare,
1041         .rpc_call_done = nfs40_call_sync_done,
1042 };
1043
1044 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1045                                    struct nfs_server *server,
1046                                    struct rpc_message *msg,
1047                                    struct nfs4_sequence_args *args,
1048                                    struct nfs4_sequence_res *res)
1049 {
1050         int ret;
1051         struct rpc_task *task;
1052         struct nfs_client *clp = server->nfs_client;
1053         struct nfs4_call_sync_data data = {
1054                 .seq_server = server,
1055                 .seq_args = args,
1056                 .seq_res = res,
1057         };
1058         struct rpc_task_setup task_setup = {
1059                 .rpc_client = clnt,
1060                 .rpc_message = msg,
1061                 .callback_ops = clp->cl_mvops->call_sync_ops,
1062                 .callback_data = &data
1063         };
1064
1065         task = rpc_run_task(&task_setup);
1066         if (IS_ERR(task))
1067                 ret = PTR_ERR(task);
1068         else {
1069                 ret = task->tk_status;
1070                 rpc_put_task(task);
1071         }
1072         return ret;
1073 }
1074
1075 int nfs4_call_sync(struct rpc_clnt *clnt,
1076                    struct nfs_server *server,
1077                    struct rpc_message *msg,
1078                    struct nfs4_sequence_args *args,
1079                    struct nfs4_sequence_res *res,
1080                    int cache_reply)
1081 {
1082         nfs4_init_sequence(args, res, cache_reply);
1083         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1084 }
1085
1086 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
1087 {
1088         struct nfs_inode *nfsi = NFS_I(dir);
1089
1090         spin_lock(&dir->i_lock);
1091         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1092         if (!cinfo->atomic || cinfo->before != dir->i_version)
1093                 nfs_force_lookup_revalidate(dir);
1094         dir->i_version = cinfo->after;
1095         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1096         nfs_fscache_invalidate(dir);
1097         spin_unlock(&dir->i_lock);
1098 }
1099
1100 struct nfs4_opendata {
1101         struct kref kref;
1102         struct nfs_openargs o_arg;
1103         struct nfs_openres o_res;
1104         struct nfs_open_confirmargs c_arg;
1105         struct nfs_open_confirmres c_res;
1106         struct nfs4_string owner_name;
1107         struct nfs4_string group_name;
1108         struct nfs4_label *a_label;
1109         struct nfs_fattr f_attr;
1110         struct nfs4_label *f_label;
1111         struct dentry *dir;
1112         struct dentry *dentry;
1113         struct nfs4_state_owner *owner;
1114         struct nfs4_state *state;
1115         struct iattr attrs;
1116         unsigned long timestamp;
1117         unsigned int rpc_done : 1;
1118         unsigned int file_created : 1;
1119         unsigned int is_recover : 1;
1120         int rpc_status;
1121         int cancelled;
1122 };
1123
1124 struct nfs4_open_createattrs {
1125         struct nfs4_label *label;
1126         struct iattr *sattr;
1127         const __u32 verf[2];
1128 };
1129
1130 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1131                 int err, struct nfs4_exception *exception)
1132 {
1133         if (err != -EINVAL)
1134                 return false;
1135         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1136                 return false;
1137         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1138         exception->retry = 1;
1139         return true;
1140 }
1141
1142 static u32
1143 nfs4_map_atomic_open_share(struct nfs_server *server,
1144                 fmode_t fmode, int openflags)
1145 {
1146         u32 res = 0;
1147
1148         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1149         case FMODE_READ:
1150                 res = NFS4_SHARE_ACCESS_READ;
1151                 break;
1152         case FMODE_WRITE:
1153                 res = NFS4_SHARE_ACCESS_WRITE;
1154                 break;
1155         case FMODE_READ|FMODE_WRITE:
1156                 res = NFS4_SHARE_ACCESS_BOTH;
1157         }
1158         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1159                 goto out;
1160         /* Want no delegation if we're using O_DIRECT */
1161         if (openflags & O_DIRECT)
1162                 res |= NFS4_SHARE_WANT_NO_DELEG;
1163 out:
1164         return res;
1165 }
1166
1167 static enum open_claim_type4
1168 nfs4_map_atomic_open_claim(struct nfs_server *server,
1169                 enum open_claim_type4 claim)
1170 {
1171         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1172                 return claim;
1173         switch (claim) {
1174         default:
1175                 return claim;
1176         case NFS4_OPEN_CLAIM_FH:
1177                 return NFS4_OPEN_CLAIM_NULL;
1178         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1179                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1180         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1181                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1182         }
1183 }
1184
1185 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1186 {
1187         p->o_res.f_attr = &p->f_attr;
1188         p->o_res.f_label = p->f_label;
1189         p->o_res.seqid = p->o_arg.seqid;
1190         p->c_res.seqid = p->c_arg.seqid;
1191         p->o_res.server = p->o_arg.server;
1192         p->o_res.access_request = p->o_arg.access;
1193         nfs_fattr_init(&p->f_attr);
1194         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1195 }
1196
1197 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1198                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1199                 const struct nfs4_open_createattrs *c,
1200                 enum open_claim_type4 claim,
1201                 gfp_t gfp_mask)
1202 {
1203         struct dentry *parent = dget_parent(dentry);
1204         struct inode *dir = d_inode(parent);
1205         struct nfs_server *server = NFS_SERVER(dir);
1206         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1207         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1208         struct nfs4_opendata *p;
1209
1210         p = kzalloc(sizeof(*p), gfp_mask);
1211         if (p == NULL)
1212                 goto err;
1213
1214         p->f_label = nfs4_label_alloc(server, gfp_mask);
1215         if (IS_ERR(p->f_label))
1216                 goto err_free_p;
1217
1218         p->a_label = nfs4_label_alloc(server, gfp_mask);
1219         if (IS_ERR(p->a_label))
1220                 goto err_free_f;
1221
1222         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1223         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1224         if (IS_ERR(p->o_arg.seqid))
1225                 goto err_free_label;
1226         nfs_sb_active(dentry->d_sb);
1227         p->dentry = dget(dentry);
1228         p->dir = parent;
1229         p->owner = sp;
1230         atomic_inc(&sp->so_count);
1231         p->o_arg.open_flags = flags;
1232         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1233         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1234                         fmode, flags);
1235         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1236          * will return permission denied for all bits until close */
1237         if (!(flags & O_EXCL)) {
1238                 /* ask server to check for all possible rights as results
1239                  * are cached */
1240                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1241                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1242         }
1243         p->o_arg.clientid = server->nfs_client->cl_clientid;
1244         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1245         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1246         p->o_arg.name = &dentry->d_name;
1247         p->o_arg.server = server;
1248         p->o_arg.bitmask = nfs4_bitmask(server, label);
1249         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1250         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1251         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1252         switch (p->o_arg.claim) {
1253         case NFS4_OPEN_CLAIM_NULL:
1254         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1255         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1256                 p->o_arg.fh = NFS_FH(dir);
1257                 break;
1258         case NFS4_OPEN_CLAIM_PREVIOUS:
1259         case NFS4_OPEN_CLAIM_FH:
1260         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1261         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1262                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1263         }
1264         if (c != NULL && c->sattr != NULL && c->sattr->ia_valid != 0) {
1265                 p->o_arg.u.attrs = &p->attrs;
1266                 memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1267
1268                 memcpy(p->o_arg.u.verifier.data, c->verf,
1269                                 sizeof(p->o_arg.u.verifier.data));
1270         }
1271         p->c_arg.fh = &p->o_res.fh;
1272         p->c_arg.stateid = &p->o_res.stateid;
1273         p->c_arg.seqid = p->o_arg.seqid;
1274         nfs4_init_opendata_res(p);
1275         kref_init(&p->kref);
1276         return p;
1277
1278 err_free_label:
1279         nfs4_label_free(p->a_label);
1280 err_free_f:
1281         nfs4_label_free(p->f_label);
1282 err_free_p:
1283         kfree(p);
1284 err:
1285         dput(parent);
1286         return NULL;
1287 }
1288
1289 static void nfs4_opendata_free(struct kref *kref)
1290 {
1291         struct nfs4_opendata *p = container_of(kref,
1292                         struct nfs4_opendata, kref);
1293         struct super_block *sb = p->dentry->d_sb;
1294
1295         nfs_free_seqid(p->o_arg.seqid);
1296         nfs4_sequence_free_slot(&p->o_res.seq_res);
1297         if (p->state != NULL)
1298                 nfs4_put_open_state(p->state);
1299         nfs4_put_state_owner(p->owner);
1300
1301         nfs4_label_free(p->a_label);
1302         nfs4_label_free(p->f_label);
1303
1304         dput(p->dir);
1305         dput(p->dentry);
1306         nfs_sb_deactive(sb);
1307         nfs_fattr_free_names(&p->f_attr);
1308         kfree(p->f_attr.mdsthreshold);
1309         kfree(p);
1310 }
1311
1312 static void nfs4_opendata_put(struct nfs4_opendata *p)
1313 {
1314         if (p != NULL)
1315                 kref_put(&p->kref, nfs4_opendata_free);
1316 }
1317
1318 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1319 {
1320         int ret;
1321
1322         ret = rpc_wait_for_completion_task(task);
1323         return ret;
1324 }
1325
1326 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1327                 fmode_t fmode)
1328 {
1329         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1330         case FMODE_READ|FMODE_WRITE:
1331                 return state->n_rdwr != 0;
1332         case FMODE_WRITE:
1333                 return state->n_wronly != 0;
1334         case FMODE_READ:
1335                 return state->n_rdonly != 0;
1336         }
1337         WARN_ON_ONCE(1);
1338         return false;
1339 }
1340
1341 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1342 {
1343         int ret = 0;
1344
1345         if (open_mode & (O_EXCL|O_TRUNC))
1346                 goto out;
1347         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1348                 case FMODE_READ:
1349                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1350                                 && state->n_rdonly != 0;
1351                         break;
1352                 case FMODE_WRITE:
1353                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1354                                 && state->n_wronly != 0;
1355                         break;
1356                 case FMODE_READ|FMODE_WRITE:
1357                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1358                                 && state->n_rdwr != 0;
1359         }
1360 out:
1361         return ret;
1362 }
1363
1364 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1365                 enum open_claim_type4 claim)
1366 {
1367         if (delegation == NULL)
1368                 return 0;
1369         if ((delegation->type & fmode) != fmode)
1370                 return 0;
1371         switch (claim) {
1372         case NFS4_OPEN_CLAIM_NULL:
1373         case NFS4_OPEN_CLAIM_FH:
1374                 break;
1375         case NFS4_OPEN_CLAIM_PREVIOUS:
1376                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1377                         break;
1378         default:
1379                 return 0;
1380         }
1381         nfs_mark_delegation_referenced(delegation);
1382         return 1;
1383 }
1384
1385 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1386 {
1387         switch (fmode) {
1388                 case FMODE_WRITE:
1389                         state->n_wronly++;
1390                         break;
1391                 case FMODE_READ:
1392                         state->n_rdonly++;
1393                         break;
1394                 case FMODE_READ|FMODE_WRITE:
1395                         state->n_rdwr++;
1396         }
1397         nfs4_state_set_mode_locked(state, state->state | fmode);
1398 }
1399
1400 #ifdef CONFIG_NFS_V4_1
1401 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1402 {
1403         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1404                 return true;
1405         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1406                 return true;
1407         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1408                 return true;
1409         return false;
1410 }
1411 #endif /* CONFIG_NFS_V4_1 */
1412
1413 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1414 {
1415         struct nfs_client *clp = state->owner->so_server->nfs_client;
1416         bool need_recover = false;
1417
1418         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1419                 need_recover = true;
1420         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1421                 need_recover = true;
1422         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1423                 need_recover = true;
1424         if (need_recover)
1425                 nfs4_state_mark_reclaim_nograce(clp, state);
1426 }
1427
1428 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1429                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1430 {
1431         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1432                 return true;
1433         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1434                 nfs4_stateid_copy(freeme, &state->open_stateid);
1435                 nfs_test_and_clear_all_open_stateid(state);
1436                 return true;
1437         }
1438         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1439                 return true;
1440         return false;
1441 }
1442
1443 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1444 {
1445         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1446                 return;
1447         if (state->n_wronly)
1448                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1449         if (state->n_rdonly)
1450                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1451         if (state->n_rdwr)
1452                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1453         set_bit(NFS_OPEN_STATE, &state->flags);
1454 }
1455
1456 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1457                 nfs4_stateid *stateid, fmode_t fmode)
1458 {
1459         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1460         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1461         case FMODE_WRITE:
1462                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1463                 break;
1464         case FMODE_READ:
1465                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1466                 break;
1467         case 0:
1468                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1469                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1470                 clear_bit(NFS_OPEN_STATE, &state->flags);
1471         }
1472         if (stateid == NULL)
1473                 return;
1474         /* Handle OPEN+OPEN_DOWNGRADE races */
1475         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1476             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1477                 nfs_resync_open_stateid_locked(state);
1478                 return;
1479         }
1480         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1481                 nfs4_stateid_copy(&state->stateid, stateid);
1482         nfs4_stateid_copy(&state->open_stateid, stateid);
1483 }
1484
1485 static void nfs_clear_open_stateid(struct nfs4_state *state,
1486         nfs4_stateid *arg_stateid,
1487         nfs4_stateid *stateid, fmode_t fmode)
1488 {
1489         write_seqlock(&state->seqlock);
1490         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1491         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1492                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1493         write_sequnlock(&state->seqlock);
1494         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1495                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1496 }
1497
1498 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1499                 const nfs4_stateid *stateid, fmode_t fmode,
1500                 nfs4_stateid *freeme)
1501 {
1502         switch (fmode) {
1503                 case FMODE_READ:
1504                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1505                         break;
1506                 case FMODE_WRITE:
1507                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1508                         break;
1509                 case FMODE_READ|FMODE_WRITE:
1510                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1511         }
1512         if (!nfs_need_update_open_stateid(state, stateid, freeme))
1513                 return;
1514         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1515                 nfs4_stateid_copy(&state->stateid, stateid);
1516         nfs4_stateid_copy(&state->open_stateid, stateid);
1517 }
1518
1519 static void __update_open_stateid(struct nfs4_state *state,
1520                 const nfs4_stateid *open_stateid,
1521                 const nfs4_stateid *deleg_stateid,
1522                 fmode_t fmode,
1523                 nfs4_stateid *freeme)
1524 {
1525         /*
1526          * Protect the call to nfs4_state_set_mode_locked and
1527          * serialise the stateid update
1528          */
1529         spin_lock(&state->owner->so_lock);
1530         write_seqlock(&state->seqlock);
1531         if (deleg_stateid != NULL) {
1532                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1533                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1534         }
1535         if (open_stateid != NULL)
1536                 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1537         write_sequnlock(&state->seqlock);
1538         update_open_stateflags(state, fmode);
1539         spin_unlock(&state->owner->so_lock);
1540 }
1541
1542 static int update_open_stateid(struct nfs4_state *state,
1543                 const nfs4_stateid *open_stateid,
1544                 const nfs4_stateid *delegation,
1545                 fmode_t fmode)
1546 {
1547         struct nfs_server *server = NFS_SERVER(state->inode);
1548         struct nfs_client *clp = server->nfs_client;
1549         struct nfs_inode *nfsi = NFS_I(state->inode);
1550         struct nfs_delegation *deleg_cur;
1551         nfs4_stateid freeme = { };
1552         int ret = 0;
1553
1554         fmode &= (FMODE_READ|FMODE_WRITE);
1555
1556         rcu_read_lock();
1557         deleg_cur = rcu_dereference(nfsi->delegation);
1558         if (deleg_cur == NULL)
1559                 goto no_delegation;
1560
1561         spin_lock(&deleg_cur->lock);
1562         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1563            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1564             (deleg_cur->type & fmode) != fmode)
1565                 goto no_delegation_unlock;
1566
1567         if (delegation == NULL)
1568                 delegation = &deleg_cur->stateid;
1569         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1570                 goto no_delegation_unlock;
1571
1572         nfs_mark_delegation_referenced(deleg_cur);
1573         __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1574                         fmode, &freeme);
1575         ret = 1;
1576 no_delegation_unlock:
1577         spin_unlock(&deleg_cur->lock);
1578 no_delegation:
1579         rcu_read_unlock();
1580
1581         if (!ret && open_stateid != NULL) {
1582                 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1583                 ret = 1;
1584         }
1585         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1586                 nfs4_schedule_state_manager(clp);
1587         if (freeme.type != 0)
1588                 nfs4_test_and_free_stateid(server, &freeme,
1589                                 state->owner->so_cred);
1590
1591         return ret;
1592 }
1593
1594 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1595                 const nfs4_stateid *stateid)
1596 {
1597         struct nfs4_state *state = lsp->ls_state;
1598         bool ret = false;
1599
1600         spin_lock(&state->state_lock);
1601         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1602                 goto out_noupdate;
1603         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1604                 goto out_noupdate;
1605         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1606         ret = true;
1607 out_noupdate:
1608         spin_unlock(&state->state_lock);
1609         return ret;
1610 }
1611
1612 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1613 {
1614         struct nfs_delegation *delegation;
1615
1616         rcu_read_lock();
1617         delegation = rcu_dereference(NFS_I(inode)->delegation);
1618         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1619                 rcu_read_unlock();
1620                 return;
1621         }
1622         rcu_read_unlock();
1623         nfs4_inode_return_delegation(inode);
1624 }
1625
1626 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1627 {
1628         struct nfs4_state *state = opendata->state;
1629         struct nfs_delegation *delegation;
1630         int open_mode = opendata->o_arg.open_flags;
1631         fmode_t fmode = opendata->o_arg.fmode;
1632         enum open_claim_type4 claim = opendata->o_arg.claim;
1633         nfs4_stateid stateid;
1634         int ret = -EAGAIN;
1635
1636         for (;;) {
1637                 spin_lock(&state->owner->so_lock);
1638                 if (can_open_cached(state, fmode, open_mode)) {
1639                         update_open_stateflags(state, fmode);
1640                         spin_unlock(&state->owner->so_lock);
1641                         goto out_return_state;
1642                 }
1643                 spin_unlock(&state->owner->so_lock);
1644                 rcu_read_lock();
1645                 delegation = nfs4_get_valid_delegation(state->inode);
1646                 if (!can_open_delegated(delegation, fmode, claim)) {
1647                         rcu_read_unlock();
1648                         break;
1649                 }
1650                 /* Save the delegation */
1651                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1652                 rcu_read_unlock();
1653                 nfs_release_seqid(opendata->o_arg.seqid);
1654                 if (!opendata->is_recover) {
1655                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1656                         if (ret != 0)
1657                                 goto out;
1658                 }
1659                 ret = -EAGAIN;
1660
1661                 /* Try to update the stateid using the delegation */
1662                 if (update_open_stateid(state, NULL, &stateid, fmode))
1663                         goto out_return_state;
1664         }
1665 out:
1666         return ERR_PTR(ret);
1667 out_return_state:
1668         atomic_inc(&state->count);
1669         return state;
1670 }
1671
1672 static void
1673 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1674 {
1675         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1676         struct nfs_delegation *delegation;
1677         int delegation_flags = 0;
1678
1679         rcu_read_lock();
1680         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1681         if (delegation)
1682                 delegation_flags = delegation->flags;
1683         rcu_read_unlock();
1684         switch (data->o_arg.claim) {
1685         default:
1686                 break;
1687         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1688         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1689                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1690                                    "returning a delegation for "
1691                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1692                                    clp->cl_hostname);
1693                 return;
1694         }
1695         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1696                 nfs_inode_set_delegation(state->inode,
1697                                          data->owner->so_cred,
1698                                          &data->o_res);
1699         else
1700                 nfs_inode_reclaim_delegation(state->inode,
1701                                              data->owner->so_cred,
1702                                              &data->o_res);
1703 }
1704
1705 /*
1706  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1707  * and update the nfs4_state.
1708  */
1709 static struct nfs4_state *
1710 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1711 {
1712         struct inode *inode = data->state->inode;
1713         struct nfs4_state *state = data->state;
1714         int ret;
1715
1716         if (!data->rpc_done) {
1717                 if (data->rpc_status) {
1718                         ret = data->rpc_status;
1719                         goto err;
1720                 }
1721                 /* cached opens have already been processed */
1722                 goto update;
1723         }
1724
1725         ret = nfs_refresh_inode(inode, &data->f_attr);
1726         if (ret)
1727                 goto err;
1728
1729         if (data->o_res.delegation_type != 0)
1730                 nfs4_opendata_check_deleg(data, state);
1731 update:
1732         update_open_stateid(state, &data->o_res.stateid, NULL,
1733                             data->o_arg.fmode);
1734         atomic_inc(&state->count);
1735
1736         return state;
1737 err:
1738         return ERR_PTR(ret);
1739
1740 }
1741
1742 static struct nfs4_state *
1743 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1744 {
1745         struct inode *inode;
1746         struct nfs4_state *state = NULL;
1747         int ret;
1748
1749         if (!data->rpc_done) {
1750                 state = nfs4_try_open_cached(data);
1751                 trace_nfs4_cached_open(data->state);
1752                 goto out;
1753         }
1754
1755         ret = -EAGAIN;
1756         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1757                 goto err;
1758         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1759         ret = PTR_ERR(inode);
1760         if (IS_ERR(inode))
1761                 goto err;
1762         ret = -ENOMEM;
1763         state = nfs4_get_open_state(inode, data->owner);
1764         if (state == NULL)
1765                 goto err_put_inode;
1766         if (data->o_res.delegation_type != 0)
1767                 nfs4_opendata_check_deleg(data, state);
1768         update_open_stateid(state, &data->o_res.stateid, NULL,
1769                         data->o_arg.fmode);
1770         iput(inode);
1771 out:
1772         nfs_release_seqid(data->o_arg.seqid);
1773         return state;
1774 err_put_inode:
1775         iput(inode);
1776 err:
1777         return ERR_PTR(ret);
1778 }
1779
1780 static struct nfs4_state *
1781 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1782 {
1783         struct nfs4_state *ret;
1784
1785         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1786                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1787         else
1788                 ret = _nfs4_opendata_to_nfs4_state(data);
1789         nfs4_sequence_free_slot(&data->o_res.seq_res);
1790         return ret;
1791 }
1792
1793 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1794 {
1795         struct nfs_inode *nfsi = NFS_I(state->inode);
1796         struct nfs_open_context *ctx;
1797
1798         spin_lock(&state->inode->i_lock);
1799         list_for_each_entry(ctx, &nfsi->open_files, list) {
1800                 if (ctx->state != state)
1801                         continue;
1802                 get_nfs_open_context(ctx);
1803                 spin_unlock(&state->inode->i_lock);
1804                 return ctx;
1805         }
1806         spin_unlock(&state->inode->i_lock);
1807         return ERR_PTR(-ENOENT);
1808 }
1809
1810 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1811                 struct nfs4_state *state, enum open_claim_type4 claim)
1812 {
1813         struct nfs4_opendata *opendata;
1814
1815         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1816                         NULL, claim, GFP_NOFS);
1817         if (opendata == NULL)
1818                 return ERR_PTR(-ENOMEM);
1819         opendata->state = state;
1820         atomic_inc(&state->count);
1821         return opendata;
1822 }
1823
1824 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1825                 fmode_t fmode)
1826 {
1827         struct nfs4_state *newstate;
1828         int ret;
1829
1830         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1831                 return 0;
1832         opendata->o_arg.open_flags = 0;
1833         opendata->o_arg.fmode = fmode;
1834         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1835                         NFS_SB(opendata->dentry->d_sb),
1836                         fmode, 0);
1837         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1838         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1839         nfs4_init_opendata_res(opendata);
1840         ret = _nfs4_recover_proc_open(opendata);
1841         if (ret != 0)
1842                 return ret;
1843         newstate = nfs4_opendata_to_nfs4_state(opendata);
1844         if (IS_ERR(newstate))
1845                 return PTR_ERR(newstate);
1846         if (newstate != opendata->state)
1847                 ret = -ESTALE;
1848         nfs4_close_state(newstate, fmode);
1849         return ret;
1850 }
1851
1852 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1853 {
1854         int ret;
1855
1856         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1857         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1858         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1859         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1860         /* memory barrier prior to reading state->n_* */
1861         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1862         clear_bit(NFS_OPEN_STATE, &state->flags);
1863         smp_rmb();
1864         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1865         if (ret != 0)
1866                 return ret;
1867         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1868         if (ret != 0)
1869                 return ret;
1870         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1871         if (ret != 0)
1872                 return ret;
1873         /*
1874          * We may have performed cached opens for all three recoveries.
1875          * Check if we need to update the current stateid.
1876          */
1877         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1878             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1879                 write_seqlock(&state->seqlock);
1880                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1881                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1882                 write_sequnlock(&state->seqlock);
1883         }
1884         return 0;
1885 }
1886
1887 /*
1888  * OPEN_RECLAIM:
1889  *      reclaim state on the server after a reboot.
1890  */
1891 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1892 {
1893         struct nfs_delegation *delegation;
1894         struct nfs4_opendata *opendata;
1895         fmode_t delegation_type = 0;
1896         int status;
1897
1898         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1899                         NFS4_OPEN_CLAIM_PREVIOUS);
1900         if (IS_ERR(opendata))
1901                 return PTR_ERR(opendata);
1902         rcu_read_lock();
1903         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1904         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1905                 delegation_type = delegation->type;
1906         rcu_read_unlock();
1907         opendata->o_arg.u.delegation_type = delegation_type;
1908         status = nfs4_open_recover(opendata, state);
1909         nfs4_opendata_put(opendata);
1910         return status;
1911 }
1912
1913 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1914 {
1915         struct nfs_server *server = NFS_SERVER(state->inode);
1916         struct nfs4_exception exception = { };
1917         int err;
1918         do {
1919                 err = _nfs4_do_open_reclaim(ctx, state);
1920                 trace_nfs4_open_reclaim(ctx, 0, err);
1921                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1922                         continue;
1923                 if (err != -NFS4ERR_DELAY)
1924                         break;
1925                 nfs4_handle_exception(server, err, &exception);
1926         } while (exception.retry);
1927         return err;
1928 }
1929
1930 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1931 {
1932         struct nfs_open_context *ctx;
1933         int ret;
1934
1935         ctx = nfs4_state_find_open_context(state);
1936         if (IS_ERR(ctx))
1937                 return -EAGAIN;
1938         ret = nfs4_do_open_reclaim(ctx, state);
1939         put_nfs_open_context(ctx);
1940         return ret;
1941 }
1942
1943 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
1944 {
1945         switch (err) {
1946                 default:
1947                         printk(KERN_ERR "NFS: %s: unhandled error "
1948                                         "%d.\n", __func__, err);
1949                 case 0:
1950                 case -ENOENT:
1951                 case -EAGAIN:
1952                 case -ESTALE:
1953                         break;
1954                 case -NFS4ERR_BADSESSION:
1955                 case -NFS4ERR_BADSLOT:
1956                 case -NFS4ERR_BAD_HIGH_SLOT:
1957                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1958                 case -NFS4ERR_DEADSESSION:
1959                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1960                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1961                         return -EAGAIN;
1962                 case -NFS4ERR_STALE_CLIENTID:
1963                 case -NFS4ERR_STALE_STATEID:
1964                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1965                         /* Don't recall a delegation if it was lost */
1966                         nfs4_schedule_lease_recovery(server->nfs_client);
1967                         return -EAGAIN;
1968                 case -NFS4ERR_MOVED:
1969                         nfs4_schedule_migration_recovery(server);
1970                         return -EAGAIN;
1971                 case -NFS4ERR_LEASE_MOVED:
1972                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1973                         return -EAGAIN;
1974                 case -NFS4ERR_DELEG_REVOKED:
1975                 case -NFS4ERR_ADMIN_REVOKED:
1976                 case -NFS4ERR_EXPIRED:
1977                 case -NFS4ERR_BAD_STATEID:
1978                 case -NFS4ERR_OPENMODE:
1979                         nfs_inode_find_state_and_recover(state->inode,
1980                                         stateid);
1981                         nfs4_schedule_stateid_recovery(server, state);
1982                         return -EAGAIN;
1983                 case -NFS4ERR_DELAY:
1984                 case -NFS4ERR_GRACE:
1985                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1986                         ssleep(1);
1987                         return -EAGAIN;
1988                 case -ENOMEM:
1989                 case -NFS4ERR_DENIED:
1990                         if (fl) {
1991                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
1992                                 if (lsp)
1993                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
1994                         }
1995                         return 0;
1996         }
1997         return err;
1998 }
1999
2000 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2001                 struct nfs4_state *state, const nfs4_stateid *stateid,
2002                 fmode_t type)
2003 {
2004         struct nfs_server *server = NFS_SERVER(state->inode);
2005         struct nfs4_opendata *opendata;
2006         int err = 0;
2007
2008         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2009                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2010         if (IS_ERR(opendata))
2011                 return PTR_ERR(opendata);
2012         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2013         write_seqlock(&state->seqlock);
2014         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2015         write_sequnlock(&state->seqlock);
2016         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2017         switch (type & (FMODE_READ|FMODE_WRITE)) {
2018         case FMODE_READ|FMODE_WRITE:
2019         case FMODE_WRITE:
2020                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2021                 if (err)
2022                         break;
2023                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2024                 if (err)
2025                         break;
2026         case FMODE_READ:
2027                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2028         }
2029         nfs4_opendata_put(opendata);
2030         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2031 }
2032
2033 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2034 {
2035         struct nfs4_opendata *data = calldata;
2036
2037         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
2038                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
2039 }
2040
2041 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2042 {
2043         struct nfs4_opendata *data = calldata;
2044
2045         nfs40_sequence_done(task, &data->c_res.seq_res);
2046
2047         data->rpc_status = task->tk_status;
2048         if (data->rpc_status == 0) {
2049                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2050                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2051                 renew_lease(data->o_res.server, data->timestamp);
2052                 data->rpc_done = 1;
2053         }
2054 }
2055
2056 static void nfs4_open_confirm_release(void *calldata)
2057 {
2058         struct nfs4_opendata *data = calldata;
2059         struct nfs4_state *state = NULL;
2060
2061         /* If this request hasn't been cancelled, do nothing */
2062         if (data->cancelled == 0)
2063                 goto out_free;
2064         /* In case of error, no cleanup! */
2065         if (!data->rpc_done)
2066                 goto out_free;
2067         state = nfs4_opendata_to_nfs4_state(data);
2068         if (!IS_ERR(state))
2069                 nfs4_close_state(state, data->o_arg.fmode);
2070 out_free:
2071         nfs4_opendata_put(data);
2072 }
2073
2074 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2075         .rpc_call_prepare = nfs4_open_confirm_prepare,
2076         .rpc_call_done = nfs4_open_confirm_done,
2077         .rpc_release = nfs4_open_confirm_release,
2078 };
2079
2080 /*
2081  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2082  */
2083 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2084 {
2085         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2086         struct rpc_task *task;
2087         struct  rpc_message msg = {
2088                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2089                 .rpc_argp = &data->c_arg,
2090                 .rpc_resp = &data->c_res,
2091                 .rpc_cred = data->owner->so_cred,
2092         };
2093         struct rpc_task_setup task_setup_data = {
2094                 .rpc_client = server->client,
2095                 .rpc_message = &msg,
2096                 .callback_ops = &nfs4_open_confirm_ops,
2097                 .callback_data = data,
2098                 .workqueue = nfsiod_workqueue,
2099                 .flags = RPC_TASK_ASYNC,
2100         };
2101         int status;
2102
2103         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2104         kref_get(&data->kref);
2105         data->rpc_done = 0;
2106         data->rpc_status = 0;
2107         data->timestamp = jiffies;
2108         if (data->is_recover)
2109                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2110         task = rpc_run_task(&task_setup_data);
2111         if (IS_ERR(task))
2112                 return PTR_ERR(task);
2113         status = nfs4_wait_for_completion_rpc_task(task);
2114         if (status != 0) {
2115                 data->cancelled = 1;
2116                 smp_wmb();
2117         } else
2118                 status = data->rpc_status;
2119         rpc_put_task(task);
2120         return status;
2121 }
2122
2123 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2124 {
2125         struct nfs4_opendata *data = calldata;
2126         struct nfs4_state_owner *sp = data->owner;
2127         struct nfs_client *clp = sp->so_server->nfs_client;
2128         enum open_claim_type4 claim = data->o_arg.claim;
2129
2130         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2131                 goto out_wait;
2132         /*
2133          * Check if we still need to send an OPEN call, or if we can use
2134          * a delegation instead.
2135          */
2136         if (data->state != NULL) {
2137                 struct nfs_delegation *delegation;
2138
2139                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2140                         goto out_no_action;
2141                 rcu_read_lock();
2142                 delegation = nfs4_get_valid_delegation(data->state->inode);
2143                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2144                         goto unlock_no_action;
2145                 rcu_read_unlock();
2146         }
2147         /* Update client id. */
2148         data->o_arg.clientid = clp->cl_clientid;
2149         switch (claim) {
2150         default:
2151                 break;
2152         case NFS4_OPEN_CLAIM_PREVIOUS:
2153         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2154         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2155                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2156         case NFS4_OPEN_CLAIM_FH:
2157                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2158                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2159         }
2160         data->timestamp = jiffies;
2161         if (nfs4_setup_sequence(data->o_arg.server,
2162                                 &data->o_arg.seq_args,
2163                                 &data->o_res.seq_res,
2164                                 task) != 0)
2165                 nfs_release_seqid(data->o_arg.seqid);
2166
2167         /* Set the create mode (note dependency on the session type) */
2168         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2169         if (data->o_arg.open_flags & O_EXCL) {
2170                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2171                 if (nfs4_has_persistent_session(clp))
2172                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2173                 else if (clp->cl_mvops->minor_version > 0)
2174                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2175         }
2176         return;
2177 unlock_no_action:
2178         trace_nfs4_cached_open(data->state);
2179         rcu_read_unlock();
2180 out_no_action:
2181         task->tk_action = NULL;
2182 out_wait:
2183         nfs4_sequence_done(task, &data->o_res.seq_res);
2184 }
2185
2186 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2187 {
2188         struct nfs4_opendata *data = calldata;
2189
2190         data->rpc_status = task->tk_status;
2191
2192         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2193                 return;
2194
2195         if (task->tk_status == 0) {
2196                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2197                         switch (data->o_res.f_attr->mode & S_IFMT) {
2198                         case S_IFREG:
2199                                 break;
2200                         case S_IFLNK:
2201                                 data->rpc_status = -ELOOP;
2202                                 break;
2203                         case S_IFDIR:
2204                                 data->rpc_status = -EISDIR;
2205                                 break;
2206                         default:
2207                                 data->rpc_status = -ENOTDIR;
2208                         }
2209                 }
2210                 renew_lease(data->o_res.server, data->timestamp);
2211                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2212                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2213         }
2214         data->rpc_done = 1;
2215 }
2216
2217 static void nfs4_open_release(void *calldata)
2218 {
2219         struct nfs4_opendata *data = calldata;
2220         struct nfs4_state *state = NULL;
2221
2222         /* If this request hasn't been cancelled, do nothing */
2223         if (data->cancelled == 0)
2224                 goto out_free;
2225         /* In case of error, no cleanup! */
2226         if (data->rpc_status != 0 || !data->rpc_done)
2227                 goto out_free;
2228         /* In case we need an open_confirm, no cleanup! */
2229         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2230                 goto out_free;
2231         state = nfs4_opendata_to_nfs4_state(data);
2232         if (!IS_ERR(state))
2233                 nfs4_close_state(state, data->o_arg.fmode);
2234 out_free:
2235         nfs4_opendata_put(data);
2236 }
2237
2238 static const struct rpc_call_ops nfs4_open_ops = {
2239         .rpc_call_prepare = nfs4_open_prepare,
2240         .rpc_call_done = nfs4_open_done,
2241         .rpc_release = nfs4_open_release,
2242 };
2243
2244 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2245 {
2246         struct inode *dir = d_inode(data->dir);
2247         struct nfs_server *server = NFS_SERVER(dir);
2248         struct nfs_openargs *o_arg = &data->o_arg;
2249         struct nfs_openres *o_res = &data->o_res;
2250         struct rpc_task *task;
2251         struct rpc_message msg = {
2252                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2253                 .rpc_argp = o_arg,
2254                 .rpc_resp = o_res,
2255                 .rpc_cred = data->owner->so_cred,
2256         };
2257         struct rpc_task_setup task_setup_data = {
2258                 .rpc_client = server->client,
2259                 .rpc_message = &msg,
2260                 .callback_ops = &nfs4_open_ops,
2261                 .callback_data = data,
2262                 .workqueue = nfsiod_workqueue,
2263                 .flags = RPC_TASK_ASYNC,
2264         };
2265         int status;
2266
2267         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2268         kref_get(&data->kref);
2269         data->rpc_done = 0;
2270         data->rpc_status = 0;
2271         data->cancelled = 0;
2272         data->is_recover = 0;
2273         if (isrecover) {
2274                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2275                 data->is_recover = 1;
2276         }
2277         task = rpc_run_task(&task_setup_data);
2278         if (IS_ERR(task))
2279                 return PTR_ERR(task);
2280         status = nfs4_wait_for_completion_rpc_task(task);
2281         if (status != 0) {
2282                 data->cancelled = 1;
2283                 smp_wmb();
2284         } else
2285                 status = data->rpc_status;
2286         rpc_put_task(task);
2287
2288         return status;
2289 }
2290
2291 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2292 {
2293         struct inode *dir = d_inode(data->dir);
2294         struct nfs_openres *o_res = &data->o_res;
2295         int status;
2296
2297         status = nfs4_run_open_task(data, 1);
2298         if (status != 0 || !data->rpc_done)
2299                 return status;
2300
2301         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2302
2303         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2304                 status = _nfs4_proc_open_confirm(data);
2305                 if (status != 0)
2306                         return status;
2307         }
2308
2309         return status;
2310 }
2311
2312 /*
2313  * Additional permission checks in order to distinguish between an
2314  * open for read, and an open for execute. This works around the
2315  * fact that NFSv4 OPEN treats read and execute permissions as being
2316  * the same.
2317  * Note that in the non-execute case, we want to turn off permission
2318  * checking if we just created a new file (POSIX open() semantics).
2319  */
2320 static int nfs4_opendata_access(struct rpc_cred *cred,
2321                                 struct nfs4_opendata *opendata,
2322                                 struct nfs4_state *state, fmode_t fmode,
2323                                 int openflags)
2324 {
2325         struct nfs_access_entry cache;
2326         u32 mask;
2327
2328         /* access call failed or for some reason the server doesn't
2329          * support any access modes -- defer access call until later */
2330         if (opendata->o_res.access_supported == 0)
2331                 return 0;
2332
2333         mask = 0;
2334         /*
2335          * Use openflags to check for exec, because fmode won't
2336          * always have FMODE_EXEC set when file open for exec.
2337          */
2338         if (openflags & __FMODE_EXEC) {
2339                 /* ONLY check for exec rights */
2340                 mask = MAY_EXEC;
2341         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2342                 mask = MAY_READ;
2343
2344         cache.cred = cred;
2345         cache.jiffies = jiffies;
2346         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2347         nfs_access_add_cache(state->inode, &cache);
2348
2349         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2350                 return 0;
2351
2352         return -EACCES;
2353 }
2354
2355 /*
2356  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2357  */
2358 static int _nfs4_proc_open(struct nfs4_opendata *data)
2359 {
2360         struct inode *dir = d_inode(data->dir);
2361         struct nfs_server *server = NFS_SERVER(dir);
2362         struct nfs_openargs *o_arg = &data->o_arg;
2363         struct nfs_openres *o_res = &data->o_res;
2364         int status;
2365
2366         status = nfs4_run_open_task(data, 0);
2367         if (!data->rpc_done)
2368                 return status;
2369         if (status != 0) {
2370                 if (status == -NFS4ERR_BADNAME &&
2371                                 !(o_arg->open_flags & O_CREAT))
2372                         return -ENOENT;
2373                 return status;
2374         }
2375
2376         nfs_fattr_map_and_free_names(server, &data->f_attr);
2377
2378         if (o_arg->open_flags & O_CREAT) {
2379                 update_changeattr(dir, &o_res->cinfo);
2380                 if (o_arg->open_flags & O_EXCL)
2381                         data->file_created = 1;
2382                 else if (o_res->cinfo.before != o_res->cinfo.after)
2383                         data->file_created = 1;
2384         }
2385         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2386                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2387         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2388                 status = _nfs4_proc_open_confirm(data);
2389                 if (status != 0)
2390                         return status;
2391         }
2392         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2393                 nfs4_sequence_free_slot(&o_res->seq_res);
2394                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2395         }
2396         return 0;
2397 }
2398
2399 static int nfs4_recover_expired_lease(struct nfs_server *server)
2400 {
2401         return nfs4_client_recover_expired_lease(server->nfs_client);
2402 }
2403
2404 /*
2405  * OPEN_EXPIRED:
2406  *      reclaim state on the server after a network partition.
2407  *      Assumes caller holds the appropriate lock
2408  */
2409 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2410 {
2411         struct nfs4_opendata *opendata;
2412         int ret;
2413
2414         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2415                         NFS4_OPEN_CLAIM_FH);
2416         if (IS_ERR(opendata))
2417                 return PTR_ERR(opendata);
2418         ret = nfs4_open_recover(opendata, state);
2419         if (ret == -ESTALE)
2420                 d_drop(ctx->dentry);
2421         nfs4_opendata_put(opendata);
2422         return ret;
2423 }
2424
2425 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2426 {
2427         struct nfs_server *server = NFS_SERVER(state->inode);
2428         struct nfs4_exception exception = { };
2429         int err;
2430
2431         do {
2432                 err = _nfs4_open_expired(ctx, state);
2433                 trace_nfs4_open_expired(ctx, 0, err);
2434                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2435                         continue;
2436                 switch (err) {
2437                 default:
2438                         goto out;
2439                 case -NFS4ERR_GRACE:
2440                 case -NFS4ERR_DELAY:
2441                         nfs4_handle_exception(server, err, &exception);
2442                         err = 0;
2443                 }
2444         } while (exception.retry);
2445 out:
2446         return err;
2447 }
2448
2449 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2450 {
2451         struct nfs_open_context *ctx;
2452         int ret;
2453
2454         ctx = nfs4_state_find_open_context(state);
2455         if (IS_ERR(ctx))
2456                 return -EAGAIN;
2457         ret = nfs4_do_open_expired(ctx, state);
2458         put_nfs_open_context(ctx);
2459         return ret;
2460 }
2461
2462 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2463                 const nfs4_stateid *stateid)
2464 {
2465         nfs_remove_bad_delegation(state->inode, stateid);
2466         write_seqlock(&state->seqlock);
2467         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2468         write_sequnlock(&state->seqlock);
2469         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2470 }
2471
2472 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2473 {
2474         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2475                 nfs_finish_clear_delegation_stateid(state, NULL);
2476 }
2477
2478 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2479 {
2480         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2481         nfs40_clear_delegation_stateid(state);
2482         return nfs4_open_expired(sp, state);
2483 }
2484
2485 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2486                 nfs4_stateid *stateid,
2487                 struct rpc_cred *cred)
2488 {
2489         return -NFS4ERR_BAD_STATEID;
2490 }
2491
2492 #if defined(CONFIG_NFS_V4_1)
2493 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2494                 nfs4_stateid *stateid,
2495                 struct rpc_cred *cred)
2496 {
2497         int status;
2498
2499         switch (stateid->type) {
2500         default:
2501                 break;
2502         case NFS4_INVALID_STATEID_TYPE:
2503         case NFS4_SPECIAL_STATEID_TYPE:
2504                 return -NFS4ERR_BAD_STATEID;
2505         case NFS4_REVOKED_STATEID_TYPE:
2506                 goto out_free;
2507         }
2508
2509         status = nfs41_test_stateid(server, stateid, cred);
2510         switch (status) {
2511         case -NFS4ERR_EXPIRED:
2512         case -NFS4ERR_ADMIN_REVOKED:
2513         case -NFS4ERR_DELEG_REVOKED:
2514                 break;
2515         default:
2516                 return status;
2517         }
2518 out_free:
2519         /* Ack the revoked state to the server */
2520         nfs41_free_stateid(server, stateid, cred, true);
2521         return -NFS4ERR_EXPIRED;
2522 }
2523
2524 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2525 {
2526         struct nfs_server *server = NFS_SERVER(state->inode);
2527         nfs4_stateid stateid;
2528         struct nfs_delegation *delegation;
2529         struct rpc_cred *cred;
2530         int status;
2531
2532         /* Get the delegation credential for use by test/free_stateid */
2533         rcu_read_lock();
2534         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2535         if (delegation == NULL) {
2536                 rcu_read_unlock();
2537                 return;
2538         }
2539
2540         nfs4_stateid_copy(&stateid, &delegation->stateid);
2541         if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags)) {
2542                 rcu_read_unlock();
2543                 nfs_finish_clear_delegation_stateid(state, &stateid);
2544                 return;
2545         }
2546
2547         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2548                                 &delegation->flags)) {
2549                 rcu_read_unlock();
2550                 return;
2551         }
2552
2553         cred = get_rpccred(delegation->cred);
2554         rcu_read_unlock();
2555         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2556         trace_nfs4_test_delegation_stateid(state, NULL, status);
2557         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2558                 nfs_finish_clear_delegation_stateid(state, &stateid);
2559
2560         put_rpccred(cred);
2561 }
2562
2563 /**
2564  * nfs41_check_expired_locks - possibly free a lock stateid
2565  *
2566  * @state: NFSv4 state for an inode
2567  *
2568  * Returns NFS_OK if recovery for this stateid is now finished.
2569  * Otherwise a negative NFS4ERR value is returned.
2570  */
2571 static int nfs41_check_expired_locks(struct nfs4_state *state)
2572 {
2573         int status, ret = NFS_OK;
2574         struct nfs4_lock_state *lsp, *prev = NULL;
2575         struct nfs_server *server = NFS_SERVER(state->inode);
2576
2577         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2578                 goto out;
2579
2580         spin_lock(&state->state_lock);
2581         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2582                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2583                         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2584
2585                         atomic_inc(&lsp->ls_count);
2586                         spin_unlock(&state->state_lock);
2587
2588                         nfs4_put_lock_state(prev);
2589                         prev = lsp;
2590
2591                         status = nfs41_test_and_free_expired_stateid(server,
2592                                         &lsp->ls_stateid,
2593                                         cred);
2594                         trace_nfs4_test_lock_stateid(state, lsp, status);
2595                         if (status == -NFS4ERR_EXPIRED ||
2596                             status == -NFS4ERR_BAD_STATEID) {
2597                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2598                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2599                                 if (!recover_lost_locks)
2600                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2601                         } else if (status != NFS_OK) {
2602                                 ret = status;
2603                                 nfs4_put_lock_state(prev);
2604                                 goto out;
2605                         }
2606                         spin_lock(&state->state_lock);
2607                 }
2608         }
2609         spin_unlock(&state->state_lock);
2610         nfs4_put_lock_state(prev);
2611 out:
2612         return ret;
2613 }
2614
2615 /**
2616  * nfs41_check_open_stateid - possibly free an open stateid
2617  *
2618  * @state: NFSv4 state for an inode
2619  *
2620  * Returns NFS_OK if recovery for this stateid is now finished.
2621  * Otherwise a negative NFS4ERR value is returned.
2622  */
2623 static int nfs41_check_open_stateid(struct nfs4_state *state)
2624 {
2625         struct nfs_server *server = NFS_SERVER(state->inode);
2626         nfs4_stateid *stateid = &state->open_stateid;
2627         struct rpc_cred *cred = state->owner->so_cred;
2628         int status;
2629
2630         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2631                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)  {
2632                         if (nfs4_have_delegation(state->inode, state->state))
2633                                 return NFS_OK;
2634                         return -NFS4ERR_OPENMODE;
2635                 }
2636                 return -NFS4ERR_BAD_STATEID;
2637         }
2638         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2639         trace_nfs4_test_open_stateid(state, NULL, status);
2640         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2641                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2642                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2643                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2644                 clear_bit(NFS_OPEN_STATE, &state->flags);
2645                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2646         }
2647         if (status != NFS_OK)
2648                 return status;
2649         if (nfs_open_stateid_recover_openmode(state))
2650                 return -NFS4ERR_OPENMODE;
2651         return NFS_OK;
2652 }
2653
2654 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2655 {
2656         int status;
2657
2658         nfs41_check_delegation_stateid(state);
2659         status = nfs41_check_expired_locks(state);
2660         if (status != NFS_OK)
2661                 return status;
2662         status = nfs41_check_open_stateid(state);
2663         if (status != NFS_OK)
2664                 status = nfs4_open_expired(sp, state);
2665         return status;
2666 }
2667 #endif
2668
2669 /*
2670  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2671  * fields corresponding to attributes that were used to store the verifier.
2672  * Make sure we clobber those fields in the later setattr call
2673  */
2674 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2675                                 struct iattr *sattr, struct nfs4_label **label)
2676 {
2677         const u32 *attrset = opendata->o_res.attrset;
2678
2679         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2680             !(sattr->ia_valid & ATTR_ATIME_SET))
2681                 sattr->ia_valid |= ATTR_ATIME;
2682
2683         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2684             !(sattr->ia_valid & ATTR_MTIME_SET))
2685                 sattr->ia_valid |= ATTR_MTIME;
2686
2687         /* Except MODE, it seems harmless of setting twice. */
2688         if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2689                 attrset[1] & FATTR4_WORD1_MODE)
2690                 sattr->ia_valid &= ~ATTR_MODE;
2691
2692         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2693                 *label = NULL;
2694 }
2695
2696 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2697                 fmode_t fmode,
2698                 int flags,
2699                 struct nfs_open_context *ctx)
2700 {
2701         struct nfs4_state_owner *sp = opendata->owner;
2702         struct nfs_server *server = sp->so_server;
2703         struct dentry *dentry;
2704         struct nfs4_state *state;
2705         unsigned int seq;
2706         int ret;
2707
2708         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2709
2710         ret = _nfs4_proc_open(opendata);
2711         if (ret != 0)
2712                 goto out;
2713
2714         state = _nfs4_opendata_to_nfs4_state(opendata);
2715         ret = PTR_ERR(state);
2716         if (IS_ERR(state))
2717                 goto out;
2718         ctx->state = state;
2719         if (server->caps & NFS_CAP_POSIX_LOCK)
2720                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2721         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2722                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2723
2724         dentry = opendata->dentry;
2725         if (d_really_is_negative(dentry)) {
2726                 struct dentry *alias;
2727                 d_drop(dentry);
2728                 alias = d_exact_alias(dentry, state->inode);
2729                 if (!alias)
2730                         alias = d_splice_alias(igrab(state->inode), dentry);
2731                 /* d_splice_alias() can't fail here - it's a non-directory */
2732                 if (alias) {
2733                         dput(ctx->dentry);
2734                         ctx->dentry = dentry = alias;
2735                 }
2736                 nfs_set_verifier(dentry,
2737                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2738         }
2739
2740         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2741         if (ret != 0)
2742                 goto out;
2743
2744         if (d_inode(dentry) == state->inode) {
2745                 nfs_inode_attach_open_context(ctx);
2746                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2747                         nfs4_schedule_stateid_recovery(server, state);
2748         }
2749 out:
2750         if (!opendata->cancelled)
2751                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
2752         return ret;
2753 }
2754
2755 /*
2756  * Returns a referenced nfs4_state
2757  */
2758 static int _nfs4_do_open(struct inode *dir,
2759                         struct nfs_open_context *ctx,
2760                         int flags,
2761                         const struct nfs4_open_createattrs *c,
2762                         int *opened)
2763 {
2764         struct nfs4_state_owner  *sp;
2765         struct nfs4_state     *state = NULL;
2766         struct nfs_server       *server = NFS_SERVER(dir);
2767         struct nfs4_opendata *opendata;
2768         struct dentry *dentry = ctx->dentry;
2769         struct rpc_cred *cred = ctx->cred;
2770         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2771         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2772         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2773         struct iattr *sattr = c->sattr;
2774         struct nfs4_label *label = c->label;
2775         struct nfs4_label *olabel = NULL;
2776         int status;
2777
2778         /* Protect against reboot recovery conflicts */
2779         status = -ENOMEM;
2780         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2781         if (sp == NULL) {
2782                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2783                 goto out_err;
2784         }
2785         status = nfs4_recover_expired_lease(server);
2786         if (status != 0)
2787                 goto err_put_state_owner;
2788         if (d_really_is_positive(dentry))
2789                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2790         status = -ENOMEM;
2791         if (d_really_is_positive(dentry))
2792                 claim = NFS4_OPEN_CLAIM_FH;
2793         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
2794                         c, claim, GFP_KERNEL);
2795         if (opendata == NULL)
2796                 goto err_put_state_owner;
2797
2798         if (label) {
2799                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2800                 if (IS_ERR(olabel)) {
2801                         status = PTR_ERR(olabel);
2802                         goto err_opendata_put;
2803                 }
2804         }
2805
2806         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2807                 if (!opendata->f_attr.mdsthreshold) {
2808                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2809                         if (!opendata->f_attr.mdsthreshold)
2810                                 goto err_free_label;
2811                 }
2812                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2813         }
2814         if (d_really_is_positive(dentry))
2815                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2816
2817         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2818         if (status != 0)
2819                 goto err_free_label;
2820         state = ctx->state;
2821
2822         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2823             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2824                 nfs4_exclusive_attrset(opendata, sattr, &label);
2825                 /*
2826                  * send create attributes which was not set by open
2827                  * with an extra setattr.
2828                  */
2829                 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2830                         nfs_fattr_init(opendata->o_res.f_attr);
2831                         status = nfs4_do_setattr(state->inode, cred,
2832                                         opendata->o_res.f_attr, sattr,
2833                                         state, label, olabel);
2834                         if (status == 0) {
2835                                 nfs_setattr_update_inode(state->inode, sattr,
2836                                                 opendata->o_res.f_attr);
2837                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2838                         }
2839                 }
2840         }
2841         if (opened && opendata->file_created)
2842                 *opened |= FILE_CREATED;
2843
2844         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2845                 *ctx_th = opendata->f_attr.mdsthreshold;
2846                 opendata->f_attr.mdsthreshold = NULL;
2847         }
2848
2849         nfs4_label_free(olabel);
2850
2851         nfs4_opendata_put(opendata);
2852         nfs4_put_state_owner(sp);
2853         return 0;
2854 err_free_label:
2855         nfs4_label_free(olabel);
2856 err_opendata_put:
2857         nfs4_opendata_put(opendata);
2858 err_put_state_owner:
2859         nfs4_put_state_owner(sp);
2860 out_err:
2861         return status;
2862 }
2863
2864
2865 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2866                                         struct nfs_open_context *ctx,
2867                                         int flags,
2868                                         struct iattr *sattr,
2869                                         struct nfs4_label *label,
2870                                         int *opened)
2871 {
2872         struct nfs_server *server = NFS_SERVER(dir);
2873         struct nfs4_exception exception = { };
2874         struct nfs4_state *res;
2875         struct nfs4_open_createattrs c = {
2876                 .label = label,
2877                 .sattr = sattr,
2878                 .verf = {
2879                         [0] = (__u32)jiffies,
2880                         [1] = (__u32)current->pid,
2881                 },
2882         };
2883         int status;
2884
2885         do {
2886                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
2887                 res = ctx->state;
2888                 trace_nfs4_open_file(ctx, flags, status);
2889                 if (status == 0)
2890                         break;
2891                 /* NOTE: BAD_SEQID means the server and client disagree about the
2892                  * book-keeping w.r.t. state-changing operations
2893                  * (OPEN/CLOSE/LOCK/LOCKU...)
2894                  * It is actually a sign of a bug on the client or on the server.
2895                  *
2896                  * If we receive a BAD_SEQID error in the particular case of
2897                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2898                  * have unhashed the old state_owner for us, and that we can
2899                  * therefore safely retry using a new one. We should still warn
2900                  * the user though...
2901                  */
2902                 if (status == -NFS4ERR_BAD_SEQID) {
2903                         pr_warn_ratelimited("NFS: v4 server %s "
2904                                         " returned a bad sequence-id error!\n",
2905                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2906                         exception.retry = 1;
2907                         continue;
2908                 }
2909                 /*
2910                  * BAD_STATEID on OPEN means that the server cancelled our
2911                  * state before it received the OPEN_CONFIRM.
2912                  * Recover by retrying the request as per the discussion
2913                  * on Page 181 of RFC3530.
2914                  */
2915                 if (status == -NFS4ERR_BAD_STATEID) {
2916                         exception.retry = 1;
2917                         continue;
2918                 }
2919                 if (status == -NFS4ERR_EXPIRED) {
2920                         nfs4_schedule_lease_recovery(server->nfs_client);
2921                         exception.retry = 1;
2922                         continue;
2923                 }
2924                 if (status == -EAGAIN) {
2925                         /* We must have found a delegation */
2926                         exception.retry = 1;
2927                         continue;
2928                 }
2929                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2930                         continue;
2931                 res = ERR_PTR(nfs4_handle_exception(server,
2932                                         status, &exception));
2933         } while (exception.retry);
2934         return res;
2935 }
2936
2937 static int _nfs4_do_setattr(struct inode *inode,
2938                             struct nfs_setattrargs *arg,
2939                             struct nfs_setattrres *res,
2940                             struct rpc_cred *cred,
2941                             struct nfs4_state *state)
2942 {
2943         struct nfs_server *server = NFS_SERVER(inode);
2944         struct rpc_message msg = {
2945                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2946                 .rpc_argp       = arg,
2947                 .rpc_resp       = res,
2948                 .rpc_cred       = cred,
2949         };
2950         struct rpc_cred *delegation_cred = NULL;
2951         unsigned long timestamp = jiffies;
2952         fmode_t fmode;
2953         bool truncate;
2954         int status;
2955
2956         nfs_fattr_init(res->fattr);
2957
2958         /* Servers should only apply open mode checks for file size changes */
2959         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2960         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2961
2962         if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2963                 /* Use that stateid */
2964         } else if (truncate && state != NULL) {
2965                 struct nfs_lockowner lockowner = {
2966                         .l_owner = current->files,
2967                         .l_pid = current->tgid,
2968                 };
2969                 if (!nfs4_valid_open_stateid(state))
2970                         return -EBADF;
2971                 if (nfs4_select_rw_stateid(state, FMODE_WRITE, &lockowner,
2972                                 &arg->stateid, &delegation_cred) == -EIO)
2973                         return -EBADF;
2974         } else
2975                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2976         if (delegation_cred)
2977                 msg.rpc_cred = delegation_cred;
2978
2979         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2980
2981         put_rpccred(delegation_cred);
2982         if (status == 0 && state != NULL)
2983                 renew_lease(server, timestamp);
2984         trace_nfs4_setattr(inode, &arg->stateid, status);
2985         return status;
2986 }
2987
2988 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2989                            struct nfs_fattr *fattr, struct iattr *sattr,
2990                            struct nfs4_state *state, struct nfs4_label *ilabel,
2991                            struct nfs4_label *olabel)
2992 {
2993         struct nfs_server *server = NFS_SERVER(inode);
2994         struct nfs_setattrargs  arg = {
2995                 .fh             = NFS_FH(inode),
2996                 .iap            = sattr,
2997                 .server         = server,
2998                 .bitmask = server->attr_bitmask,
2999                 .label          = ilabel,
3000         };
3001         struct nfs_setattrres  res = {
3002                 .fattr          = fattr,
3003                 .label          = olabel,
3004                 .server         = server,
3005         };
3006         struct nfs4_exception exception = {
3007                 .state = state,
3008                 .inode = inode,
3009                 .stateid = &arg.stateid,
3010         };
3011         int err;
3012
3013         arg.bitmask = nfs4_bitmask(server, ilabel);
3014         if (ilabel)
3015                 arg.bitmask = nfs4_bitmask(server, olabel);
3016
3017         do {
3018                 err = _nfs4_do_setattr(inode, &arg, &res, cred, state);
3019                 switch (err) {
3020                 case -NFS4ERR_OPENMODE:
3021                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3022                                 pr_warn_once("NFSv4: server %s is incorrectly "
3023                                                 "applying open mode checks to "
3024                                                 "a SETATTR that is not "
3025                                                 "changing file size.\n",
3026                                                 server->nfs_client->cl_hostname);
3027                         }
3028                         if (state && !(state->state & FMODE_WRITE)) {
3029                                 err = -EBADF;
3030                                 if (sattr->ia_valid & ATTR_OPEN)
3031                                         err = -EACCES;
3032                                 goto out;
3033                         }
3034                 }
3035                 err = nfs4_handle_exception(server, err, &exception);
3036         } while (exception.retry);
3037 out:
3038         return err;
3039 }
3040
3041 static bool
3042 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3043 {
3044         if (inode == NULL || !nfs_have_layout(inode))
3045                 return false;
3046
3047         return pnfs_wait_on_layoutreturn(inode, task);
3048 }
3049
3050 struct nfs4_closedata {
3051         struct inode *inode;
3052         struct nfs4_state *state;
3053         struct nfs_closeargs arg;
3054         struct nfs_closeres res;
3055         struct nfs_fattr fattr;
3056         unsigned long timestamp;
3057         bool roc;
3058         u32 roc_barrier;
3059 };
3060
3061 static void nfs4_free_closedata(void *data)
3062 {
3063         struct nfs4_closedata *calldata = data;
3064         struct nfs4_state_owner *sp = calldata->state->owner;
3065         struct super_block *sb = calldata->state->inode->i_sb;
3066
3067         if (calldata->roc)
3068                 pnfs_roc_release(calldata->state->inode);
3069         nfs4_put_open_state(calldata->state);
3070         nfs_free_seqid(calldata->arg.seqid);
3071         nfs4_put_state_owner(sp);
3072         nfs_sb_deactive(sb);
3073         kfree(calldata);
3074 }
3075
3076 static void nfs4_close_done(struct rpc_task *task, void *data)
3077 {
3078         struct nfs4_closedata *calldata = data;
3079         struct nfs4_state *state = calldata->state;
3080         struct nfs_server *server = NFS_SERVER(calldata->inode);
3081         nfs4_stateid *res_stateid = NULL;
3082
3083         dprintk("%s: begin!\n", __func__);
3084         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3085                 return;
3086         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3087         /* hmm. we are done with the inode, and in the process of freeing
3088          * the state_owner. we keep this around to process errors
3089          */
3090         switch (task->tk_status) {
3091                 case 0:
3092                         res_stateid = &calldata->res.stateid;
3093                         if (calldata->roc)
3094                                 pnfs_roc_set_barrier(state->inode,
3095                                                      calldata->roc_barrier);
3096                         renew_lease(server, calldata->timestamp);
3097                         break;
3098                 case -NFS4ERR_ADMIN_REVOKED:
3099                 case -NFS4ERR_STALE_STATEID:
3100                 case -NFS4ERR_EXPIRED:
3101                         nfs4_free_revoked_stateid(server,
3102                                         &calldata->arg.stateid,
3103                                         task->tk_msg.rpc_cred);
3104                 case -NFS4ERR_OLD_STATEID:
3105                 case -NFS4ERR_BAD_STATEID:
3106                         if (!nfs4_stateid_match(&calldata->arg.stateid,
3107                                                 &state->open_stateid)) {
3108                                 rpc_restart_call_prepare(task);
3109                                 goto out_release;
3110                         }
3111                         if (calldata->arg.fmode == 0)
3112                                 break;
3113                 default:
3114                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3115                                 rpc_restart_call_prepare(task);
3116                                 goto out_release;
3117                         }
3118         }
3119         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3120                         res_stateid, calldata->arg.fmode);
3121 out_release:
3122         nfs_release_seqid(calldata->arg.seqid);
3123         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
3124         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3125 }
3126
3127 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3128 {
3129         struct nfs4_closedata *calldata = data;
3130         struct nfs4_state *state = calldata->state;
3131         struct inode *inode = calldata->inode;
3132         bool is_rdonly, is_wronly, is_rdwr;
3133         int call_close = 0;
3134
3135         dprintk("%s: begin!\n", __func__);
3136         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3137                 goto out_wait;
3138
3139         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3140         spin_lock(&state->owner->so_lock);
3141         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3142         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3143         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3144         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3145         /* Calculate the change in open mode */
3146         calldata->arg.fmode = 0;
3147         if (state->n_rdwr == 0) {
3148                 if (state->n_rdonly == 0)
3149                         call_close |= is_rdonly;
3150                 else if (is_rdonly)
3151                         calldata->arg.fmode |= FMODE_READ;
3152                 if (state->n_wronly == 0)
3153                         call_close |= is_wronly;
3154                 else if (is_wronly)
3155                         calldata->arg.fmode |= FMODE_WRITE;
3156                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3157                         call_close |= is_rdwr;
3158         } else if (is_rdwr)
3159                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3160
3161         if (!nfs4_valid_open_stateid(state) ||
3162             test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3163                 call_close = 0;
3164         spin_unlock(&state->owner->so_lock);
3165
3166         if (!call_close) {
3167                 /* Note: exit _without_ calling nfs4_close_done */
3168                 goto out_no_action;
3169         }
3170
3171         if (nfs4_wait_on_layoutreturn(inode, task)) {
3172                 nfs_release_seqid(calldata->arg.seqid);
3173                 goto out_wait;
3174         }
3175
3176         if (calldata->arg.fmode == 0)
3177                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3178         if (calldata->roc)
3179                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
3180
3181         calldata->arg.share_access =
3182                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3183                                 calldata->arg.fmode, 0);
3184
3185         nfs_fattr_init(calldata->res.fattr);
3186         calldata->timestamp = jiffies;
3187         if (nfs4_setup_sequence(NFS_SERVER(inode),
3188                                 &calldata->arg.seq_args,
3189                                 &calldata->res.seq_res,
3190                                 task) != 0)
3191                 nfs_release_seqid(calldata->arg.seqid);
3192         dprintk("%s: done!\n", __func__);
3193         return;
3194 out_no_action:
3195         task->tk_action = NULL;
3196 out_wait:
3197         nfs4_sequence_done(task, &calldata->res.seq_res);
3198 }
3199
3200 static const struct rpc_call_ops nfs4_close_ops = {
3201         .rpc_call_prepare = nfs4_close_prepare,
3202         .rpc_call_done = nfs4_close_done,
3203         .rpc_release = nfs4_free_closedata,
3204 };
3205
3206 static bool nfs4_roc(struct inode *inode)
3207 {
3208         if (!nfs_have_layout(inode))
3209                 return false;
3210         return pnfs_roc(inode);
3211 }
3212
3213 /*
3214  * It is possible for data to be read/written from a mem-mapped file
3215  * after the sys_close call (which hits the vfs layer as a flush).
3216  * This means that we can't safely call nfsv4 close on a file until
3217  * the inode is cleared. This in turn means that we are not good
3218  * NFSv4 citizens - we do not indicate to the server to update the file's
3219  * share state even when we are done with one of the three share
3220  * stateid's in the inode.
3221  *
3222  * NOTE: Caller must be holding the sp->so_owner semaphore!
3223  */
3224 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3225 {
3226         struct nfs_server *server = NFS_SERVER(state->inode);
3227         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3228         struct nfs4_closedata *calldata;
3229         struct nfs4_state_owner *sp = state->owner;
3230         struct rpc_task *task;
3231         struct rpc_message msg = {
3232                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3233                 .rpc_cred = state->owner->so_cred,
3234         };
3235         struct rpc_task_setup task_setup_data = {
3236                 .rpc_client = server->client,
3237                 .rpc_message = &msg,
3238                 .callback_ops = &nfs4_close_ops,
3239                 .workqueue = nfsiod_workqueue,
3240                 .flags = RPC_TASK_ASYNC,
3241         };
3242         int status = -ENOMEM;
3243
3244         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3245                 &task_setup_data.rpc_client, &msg);
3246
3247         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3248         if (calldata == NULL)
3249                 goto out;
3250         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3251         calldata->inode = state->inode;
3252         calldata->state = state;
3253         calldata->arg.fh = NFS_FH(state->inode);
3254         /* Serialization for the sequence id */
3255         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3256         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3257         if (IS_ERR(calldata->arg.seqid))
3258                 goto out_free_calldata;
3259         calldata->arg.fmode = 0;
3260         calldata->arg.bitmask = server->cache_consistency_bitmask;
3261         calldata->res.fattr = &calldata->fattr;
3262         calldata->res.seqid = calldata->arg.seqid;
3263         calldata->res.server = server;
3264         calldata->roc = nfs4_roc(state->inode);
3265         nfs_sb_active(calldata->inode->i_sb);
3266
3267         msg.rpc_argp = &calldata->arg;
3268         msg.rpc_resp = &calldata->res;
3269         task_setup_data.callback_data = calldata;
3270         task = rpc_run_task(&task_setup_data);
3271         if (IS_ERR(task))
3272                 return PTR_ERR(task);
3273         status = 0;
3274         if (wait)
3275                 status = rpc_wait_for_completion_task(task);
3276         rpc_put_task(task);
3277         return status;
3278 out_free_calldata:
3279         kfree(calldata);
3280 out:
3281         nfs4_put_open_state(state);
3282         nfs4_put_state_owner(sp);
3283         return status;
3284 }
3285
3286 static struct inode *
3287 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3288                 int open_flags, struct iattr *attr, int *opened)
3289 {
3290         struct nfs4_state *state;
3291         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3292
3293         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3294
3295         /* Protect against concurrent sillydeletes */
3296         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3297
3298         nfs4_label_release_security(label);
3299
3300         if (IS_ERR(state))
3301                 return ERR_CAST(state);
3302         return state->inode;
3303 }
3304
3305 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3306 {
3307         if (ctx->state == NULL)
3308                 return;
3309         if (is_sync)
3310                 nfs4_close_sync(ctx->state, ctx->mode);
3311         else
3312                 nfs4_close_state(ctx->state, ctx->mode);
3313 }
3314
3315 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3316 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3317 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3318
3319 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3320 {
3321         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3322         struct nfs4_server_caps_arg args = {
3323                 .fhandle = fhandle,
3324                 .bitmask = bitmask,
3325         };
3326         struct nfs4_server_caps_res res = {};
3327         struct rpc_message msg = {
3328                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3329                 .rpc_argp = &args,
3330                 .rpc_resp = &res,
3331         };
3332         int status;
3333         int i;
3334
3335         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3336                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3337                      FATTR4_WORD0_LINK_SUPPORT |
3338                      FATTR4_WORD0_SYMLINK_SUPPORT |
3339                      FATTR4_WORD0_ACLSUPPORT;
3340         if (minorversion)
3341                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3342
3343         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3344         if (status == 0) {
3345                 /* Sanity check the server answers */
3346                 switch (minorversion) {
3347                 case 0:
3348                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3349                         res.attr_bitmask[2] = 0;
3350                         break;
3351                 case 1:
3352                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3353                         break;
3354                 case 2:
3355                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3356                 }
3357                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3358                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3359                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3360                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3361                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3362                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3363                                 NFS_CAP_SECURITY_LABEL);
3364                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3365                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3366                         server->caps |= NFS_CAP_ACLS;
3367                 if (res.has_links != 0)
3368                         server->caps |= NFS_CAP_HARDLINKS;
3369                 if (res.has_symlinks != 0)
3370                         server->caps |= NFS_CAP_SYMLINKS;
3371                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3372                         server->caps |= NFS_CAP_FILEID;
3373                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3374                         server->caps |= NFS_CAP_MODE;
3375                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3376                         server->caps |= NFS_CAP_NLINK;
3377                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3378                         server->caps |= NFS_CAP_OWNER;
3379                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3380                         server->caps |= NFS_CAP_OWNER_GROUP;
3381                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3382                         server->caps |= NFS_CAP_ATIME;
3383                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3384                         server->caps |= NFS_CAP_CTIME;
3385                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3386                         server->caps |= NFS_CAP_MTIME;
3387 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3388                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3389                         server->caps |= NFS_CAP_SECURITY_LABEL;
3390 #endif
3391                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3392                                 sizeof(server->attr_bitmask));
3393                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3394
3395                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3396                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3397                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3398                 server->cache_consistency_bitmask[2] = 0;
3399
3400                 /* Avoid a regression due to buggy server */
3401                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3402                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3403                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3404                         sizeof(server->exclcreat_bitmask));
3405
3406                 server->acl_bitmask = res.acl_bitmask;
3407                 server->fh_expire_type = res.fh_expire_type;
3408         }
3409
3410         return status;
3411 }
3412
3413 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3414 {
3415         struct nfs4_exception exception = { };
3416         int err;
3417         do {
3418                 err = nfs4_handle_exception(server,
3419                                 _nfs4_server_capabilities(server, fhandle),
3420                                 &exception);
3421         } while (exception.retry);
3422         return err;
3423 }
3424
3425 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3426                 struct nfs_fsinfo *info)
3427 {
3428         u32 bitmask[3];
3429         struct nfs4_lookup_root_arg args = {
3430                 .bitmask = bitmask,
3431         };
3432         struct nfs4_lookup_res res = {
3433                 .server = server,
3434                 .fattr = info->fattr,
3435                 .fh = fhandle,
3436         };
3437         struct rpc_message msg = {
3438                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3439                 .rpc_argp = &args,
3440                 .rpc_resp = &res,
3441         };
3442
3443         bitmask[0] = nfs4_fattr_bitmap[0];
3444         bitmask[1] = nfs4_fattr_bitmap[1];
3445         /*
3446          * Process the label in the upcoming getfattr
3447          */
3448         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3449
3450         nfs_fattr_init(info->fattr);
3451         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3452 }
3453
3454 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3455                 struct nfs_fsinfo *info)
3456 {
3457         struct nfs4_exception exception = { };
3458         int err;
3459         do {
3460                 err = _nfs4_lookup_root(server, fhandle, info);
3461                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3462                 switch (err) {
3463                 case 0:
3464                 case -NFS4ERR_WRONGSEC:
3465                         goto out;
3466                 default:
3467                         err = nfs4_handle_exception(server, err, &exception);
3468                 }
3469         } while (exception.retry);
3470 out:
3471         return err;
3472 }
3473
3474 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3475                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3476 {
3477         struct rpc_auth_create_args auth_args = {
3478                 .pseudoflavor = flavor,
3479         };
3480         struct rpc_auth *auth;
3481         int ret;
3482
3483         auth = rpcauth_create(&auth_args, server->client);
3484         if (IS_ERR(auth)) {
3485                 ret = -EACCES;
3486                 goto out;
3487         }
3488         ret = nfs4_lookup_root(server, fhandle, info);
3489 out:
3490         return ret;
3491 }
3492
3493 /*
3494  * Retry pseudoroot lookup with various security flavors.  We do this when:
3495  *
3496  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3497  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3498  *
3499  * Returns zero on success, or a negative NFS4ERR value, or a
3500  * negative errno value.
3501  */
3502 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3503                               struct nfs_fsinfo *info)
3504 {
3505         /* Per 3530bis 15.33.5 */
3506         static const rpc_authflavor_t flav_array[] = {
3507                 RPC_AUTH_GSS_KRB5P,
3508                 RPC_AUTH_GSS_KRB5I,
3509                 RPC_AUTH_GSS_KRB5,
3510                 RPC_AUTH_UNIX,                  /* courtesy */
3511                 RPC_AUTH_NULL,
3512         };
3513         int status = -EPERM;
3514         size_t i;
3515
3516         if (server->auth_info.flavor_len > 0) {
3517                 /* try each flavor specified by user */
3518                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3519                         status = nfs4_lookup_root_sec(server, fhandle, info,
3520                                                 server->auth_info.flavors[i]);
3521                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3522                                 continue;
3523                         break;
3524                 }
3525         } else {
3526                 /* no flavors specified by user, try default list */
3527                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3528                         status = nfs4_lookup_root_sec(server, fhandle, info,
3529                                                       flav_array[i]);
3530                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3531                                 continue;
3532                         break;
3533                 }
3534         }
3535
3536         /*
3537          * -EACCESS could mean that the user doesn't have correct permissions
3538          * to access the mount.  It could also mean that we tried to mount
3539          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3540          * existing mount programs don't handle -EACCES very well so it should
3541          * be mapped to -EPERM instead.
3542          */
3543         if (status == -EACCES)
3544                 status = -EPERM;
3545         return status;
3546 }
3547
3548 /**
3549  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3550  * @server: initialized nfs_server handle
3551  * @fhandle: we fill in the pseudo-fs root file handle
3552  * @info: we fill in an FSINFO struct
3553  * @auth_probe: probe the auth flavours
3554  *
3555  * Returns zero on success, or a negative errno.
3556  */
3557 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3558                          struct nfs_fsinfo *info,
3559                          bool auth_probe)
3560 {
3561         int status = 0;
3562
3563         if (!auth_probe)
3564                 status = nfs4_lookup_root(server, fhandle, info);
3565
3566         if (auth_probe || status == NFS4ERR_WRONGSEC)
3567                 status = server->nfs_client->cl_mvops->find_root_sec(server,
3568                                 fhandle, info);
3569
3570         if (status == 0)
3571                 status = nfs4_server_capabilities(server, fhandle);
3572         if (status == 0)
3573                 status = nfs4_do_fsinfo(server, fhandle, info);
3574
3575         return nfs4_map_errors(status);
3576 }
3577
3578 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3579                               struct nfs_fsinfo *info)
3580 {
3581         int error;
3582         struct nfs_fattr *fattr = info->fattr;
3583         struct nfs4_label *label = NULL;
3584
3585         error = nfs4_server_capabilities(server, mntfh);
3586         if (error < 0) {
3587                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3588                 return error;
3589         }
3590
3591         label = nfs4_label_alloc(server, GFP_KERNEL);
3592         if (IS_ERR(label))
3593                 return PTR_ERR(label);
3594
3595         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3596         if (error < 0) {
3597                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3598                 goto err_free_label;
3599         }
3600
3601         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3602             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3603                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3604
3605 err_free_label:
3606         nfs4_label_free(label);
3607
3608         return error;
3609 }
3610
3611 /*
3612  * Get locations and (maybe) other attributes of a referral.
3613  * Note that we'll actually follow the referral later when
3614  * we detect fsid mismatch in inode revalidation
3615  */
3616 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3617                              const struct qstr *name, struct nfs_fattr *fattr,
3618                              struct nfs_fh *fhandle)
3619 {
3620         int status = -ENOMEM;
3621         struct page *page = NULL;
3622         struct nfs4_fs_locations *locations = NULL;
3623
3624         page = alloc_page(GFP_KERNEL);
3625         if (page == NULL)
3626                 goto out;
3627         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3628         if (locations == NULL)
3629                 goto out;
3630
3631         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3632         if (status != 0)
3633                 goto out;
3634
3635         /*
3636          * If the fsid didn't change, this is a migration event, not a
3637          * referral.  Cause us to drop into the exception handler, which
3638          * will kick off migration recovery.
3639          */
3640         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3641                 dprintk("%s: server did not return a different fsid for"
3642                         " a referral at %s\n", __func__, name->name);
3643                 status = -NFS4ERR_MOVED;
3644                 goto out;
3645         }
3646         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3647         nfs_fixup_referral_attributes(&locations->fattr);
3648
3649         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3650         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3651         memset(fhandle, 0, sizeof(struct nfs_fh));
3652 out:
3653         if (page)
3654                 __free_page(page);
3655         kfree(locations);
3656         return status;
3657 }
3658
3659 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3660                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3661 {
3662         struct nfs4_getattr_arg args = {
3663                 .fh = fhandle,
3664                 .bitmask = server->attr_bitmask,
3665         };
3666         struct nfs4_getattr_res res = {
3667                 .fattr = fattr,
3668                 .label = label,
3669                 .server = server,
3670         };
3671         struct rpc_message msg = {
3672                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3673                 .rpc_argp = &args,
3674                 .rpc_resp = &res,
3675         };
3676
3677         args.bitmask = nfs4_bitmask(server, label);
3678
3679         nfs_fattr_init(fattr);
3680         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3681 }
3682
3683 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3684                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3685 {
3686         struct nfs4_exception exception = { };
3687         int err;
3688         do {
3689                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3690                 trace_nfs4_getattr(server, fhandle, fattr, err);
3691                 err = nfs4_handle_exception(server, err,
3692                                 &exception);
3693         } while (exception.retry);
3694         return err;
3695 }
3696
3697 /*
3698  * The file is not closed if it is opened due to the a request to change
3699  * the size of the file. The open call will not be needed once the
3700  * VFS layer lookup-intents are implemented.
3701  *
3702  * Close is called when the inode is destroyed.
3703  * If we haven't opened the file for O_WRONLY, we
3704  * need to in the size_change case to obtain a stateid.
3705  *
3706  * Got race?
3707  * Because OPEN is always done by name in nfsv4, it is
3708  * possible that we opened a different file by the same
3709  * name.  We can recognize this race condition, but we
3710  * can't do anything about it besides returning an error.
3711  *
3712  * This will be fixed with VFS changes (lookup-intent).
3713  */
3714 static int
3715 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3716                   struct iattr *sattr)
3717 {
3718         struct inode *inode = d_inode(dentry);
3719         struct rpc_cred *cred = NULL;
3720         struct nfs4_state *state = NULL;
3721         struct nfs4_label *label = NULL;
3722         int status;
3723
3724         if (pnfs_ld_layoutret_on_setattr(inode) &&
3725             sattr->ia_valid & ATTR_SIZE &&
3726             sattr->ia_size < i_size_read(inode))
3727                 pnfs_commit_and_return_layout(inode);
3728
3729         nfs_fattr_init(fattr);
3730
3731         /* Deal with open(O_TRUNC) */
3732         if (sattr->ia_valid & ATTR_OPEN)
3733                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3734
3735         /* Optimization: if the end result is no change, don't RPC */
3736         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3737                 return 0;
3738
3739         /* Search for an existing open(O_WRITE) file */
3740         if (sattr->ia_valid & ATTR_FILE) {
3741                 struct nfs_open_context *ctx;
3742
3743                 ctx = nfs_file_open_context(sattr->ia_file);
3744                 if (ctx) {
3745                         cred = ctx->cred;
3746                         state = ctx->state;
3747                 }
3748         }
3749
3750         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3751         if (IS_ERR(label))
3752                 return PTR_ERR(label);
3753
3754         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3755         if (status == 0) {
3756                 nfs_setattr_update_inode(inode, sattr, fattr);
3757                 nfs_setsecurity(inode, fattr, label);
3758         }
3759         nfs4_label_free(label);
3760         return status;
3761 }
3762
3763 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3764                 const struct qstr *name, struct nfs_fh *fhandle,
3765                 struct nfs_fattr *fattr, struct nfs4_label *label)
3766 {
3767         struct nfs_server *server = NFS_SERVER(dir);
3768         int                    status;
3769         struct nfs4_lookup_arg args = {
3770                 .bitmask = server->attr_bitmask,
3771                 .dir_fh = NFS_FH(dir),
3772                 .name = name,
3773         };
3774         struct nfs4_lookup_res res = {
3775                 .server = server,
3776                 .fattr = fattr,
3777                 .label = label,
3778                 .fh = fhandle,
3779         };
3780         struct rpc_message msg = {
3781                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3782                 .rpc_argp = &args,
3783                 .rpc_resp = &res,
3784         };
3785
3786         args.bitmask = nfs4_bitmask(server, label);
3787
3788         nfs_fattr_init(fattr);
3789
3790         dprintk("NFS call  lookup %s\n", name->name);
3791         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3792         dprintk("NFS reply lookup: %d\n", status);
3793         return status;
3794 }
3795
3796 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3797 {
3798         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3799                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3800         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3801         fattr->nlink = 2;
3802 }
3803
3804 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3805                                    const struct qstr *name, struct nfs_fh *fhandle,
3806                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3807 {
3808         struct nfs4_exception exception = { };
3809         struct rpc_clnt *client = *clnt;
3810         int err;
3811         do {
3812                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3813                 trace_nfs4_lookup(dir, name, err);
3814                 switch (err) {
3815                 case -NFS4ERR_BADNAME:
3816                         err = -ENOENT;
3817                         goto out;
3818                 case -NFS4ERR_MOVED:
3819                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3820                         if (err == -NFS4ERR_MOVED)
3821                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3822                         goto out;
3823                 case -NFS4ERR_WRONGSEC:
3824                         err = -EPERM;
3825                         if (client != *clnt)
3826                                 goto out;
3827                         client = nfs4_negotiate_security(client, dir, name);
3828                         if (IS_ERR(client))
3829                                 return PTR_ERR(client);
3830
3831                         exception.retry = 1;
3832                         break;
3833                 default:
3834                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3835                 }
3836         } while (exception.retry);
3837
3838 out:
3839         if (err == 0)
3840                 *clnt = client;
3841         else if (client != *clnt)
3842                 rpc_shutdown_client(client);
3843
3844         return err;
3845 }
3846
3847 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3848                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3849                             struct nfs4_label *label)
3850 {
3851         int status;
3852         struct rpc_clnt *client = NFS_CLIENT(dir);
3853
3854         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3855         if (client != NFS_CLIENT(dir)) {
3856                 rpc_shutdown_client(client);
3857                 nfs_fixup_secinfo_attributes(fattr);
3858         }
3859         return status;
3860 }
3861
3862 struct rpc_clnt *
3863 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3864                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3865 {
3866         struct rpc_clnt *client = NFS_CLIENT(dir);
3867         int status;
3868
3869         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3870         if (status < 0)
3871                 return ERR_PTR(status);
3872         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3873 }
3874
3875 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3876 {
3877         struct nfs_server *server = NFS_SERVER(inode);
3878         struct nfs4_accessargs args = {
3879                 .fh = NFS_FH(inode),
3880                 .bitmask = server->cache_consistency_bitmask,
3881         };
3882         struct nfs4_accessres res = {
3883                 .server = server,
3884         };
3885         struct rpc_message msg = {
3886                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3887                 .rpc_argp = &args,
3888                 .rpc_resp = &res,
3889                 .rpc_cred = entry->cred,
3890         };
3891         int mode = entry->mask;
3892         int status = 0;
3893
3894         /*
3895          * Determine which access bits we want to ask for...
3896          */
3897         if (mode & MAY_READ)
3898                 args.access |= NFS4_ACCESS_READ;
3899         if (S_ISDIR(inode->i_mode)) {
3900                 if (mode & MAY_WRITE)
3901                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3902                 if (mode & MAY_EXEC)
3903                         args.access |= NFS4_ACCESS_LOOKUP;
3904         } else {
3905                 if (mode & MAY_WRITE)
3906                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3907                 if (mode & MAY_EXEC)
3908                         args.access |= NFS4_ACCESS_EXECUTE;
3909         }
3910
3911         res.fattr = nfs_alloc_fattr();
3912         if (res.fattr == NULL)
3913                 return -ENOMEM;
3914
3915         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3916         if (!status) {
3917                 nfs_access_set_mask(entry, res.access);
3918                 nfs_refresh_inode(inode, res.fattr);
3919         }
3920         nfs_free_fattr(res.fattr);
3921         return status;
3922 }
3923
3924 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3925 {
3926         struct nfs4_exception exception = { };
3927         int err;
3928         do {
3929                 err = _nfs4_proc_access(inode, entry);
3930                 trace_nfs4_access(inode, err);
3931                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3932                                 &exception);
3933         } while (exception.retry);
3934         return err;
3935 }
3936
3937 /*
3938  * TODO: For the time being, we don't try to get any attributes
3939  * along with any of the zero-copy operations READ, READDIR,
3940  * READLINK, WRITE.
3941  *
3942  * In the case of the first three, we want to put the GETATTR
3943  * after the read-type operation -- this is because it is hard
3944  * to predict the length of a GETATTR response in v4, and thus
3945  * align the READ data correctly.  This means that the GETATTR
3946  * may end up partially falling into the page cache, and we should
3947  * shift it into the 'tail' of the xdr_buf before processing.
3948  * To do this efficiently, we need to know the total length
3949  * of data received, which doesn't seem to be available outside
3950  * of the RPC layer.
3951  *
3952  * In the case of WRITE, we also want to put the GETATTR after
3953  * the operation -- in this case because we want to make sure
3954  * we get the post-operation mtime and size.
3955  *
3956  * Both of these changes to the XDR layer would in fact be quite
3957  * minor, but I decided to leave them for a subsequent patch.
3958  */
3959 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3960                 unsigned int pgbase, unsigned int pglen)
3961 {
3962         struct nfs4_readlink args = {
3963                 .fh       = NFS_FH(inode),
3964                 .pgbase   = pgbase,
3965                 .pglen    = pglen,
3966                 .pages    = &page,
3967         };
3968         struct nfs4_readlink_res res;
3969         struct rpc_message msg = {
3970                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3971                 .rpc_argp = &args,
3972                 .rpc_resp = &res,
3973         };
3974
3975         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3976 }
3977
3978 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3979                 unsigned int pgbase, unsigned int pglen)
3980 {
3981         struct nfs4_exception exception = { };
3982         int err;
3983         do {
3984                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3985                 trace_nfs4_readlink(inode, err);
3986                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3987                                 &exception);
3988         } while (exception.retry);
3989         return err;
3990 }
3991
3992 /*
3993  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3994  */
3995 static int
3996 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3997                  int flags)
3998 {
3999         struct nfs4_label l, *ilabel = NULL;
4000         struct nfs_open_context *ctx;
4001         struct nfs4_state *state;
4002         int status = 0;
4003
4004         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
4005         if (IS_ERR(ctx))
4006                 return PTR_ERR(ctx);
4007
4008         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4009
4010         sattr->ia_mode &= ~current_umask();
4011         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4012         if (IS_ERR(state)) {
4013                 status = PTR_ERR(state);
4014                 goto out;
4015         }
4016 out:
4017         nfs4_label_release_security(ilabel);
4018         put_nfs_open_context(ctx);
4019         return status;
4020 }
4021
4022 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4023 {
4024         struct nfs_server *server = NFS_SERVER(dir);
4025         struct nfs_removeargs args = {
4026                 .fh = NFS_FH(dir),
4027                 .name = *name,
4028         };
4029         struct nfs_removeres res = {
4030                 .server = server,
4031         };
4032         struct rpc_message msg = {
4033                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4034                 .rpc_argp = &args,
4035                 .rpc_resp = &res,
4036         };
4037         int status;
4038
4039         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4040         if (status == 0)
4041                 update_changeattr(dir, &res.cinfo);
4042         return status;
4043 }
4044
4045 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
4046 {
4047         struct nfs4_exception exception = { };
4048         int err;
4049         do {
4050                 err = _nfs4_proc_remove(dir, name);
4051                 trace_nfs4_remove(dir, name, err);
4052                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4053                                 &exception);
4054         } while (exception.retry);
4055         return err;
4056 }
4057
4058 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
4059 {
4060         struct nfs_server *server = NFS_SERVER(dir);
4061         struct nfs_removeargs *args = msg->rpc_argp;
4062         struct nfs_removeres *res = msg->rpc_resp;
4063
4064         res->server = server;
4065         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4066         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
4067
4068         nfs_fattr_init(res->dir_attr);
4069 }
4070
4071 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4072 {
4073         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb),
4074                         &data->args.seq_args,
4075                         &data->res.seq_res,
4076                         task);
4077 }
4078
4079 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4080 {
4081         struct nfs_unlinkdata *data = task->tk_calldata;
4082         struct nfs_removeres *res = &data->res;
4083
4084         if (!nfs4_sequence_done(task, &res->seq_res))
4085                 return 0;
4086         if (nfs4_async_handle_error(task, res->server, NULL,
4087                                     &data->timeout) == -EAGAIN)
4088                 return 0;
4089         update_changeattr(dir, &res->cinfo);
4090         return 1;
4091 }
4092
4093 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4094 {
4095         struct nfs_server *server = NFS_SERVER(dir);
4096         struct nfs_renameargs *arg = msg->rpc_argp;
4097         struct nfs_renameres *res = msg->rpc_resp;
4098
4099         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4100         res->server = server;
4101         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4102 }
4103
4104 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4105 {
4106         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
4107                         &data->args.seq_args,
4108                         &data->res.seq_res,
4109                         task);
4110 }
4111
4112 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4113                                  struct inode *new_dir)
4114 {
4115         struct nfs_renamedata *data = task->tk_calldata;
4116         struct nfs_renameres *res = &data->res;
4117
4118         if (!nfs4_sequence_done(task, &res->seq_res))
4119                 return 0;
4120         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4121                 return 0;
4122
4123         update_changeattr(old_dir, &res->old_cinfo);
4124         update_changeattr(new_dir, &res->new_cinfo);
4125         return 1;
4126 }
4127
4128 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4129 {
4130         struct nfs_server *server = NFS_SERVER(inode);
4131         struct nfs4_link_arg arg = {
4132                 .fh     = NFS_FH(inode),
4133                 .dir_fh = NFS_FH(dir),
4134                 .name   = name,
4135                 .bitmask = server->attr_bitmask,
4136         };
4137         struct nfs4_link_res res = {
4138                 .server = server,
4139                 .label = NULL,
4140         };
4141         struct rpc_message msg = {
4142                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4143                 .rpc_argp = &arg,
4144                 .rpc_resp = &res,
4145         };
4146         int status = -ENOMEM;
4147
4148         res.fattr = nfs_alloc_fattr();
4149         if (res.fattr == NULL)
4150                 goto out;
4151
4152         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4153         if (IS_ERR(res.label)) {
4154                 status = PTR_ERR(res.label);
4155                 goto out;
4156         }
4157         arg.bitmask = nfs4_bitmask(server, res.label);
4158
4159         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4160         if (!status) {
4161                 update_changeattr(dir, &res.cinfo);
4162                 status = nfs_post_op_update_inode(inode, res.fattr);
4163                 if (!status)
4164                         nfs_setsecurity(inode, res.fattr, res.label);
4165         }
4166
4167
4168         nfs4_label_free(res.label);
4169
4170 out:
4171         nfs_free_fattr(res.fattr);
4172         return status;
4173 }
4174
4175 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4176 {
4177         struct nfs4_exception exception = { };
4178         int err;
4179         do {
4180                 err = nfs4_handle_exception(NFS_SERVER(inode),
4181                                 _nfs4_proc_link(inode, dir, name),
4182                                 &exception);
4183         } while (exception.retry);
4184         return err;
4185 }
4186
4187 struct nfs4_createdata {
4188         struct rpc_message msg;
4189         struct nfs4_create_arg arg;
4190         struct nfs4_create_res res;
4191         struct nfs_fh fh;
4192         struct nfs_fattr fattr;
4193         struct nfs4_label *label;
4194 };
4195
4196 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4197                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4198 {
4199         struct nfs4_createdata *data;
4200
4201         data = kzalloc(sizeof(*data), GFP_KERNEL);
4202         if (data != NULL) {
4203                 struct nfs_server *server = NFS_SERVER(dir);
4204
4205                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4206                 if (IS_ERR(data->label))
4207                         goto out_free;
4208
4209                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4210                 data->msg.rpc_argp = &data->arg;
4211                 data->msg.rpc_resp = &data->res;
4212                 data->arg.dir_fh = NFS_FH(dir);
4213                 data->arg.server = server;
4214                 data->arg.name = name;
4215                 data->arg.attrs = sattr;
4216                 data->arg.ftype = ftype;
4217                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4218                 data->res.server = server;
4219                 data->res.fh = &data->fh;
4220                 data->res.fattr = &data->fattr;
4221                 data->res.label = data->label;
4222                 nfs_fattr_init(data->res.fattr);
4223         }
4224         return data;
4225 out_free:
4226         kfree(data);
4227         return NULL;
4228 }
4229
4230 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4231 {
4232         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4233                                     &data->arg.seq_args, &data->res.seq_res, 1);
4234         if (status == 0) {
4235                 update_changeattr(dir, &data->res.dir_cinfo);
4236                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4237         }
4238         return status;
4239 }
4240
4241 static void nfs4_free_createdata(struct nfs4_createdata *data)
4242 {
4243         nfs4_label_free(data->label);
4244         kfree(data);
4245 }
4246
4247 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4248                 struct page *page, unsigned int len, struct iattr *sattr,
4249                 struct nfs4_label *label)
4250 {
4251         struct nfs4_createdata *data;
4252         int status = -ENAMETOOLONG;
4253
4254         if (len > NFS4_MAXPATHLEN)
4255                 goto out;
4256
4257         status = -ENOMEM;
4258         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4259         if (data == NULL)
4260                 goto out;
4261
4262         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4263         data->arg.u.symlink.pages = &page;
4264         data->arg.u.symlink.len = len;
4265         data->arg.label = label;
4266
4267         status = nfs4_do_create(dir, dentry, data);
4268
4269         nfs4_free_createdata(data);
4270 out:
4271         return status;
4272 }
4273
4274 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4275                 struct page *page, unsigned int len, struct iattr *sattr)
4276 {
4277         struct nfs4_exception exception = { };
4278         struct nfs4_label l, *label = NULL;
4279         int err;
4280
4281         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4282
4283         do {
4284                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4285                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4286                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4287                                 &exception);
4288         } while (exception.retry);
4289
4290         nfs4_label_release_security(label);
4291         return err;
4292 }
4293
4294 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4295                 struct iattr *sattr, struct nfs4_label *label)
4296 {
4297         struct nfs4_createdata *data;
4298         int status = -ENOMEM;
4299
4300         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4301         if (data == NULL)
4302                 goto out;
4303
4304         data->arg.label = label;
4305         status = nfs4_do_create(dir, dentry, data);
4306
4307         nfs4_free_createdata(data);
4308 out:
4309         return status;
4310 }
4311
4312 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4313                 struct iattr *sattr)
4314 {
4315         struct nfs4_exception exception = { };
4316         struct nfs4_label l, *label = NULL;
4317         int err;
4318
4319         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4320
4321         sattr->ia_mode &= ~current_umask();
4322         do {
4323                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4324                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4325                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4326                                 &exception);
4327         } while (exception.retry);
4328         nfs4_label_release_security(label);
4329
4330         return err;
4331 }
4332
4333 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4334                 u64 cookie, struct page **pages, unsigned int count, int plus)
4335 {
4336         struct inode            *dir = d_inode(dentry);
4337         struct nfs_server       *server = NFS_SERVER(dir);
4338         struct nfs4_readdir_arg args = {
4339                 .fh = NFS_FH(dir),
4340                 .pages = pages,
4341                 .pgbase = 0,
4342                 .count = count,
4343                 .plus = plus,
4344         };
4345         struct nfs4_readdir_res res;
4346         struct rpc_message msg = {
4347                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4348                 .rpc_argp = &args,
4349                 .rpc_resp = &res,
4350                 .rpc_cred = cred,
4351         };
4352         int                     status;
4353
4354         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4355                         dentry,
4356                         (unsigned long long)cookie);
4357         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
4358                 args.bitmask = server->attr_bitmask_nl;
4359         else
4360                 args.bitmask = server->attr_bitmask;
4361
4362         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4363         res.pgbase = args.pgbase;
4364         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
4365                         &res.seq_res, 0);
4366         if (status >= 0) {
4367                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4368                 status += args.pgbase;
4369         }
4370
4371         nfs_invalidate_atime(dir);
4372
4373         dprintk("%s: returns %d\n", __func__, status);
4374         return status;
4375 }
4376
4377 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4378                 u64 cookie, struct page **pages, unsigned int count, int plus)
4379 {
4380         struct nfs4_exception exception = { };
4381         int err;
4382         do {
4383                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4384                                 pages, count, plus);
4385                 trace_nfs4_readdir(d_inode(dentry), err);
4386                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4387                                 &exception);
4388         } while (exception.retry);
4389         return err;
4390 }
4391
4392 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4393                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4394 {
4395         struct nfs4_createdata *data;
4396         int mode = sattr->ia_mode;
4397         int status = -ENOMEM;
4398
4399         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4400         if (data == NULL)
4401                 goto out;
4402
4403         if (S_ISFIFO(mode))
4404                 data->arg.ftype = NF4FIFO;
4405         else if (S_ISBLK(mode)) {
4406                 data->arg.ftype = NF4BLK;
4407                 data->arg.u.device.specdata1 = MAJOR(rdev);
4408                 data->arg.u.device.specdata2 = MINOR(rdev);
4409         }
4410         else if (S_ISCHR(mode)) {
4411                 data->arg.ftype = NF4CHR;
4412                 data->arg.u.device.specdata1 = MAJOR(rdev);
4413                 data->arg.u.device.specdata2 = MINOR(rdev);
4414         } else if (!S_ISSOCK(mode)) {
4415                 status = -EINVAL;
4416                 goto out_free;
4417         }
4418
4419         data->arg.label = label;
4420         status = nfs4_do_create(dir, dentry, data);
4421 out_free:
4422         nfs4_free_createdata(data);
4423 out:
4424         return status;
4425 }
4426
4427 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4428                 struct iattr *sattr, dev_t rdev)
4429 {
4430         struct nfs4_exception exception = { };
4431         struct nfs4_label l, *label = NULL;
4432         int err;
4433
4434         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4435
4436         sattr->ia_mode &= ~current_umask();
4437         do {
4438                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4439                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4440                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4441                                 &exception);
4442         } while (exception.retry);
4443
4444         nfs4_label_release_security(label);
4445
4446         return err;
4447 }
4448
4449 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4450                  struct nfs_fsstat *fsstat)
4451 {
4452         struct nfs4_statfs_arg args = {
4453                 .fh = fhandle,
4454                 .bitmask = server->attr_bitmask,
4455         };
4456         struct nfs4_statfs_res res = {
4457                 .fsstat = fsstat,
4458         };
4459         struct rpc_message msg = {
4460                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4461                 .rpc_argp = &args,
4462                 .rpc_resp = &res,
4463         };
4464
4465         nfs_fattr_init(fsstat->fattr);
4466         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4467 }
4468
4469 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4470 {
4471         struct nfs4_exception exception = { };
4472         int err;
4473         do {
4474                 err = nfs4_handle_exception(server,
4475                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4476                                 &exception);
4477         } while (exception.retry);
4478         return err;
4479 }
4480
4481 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4482                 struct nfs_fsinfo *fsinfo)
4483 {
4484         struct nfs4_fsinfo_arg args = {
4485                 .fh = fhandle,
4486                 .bitmask = server->attr_bitmask,
4487         };
4488         struct nfs4_fsinfo_res res = {
4489                 .fsinfo = fsinfo,
4490         };
4491         struct rpc_message msg = {
4492                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4493                 .rpc_argp = &args,
4494                 .rpc_resp = &res,
4495         };
4496
4497         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4498 }
4499
4500 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4501 {
4502         struct nfs4_exception exception = { };
4503         unsigned long now = jiffies;
4504         int err;
4505
4506         do {
4507                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4508                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4509                 if (err == 0) {
4510                         nfs4_set_lease_period(server->nfs_client,
4511                                         fsinfo->lease_time * HZ,
4512                                         now);
4513                         break;
4514                 }
4515                 err = nfs4_handle_exception(server, err, &exception);
4516         } while (exception.retry);
4517         return err;
4518 }
4519
4520 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4521 {
4522         int error;
4523
4524         nfs_fattr_init(fsinfo->fattr);
4525         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4526         if (error == 0) {
4527                 /* block layout checks this! */
4528                 server->pnfs_blksize = fsinfo->blksize;
4529                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4530         }
4531
4532         return error;
4533 }
4534
4535 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4536                 struct nfs_pathconf *pathconf)
4537 {
4538         struct nfs4_pathconf_arg args = {
4539                 .fh = fhandle,
4540                 .bitmask = server->attr_bitmask,
4541         };
4542         struct nfs4_pathconf_res res = {
4543                 .pathconf = pathconf,
4544         };
4545         struct rpc_message msg = {
4546                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4547                 .rpc_argp = &args,
4548                 .rpc_resp = &res,
4549         };
4550
4551         /* None of the pathconf attributes are mandatory to implement */
4552         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4553                 memset(pathconf, 0, sizeof(*pathconf));
4554                 return 0;
4555         }
4556
4557         nfs_fattr_init(pathconf->fattr);
4558         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4559 }
4560
4561 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4562                 struct nfs_pathconf *pathconf)
4563 {
4564         struct nfs4_exception exception = { };
4565         int err;
4566
4567         do {
4568                 err = nfs4_handle_exception(server,
4569                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4570                                 &exception);
4571         } while (exception.retry);
4572         return err;
4573 }
4574
4575 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4576                 const struct nfs_open_context *ctx,
4577                 const struct nfs_lock_context *l_ctx,
4578                 fmode_t fmode)
4579 {
4580         const struct nfs_lockowner *lockowner = NULL;
4581
4582         if (l_ctx != NULL)
4583                 lockowner = &l_ctx->lockowner;
4584         return nfs4_select_rw_stateid(ctx->state, fmode, lockowner, stateid, NULL);
4585 }
4586 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4587
4588 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4589                 const struct nfs_open_context *ctx,
4590                 const struct nfs_lock_context *l_ctx,
4591                 fmode_t fmode)
4592 {
4593         nfs4_stateid current_stateid;
4594
4595         /* If the current stateid represents a lost lock, then exit */
4596         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4597                 return true;
4598         return nfs4_stateid_match(stateid, &current_stateid);
4599 }
4600
4601 static bool nfs4_error_stateid_expired(int err)
4602 {
4603         switch (err) {
4604         case -NFS4ERR_DELEG_REVOKED:
4605         case -NFS4ERR_ADMIN_REVOKED:
4606         case -NFS4ERR_BAD_STATEID:
4607         case -NFS4ERR_STALE_STATEID:
4608         case -NFS4ERR_OLD_STATEID:
4609         case -NFS4ERR_OPENMODE:
4610         case -NFS4ERR_EXPIRED:
4611                 return true;
4612         }
4613         return false;
4614 }
4615
4616 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4617 {
4618         struct nfs_server *server = NFS_SERVER(hdr->inode);
4619
4620         trace_nfs4_read(hdr, task->tk_status);
4621         if (task->tk_status < 0) {
4622                 struct nfs4_exception exception = {
4623                         .inode = hdr->inode,
4624                         .state = hdr->args.context->state,
4625                         .stateid = &hdr->args.stateid,
4626                 };
4627                 task->tk_status = nfs4_async_handle_exception(task,
4628                                 server, task->tk_status, &exception);
4629                 if (exception.retry) {
4630                         rpc_restart_call_prepare(task);
4631                         return -EAGAIN;
4632                 }
4633         }
4634
4635         if (task->tk_status > 0)
4636                 renew_lease(server, hdr->timestamp);
4637         return 0;
4638 }
4639
4640 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4641                 struct nfs_pgio_args *args)
4642 {
4643
4644         if (!nfs4_error_stateid_expired(task->tk_status) ||
4645                 nfs4_stateid_is_current(&args->stateid,
4646                                 args->context,
4647                                 args->lock_context,
4648                                 FMODE_READ))
4649                 return false;
4650         rpc_restart_call_prepare(task);
4651         return true;
4652 }
4653
4654 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4655 {
4656
4657         dprintk("--> %s\n", __func__);
4658
4659         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4660                 return -EAGAIN;
4661         if (nfs4_read_stateid_changed(task, &hdr->args))
4662                 return -EAGAIN;
4663         if (task->tk_status > 0)
4664                 nfs_invalidate_atime(hdr->inode);
4665         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4666                                     nfs4_read_done_cb(task, hdr);
4667 }
4668
4669 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4670                                  struct rpc_message *msg)
4671 {
4672         hdr->timestamp   = jiffies;
4673         if (!hdr->pgio_done_cb)
4674                 hdr->pgio_done_cb = nfs4_read_done_cb;
4675         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4676         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4677 }
4678
4679 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4680                                       struct nfs_pgio_header *hdr)
4681 {
4682         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4683                         &hdr->args.seq_args,
4684                         &hdr->res.seq_res,
4685                         task))
4686                 return 0;
4687         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4688                                 hdr->args.lock_context,
4689                                 hdr->rw_ops->rw_mode) == -EIO)
4690                 return -EIO;
4691         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4692                 return -EIO;
4693         return 0;
4694 }
4695
4696 static int nfs4_write_done_cb(struct rpc_task *task,
4697                               struct nfs_pgio_header *hdr)
4698 {
4699         struct inode *inode = hdr->inode;
4700
4701         trace_nfs4_write(hdr, task->tk_status);
4702         if (task->tk_status < 0) {
4703                 struct nfs4_exception exception = {
4704                         .inode = hdr->inode,
4705                         .state = hdr->args.context->state,
4706                         .stateid = &hdr->args.stateid,
4707                 };
4708                 task->tk_status = nfs4_async_handle_exception(task,
4709                                 NFS_SERVER(inode), task->tk_status,
4710                                 &exception);
4711                 if (exception.retry) {
4712                         rpc_restart_call_prepare(task);
4713                         return -EAGAIN;
4714                 }
4715         }
4716         if (task->tk_status >= 0) {
4717                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4718                 nfs_writeback_update_inode(hdr);
4719         }
4720         return 0;
4721 }
4722
4723 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4724                 struct nfs_pgio_args *args)
4725 {
4726
4727         if (!nfs4_error_stateid_expired(task->tk_status) ||
4728                 nfs4_stateid_is_current(&args->stateid,
4729                                 args->context,
4730                                 args->lock_context,
4731                                 FMODE_WRITE))
4732                 return false;
4733         rpc_restart_call_prepare(task);
4734         return true;
4735 }
4736
4737 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4738 {
4739         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4740                 return -EAGAIN;
4741         if (nfs4_write_stateid_changed(task, &hdr->args))
4742                 return -EAGAIN;
4743         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4744                 nfs4_write_done_cb(task, hdr);
4745 }
4746
4747 static
4748 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4749 {
4750         /* Don't request attributes for pNFS or O_DIRECT writes */
4751         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4752                 return false;
4753         /* Otherwise, request attributes if and only if we don't hold
4754          * a delegation
4755          */
4756         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4757 }
4758
4759 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4760                                   struct rpc_message *msg)
4761 {
4762         struct nfs_server *server = NFS_SERVER(hdr->inode);
4763
4764         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4765                 hdr->args.bitmask = NULL;
4766                 hdr->res.fattr = NULL;
4767         } else
4768                 hdr->args.bitmask = server->cache_consistency_bitmask;
4769
4770         if (!hdr->pgio_done_cb)
4771                 hdr->pgio_done_cb = nfs4_write_done_cb;
4772         hdr->res.server = server;
4773         hdr->timestamp   = jiffies;
4774
4775         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4776         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4777 }
4778
4779 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4780 {
4781         nfs4_setup_sequence(NFS_SERVER(data->inode),
4782                         &data->args.seq_args,
4783                         &data->res.seq_res,
4784                         task);
4785 }
4786
4787 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4788 {
4789         struct inode *inode = data->inode;
4790
4791         trace_nfs4_commit(data, task->tk_status);
4792         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4793                                     NULL, NULL) == -EAGAIN) {
4794                 rpc_restart_call_prepare(task);
4795                 return -EAGAIN;
4796         }
4797         return 0;
4798 }
4799
4800 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4801 {
4802         if (!nfs4_sequence_done(task, &data->res.seq_res))
4803                 return -EAGAIN;
4804         return data->commit_done_cb(task, data);
4805 }
4806
4807 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4808 {
4809         struct nfs_server *server = NFS_SERVER(data->inode);
4810
4811         if (data->commit_done_cb == NULL)
4812                 data->commit_done_cb = nfs4_commit_done_cb;
4813         data->res.server = server;
4814         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4815         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4816 }
4817
4818 struct nfs4_renewdata {
4819         struct nfs_client       *client;
4820         unsigned long           timestamp;
4821 };
4822
4823 /*
4824  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4825  * standalone procedure for queueing an asynchronous RENEW.
4826  */
4827 static void nfs4_renew_release(void *calldata)
4828 {
4829         struct nfs4_renewdata *data = calldata;
4830         struct nfs_client *clp = data->client;
4831
4832         if (atomic_read(&clp->cl_count) > 1)
4833                 nfs4_schedule_state_renewal(clp);
4834         nfs_put_client(clp);
4835         kfree(data);
4836 }
4837
4838 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4839 {
4840         struct nfs4_renewdata *data = calldata;
4841         struct nfs_client *clp = data->client;
4842         unsigned long timestamp = data->timestamp;
4843
4844         trace_nfs4_renew_async(clp, task->tk_status);
4845         switch (task->tk_status) {
4846         case 0:
4847                 break;
4848         case -NFS4ERR_LEASE_MOVED:
4849                 nfs4_schedule_lease_moved_recovery(clp);
4850                 break;
4851         default:
4852                 /* Unless we're shutting down, schedule state recovery! */
4853                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4854                         return;
4855                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4856                         nfs4_schedule_lease_recovery(clp);
4857                         return;
4858                 }
4859                 nfs4_schedule_path_down_recovery(clp);
4860         }
4861         do_renew_lease(clp, timestamp);
4862 }
4863
4864 static const struct rpc_call_ops nfs4_renew_ops = {
4865         .rpc_call_done = nfs4_renew_done,
4866         .rpc_release = nfs4_renew_release,
4867 };
4868
4869 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4870 {
4871         struct rpc_message msg = {
4872                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4873                 .rpc_argp       = clp,
4874                 .rpc_cred       = cred,
4875         };
4876         struct nfs4_renewdata *data;
4877
4878         if (renew_flags == 0)
4879                 return 0;
4880         if (!atomic_inc_not_zero(&clp->cl_count))
4881                 return -EIO;
4882         data = kmalloc(sizeof(*data), GFP_NOFS);
4883         if (data == NULL)
4884                 return -ENOMEM;
4885         data->client = clp;
4886         data->timestamp = jiffies;
4887         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4888                         &nfs4_renew_ops, data);
4889 }
4890
4891 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4892 {
4893         struct rpc_message msg = {
4894                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4895                 .rpc_argp       = clp,
4896                 .rpc_cred       = cred,
4897         };
4898         unsigned long now = jiffies;
4899         int status;
4900
4901         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4902         if (status < 0)
4903                 return status;
4904         do_renew_lease(clp, now);
4905         return 0;
4906 }
4907
4908 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4909 {
4910         return server->caps & NFS_CAP_ACLS;
4911 }
4912
4913 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4914  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4915  * the stack.
4916  */
4917 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4918
4919 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4920                 struct page **pages)
4921 {
4922         struct page *newpage, **spages;
4923         int rc = 0;
4924         size_t len;
4925         spages = pages;
4926
4927         do {
4928                 len = min_t(size_t, PAGE_SIZE, buflen);
4929                 newpage = alloc_page(GFP_KERNEL);
4930
4931                 if (newpage == NULL)
4932                         goto unwind;
4933                 memcpy(page_address(newpage), buf, len);
4934                 buf += len;
4935                 buflen -= len;
4936                 *pages++ = newpage;
4937                 rc++;
4938         } while (buflen != 0);
4939
4940         return rc;
4941
4942 unwind:
4943         for(; rc > 0; rc--)
4944                 __free_page(spages[rc-1]);
4945         return -ENOMEM;
4946 }
4947
4948 struct nfs4_cached_acl {
4949         int cached;
4950         size_t len;
4951         char data[0];
4952 };
4953
4954 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4955 {
4956         struct nfs_inode *nfsi = NFS_I(inode);
4957
4958         spin_lock(&inode->i_lock);
4959         kfree(nfsi->nfs4_acl);
4960         nfsi->nfs4_acl = acl;
4961         spin_unlock(&inode->i_lock);
4962 }
4963
4964 static void nfs4_zap_acl_attr(struct inode *inode)
4965 {
4966         nfs4_set_cached_acl(inode, NULL);
4967 }
4968
4969 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4970 {
4971         struct nfs_inode *nfsi = NFS_I(inode);
4972         struct nfs4_cached_acl *acl;
4973         int ret = -ENOENT;
4974
4975         spin_lock(&inode->i_lock);
4976         acl = nfsi->nfs4_acl;
4977         if (acl == NULL)
4978                 goto out;
4979         if (buf == NULL) /* user is just asking for length */
4980                 goto out_len;
4981         if (acl->cached == 0)
4982                 goto out;
4983         ret = -ERANGE; /* see getxattr(2) man page */
4984         if (acl->len > buflen)
4985                 goto out;
4986         memcpy(buf, acl->data, acl->len);
4987 out_len:
4988         ret = acl->len;
4989 out:
4990         spin_unlock(&inode->i_lock);
4991         return ret;
4992 }
4993
4994 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4995 {
4996         struct nfs4_cached_acl *acl;
4997         size_t buflen = sizeof(*acl) + acl_len;
4998
4999         if (buflen <= PAGE_SIZE) {
5000                 acl = kmalloc(buflen, GFP_KERNEL);
5001                 if (acl == NULL)
5002                         goto out;
5003                 acl->cached = 1;
5004                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5005         } else {
5006                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5007                 if (acl == NULL)
5008                         goto out;
5009                 acl->cached = 0;
5010         }
5011         acl->len = acl_len;
5012 out:
5013         nfs4_set_cached_acl(inode, acl);
5014 }
5015
5016 /*
5017  * The getxattr API returns the required buffer length when called with a
5018  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5019  * the required buf.  On a NULL buf, we send a page of data to the server
5020  * guessing that the ACL request can be serviced by a page. If so, we cache
5021  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5022  * the cache. If not so, we throw away the page, and cache the required
5023  * length. The next getxattr call will then produce another round trip to
5024  * the server, this time with the input buf of the required size.
5025  */
5026 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5027 {
5028         struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
5029         struct nfs_getaclargs args = {
5030                 .fh = NFS_FH(inode),
5031                 .acl_pages = pages,
5032                 .acl_len = buflen,
5033         };
5034         struct nfs_getaclres res = {
5035                 .acl_len = buflen,
5036         };
5037         struct rpc_message msg = {
5038                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5039                 .rpc_argp = &args,
5040                 .rpc_resp = &res,
5041         };
5042         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5043         int ret = -ENOMEM, i;
5044
5045         if (npages > ARRAY_SIZE(pages))
5046                 return -ERANGE;
5047
5048         for (i = 0; i < npages; i++) {
5049                 pages[i] = alloc_page(GFP_KERNEL);
5050                 if (!pages[i])
5051                         goto out_free;
5052         }
5053
5054         /* for decoding across pages */
5055         res.acl_scratch = alloc_page(GFP_KERNEL);
5056         if (!res.acl_scratch)
5057                 goto out_free;
5058
5059         args.acl_len = npages * PAGE_SIZE;
5060
5061         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5062                 __func__, buf, buflen, npages, args.acl_len);
5063         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5064                              &msg, &args.seq_args, &res.seq_res, 0);
5065         if (ret)
5066                 goto out_free;
5067
5068         /* Handle the case where the passed-in buffer is too short */
5069         if (res.acl_flags & NFS4_ACL_TRUNC) {
5070                 /* Did the user only issue a request for the acl length? */
5071                 if (buf == NULL)
5072                         goto out_ok;
5073                 ret = -ERANGE;
5074                 goto out_free;
5075         }
5076         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5077         if (buf) {
5078                 if (res.acl_len > buflen) {
5079                         ret = -ERANGE;
5080                         goto out_free;
5081                 }
5082                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5083         }
5084 out_ok:
5085         ret = res.acl_len;
5086 out_free:
5087         for (i = 0; i < npages; i++)
5088                 if (pages[i])
5089                         __free_page(pages[i]);
5090         if (res.acl_scratch)
5091                 __free_page(res.acl_scratch);
5092         return ret;
5093 }
5094
5095 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5096 {
5097         struct nfs4_exception exception = { };
5098         ssize_t ret;
5099         do {
5100                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5101                 trace_nfs4_get_acl(inode, ret);
5102                 if (ret >= 0)
5103                         break;
5104                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5105         } while (exception.retry);
5106         return ret;
5107 }
5108
5109 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5110 {
5111         struct nfs_server *server = NFS_SERVER(inode);
5112         int ret;
5113
5114         if (!nfs4_server_supports_acls(server))
5115                 return -EOPNOTSUPP;
5116         ret = nfs_revalidate_inode(server, inode);
5117         if (ret < 0)
5118                 return ret;
5119         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5120                 nfs_zap_acl_cache(inode);
5121         ret = nfs4_read_cached_acl(inode, buf, buflen);
5122         if (ret != -ENOENT)
5123                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5124                  * but no cached acl data, just the acl length */
5125                 return ret;
5126         return nfs4_get_acl_uncached(inode, buf, buflen);
5127 }
5128
5129 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5130 {
5131         struct nfs_server *server = NFS_SERVER(inode);
5132         struct page *pages[NFS4ACL_MAXPAGES];
5133         struct nfs_setaclargs arg = {
5134                 .fh             = NFS_FH(inode),
5135                 .acl_pages      = pages,
5136                 .acl_len        = buflen,
5137         };
5138         struct nfs_setaclres res;
5139         struct rpc_message msg = {
5140                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5141                 .rpc_argp       = &arg,
5142                 .rpc_resp       = &res,
5143         };
5144         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5145         int ret, i;
5146
5147         /* You can't remove system.nfs4_acl: */
5148         if (buflen == 0)
5149                 return -EINVAL;
5150         if (!nfs4_server_supports_acls(server))
5151                 return -EOPNOTSUPP;
5152         if (npages > ARRAY_SIZE(pages))
5153                 return -ERANGE;
5154         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5155         if (i < 0)
5156                 return i;
5157         nfs4_inode_return_delegation(inode);
5158         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5159
5160         /*
5161          * Free each page after tx, so the only ref left is
5162          * held by the network stack
5163          */
5164         for (; i > 0; i--)
5165                 put_page(pages[i-1]);
5166
5167         /*
5168          * Acl update can result in inode attribute update.
5169          * so mark the attribute cache invalid.
5170          */
5171         spin_lock(&inode->i_lock);
5172         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5173         spin_unlock(&inode->i_lock);
5174         nfs_access_zap_cache(inode);
5175         nfs_zap_acl_cache(inode);
5176         return ret;
5177 }
5178
5179 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5180 {
5181         struct nfs4_exception exception = { };
5182         int err;
5183         do {
5184                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5185                 trace_nfs4_set_acl(inode, err);
5186                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
5187                         /*
5188                          * no need to retry since the kernel
5189                          * isn't involved in encoding the ACEs.
5190                          */
5191                         err = -EINVAL;
5192                         break;
5193                 }
5194                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5195                                 &exception);
5196         } while (exception.retry);
5197         return err;
5198 }
5199
5200 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5201 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5202                                         size_t buflen)
5203 {
5204         struct nfs_server *server = NFS_SERVER(inode);
5205         struct nfs_fattr fattr;
5206         struct nfs4_label label = {0, 0, buflen, buf};
5207
5208         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5209         struct nfs4_getattr_arg arg = {
5210                 .fh             = NFS_FH(inode),
5211                 .bitmask        = bitmask,
5212         };
5213         struct nfs4_getattr_res res = {
5214                 .fattr          = &fattr,
5215                 .label          = &label,
5216                 .server         = server,
5217         };
5218         struct rpc_message msg = {
5219                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5220                 .rpc_argp       = &arg,
5221                 .rpc_resp       = &res,
5222         };
5223         int ret;
5224
5225         nfs_fattr_init(&fattr);
5226
5227         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5228         if (ret)
5229                 return ret;
5230         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5231                 return -ENOENT;
5232         return label.len;
5233 }
5234
5235 static int nfs4_get_security_label(struct inode *inode, void *buf,
5236                                         size_t buflen)
5237 {
5238         struct nfs4_exception exception = { };
5239         int err;
5240
5241         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5242                 return -EOPNOTSUPP;
5243
5244         do {
5245                 err = _nfs4_get_security_label(inode, buf, buflen);
5246                 trace_nfs4_get_security_label(inode, err);
5247                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5248                                 &exception);
5249         } while (exception.retry);
5250         return err;
5251 }
5252
5253 static int _nfs4_do_set_security_label(struct inode *inode,
5254                 struct nfs4_label *ilabel,
5255                 struct nfs_fattr *fattr,
5256                 struct nfs4_label *olabel)
5257 {
5258
5259         struct iattr sattr = {0};
5260         struct nfs_server *server = NFS_SERVER(inode);
5261         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5262         struct nfs_setattrargs arg = {
5263                 .fh             = NFS_FH(inode),
5264                 .iap            = &sattr,
5265                 .server         = server,
5266                 .bitmask        = bitmask,
5267                 .label          = ilabel,
5268         };
5269         struct nfs_setattrres res = {
5270                 .fattr          = fattr,
5271                 .label          = olabel,
5272                 .server         = server,
5273         };
5274         struct rpc_message msg = {
5275                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5276                 .rpc_argp       = &arg,
5277                 .rpc_resp       = &res,
5278         };
5279         int status;
5280
5281         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5282
5283         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5284         if (status)
5285                 dprintk("%s failed: %d\n", __func__, status);
5286
5287         return status;
5288 }
5289
5290 static int nfs4_do_set_security_label(struct inode *inode,
5291                 struct nfs4_label *ilabel,
5292                 struct nfs_fattr *fattr,
5293                 struct nfs4_label *olabel)
5294 {
5295         struct nfs4_exception exception = { };
5296         int err;
5297
5298         do {
5299                 err = _nfs4_do_set_security_label(inode, ilabel,
5300                                 fattr, olabel);
5301                 trace_nfs4_set_security_label(inode, err);
5302                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5303                                 &exception);
5304         } while (exception.retry);
5305         return err;
5306 }
5307
5308 static int
5309 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5310 {
5311         struct nfs4_label ilabel, *olabel = NULL;
5312         struct nfs_fattr fattr;
5313         struct rpc_cred *cred;
5314         int status;
5315
5316         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5317                 return -EOPNOTSUPP;
5318
5319         nfs_fattr_init(&fattr);
5320
5321         ilabel.pi = 0;
5322         ilabel.lfs = 0;
5323         ilabel.label = (char *)buf;
5324         ilabel.len = buflen;
5325
5326         cred = rpc_lookup_cred();
5327         if (IS_ERR(cred))
5328                 return PTR_ERR(cred);
5329
5330         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5331         if (IS_ERR(olabel)) {
5332                 status = -PTR_ERR(olabel);
5333                 goto out;
5334         }
5335
5336         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5337         if (status == 0)
5338                 nfs_setsecurity(inode, &fattr, olabel);
5339
5340         nfs4_label_free(olabel);
5341 out:
5342         put_rpccred(cred);
5343         return status;
5344 }
5345 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5346
5347
5348 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5349                                     nfs4_verifier *bootverf)
5350 {
5351         __be32 verf[2];
5352
5353         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5354                 /* An impossible timestamp guarantees this value
5355                  * will never match a generated boot time. */
5356                 verf[0] = cpu_to_be32(U32_MAX);
5357                 verf[1] = cpu_to_be32(U32_MAX);
5358         } else {
5359                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5360                 u64 ns = ktime_to_ns(nn->boot_time);
5361
5362                 verf[0] = cpu_to_be32(ns >> 32);
5363                 verf[1] = cpu_to_be32(ns);
5364         }
5365         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5366 }
5367
5368 static int
5369 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5370 {
5371         size_t len;
5372         char *str;
5373
5374         if (clp->cl_owner_id != NULL)
5375                 return 0;
5376
5377         rcu_read_lock();
5378         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5379                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5380                 1 +
5381                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5382                 1;
5383         rcu_read_unlock();
5384
5385         if (len > NFS4_OPAQUE_LIMIT + 1)
5386                 return -EINVAL;
5387
5388         /*
5389          * Since this string is allocated at mount time, and held until the
5390          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5391          * about a memory-reclaim deadlock.
5392          */
5393         str = kmalloc(len, GFP_KERNEL);
5394         if (!str)
5395                 return -ENOMEM;
5396
5397         rcu_read_lock();
5398         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5399                         clp->cl_ipaddr,
5400                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5401                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5402         rcu_read_unlock();
5403
5404         clp->cl_owner_id = str;
5405         return 0;
5406 }
5407
5408 static int
5409 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5410 {
5411         size_t len;
5412         char *str;
5413
5414         len = 10 + 10 + 1 + 10 + 1 +
5415                 strlen(nfs4_client_id_uniquifier) + 1 +
5416                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5417
5418         if (len > NFS4_OPAQUE_LIMIT + 1)
5419                 return -EINVAL;
5420
5421         /*
5422          * Since this string is allocated at mount time, and held until the
5423          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5424          * about a memory-reclaim deadlock.
5425          */
5426         str = kmalloc(len, GFP_KERNEL);
5427         if (!str)
5428                 return -ENOMEM;
5429
5430         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5431                         clp->rpc_ops->version, clp->cl_minorversion,
5432                         nfs4_client_id_uniquifier,
5433                         clp->cl_rpcclient->cl_nodename);
5434         clp->cl_owner_id = str;
5435         return 0;
5436 }
5437
5438 static int
5439 nfs4_init_uniform_client_string(struct nfs_client *clp)
5440 {
5441         size_t len;
5442         char *str;
5443
5444         if (clp->cl_owner_id != NULL)
5445                 return 0;
5446
5447         if (nfs4_client_id_uniquifier[0] != '\0')
5448                 return nfs4_init_uniquifier_client_string(clp);
5449
5450         len = 10 + 10 + 1 + 10 + 1 +
5451                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5452
5453         if (len > NFS4_OPAQUE_LIMIT + 1)
5454                 return -EINVAL;
5455
5456         /*
5457          * Since this string is allocated at mount time, and held until the
5458          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5459          * about a memory-reclaim deadlock.
5460          */
5461         str = kmalloc(len, GFP_KERNEL);
5462         if (!str)
5463                 return -ENOMEM;
5464
5465         scnprintf(str, len, "Linux NFSv%u.%u %s",
5466                         clp->rpc_ops->version, clp->cl_minorversion,
5467                         clp->cl_rpcclient->cl_nodename);
5468         clp->cl_owner_id = str;
5469         return 0;
5470 }
5471
5472 /*
5473  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5474  * services.  Advertise one based on the address family of the
5475  * clientaddr.
5476  */
5477 static unsigned int
5478 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5479 {
5480         if (strchr(clp->cl_ipaddr, ':') != NULL)
5481                 return scnprintf(buf, len, "tcp6");
5482         else
5483                 return scnprintf(buf, len, "tcp");
5484 }
5485
5486 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5487 {
5488         struct nfs4_setclientid *sc = calldata;
5489
5490         if (task->tk_status == 0)
5491                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5492 }
5493
5494 static const struct rpc_call_ops nfs4_setclientid_ops = {
5495         .rpc_call_done = nfs4_setclientid_done,
5496 };
5497
5498 /**
5499  * nfs4_proc_setclientid - Negotiate client ID
5500  * @clp: state data structure
5501  * @program: RPC program for NFSv4 callback service
5502  * @port: IP port number for NFS4 callback service
5503  * @cred: RPC credential to use for this call
5504  * @res: where to place the result
5505  *
5506  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5507  */
5508 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5509                 unsigned short port, struct rpc_cred *cred,
5510                 struct nfs4_setclientid_res *res)
5511 {
5512         nfs4_verifier sc_verifier;
5513         struct nfs4_setclientid setclientid = {
5514                 .sc_verifier = &sc_verifier,
5515                 .sc_prog = program,
5516                 .sc_clnt = clp,
5517         };
5518         struct rpc_message msg = {
5519                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5520                 .rpc_argp = &setclientid,
5521                 .rpc_resp = res,
5522                 .rpc_cred = cred,
5523         };
5524         struct rpc_task *task;
5525         struct rpc_task_setup task_setup_data = {
5526                 .rpc_client = clp->cl_rpcclient,
5527                 .rpc_message = &msg,
5528                 .callback_ops = &nfs4_setclientid_ops,
5529                 .callback_data = &setclientid,
5530                 .flags = RPC_TASK_TIMEOUT,
5531         };
5532         int status;
5533
5534         /* nfs_client_id4 */
5535         nfs4_init_boot_verifier(clp, &sc_verifier);
5536
5537         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5538                 status = nfs4_init_uniform_client_string(clp);
5539         else
5540                 status = nfs4_init_nonuniform_client_string(clp);
5541
5542         if (status)
5543                 goto out;
5544
5545         /* cb_client4 */
5546         setclientid.sc_netid_len =
5547                                 nfs4_init_callback_netid(clp,
5548                                                 setclientid.sc_netid,
5549                                                 sizeof(setclientid.sc_netid));
5550         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5551                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5552                                 clp->cl_ipaddr, port >> 8, port & 255);
5553
5554         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5555                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5556                 clp->cl_owner_id);
5557         task = rpc_run_task(&task_setup_data);
5558         if (IS_ERR(task)) {
5559                 status = PTR_ERR(task);
5560                 goto out;
5561         }
5562         status = task->tk_status;
5563         if (setclientid.sc_cred) {
5564                 kfree(clp->cl_acceptor);
5565                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5566                 put_rpccred(setclientid.sc_cred);
5567         }
5568         rpc_put_task(task);
5569 out:
5570         trace_nfs4_setclientid(clp, status);
5571         dprintk("NFS reply setclientid: %d\n", status);
5572         return status;
5573 }
5574
5575 /**
5576  * nfs4_proc_setclientid_confirm - Confirm client ID
5577  * @clp: state data structure
5578  * @res: result of a previous SETCLIENTID
5579  * @cred: RPC credential to use for this call
5580  *
5581  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5582  */
5583 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5584                 struct nfs4_setclientid_res *arg,
5585                 struct rpc_cred *cred)
5586 {
5587         struct rpc_message msg = {
5588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5589                 .rpc_argp = arg,
5590                 .rpc_cred = cred,
5591         };
5592         int status;
5593
5594         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5595                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5596                 clp->cl_clientid);
5597         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5598         trace_nfs4_setclientid_confirm(clp, status);
5599         dprintk("NFS reply setclientid_confirm: %d\n", status);
5600         return status;
5601 }
5602
5603 struct nfs4_delegreturndata {
5604         struct nfs4_delegreturnargs args;
5605         struct nfs4_delegreturnres res;
5606         struct nfs_fh fh;
5607         nfs4_stateid stateid;
5608         unsigned long timestamp;
5609         struct nfs_fattr fattr;
5610         int rpc_status;
5611         struct inode *inode;
5612         bool roc;
5613         u32 roc_barrier;
5614 };
5615
5616 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5617 {
5618         struct nfs4_delegreturndata *data = calldata;
5619
5620         if (!nfs4_sequence_done(task, &data->res.seq_res))
5621                 return;
5622
5623         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5624         switch (task->tk_status) {
5625         case 0:
5626                 renew_lease(data->res.server, data->timestamp);
5627                 break;
5628         case -NFS4ERR_ADMIN_REVOKED:
5629         case -NFS4ERR_DELEG_REVOKED:
5630         case -NFS4ERR_EXPIRED:
5631                 nfs4_free_revoked_stateid(data->res.server,
5632                                 data->args.stateid,
5633                                 task->tk_msg.rpc_cred);
5634         case -NFS4ERR_BAD_STATEID:
5635         case -NFS4ERR_OLD_STATEID:
5636         case -NFS4ERR_STALE_STATEID:
5637                 task->tk_status = 0;
5638                 break;
5639         default:
5640                 if (nfs4_async_handle_error(task, data->res.server,
5641                                             NULL, NULL) == -EAGAIN) {
5642                         rpc_restart_call_prepare(task);
5643                         return;
5644                 }
5645         }
5646         data->rpc_status = task->tk_status;
5647         if (data->roc && data->rpc_status == 0)
5648                 pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5649 }
5650
5651 static void nfs4_delegreturn_release(void *calldata)
5652 {
5653         struct nfs4_delegreturndata *data = calldata;
5654         struct inode *inode = data->inode;
5655
5656         if (inode) {
5657                 if (data->roc)
5658                         pnfs_roc_release(inode);
5659                 nfs_iput_and_deactive(inode);
5660         }
5661         kfree(calldata);
5662 }
5663
5664 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5665 {
5666         struct nfs4_delegreturndata *d_data;
5667
5668         d_data = (struct nfs4_delegreturndata *)data;
5669
5670         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5671                 return;
5672
5673         if (d_data->roc)
5674                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5675
5676         nfs4_setup_sequence(d_data->res.server,
5677                         &d_data->args.seq_args,
5678                         &d_data->res.seq_res,
5679                         task);
5680 }
5681
5682 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5683         .rpc_call_prepare = nfs4_delegreturn_prepare,
5684         .rpc_call_done = nfs4_delegreturn_done,
5685         .rpc_release = nfs4_delegreturn_release,
5686 };
5687
5688 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5689 {
5690         struct nfs4_delegreturndata *data;
5691         struct nfs_server *server = NFS_SERVER(inode);
5692         struct rpc_task *task;
5693         struct rpc_message msg = {
5694                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5695                 .rpc_cred = cred,
5696         };
5697         struct rpc_task_setup task_setup_data = {
5698                 .rpc_client = server->client,
5699                 .rpc_message = &msg,
5700                 .callback_ops = &nfs4_delegreturn_ops,
5701                 .flags = RPC_TASK_ASYNC,
5702         };
5703         int status = 0;
5704
5705         data = kzalloc(sizeof(*data), GFP_NOFS);
5706         if (data == NULL)
5707                 return -ENOMEM;
5708         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5709
5710         nfs4_state_protect(server->nfs_client,
5711                         NFS_SP4_MACH_CRED_CLEANUP,
5712                         &task_setup_data.rpc_client, &msg);
5713
5714         data->args.fhandle = &data->fh;
5715         data->args.stateid = &data->stateid;
5716         data->args.bitmask = server->cache_consistency_bitmask;
5717         nfs_copy_fh(&data->fh, NFS_FH(inode));
5718         nfs4_stateid_copy(&data->stateid, stateid);
5719         data->res.fattr = &data->fattr;
5720         data->res.server = server;
5721         nfs_fattr_init(data->res.fattr);
5722         data->timestamp = jiffies;
5723         data->rpc_status = 0;
5724         data->inode = nfs_igrab_and_active(inode);
5725         if (data->inode)
5726                 data->roc = nfs4_roc(inode);
5727
5728         task_setup_data.callback_data = data;
5729         msg.rpc_argp = &data->args;
5730         msg.rpc_resp = &data->res;
5731         task = rpc_run_task(&task_setup_data);
5732         if (IS_ERR(task))
5733                 return PTR_ERR(task);
5734         if (!issync)
5735                 goto out;
5736         status = nfs4_wait_for_completion_rpc_task(task);
5737         if (status != 0)
5738                 goto out;
5739         status = data->rpc_status;
5740         if (status == 0)
5741                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5742         else
5743                 nfs_refresh_inode(inode, &data->fattr);
5744 out:
5745         rpc_put_task(task);
5746         return status;
5747 }
5748
5749 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5750 {
5751         struct nfs_server *server = NFS_SERVER(inode);
5752         struct nfs4_exception exception = { };
5753         int err;
5754         do {
5755                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5756                 trace_nfs4_delegreturn(inode, stateid, err);
5757                 switch (err) {
5758                         case -NFS4ERR_STALE_STATEID:
5759                         case -NFS4ERR_EXPIRED:
5760                         case 0:
5761                                 return 0;
5762                 }
5763                 err = nfs4_handle_exception(server, err, &exception);
5764         } while (exception.retry);
5765         return err;
5766 }
5767
5768 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5769 {
5770         struct inode *inode = state->inode;
5771         struct nfs_server *server = NFS_SERVER(inode);
5772         struct nfs_client *clp = server->nfs_client;
5773         struct nfs_lockt_args arg = {
5774                 .fh = NFS_FH(inode),
5775                 .fl = request,
5776         };
5777         struct nfs_lockt_res res = {
5778                 .denied = request,
5779         };
5780         struct rpc_message msg = {
5781                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5782                 .rpc_argp       = &arg,
5783                 .rpc_resp       = &res,
5784                 .rpc_cred       = state->owner->so_cred,
5785         };
5786         struct nfs4_lock_state *lsp;
5787         int status;
5788
5789         arg.lock_owner.clientid = clp->cl_clientid;
5790         status = nfs4_set_lock_state(state, request);
5791         if (status != 0)
5792                 goto out;
5793         lsp = request->fl_u.nfs4_fl.owner;
5794         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5795         arg.lock_owner.s_dev = server->s_dev;
5796         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5797         switch (status) {
5798                 case 0:
5799                         request->fl_type = F_UNLCK;
5800                         break;
5801                 case -NFS4ERR_DENIED:
5802                         status = 0;
5803         }
5804         request->fl_ops->fl_release_private(request);
5805         request->fl_ops = NULL;
5806 out:
5807         return status;
5808 }
5809
5810 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5811 {
5812         struct nfs4_exception exception = { };
5813         int err;
5814
5815         do {
5816                 err = _nfs4_proc_getlk(state, cmd, request);
5817                 trace_nfs4_get_lock(request, state, cmd, err);
5818                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5819                                 &exception);
5820         } while (exception.retry);
5821         return err;
5822 }
5823
5824 struct nfs4_unlockdata {
5825         struct nfs_locku_args arg;
5826         struct nfs_locku_res res;
5827         struct nfs4_lock_state *lsp;
5828         struct nfs_open_context *ctx;
5829         struct file_lock fl;
5830         struct nfs_server *server;
5831         unsigned long timestamp;
5832 };
5833
5834 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5835                 struct nfs_open_context *ctx,
5836                 struct nfs4_lock_state *lsp,
5837                 struct nfs_seqid *seqid)
5838 {
5839         struct nfs4_unlockdata *p;
5840         struct inode *inode = lsp->ls_state->inode;
5841
5842         p = kzalloc(sizeof(*p), GFP_NOFS);
5843         if (p == NULL)
5844                 return NULL;
5845         p->arg.fh = NFS_FH(inode);
5846         p->arg.fl = &p->fl;
5847         p->arg.seqid = seqid;
5848         p->res.seqid = seqid;
5849         p->lsp = lsp;
5850         atomic_inc(&lsp->ls_count);
5851         /* Ensure we don't close file until we're done freeing locks! */
5852         p->ctx = get_nfs_open_context(ctx);
5853         memcpy(&p->fl, fl, sizeof(p->fl));
5854         p->server = NFS_SERVER(inode);
5855         return p;
5856 }
5857
5858 static void nfs4_locku_release_calldata(void *data)
5859 {
5860         struct nfs4_unlockdata *calldata = data;
5861         nfs_free_seqid(calldata->arg.seqid);
5862         nfs4_put_lock_state(calldata->lsp);
5863         put_nfs_open_context(calldata->ctx);
5864         kfree(calldata);
5865 }
5866
5867 static void nfs4_locku_done(struct rpc_task *task, void *data)
5868 {
5869         struct nfs4_unlockdata *calldata = data;
5870
5871         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5872                 return;
5873         switch (task->tk_status) {
5874                 case 0:
5875                         renew_lease(calldata->server, calldata->timestamp);
5876                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5877                         if (nfs4_update_lock_stateid(calldata->lsp,
5878                                         &calldata->res.stateid))
5879                                 break;
5880                 case -NFS4ERR_ADMIN_REVOKED:
5881                 case -NFS4ERR_EXPIRED:
5882                         nfs4_free_revoked_stateid(calldata->server,
5883                                         &calldata->arg.stateid,
5884                                         task->tk_msg.rpc_cred);
5885                 case -NFS4ERR_BAD_STATEID:
5886                 case -NFS4ERR_OLD_STATEID:
5887                 case -NFS4ERR_STALE_STATEID:
5888                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5889                                                 &calldata->lsp->ls_stateid))
5890                                 rpc_restart_call_prepare(task);
5891                         break;
5892                 default:
5893                         if (nfs4_async_handle_error(task, calldata->server,
5894                                                     NULL, NULL) == -EAGAIN)
5895                                 rpc_restart_call_prepare(task);
5896         }
5897         nfs_release_seqid(calldata->arg.seqid);
5898 }
5899
5900 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5901 {
5902         struct nfs4_unlockdata *calldata = data;
5903
5904         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5905                 goto out_wait;
5906         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5907         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5908                 /* Note: exit _without_ running nfs4_locku_done */
5909                 goto out_no_action;
5910         }
5911         calldata->timestamp = jiffies;
5912         if (nfs4_setup_sequence(calldata->server,
5913                                 &calldata->arg.seq_args,
5914                                 &calldata->res.seq_res,
5915                                 task) != 0)
5916                 nfs_release_seqid(calldata->arg.seqid);
5917         return;
5918 out_no_action:
5919         task->tk_action = NULL;
5920 out_wait:
5921         nfs4_sequence_done(task, &calldata->res.seq_res);
5922 }
5923
5924 static const struct rpc_call_ops nfs4_locku_ops = {
5925         .rpc_call_prepare = nfs4_locku_prepare,
5926         .rpc_call_done = nfs4_locku_done,
5927         .rpc_release = nfs4_locku_release_calldata,
5928 };
5929
5930 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5931                 struct nfs_open_context *ctx,
5932                 struct nfs4_lock_state *lsp,
5933                 struct nfs_seqid *seqid)
5934 {
5935         struct nfs4_unlockdata *data;
5936         struct rpc_message msg = {
5937                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5938                 .rpc_cred = ctx->cred,
5939         };
5940         struct rpc_task_setup task_setup_data = {
5941                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5942                 .rpc_message = &msg,
5943                 .callback_ops = &nfs4_locku_ops,
5944                 .workqueue = nfsiod_workqueue,
5945                 .flags = RPC_TASK_ASYNC,
5946         };
5947
5948         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5949                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5950
5951         /* Ensure this is an unlock - when canceling a lock, the
5952          * canceled lock is passed in, and it won't be an unlock.
5953          */
5954         fl->fl_type = F_UNLCK;
5955
5956         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5957         if (data == NULL) {
5958                 nfs_free_seqid(seqid);
5959                 return ERR_PTR(-ENOMEM);
5960         }
5961
5962         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5963         msg.rpc_argp = &data->arg;
5964         msg.rpc_resp = &data->res;
5965         task_setup_data.callback_data = data;
5966         return rpc_run_task(&task_setup_data);
5967 }
5968
5969 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5970 {
5971         struct inode *inode = state->inode;
5972         struct nfs4_state_owner *sp = state->owner;
5973         struct nfs_inode *nfsi = NFS_I(inode);
5974         struct nfs_seqid *seqid;
5975         struct nfs4_lock_state *lsp;
5976         struct rpc_task *task;
5977         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5978         int status = 0;
5979         unsigned char fl_flags = request->fl_flags;
5980
5981         status = nfs4_set_lock_state(state, request);
5982         /* Unlock _before_ we do the RPC call */
5983         request->fl_flags |= FL_EXISTS;
5984         /* Exclude nfs_delegation_claim_locks() */
5985         mutex_lock(&sp->so_delegreturn_mutex);
5986         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5987         down_read(&nfsi->rwsem);
5988         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
5989                 up_read(&nfsi->rwsem);
5990                 mutex_unlock(&sp->so_delegreturn_mutex);
5991                 goto out;
5992         }
5993         up_read(&nfsi->rwsem);
5994         mutex_unlock(&sp->so_delegreturn_mutex);
5995         if (status != 0)
5996                 goto out;
5997         /* Is this a delegated lock? */
5998         lsp = request->fl_u.nfs4_fl.owner;
5999         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6000                 goto out;
6001         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6002         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6003         status = -ENOMEM;
6004         if (IS_ERR(seqid))
6005                 goto out;
6006         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6007         status = PTR_ERR(task);
6008         if (IS_ERR(task))
6009                 goto out;
6010         status = nfs4_wait_for_completion_rpc_task(task);
6011         rpc_put_task(task);
6012 out:
6013         request->fl_flags = fl_flags;
6014         trace_nfs4_unlock(request, state, F_SETLK, status);
6015         return status;
6016 }
6017
6018 struct nfs4_lockdata {
6019         struct nfs_lock_args arg;
6020         struct nfs_lock_res res;
6021         struct nfs4_lock_state *lsp;
6022         struct nfs_open_context *ctx;
6023         struct file_lock fl;
6024         unsigned long timestamp;
6025         int rpc_status;
6026         int cancelled;
6027         struct nfs_server *server;
6028 };
6029
6030 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6031                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6032                 gfp_t gfp_mask)
6033 {
6034         struct nfs4_lockdata *p;
6035         struct inode *inode = lsp->ls_state->inode;
6036         struct nfs_server *server = NFS_SERVER(inode);
6037         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6038
6039         p = kzalloc(sizeof(*p), gfp_mask);
6040         if (p == NULL)
6041                 return NULL;
6042
6043         p->arg.fh = NFS_FH(inode);
6044         p->arg.fl = &p->fl;
6045         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6046         if (IS_ERR(p->arg.open_seqid))
6047                 goto out_free;
6048         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6049         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6050         if (IS_ERR(p->arg.lock_seqid))
6051                 goto out_free_seqid;
6052         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6053         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6054         p->arg.lock_owner.s_dev = server->s_dev;
6055         p->res.lock_seqid = p->arg.lock_seqid;
6056         p->lsp = lsp;
6057         p->server = server;
6058         atomic_inc(&lsp->ls_count);
6059         p->ctx = get_nfs_open_context(ctx);
6060         memcpy(&p->fl, fl, sizeof(p->fl));
6061         return p;
6062 out_free_seqid:
6063         nfs_free_seqid(p->arg.open_seqid);
6064 out_free:
6065         kfree(p);
6066         return NULL;
6067 }
6068
6069 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6070 {
6071         struct nfs4_lockdata *data = calldata;
6072         struct nfs4_state *state = data->lsp->ls_state;
6073
6074         dprintk("%s: begin!\n", __func__);
6075         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6076                 goto out_wait;
6077         /* Do we need to do an open_to_lock_owner? */
6078         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6079                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6080                         goto out_release_lock_seqid;
6081                 }
6082                 nfs4_stateid_copy(&data->arg.open_stateid,
6083                                 &state->open_stateid);
6084                 data->arg.new_lock_owner = 1;
6085                 data->res.open_seqid = data->arg.open_seqid;
6086         } else {
6087                 data->arg.new_lock_owner = 0;
6088                 nfs4_stateid_copy(&data->arg.lock_stateid,
6089                                 &data->lsp->ls_stateid);
6090         }
6091         if (!nfs4_valid_open_stateid(state)) {
6092                 data->rpc_status = -EBADF;
6093                 task->tk_action = NULL;
6094                 goto out_release_open_seqid;
6095         }
6096         data->timestamp = jiffies;
6097         if (nfs4_setup_sequence(data->server,
6098                                 &data->arg.seq_args,
6099                                 &data->res.seq_res,
6100                                 task) == 0)
6101                 return;
6102 out_release_open_seqid:
6103         nfs_release_seqid(data->arg.open_seqid);
6104 out_release_lock_seqid:
6105         nfs_release_seqid(data->arg.lock_seqid);
6106 out_wait:
6107         nfs4_sequence_done(task, &data->res.seq_res);
6108         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6109 }
6110
6111 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6112 {
6113         struct nfs4_lockdata *data = calldata;
6114         struct nfs4_lock_state *lsp = data->lsp;
6115
6116         dprintk("%s: begin!\n", __func__);
6117
6118         if (!nfs4_sequence_done(task, &data->res.seq_res))
6119                 return;
6120
6121         data->rpc_status = task->tk_status;
6122         switch (task->tk_status) {
6123         case 0:
6124                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6125                                 data->timestamp);
6126                 if (data->arg.new_lock) {
6127                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6128                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0) {
6129                                 rpc_restart_call_prepare(task);
6130                                 break;
6131                         }
6132                 }
6133                 if (data->arg.new_lock_owner != 0) {
6134                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6135                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6136                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6137                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6138                         rpc_restart_call_prepare(task);
6139                 break;
6140         case -NFS4ERR_BAD_STATEID:
6141         case -NFS4ERR_OLD_STATEID:
6142         case -NFS4ERR_STALE_STATEID:
6143         case -NFS4ERR_EXPIRED:
6144                 if (data->arg.new_lock_owner != 0) {
6145                         if (!nfs4_stateid_match(&data->arg.open_stateid,
6146                                                 &lsp->ls_state->open_stateid))
6147                                 rpc_restart_call_prepare(task);
6148                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
6149                                                 &lsp->ls_stateid))
6150                                 rpc_restart_call_prepare(task);
6151         }
6152         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
6153 }
6154
6155 static void nfs4_lock_release(void *calldata)
6156 {
6157         struct nfs4_lockdata *data = calldata;
6158
6159         dprintk("%s: begin!\n", __func__);
6160         nfs_free_seqid(data->arg.open_seqid);
6161         if (data->cancelled != 0) {
6162                 struct rpc_task *task;
6163                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
6164                                 data->arg.lock_seqid);
6165                 if (!IS_ERR(task))
6166                         rpc_put_task_async(task);
6167                 dprintk("%s: cancelling lock!\n", __func__);
6168         } else
6169                 nfs_free_seqid(data->arg.lock_seqid);
6170         nfs4_put_lock_state(data->lsp);
6171         put_nfs_open_context(data->ctx);
6172         kfree(data);
6173         dprintk("%s: done!\n", __func__);
6174 }
6175
6176 static const struct rpc_call_ops nfs4_lock_ops = {
6177         .rpc_call_prepare = nfs4_lock_prepare,
6178         .rpc_call_done = nfs4_lock_done,
6179         .rpc_release = nfs4_lock_release,
6180 };
6181
6182 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
6183 {
6184         switch (error) {
6185         case -NFS4ERR_ADMIN_REVOKED:
6186         case -NFS4ERR_EXPIRED:
6187         case -NFS4ERR_BAD_STATEID:
6188                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6189                 if (new_lock_owner != 0 ||
6190                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
6191                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
6192                 break;
6193         case -NFS4ERR_STALE_STATEID:
6194                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
6195                 nfs4_schedule_lease_recovery(server->nfs_client);
6196         };
6197 }
6198
6199 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
6200 {
6201         struct nfs4_lockdata *data;
6202         struct rpc_task *task;
6203         struct rpc_message msg = {
6204                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
6205                 .rpc_cred = state->owner->so_cred,
6206         };
6207         struct rpc_task_setup task_setup_data = {
6208                 .rpc_client = NFS_CLIENT(state->inode),
6209                 .rpc_message = &msg,
6210                 .callback_ops = &nfs4_lock_ops,
6211                 .workqueue = nfsiod_workqueue,
6212                 .flags = RPC_TASK_ASYNC,
6213         };
6214         int ret;
6215
6216         dprintk("%s: begin!\n", __func__);
6217         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
6218                         fl->fl_u.nfs4_fl.owner,
6219                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
6220         if (data == NULL)
6221                 return -ENOMEM;
6222         if (IS_SETLKW(cmd))
6223                 data->arg.block = 1;
6224         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
6225         msg.rpc_argp = &data->arg;
6226         msg.rpc_resp = &data->res;
6227         task_setup_data.callback_data = data;
6228         if (recovery_type > NFS_LOCK_NEW) {
6229                 if (recovery_type == NFS_LOCK_RECLAIM)
6230                         data->arg.reclaim = NFS_LOCK_RECLAIM;
6231                 nfs4_set_sequence_privileged(&data->arg.seq_args);
6232         } else
6233                 data->arg.new_lock = 1;
6234         task = rpc_run_task(&task_setup_data);
6235         if (IS_ERR(task))
6236                 return PTR_ERR(task);
6237         ret = nfs4_wait_for_completion_rpc_task(task);
6238         if (ret == 0) {
6239                 ret = data->rpc_status;
6240                 if (ret)
6241                         nfs4_handle_setlk_error(data->server, data->lsp,
6242                                         data->arg.new_lock_owner, ret);
6243         } else
6244                 data->cancelled = 1;
6245         rpc_put_task(task);
6246         dprintk("%s: done, ret = %d!\n", __func__, ret);
6247         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
6248         return ret;
6249 }
6250
6251 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
6252 {
6253         struct nfs_server *server = NFS_SERVER(state->inode);
6254         struct nfs4_exception exception = {
6255                 .inode = state->inode,
6256         };
6257         int err;
6258
6259         do {
6260                 /* Cache the lock if possible... */
6261                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6262                         return 0;
6263                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
6264                 if (err != -NFS4ERR_DELAY)
6265                         break;
6266                 nfs4_handle_exception(server, err, &exception);
6267         } while (exception.retry);
6268         return err;
6269 }
6270
6271 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
6272 {
6273         struct nfs_server *server = NFS_SERVER(state->inode);
6274         struct nfs4_exception exception = {
6275                 .inode = state->inode,
6276         };
6277         int err;
6278
6279         err = nfs4_set_lock_state(state, request);
6280         if (err != 0)
6281                 return err;
6282         if (!recover_lost_locks) {
6283                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
6284                 return 0;
6285         }
6286         do {
6287                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
6288                         return 0;
6289                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
6290                 switch (err) {
6291                 default:
6292                         goto out;
6293                 case -NFS4ERR_GRACE:
6294                 case -NFS4ERR_DELAY:
6295                         nfs4_handle_exception(server, err, &exception);
6296                         err = 0;
6297                 }
6298         } while (exception.retry);
6299 out:
6300         return err;
6301 }
6302
6303 #if defined(CONFIG_NFS_V4_1)
6304 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
6305 {
6306         struct nfs4_lock_state *lsp;
6307         int status;
6308
6309         status = nfs4_set_lock_state(state, request);
6310         if (status != 0)
6311                 return status;
6312         lsp = request->fl_u.nfs4_fl.owner;
6313         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
6314             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
6315                 return 0;
6316         status = nfs4_lock_expired(state, request);
6317         return status;
6318 }
6319 #endif
6320
6321 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6322 {
6323         struct nfs_inode *nfsi = NFS_I(state->inode);
6324         struct nfs4_state_owner *sp = state->owner;
6325         unsigned char fl_flags = request->fl_flags;
6326         int status;
6327
6328         request->fl_flags |= FL_ACCESS;
6329         status = locks_lock_inode_wait(state->inode, request);
6330         if (status < 0)
6331                 goto out;
6332         mutex_lock(&sp->so_delegreturn_mutex);
6333         down_read(&nfsi->rwsem);
6334         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
6335                 /* Yes: cache locks! */
6336                 /* ...but avoid races with delegation recall... */
6337                 request->fl_flags = fl_flags & ~FL_SLEEP;
6338                 status = locks_lock_inode_wait(state->inode, request);
6339                 up_read(&nfsi->rwsem);
6340                 mutex_unlock(&sp->so_delegreturn_mutex);
6341                 goto out;
6342         }
6343         up_read(&nfsi->rwsem);
6344         mutex_unlock(&sp->so_delegreturn_mutex);
6345         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
6346 out:
6347         request->fl_flags = fl_flags;
6348         return status;
6349 }
6350
6351 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6352 {
6353         struct nfs4_exception exception = {
6354                 .state = state,
6355                 .inode = state->inode,
6356         };
6357         int err;
6358
6359         do {
6360                 err = _nfs4_proc_setlk(state, cmd, request);
6361                 if (err == -NFS4ERR_DENIED)
6362                         err = -EAGAIN;
6363                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
6364                                 err, &exception);
6365         } while (exception.retry);
6366         return err;
6367 }
6368
6369 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
6370 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
6371
6372 static int
6373 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
6374                         struct file_lock *request)
6375 {
6376         int             status = -ERESTARTSYS;
6377         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
6378
6379         while(!signalled()) {
6380                 status = nfs4_proc_setlk(state, cmd, request);
6381                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6382                         break;
6383                 freezable_schedule_timeout_interruptible(timeout);
6384                 timeout *= 2;
6385                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
6386                 status = -ERESTARTSYS;
6387         }
6388         return status;
6389 }
6390
6391 #ifdef CONFIG_NFS_V4_1
6392 struct nfs4_lock_waiter {
6393         struct task_struct      *task;
6394         struct inode            *inode;
6395         struct nfs_lowner       *owner;
6396         bool                    notified;
6397 };
6398
6399 static int
6400 nfs4_wake_lock_waiter(wait_queue_t *wait, unsigned int mode, int flags, void *key)
6401 {
6402         int ret;
6403         struct cb_notify_lock_args *cbnl = key;
6404         struct nfs4_lock_waiter *waiter = wait->private;
6405         struct nfs_lowner       *lowner = &cbnl->cbnl_owner,
6406                                 *wowner = waiter->owner;
6407
6408         /* Only wake if the callback was for the same owner */
6409         if (lowner->clientid != wowner->clientid ||
6410             lowner->id != wowner->id             ||
6411             lowner->s_dev != wowner->s_dev)
6412                 return 0;
6413
6414         /* Make sure it's for the right inode */
6415         if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
6416                 return 0;
6417
6418         waiter->notified = true;
6419
6420         /* override "private" so we can use default_wake_function */
6421         wait->private = waiter->task;
6422         ret = autoremove_wake_function(wait, mode, flags, key);
6423         wait->private = waiter;
6424         return ret;
6425 }
6426
6427 static int
6428 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6429 {
6430         int status = -ERESTARTSYS;
6431         unsigned long flags;
6432         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
6433         struct nfs_server *server = NFS_SERVER(state->inode);
6434         struct nfs_client *clp = server->nfs_client;
6435         wait_queue_head_t *q = &clp->cl_lock_waitq;
6436         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
6437                                     .id = lsp->ls_seqid.owner_id,
6438                                     .s_dev = server->s_dev };
6439         struct nfs4_lock_waiter waiter = { .task  = current,
6440                                            .inode = state->inode,
6441                                            .owner = &owner,
6442                                            .notified = false };
6443         wait_queue_t wait;
6444
6445         /* Don't bother with waitqueue if we don't expect a callback */
6446         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
6447                 return nfs4_retry_setlk_simple(state, cmd, request);
6448
6449         init_wait(&wait);
6450         wait.private = &waiter;
6451         wait.func = nfs4_wake_lock_waiter;
6452         add_wait_queue(q, &wait);
6453
6454         while(!signalled()) {
6455                 status = nfs4_proc_setlk(state, cmd, request);
6456                 if ((status != -EAGAIN) || IS_SETLK(cmd))
6457                         break;
6458
6459                 status = -ERESTARTSYS;
6460                 spin_lock_irqsave(&q->lock, flags);
6461                 if (waiter.notified) {
6462                         spin_unlock_irqrestore(&q->lock, flags);
6463                         continue;
6464                 }
6465                 set_current_state(TASK_INTERRUPTIBLE);
6466                 spin_unlock_irqrestore(&q->lock, flags);
6467
6468                 freezable_schedule_timeout(NFS4_LOCK_MAXTIMEOUT);
6469         }
6470
6471         finish_wait(q, &wait);
6472         return status;
6473 }
6474 #else /* !CONFIG_NFS_V4_1 */
6475 static inline int
6476 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6477 {
6478         return nfs4_retry_setlk_simple(state, cmd, request);
6479 }
6480 #endif
6481
6482 static int
6483 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
6484 {
6485         struct nfs_open_context *ctx;
6486         struct nfs4_state *state;
6487         int status;
6488
6489         /* verify open state */
6490         ctx = nfs_file_open_context(filp);
6491         state = ctx->state;
6492
6493         if (request->fl_start < 0 || request->fl_end < 0)
6494                 return -EINVAL;
6495
6496         if (IS_GETLK(cmd)) {
6497                 if (state != NULL)
6498                         return nfs4_proc_getlk(state, F_GETLK, request);
6499                 return 0;
6500         }
6501
6502         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
6503                 return -EINVAL;
6504
6505         if (request->fl_type == F_UNLCK) {
6506                 if (state != NULL)
6507                         return nfs4_proc_unlck(state, cmd, request);
6508                 return 0;
6509         }
6510
6511         if (state == NULL)
6512                 return -ENOLCK;
6513
6514         if ((request->fl_flags & FL_POSIX) &&
6515             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
6516                 return -ENOLCK;
6517
6518         /*
6519          * Don't rely on the VFS having checked the file open mode,
6520          * since it won't do this for flock() locks.
6521          */
6522         switch (request->fl_type) {
6523         case F_RDLCK:
6524                 if (!(filp->f_mode & FMODE_READ))
6525                         return -EBADF;
6526                 break;
6527         case F_WRLCK:
6528                 if (!(filp->f_mode & FMODE_WRITE))
6529                         return -EBADF;
6530         }
6531
6532         status = nfs4_set_lock_state(state, request);
6533         if (status != 0)
6534                 return status;
6535
6536         return nfs4_retry_setlk(state, cmd, request);
6537 }
6538
6539 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
6540 {
6541         struct nfs_server *server = NFS_SERVER(state->inode);
6542         int err;
6543
6544         err = nfs4_set_lock_state(state, fl);
6545         if (err != 0)
6546                 return err;
6547         do {
6548                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
6549                 if (err != -NFS4ERR_DELAY)
6550                         break;
6551                 ssleep(1);
6552         } while (err == -NFS4ERR_DELAY);
6553         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
6554 }
6555
6556 struct nfs_release_lockowner_data {
6557         struct nfs4_lock_state *lsp;
6558         struct nfs_server *server;
6559         struct nfs_release_lockowner_args args;
6560         struct nfs_release_lockowner_res res;
6561         unsigned long timestamp;
6562 };
6563
6564 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
6565 {
6566         struct nfs_release_lockowner_data *data = calldata;
6567         struct nfs_server *server = data->server;
6568         nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
6569                              &data->args.seq_args, &data->res.seq_res, task);
6570         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6571         data->timestamp = jiffies;
6572 }
6573
6574 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
6575 {
6576         struct nfs_release_lockowner_data *data = calldata;
6577         struct nfs_server *server = data->server;
6578
6579         nfs40_sequence_done(task, &data->res.seq_res);
6580
6581         switch (task->tk_status) {
6582         case 0:
6583                 renew_lease(server, data->timestamp);
6584                 break;
6585         case -NFS4ERR_STALE_CLIENTID:
6586         case -NFS4ERR_EXPIRED:
6587                 nfs4_schedule_lease_recovery(server->nfs_client);
6588                 break;
6589         case -NFS4ERR_LEASE_MOVED:
6590         case -NFS4ERR_DELAY:
6591                 if (nfs4_async_handle_error(task, server,
6592                                             NULL, NULL) == -EAGAIN)
6593                         rpc_restart_call_prepare(task);
6594         }
6595 }
6596
6597 static void nfs4_release_lockowner_release(void *calldata)
6598 {
6599         struct nfs_release_lockowner_data *data = calldata;
6600         nfs4_free_lock_state(data->server, data->lsp);
6601         kfree(calldata);
6602 }
6603
6604 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
6605         .rpc_call_prepare = nfs4_release_lockowner_prepare,
6606         .rpc_call_done = nfs4_release_lockowner_done,
6607         .rpc_release = nfs4_release_lockowner_release,
6608 };
6609
6610 static void
6611 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
6612 {
6613         struct nfs_release_lockowner_data *data;
6614         struct rpc_message msg = {
6615                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
6616         };
6617
6618         if (server->nfs_client->cl_mvops->minor_version != 0)
6619                 return;
6620
6621         data = kmalloc(sizeof(*data), GFP_NOFS);
6622         if (!data)
6623                 return;
6624         data->lsp = lsp;
6625         data->server = server;
6626         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
6627         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
6628         data->args.lock_owner.s_dev = server->s_dev;
6629
6630         msg.rpc_argp = &data->args;
6631         msg.rpc_resp = &data->res;
6632         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
6633         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
6634 }
6635
6636 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
6637
6638 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
6639                                    struct dentry *unused, struct inode *inode,
6640                                    const char *key, const void *buf,
6641                                    size_t buflen, int flags)
6642 {
6643         return nfs4_proc_set_acl(inode, buf, buflen);
6644 }
6645
6646 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
6647                                    struct dentry *unused, struct inode *inode,
6648                                    const char *key, void *buf, size_t buflen)
6649 {
6650         return nfs4_proc_get_acl(inode, buf, buflen);
6651 }
6652
6653 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
6654 {
6655         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
6656 }
6657
6658 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6659
6660 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
6661                                      struct dentry *unused, struct inode *inode,
6662                                      const char *key, const void *buf,
6663                                      size_t buflen, int flags)
6664 {
6665         if (security_ismaclabel(key))
6666                 return nfs4_set_security_label(inode, buf, buflen);
6667
6668         return -EOPNOTSUPP;
6669 }
6670
6671 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
6672                                      struct dentry *unused, struct inode *inode,
6673                                      const char *key, void *buf, size_t buflen)
6674 {
6675         if (security_ismaclabel(key))
6676                 return nfs4_get_security_label(inode, buf, buflen);
6677         return -EOPNOTSUPP;
6678 }
6679
6680 static ssize_t
6681 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6682 {
6683         int len = 0;
6684
6685         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
6686                 len = security_inode_listsecurity(inode, list, list_len);
6687                 if (list_len && len > list_len)
6688                         return -ERANGE;
6689         }
6690         return len;
6691 }
6692
6693 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
6694         .prefix = XATTR_SECURITY_PREFIX,
6695         .get    = nfs4_xattr_get_nfs4_label,
6696         .set    = nfs4_xattr_set_nfs4_label,
6697 };
6698
6699 #else
6700
6701 static ssize_t
6702 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
6703 {
6704         return 0;
6705 }
6706
6707 #endif
6708
6709 /*
6710  * nfs_fhget will use either the mounted_on_fileid or the fileid
6711  */
6712 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
6713 {
6714         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
6715                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
6716               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
6717               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
6718                 return;
6719
6720         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
6721                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
6722         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
6723         fattr->nlink = 2;
6724 }
6725
6726 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6727                                    const struct qstr *name,
6728                                    struct nfs4_fs_locations *fs_locations,
6729                                    struct page *page)
6730 {
6731         struct nfs_server *server = NFS_SERVER(dir);
6732         u32 bitmask[3];
6733         struct nfs4_fs_locations_arg args = {
6734                 .dir_fh = NFS_FH(dir),
6735                 .name = name,
6736                 .page = page,
6737                 .bitmask = bitmask,
6738         };
6739         struct nfs4_fs_locations_res res = {
6740                 .fs_locations = fs_locations,
6741         };
6742         struct rpc_message msg = {
6743                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6744                 .rpc_argp = &args,
6745                 .rpc_resp = &res,
6746         };
6747         int status;
6748
6749         dprintk("%s: start\n", __func__);
6750
6751         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
6752         bitmask[1] = nfs4_fattr_bitmap[1];
6753
6754         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
6755          * is not supported */
6756         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
6757                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
6758         else
6759                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
6760
6761         nfs_fattr_init(&fs_locations->fattr);
6762         fs_locations->server = server;
6763         fs_locations->nlocations = 0;
6764         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
6765         dprintk("%s: returned status = %d\n", __func__, status);
6766         return status;
6767 }
6768
6769 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
6770                            const struct qstr *name,
6771                            struct nfs4_fs_locations *fs_locations,
6772                            struct page *page)
6773 {
6774         struct nfs4_exception exception = { };
6775         int err;
6776         do {
6777                 err = _nfs4_proc_fs_locations(client, dir, name,
6778                                 fs_locations, page);
6779                 trace_nfs4_get_fs_locations(dir, name, err);
6780                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
6781                                 &exception);
6782         } while (exception.retry);
6783         return err;
6784 }
6785
6786 /*
6787  * This operation also signals the server that this client is
6788  * performing migration recovery.  The server can stop returning
6789  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
6790  * appended to this compound to identify the client ID which is
6791  * performing recovery.
6792  */
6793 static int _nfs40_proc_get_locations(struct inode *inode,
6794                                      struct nfs4_fs_locations *locations,
6795                                      struct page *page, struct rpc_cred *cred)
6796 {
6797         struct nfs_server *server = NFS_SERVER(inode);
6798         struct rpc_clnt *clnt = server->client;
6799         u32 bitmask[2] = {
6800                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6801         };
6802         struct nfs4_fs_locations_arg args = {
6803                 .clientid       = server->nfs_client->cl_clientid,
6804                 .fh             = NFS_FH(inode),
6805                 .page           = page,
6806                 .bitmask        = bitmask,
6807                 .migration      = 1,            /* skip LOOKUP */
6808                 .renew          = 1,            /* append RENEW */
6809         };
6810         struct nfs4_fs_locations_res res = {
6811                 .fs_locations   = locations,
6812                 .migration      = 1,
6813                 .renew          = 1,
6814         };
6815         struct rpc_message msg = {
6816                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6817                 .rpc_argp       = &args,
6818                 .rpc_resp       = &res,
6819                 .rpc_cred       = cred,
6820         };
6821         unsigned long now = jiffies;
6822         int status;
6823
6824         nfs_fattr_init(&locations->fattr);
6825         locations->server = server;
6826         locations->nlocations = 0;
6827
6828         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6829         nfs4_set_sequence_privileged(&args.seq_args);
6830         status = nfs4_call_sync_sequence(clnt, server, &msg,
6831                                         &args.seq_args, &res.seq_res);
6832         if (status)
6833                 return status;
6834
6835         renew_lease(server, now);
6836         return 0;
6837 }
6838
6839 #ifdef CONFIG_NFS_V4_1
6840
6841 /*
6842  * This operation also signals the server that this client is
6843  * performing migration recovery.  The server can stop asserting
6844  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
6845  * performing this operation is identified in the SEQUENCE
6846  * operation in this compound.
6847  *
6848  * When the client supports GETATTR(fs_locations_info), it can
6849  * be plumbed in here.
6850  */
6851 static int _nfs41_proc_get_locations(struct inode *inode,
6852                                      struct nfs4_fs_locations *locations,
6853                                      struct page *page, struct rpc_cred *cred)
6854 {
6855         struct nfs_server *server = NFS_SERVER(inode);
6856         struct rpc_clnt *clnt = server->client;
6857         u32 bitmask[2] = {
6858                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
6859         };
6860         struct nfs4_fs_locations_arg args = {
6861                 .fh             = NFS_FH(inode),
6862                 .page           = page,
6863                 .bitmask        = bitmask,
6864                 .migration      = 1,            /* skip LOOKUP */
6865         };
6866         struct nfs4_fs_locations_res res = {
6867                 .fs_locations   = locations,
6868                 .migration      = 1,
6869         };
6870         struct rpc_message msg = {
6871                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
6872                 .rpc_argp       = &args,
6873                 .rpc_resp       = &res,
6874                 .rpc_cred       = cred,
6875         };
6876         int status;
6877
6878         nfs_fattr_init(&locations->fattr);
6879         locations->server = server;
6880         locations->nlocations = 0;
6881
6882         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6883         nfs4_set_sequence_privileged(&args.seq_args);
6884         status = nfs4_call_sync_sequence(clnt, server, &msg,
6885                                         &args.seq_args, &res.seq_res);
6886         if (status == NFS4_OK &&
6887             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
6888                 status = -NFS4ERR_LEASE_MOVED;
6889         return status;
6890 }
6891
6892 #endif  /* CONFIG_NFS_V4_1 */
6893
6894 /**
6895  * nfs4_proc_get_locations - discover locations for a migrated FSID
6896  * @inode: inode on FSID that is migrating
6897  * @locations: result of query
6898  * @page: buffer
6899  * @cred: credential to use for this operation
6900  *
6901  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
6902  * operation failed, or a negative errno if a local error occurred.
6903  *
6904  * On success, "locations" is filled in, but if the server has
6905  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
6906  * asserted.
6907  *
6908  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
6909  * from this client that require migration recovery.
6910  */
6911 int nfs4_proc_get_locations(struct inode *inode,
6912                             struct nfs4_fs_locations *locations,
6913                             struct page *page, struct rpc_cred *cred)
6914 {
6915         struct nfs_server *server = NFS_SERVER(inode);
6916         struct nfs_client *clp = server->nfs_client;
6917         const struct nfs4_mig_recovery_ops *ops =
6918                                         clp->cl_mvops->mig_recovery_ops;
6919         struct nfs4_exception exception = { };
6920         int status;
6921
6922         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
6923                 (unsigned long long)server->fsid.major,
6924                 (unsigned long long)server->fsid.minor,
6925                 clp->cl_hostname);
6926         nfs_display_fhandle(NFS_FH(inode), __func__);
6927
6928         do {
6929                 status = ops->get_locations(inode, locations, page, cred);
6930                 if (status != -NFS4ERR_DELAY)
6931                         break;
6932                 nfs4_handle_exception(server, status, &exception);
6933         } while (exception.retry);
6934         return status;
6935 }
6936
6937 /*
6938  * This operation also signals the server that this client is
6939  * performing "lease moved" recovery.  The server can stop
6940  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
6941  * is appended to this compound to identify the client ID which is
6942  * performing recovery.
6943  */
6944 static int _nfs40_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6945 {
6946         struct nfs_server *server = NFS_SERVER(inode);
6947         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
6948         struct rpc_clnt *clnt = server->client;
6949         struct nfs4_fsid_present_arg args = {
6950                 .fh             = NFS_FH(inode),
6951                 .clientid       = clp->cl_clientid,
6952                 .renew          = 1,            /* append RENEW */
6953         };
6954         struct nfs4_fsid_present_res res = {
6955                 .renew          = 1,
6956         };
6957         struct rpc_message msg = {
6958                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
6959                 .rpc_argp       = &args,
6960                 .rpc_resp       = &res,
6961                 .rpc_cred       = cred,
6962         };
6963         unsigned long now = jiffies;
6964         int status;
6965
6966         res.fh = nfs_alloc_fhandle();
6967         if (res.fh == NULL)
6968                 return -ENOMEM;
6969
6970         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
6971         nfs4_set_sequence_privileged(&args.seq_args);
6972         status = nfs4_call_sync_sequence(clnt, server, &msg,
6973                                                 &args.seq_args, &res.seq_res);
6974         nfs_free_fhandle(res.fh);
6975         if (status)
6976                 return status;
6977
6978         do_renew_lease(clp, now);
6979         return 0;
6980 }
6981
6982 #ifdef CONFIG_NFS_V4_1
6983
6984 /*
6985  * This operation also signals the server that this client is
6986  * performing "lease moved" recovery.  The server can stop asserting
6987  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
6988  * this operation is identified in the SEQUENCE operation in this
6989  * compound.
6990  */
6991 static int _nfs41_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
6992 {
6993         struct nfs_server *server = NFS_SERVER(inode);
6994         struct rpc_clnt *clnt = server->client;
6995         struct nfs4_fsid_present_arg args = {
6996                 .fh             = NFS_FH(inode),
6997         };
6998         struct nfs4_fsid_present_res res = {
6999         };
7000         struct rpc_message msg = {
7001                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7002                 .rpc_argp       = &args,
7003                 .rpc_resp       = &res,
7004                 .rpc_cred       = cred,
7005         };
7006         int status;
7007
7008         res.fh = nfs_alloc_fhandle();
7009         if (res.fh == NULL)
7010                 return -ENOMEM;
7011
7012         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
7013         nfs4_set_sequence_privileged(&args.seq_args);
7014         status = nfs4_call_sync_sequence(clnt, server, &msg,
7015                                                 &args.seq_args, &res.seq_res);
7016         nfs_free_fhandle(res.fh);
7017         if (status == NFS4_OK &&
7018             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7019                 status = -NFS4ERR_LEASE_MOVED;
7020         return status;
7021 }
7022
7023 #endif  /* CONFIG_NFS_V4_1 */
7024
7025 /**
7026  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
7027  * @inode: inode on FSID to check
7028  * @cred: credential to use for this operation
7029  *
7030  * Server indicates whether the FSID is present, moved, or not
7031  * recognized.  This operation is necessary to clear a LEASE_MOVED
7032  * condition for this client ID.
7033  *
7034  * Returns NFS4_OK if the FSID is present on this server,
7035  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
7036  *  NFS4ERR code if some error occurred on the server, or a
7037  *  negative errno if a local failure occurred.
7038  */
7039 int nfs4_proc_fsid_present(struct inode *inode, struct rpc_cred *cred)
7040 {
7041         struct nfs_server *server = NFS_SERVER(inode);
7042         struct nfs_client *clp = server->nfs_client;
7043         const struct nfs4_mig_recovery_ops *ops =
7044                                         clp->cl_mvops->mig_recovery_ops;
7045         struct nfs4_exception exception = { };
7046         int status;
7047
7048         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7049                 (unsigned long long)server->fsid.major,
7050                 (unsigned long long)server->fsid.minor,
7051                 clp->cl_hostname);
7052         nfs_display_fhandle(NFS_FH(inode), __func__);
7053
7054         do {
7055                 status = ops->fsid_present(inode, cred);
7056                 if (status != -NFS4ERR_DELAY)
7057                         break;
7058                 nfs4_handle_exception(server, status, &exception);
7059         } while (exception.retry);
7060         return status;
7061 }
7062
7063 /**
7064  * If 'use_integrity' is true and the state managment nfs_client
7065  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
7066  * and the machine credential as per RFC3530bis and RFC5661 Security
7067  * Considerations sections. Otherwise, just use the user cred with the
7068  * filesystem's rpc_client.
7069  */
7070 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
7071 {
7072         int status;
7073         struct nfs4_secinfo_arg args = {
7074                 .dir_fh = NFS_FH(dir),
7075                 .name   = name,
7076         };
7077         struct nfs4_secinfo_res res = {
7078                 .flavors     = flavors,
7079         };
7080         struct rpc_message msg = {
7081                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
7082                 .rpc_argp = &args,
7083                 .rpc_resp = &res,
7084         };
7085         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
7086         struct rpc_cred *cred = NULL;
7087
7088         if (use_integrity) {
7089                 clnt = NFS_SERVER(dir)->nfs_client->cl_rpcclient;
7090                 cred = nfs4_get_clid_cred(NFS_SERVER(dir)->nfs_client);
7091                 msg.rpc_cred = cred;
7092         }
7093
7094         dprintk("NFS call  secinfo %s\n", name->name);
7095
7096         nfs4_state_protect(NFS_SERVER(dir)->nfs_client,
7097                 NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
7098
7099         status = nfs4_call_sync(clnt, NFS_SERVER(dir), &msg, &args.seq_args,
7100                                 &res.seq_res, 0);
7101         dprintk("NFS reply  secinfo: %d\n", status);
7102
7103         if (cred)
7104                 put_rpccred(cred);
7105
7106         return status;
7107 }
7108
7109 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
7110                       struct nfs4_secinfo_flavors *flavors)
7111 {
7112         struct nfs4_exception exception = { };
7113         int err;
7114         do {
7115                 err = -NFS4ERR_WRONGSEC;
7116
7117                 /* try to use integrity protection with machine cred */
7118                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
7119                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
7120
7121                 /*
7122                  * if unable to use integrity protection, or SECINFO with
7123                  * integrity protection returns NFS4ERR_WRONGSEC (which is
7124                  * disallowed by spec, but exists in deployed servers) use
7125                  * the current filesystem's rpc_client and the user cred.
7126                  */
7127                 if (err == -NFS4ERR_WRONGSEC)
7128                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
7129
7130                 trace_nfs4_secinfo(dir, name, err);
7131                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7132                                 &exception);
7133         } while (exception.retry);
7134         return err;
7135 }
7136
7137 #ifdef CONFIG_NFS_V4_1
7138 /*
7139  * Check the exchange flags returned by the server for invalid flags, having
7140  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
7141  * DS flags set.
7142  */
7143 static int nfs4_check_cl_exchange_flags(u32 flags)
7144 {
7145         if (flags & ~EXCHGID4_FLAG_MASK_R)
7146                 goto out_inval;
7147         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
7148             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
7149                 goto out_inval;
7150         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
7151                 goto out_inval;
7152         return NFS_OK;
7153 out_inval:
7154         return -NFS4ERR_INVAL;
7155 }
7156
7157 static bool
7158 nfs41_same_server_scope(struct nfs41_server_scope *a,
7159                         struct nfs41_server_scope *b)
7160 {
7161         if (a->server_scope_sz == b->server_scope_sz &&
7162             memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0)
7163                 return true;
7164
7165         return false;
7166 }
7167
7168 static void
7169 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
7170 {
7171 }
7172
7173 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
7174         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
7175 };
7176
7177 /*
7178  * nfs4_proc_bind_one_conn_to_session()
7179  *
7180  * The 4.1 client currently uses the same TCP connection for the
7181  * fore and backchannel.
7182  */
7183 static
7184 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
7185                 struct rpc_xprt *xprt,
7186                 struct nfs_client *clp,
7187                 struct rpc_cred *cred)
7188 {
7189         int status;
7190         struct nfs41_bind_conn_to_session_args args = {
7191                 .client = clp,
7192                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
7193         };
7194         struct nfs41_bind_conn_to_session_res res;
7195         struct rpc_message msg = {
7196                 .rpc_proc =
7197                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
7198                 .rpc_argp = &args,
7199                 .rpc_resp = &res,
7200                 .rpc_cred = cred,
7201         };
7202         struct rpc_task_setup task_setup_data = {
7203                 .rpc_client = clnt,
7204                 .rpc_xprt = xprt,
7205                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
7206                 .rpc_message = &msg,
7207                 .flags = RPC_TASK_TIMEOUT,
7208         };
7209         struct rpc_task *task;
7210
7211         dprintk("--> %s\n", __func__);
7212
7213         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
7214         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
7215                 args.dir = NFS4_CDFC4_FORE;
7216
7217         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
7218         if (xprt != rcu_access_pointer(clnt->cl_xprt))
7219                 args.dir = NFS4_CDFC4_FORE;
7220
7221         task = rpc_run_task(&task_setup_data);
7222         if (!IS_ERR(task)) {
7223                 status = task->tk_status;
7224                 rpc_put_task(task);
7225         } else
7226                 status = PTR_ERR(task);
7227         trace_nfs4_bind_conn_to_session(clp, status);
7228         if (status == 0) {
7229                 if (memcmp(res.sessionid.data,
7230                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
7231                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
7232                         status = -EIO;
7233                         goto out;
7234                 }
7235                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
7236                         dprintk("NFS: %s: Unexpected direction from server\n",
7237                                 __func__);
7238                         status = -EIO;
7239                         goto out;
7240                 }
7241                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
7242                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
7243                                 __func__);
7244                         status = -EIO;
7245                         goto out;
7246                 }
7247         }
7248 out:
7249         dprintk("<-- %s status= %d\n", __func__, status);
7250         return status;
7251 }
7252
7253 struct rpc_bind_conn_calldata {
7254         struct nfs_client *clp;
7255         struct rpc_cred *cred;
7256 };
7257
7258 static int
7259 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
7260                 struct rpc_xprt *xprt,
7261                 void *calldata)
7262 {
7263         struct rpc_bind_conn_calldata *p = calldata;
7264
7265         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
7266 }
7267
7268 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, struct rpc_cred *cred)
7269 {
7270         struct rpc_bind_conn_calldata data = {
7271                 .clp = clp,
7272                 .cred = cred,
7273         };
7274         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
7275                         nfs4_proc_bind_conn_to_session_callback, &data);
7276 }
7277
7278 /*
7279  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
7280  * and operations we'd like to see to enable certain features in the allow map
7281  */
7282 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
7283         .how = SP4_MACH_CRED,
7284         .enforce.u.words = {
7285                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7286                       1 << (OP_EXCHANGE_ID - 32) |
7287                       1 << (OP_CREATE_SESSION - 32) |
7288                       1 << (OP_DESTROY_SESSION - 32) |
7289                       1 << (OP_DESTROY_CLIENTID - 32)
7290         },
7291         .allow.u.words = {
7292                 [0] = 1 << (OP_CLOSE) |
7293                       1 << (OP_OPEN_DOWNGRADE) |
7294                       1 << (OP_LOCKU) |
7295                       1 << (OP_DELEGRETURN) |
7296                       1 << (OP_COMMIT),
7297                 [1] = 1 << (OP_SECINFO - 32) |
7298                       1 << (OP_SECINFO_NO_NAME - 32) |
7299                       1 << (OP_LAYOUTRETURN - 32) |
7300                       1 << (OP_TEST_STATEID - 32) |
7301                       1 << (OP_FREE_STATEID - 32) |
7302                       1 << (OP_WRITE - 32)
7303         }
7304 };
7305
7306 /*
7307  * Select the state protection mode for client `clp' given the server results
7308  * from exchange_id in `sp'.
7309  *
7310  * Returns 0 on success, negative errno otherwise.
7311  */
7312 static int nfs4_sp4_select_mode(struct nfs_client *clp,
7313                                  struct nfs41_state_protection *sp)
7314 {
7315         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
7316                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
7317                       1 << (OP_EXCHANGE_ID - 32) |
7318                       1 << (OP_CREATE_SESSION - 32) |
7319                       1 << (OP_DESTROY_SESSION - 32) |
7320                       1 << (OP_DESTROY_CLIENTID - 32)
7321         };
7322         unsigned int i;
7323
7324         if (sp->how == SP4_MACH_CRED) {
7325                 /* Print state protect result */
7326                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
7327                 for (i = 0; i <= LAST_NFS4_OP; i++) {
7328                         if (test_bit(i, sp->enforce.u.longs))
7329                                 dfprintk(MOUNT, "  enforce op %d\n", i);
7330                         if (test_bit(i, sp->allow.u.longs))
7331                                 dfprintk(MOUNT, "  allow op %d\n", i);
7332                 }
7333
7334                 /* make sure nothing is on enforce list that isn't supported */
7335                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
7336                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
7337                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7338                                 return -EINVAL;
7339                         }
7340                 }
7341
7342                 /*
7343                  * Minimal mode - state operations are allowed to use machine
7344                  * credential.  Note this already happens by default, so the
7345                  * client doesn't have to do anything more than the negotiation.
7346                  *
7347                  * NOTE: we don't care if EXCHANGE_ID is in the list -
7348                  *       we're already using the machine cred for exchange_id
7349                  *       and will never use a different cred.
7350                  */
7351                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
7352                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
7353                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
7354                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
7355                         dfprintk(MOUNT, "sp4_mach_cred:\n");
7356                         dfprintk(MOUNT, "  minimal mode enabled\n");
7357                         set_bit(NFS_SP4_MACH_CRED_MINIMAL, &clp->cl_sp4_flags);
7358                 } else {
7359                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
7360                         return -EINVAL;
7361                 }
7362
7363                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
7364                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
7365                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
7366                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
7367                         dfprintk(MOUNT, "  cleanup mode enabled\n");
7368                         set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
7369                 }
7370
7371                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
7372                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
7373                         set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
7374                                 &clp->cl_sp4_flags);
7375                 }
7376
7377                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
7378                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
7379                         dfprintk(MOUNT, "  secinfo mode enabled\n");
7380                         set_bit(NFS_SP4_MACH_CRED_SECINFO, &clp->cl_sp4_flags);
7381                 }
7382
7383                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
7384                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
7385                         dfprintk(MOUNT, "  stateid mode enabled\n");
7386                         set_bit(NFS_SP4_MACH_CRED_STATEID, &clp->cl_sp4_flags);
7387                 }
7388
7389                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
7390                         dfprintk(MOUNT, "  write mode enabled\n");
7391                         set_bit(NFS_SP4_MACH_CRED_WRITE, &clp->cl_sp4_flags);
7392                 }
7393
7394                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
7395                         dfprintk(MOUNT, "  commit mode enabled\n");
7396                         set_bit(NFS_SP4_MACH_CRED_COMMIT, &clp->cl_sp4_flags);
7397                 }
7398         }
7399
7400         return 0;
7401 }
7402
7403 struct nfs41_exchange_id_data {
7404         struct nfs41_exchange_id_res res;
7405         struct nfs41_exchange_id_args args;
7406         struct rpc_xprt *xprt;
7407         int rpc_status;
7408 };
7409
7410 static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
7411 {
7412         struct nfs41_exchange_id_data *cdata =
7413                                         (struct nfs41_exchange_id_data *)data;
7414         struct nfs_client *clp = cdata->args.client;
7415         int status = task->tk_status;
7416
7417         trace_nfs4_exchange_id(clp, status);
7418
7419         if (status == 0)
7420                 status = nfs4_check_cl_exchange_flags(cdata->res.flags);
7421
7422         if (cdata->xprt && status == 0) {
7423                 status = nfs4_detect_session_trunking(clp, &cdata->res,
7424                                                       cdata->xprt);
7425                 goto out;
7426         }
7427
7428         if (status  == 0)
7429                 status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
7430
7431         if (status == 0) {
7432                 clp->cl_clientid = cdata->res.clientid;
7433                 clp->cl_exchange_flags = cdata->res.flags;
7434                 /* 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                         clp->cl_seqid = cdata->res.seqid;
7439                 }
7440
7441                 kfree(clp->cl_serverowner);
7442                 clp->cl_serverowner = cdata->res.server_owner;
7443                 cdata->res.server_owner = NULL;
7444
7445                 /* use the most recent implementation id */
7446                 kfree(clp->cl_implid);
7447                 clp->cl_implid = cdata->res.impl_id;
7448                 cdata->res.impl_id = NULL;
7449
7450                 if (clp->cl_serverscope != NULL &&
7451                     !nfs41_same_server_scope(clp->cl_serverscope,
7452                                         cdata->res.server_scope)) {
7453                         dprintk("%s: server_scope mismatch detected\n",
7454                                 __func__);
7455                         set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
7456                         kfree(clp->cl_serverscope);
7457                         clp->cl_serverscope = NULL;
7458                 }
7459
7460                 if (clp->cl_serverscope == NULL) {
7461                         clp->cl_serverscope = cdata->res.server_scope;
7462                         cdata->res.server_scope = NULL;
7463                 }
7464                 /* Save the EXCHANGE_ID verifier session trunk tests */
7465                 memcpy(clp->cl_confirm.data, cdata->args.verifier.data,
7466                        sizeof(clp->cl_confirm.data));
7467         }
7468 out:
7469         cdata->rpc_status = status;
7470         return;
7471 }
7472
7473 static void nfs4_exchange_id_release(void *data)
7474 {
7475         struct nfs41_exchange_id_data *cdata =
7476                                         (struct nfs41_exchange_id_data *)data;
7477
7478         if (cdata->xprt) {
7479                 xprt_put(cdata->xprt);
7480                 rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
7481         }
7482         nfs_put_client(cdata->args.client);
7483         kfree(cdata->res.impl_id);
7484         kfree(cdata->res.server_scope);
7485         kfree(cdata->res.server_owner);
7486         kfree(cdata);
7487 }
7488
7489 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
7490         .rpc_call_done = nfs4_exchange_id_done,
7491         .rpc_release = nfs4_exchange_id_release,
7492 };
7493
7494 /*
7495  * _nfs4_proc_exchange_id()
7496  *
7497  * Wrapper for EXCHANGE_ID operation.
7498  */
7499 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
7500                         u32 sp4_how, struct rpc_xprt *xprt)
7501 {
7502         struct rpc_message msg = {
7503                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
7504                 .rpc_cred = cred,
7505         };
7506         struct rpc_task_setup task_setup_data = {
7507                 .rpc_client = clp->cl_rpcclient,
7508                 .callback_ops = &nfs4_exchange_id_call_ops,
7509                 .rpc_message = &msg,
7510                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
7511         };
7512         struct nfs41_exchange_id_data *calldata;
7513         struct rpc_task *task;
7514         int status = -EIO;
7515
7516         if (!atomic_inc_not_zero(&clp->cl_count))
7517                 goto out;
7518
7519         status = -ENOMEM;
7520         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
7521         if (!calldata)
7522                 goto out;
7523
7524         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
7525
7526         status = nfs4_init_uniform_client_string(clp);
7527         if (status)
7528                 goto out_calldata;
7529
7530         dprintk("NFS call  exchange_id auth=%s, '%s'\n",
7531                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
7532                 clp->cl_owner_id);
7533
7534         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
7535                                                 GFP_NOFS);
7536         status = -ENOMEM;
7537         if (unlikely(calldata->res.server_owner == NULL))
7538                 goto out_calldata;
7539
7540         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
7541                                         GFP_NOFS);
7542         if (unlikely(calldata->res.server_scope == NULL))
7543                 goto out_server_owner;
7544
7545         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
7546         if (unlikely(calldata->res.impl_id == NULL))
7547                 goto out_server_scope;
7548
7549         switch (sp4_how) {
7550         case SP4_NONE:
7551                 calldata->args.state_protect.how = SP4_NONE;
7552                 break;
7553
7554         case SP4_MACH_CRED:
7555                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
7556                 break;
7557
7558         default:
7559                 /* unsupported! */
7560                 WARN_ON_ONCE(1);
7561                 status = -EINVAL;
7562                 goto out_impl_id;
7563         }
7564         if (xprt) {
7565                 calldata->xprt = xprt;
7566                 task_setup_data.rpc_xprt = xprt;
7567                 task_setup_data.flags =
7568                                 RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
7569                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
7570                                 sizeof(calldata->args.verifier.data));
7571         }
7572         calldata->args.client = clp;
7573 #ifdef CONFIG_NFS_V4_1_MIGRATION
7574         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7575         EXCHGID4_FLAG_BIND_PRINC_STATEID |
7576         EXCHGID4_FLAG_SUPP_MOVED_MIGR,
7577 #else
7578         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
7579         EXCHGID4_FLAG_BIND_PRINC_STATEID,
7580 #endif
7581         msg.rpc_argp = &calldata->args;
7582         msg.rpc_resp = &calldata->res;
7583         task_setup_data.callback_data = calldata;
7584
7585         task = rpc_run_task(&task_setup_data);
7586         if (IS_ERR(task))
7587                 return PTR_ERR(task);
7588
7589         if (!xprt) {
7590                 status = rpc_wait_for_completion_task(task);
7591                 if (!status)
7592                         status = calldata->rpc_status;
7593         } else  /* session trunking test */
7594                 status = calldata->rpc_status;
7595
7596         rpc_put_task(task);
7597 out:
7598         if (clp->cl_implid != NULL)
7599                 dprintk("NFS reply exchange_id: Server Implementation ID: "
7600                         "domain: %s, name: %s, date: %llu,%u\n",
7601                         clp->cl_implid->domain, clp->cl_implid->name,
7602                         clp->cl_implid->date.seconds,
7603                         clp->cl_implid->date.nseconds);
7604         dprintk("NFS reply exchange_id: %d\n", status);
7605         return status;
7606
7607 out_impl_id:
7608         kfree(calldata->res.impl_id);
7609 out_server_scope:
7610         kfree(calldata->res.server_scope);
7611 out_server_owner:
7612         kfree(calldata->res.server_owner);
7613 out_calldata:
7614         kfree(calldata);
7615         nfs_put_client(clp);
7616         goto out;
7617 }
7618
7619 /*
7620  * nfs4_proc_exchange_id()
7621  *
7622  * Returns zero, a negative errno, or a negative NFS4ERR status code.
7623  *
7624  * Since the clientid has expired, all compounds using sessions
7625  * associated with the stale clientid will be returning
7626  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
7627  * be in some phase of session reset.
7628  *
7629  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
7630  */
7631 int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
7632 {
7633         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
7634         int status;
7635
7636         /* try SP4_MACH_CRED if krb5i/p */
7637         if (authflavor == RPC_AUTH_GSS_KRB5I ||
7638             authflavor == RPC_AUTH_GSS_KRB5P) {
7639                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
7640                 if (!status)
7641                         return 0;
7642         }
7643
7644         /* try SP4_NONE */
7645         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
7646 }
7647
7648 /**
7649  * nfs4_test_session_trunk
7650  *
7651  * This is an add_xprt_test() test function called from
7652  * rpc_clnt_setup_test_and_add_xprt.
7653  *
7654  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
7655  * and is dereferrenced in nfs4_exchange_id_release
7656  *
7657  * Upon success, add the new transport to the rpc_clnt
7658  *
7659  * @clnt: struct rpc_clnt to get new transport
7660  * @xprt: the rpc_xprt to test
7661  * @data: call data for _nfs4_proc_exchange_id.
7662  */
7663 int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
7664                             void *data)
7665 {
7666         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
7667         u32 sp4_how;
7668
7669         dprintk("--> %s try %s\n", __func__,
7670                 xprt->address_strings[RPC_DISPLAY_ADDR]);
7671
7672         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
7673
7674         /* Test connection for session trunking. Async exchange_id call */
7675         return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
7676 }
7677 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
7678
7679 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
7680                 struct rpc_cred *cred)
7681 {
7682         struct rpc_message msg = {
7683                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
7684                 .rpc_argp = clp,
7685                 .rpc_cred = cred,
7686         };
7687         int status;
7688
7689         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7690         trace_nfs4_destroy_clientid(clp, status);
7691         if (status)
7692                 dprintk("NFS: Got error %d from the server %s on "
7693                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
7694         return status;
7695 }
7696
7697 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
7698                 struct rpc_cred *cred)
7699 {
7700         unsigned int loop;
7701         int ret;
7702
7703         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
7704                 ret = _nfs4_proc_destroy_clientid(clp, cred);
7705                 switch (ret) {
7706                 case -NFS4ERR_DELAY:
7707                 case -NFS4ERR_CLIENTID_BUSY:
7708                         ssleep(1);
7709                         break;
7710                 default:
7711                         return ret;
7712                 }
7713         }
7714         return 0;
7715 }
7716
7717 int nfs4_destroy_clientid(struct nfs_client *clp)
7718 {
7719         struct rpc_cred *cred;
7720         int ret = 0;
7721
7722         if (clp->cl_mvops->minor_version < 1)
7723                 goto out;
7724         if (clp->cl_exchange_flags == 0)
7725                 goto out;
7726         if (clp->cl_preserve_clid)
7727                 goto out;
7728         cred = nfs4_get_clid_cred(clp);
7729         ret = nfs4_proc_destroy_clientid(clp, cred);
7730         if (cred)
7731                 put_rpccred(cred);
7732         switch (ret) {
7733         case 0:
7734         case -NFS4ERR_STALE_CLIENTID:
7735                 clp->cl_exchange_flags = 0;
7736         }
7737 out:
7738         return ret;
7739 }
7740
7741 struct nfs4_get_lease_time_data {
7742         struct nfs4_get_lease_time_args *args;
7743         struct nfs4_get_lease_time_res *res;
7744         struct nfs_client *clp;
7745 };
7746
7747 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
7748                                         void *calldata)
7749 {
7750         struct nfs4_get_lease_time_data *data =
7751                         (struct nfs4_get_lease_time_data *)calldata;
7752
7753         dprintk("--> %s\n", __func__);
7754         /* just setup sequence, do not trigger session recovery
7755            since we're invoked within one */
7756         nfs41_setup_sequence(data->clp->cl_session,
7757                         &data->args->la_seq_args,
7758                         &data->res->lr_seq_res,
7759                         task);
7760         dprintk("<-- %s\n", __func__);
7761 }
7762
7763 /*
7764  * Called from nfs4_state_manager thread for session setup, so don't recover
7765  * from sequence operation or clientid errors.
7766  */
7767 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
7768 {
7769         struct nfs4_get_lease_time_data *data =
7770                         (struct nfs4_get_lease_time_data *)calldata;
7771
7772         dprintk("--> %s\n", __func__);
7773         if (!nfs41_sequence_done(task, &data->res->lr_seq_res))
7774                 return;
7775         switch (task->tk_status) {
7776         case -NFS4ERR_DELAY:
7777         case -NFS4ERR_GRACE:
7778                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
7779                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
7780                 task->tk_status = 0;
7781                 /* fall through */
7782         case -NFS4ERR_RETRY_UNCACHED_REP:
7783                 rpc_restart_call_prepare(task);
7784                 return;
7785         }
7786         dprintk("<-- %s\n", __func__);
7787 }
7788
7789 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
7790         .rpc_call_prepare = nfs4_get_lease_time_prepare,
7791         .rpc_call_done = nfs4_get_lease_time_done,
7792 };
7793
7794 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
7795 {
7796         struct rpc_task *task;
7797         struct nfs4_get_lease_time_args args;
7798         struct nfs4_get_lease_time_res res = {
7799                 .lr_fsinfo = fsinfo,
7800         };
7801         struct nfs4_get_lease_time_data data = {
7802                 .args = &args,
7803                 .res = &res,
7804                 .clp = clp,
7805         };
7806         struct rpc_message msg = {
7807                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
7808                 .rpc_argp = &args,
7809                 .rpc_resp = &res,
7810         };
7811         struct rpc_task_setup task_setup = {
7812                 .rpc_client = clp->cl_rpcclient,
7813                 .rpc_message = &msg,
7814                 .callback_ops = &nfs4_get_lease_time_ops,
7815                 .callback_data = &data,
7816                 .flags = RPC_TASK_TIMEOUT,
7817         };
7818         int status;
7819
7820         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0);
7821         nfs4_set_sequence_privileged(&args.la_seq_args);
7822         dprintk("--> %s\n", __func__);
7823         task = rpc_run_task(&task_setup);
7824
7825         if (IS_ERR(task))
7826                 status = PTR_ERR(task);
7827         else {
7828                 status = task->tk_status;
7829                 rpc_put_task(task);
7830         }
7831         dprintk("<-- %s return %d\n", __func__, status);
7832
7833         return status;
7834 }
7835
7836 /*
7837  * Initialize the values to be used by the client in CREATE_SESSION
7838  * If nfs4_init_session set the fore channel request and response sizes,
7839  * use them.
7840  *
7841  * Set the back channel max_resp_sz_cached to zero to force the client to
7842  * always set csa_cachethis to FALSE because the current implementation
7843  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
7844  */
7845 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
7846                                     struct rpc_clnt *clnt)
7847 {
7848         unsigned int max_rqst_sz, max_resp_sz;
7849         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
7850
7851         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
7852         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
7853
7854         /* Fore channel attributes */
7855         args->fc_attrs.max_rqst_sz = max_rqst_sz;
7856         args->fc_attrs.max_resp_sz = max_resp_sz;
7857         args->fc_attrs.max_ops = NFS4_MAX_OPS;
7858         args->fc_attrs.max_reqs = max_session_slots;
7859
7860         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
7861                 "max_ops=%u max_reqs=%u\n",
7862                 __func__,
7863                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
7864                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
7865
7866         /* Back channel attributes */
7867         args->bc_attrs.max_rqst_sz = max_bc_payload;
7868         args->bc_attrs.max_resp_sz = max_bc_payload;
7869         args->bc_attrs.max_resp_sz_cached = 0;
7870         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
7871         args->bc_attrs.max_reqs = min_t(unsigned short, max_session_cb_slots, 1);
7872
7873         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
7874                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
7875                 __func__,
7876                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
7877                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
7878                 args->bc_attrs.max_reqs);
7879 }
7880
7881 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
7882                 struct nfs41_create_session_res *res)
7883 {
7884         struct nfs4_channel_attrs *sent = &args->fc_attrs;
7885         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
7886
7887         if (rcvd->max_resp_sz > sent->max_resp_sz)
7888                 return -EINVAL;
7889         /*
7890          * Our requested max_ops is the minimum we need; we're not
7891          * prepared to break up compounds into smaller pieces than that.
7892          * So, no point even trying to continue if the server won't
7893          * cooperate:
7894          */
7895         if (rcvd->max_ops < sent->max_ops)
7896                 return -EINVAL;
7897         if (rcvd->max_reqs == 0)
7898                 return -EINVAL;
7899         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
7900                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
7901         return 0;
7902 }
7903
7904 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
7905                 struct nfs41_create_session_res *res)
7906 {
7907         struct nfs4_channel_attrs *sent = &args->bc_attrs;
7908         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
7909
7910         if (!(res->flags & SESSION4_BACK_CHAN))
7911                 goto out;
7912         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
7913                 return -EINVAL;
7914         if (rcvd->max_resp_sz < sent->max_resp_sz)
7915                 return -EINVAL;
7916         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
7917                 return -EINVAL;
7918         if (rcvd->max_ops > sent->max_ops)
7919                 return -EINVAL;
7920         if (rcvd->max_reqs > sent->max_reqs)
7921                 return -EINVAL;
7922 out:
7923         return 0;
7924 }
7925
7926 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
7927                                      struct nfs41_create_session_res *res)
7928 {
7929         int ret;
7930
7931         ret = nfs4_verify_fore_channel_attrs(args, res);
7932         if (ret)
7933                 return ret;
7934         return nfs4_verify_back_channel_attrs(args, res);
7935 }
7936
7937 static void nfs4_update_session(struct nfs4_session *session,
7938                 struct nfs41_create_session_res *res)
7939 {
7940         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
7941         /* Mark client id and session as being confirmed */
7942         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
7943         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
7944         session->flags = res->flags;
7945         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
7946         if (res->flags & SESSION4_BACK_CHAN)
7947                 memcpy(&session->bc_attrs, &res->bc_attrs,
7948                                 sizeof(session->bc_attrs));
7949 }
7950
7951 static int _nfs4_proc_create_session(struct nfs_client *clp,
7952                 struct rpc_cred *cred)
7953 {
7954         struct nfs4_session *session = clp->cl_session;
7955         struct nfs41_create_session_args args = {
7956                 .client = clp,
7957                 .clientid = clp->cl_clientid,
7958                 .seqid = clp->cl_seqid,
7959                 .cb_program = NFS4_CALLBACK,
7960         };
7961         struct nfs41_create_session_res res;
7962
7963         struct rpc_message msg = {
7964                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
7965                 .rpc_argp = &args,
7966                 .rpc_resp = &res,
7967                 .rpc_cred = cred,
7968         };
7969         int status;
7970
7971         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
7972         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
7973
7974         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
7975         trace_nfs4_create_session(clp, status);
7976
7977         switch (status) {
7978         case -NFS4ERR_STALE_CLIENTID:
7979         case -NFS4ERR_DELAY:
7980         case -ETIMEDOUT:
7981         case -EACCES:
7982         case -EAGAIN:
7983                 goto out;
7984         };
7985
7986         clp->cl_seqid++;
7987         if (!status) {
7988                 /* Verify the session's negotiated channel_attrs values */
7989                 status = nfs4_verify_channel_attrs(&args, &res);
7990                 /* Increment the clientid slot sequence id */
7991                 if (status)
7992                         goto out;
7993                 nfs4_update_session(session, &res);
7994         }
7995 out:
7996         return status;
7997 }
7998
7999 /*
8000  * Issues a CREATE_SESSION operation to the server.
8001  * It is the responsibility of the caller to verify the session is
8002  * expired before calling this routine.
8003  */
8004 int nfs4_proc_create_session(struct nfs_client *clp, struct rpc_cred *cred)
8005 {
8006         int status;
8007         unsigned *ptr;
8008         struct nfs4_session *session = clp->cl_session;
8009
8010         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8011
8012         status = _nfs4_proc_create_session(clp, cred);
8013         if (status)
8014                 goto out;
8015
8016         /* Init or reset the session slot tables */
8017         status = nfs4_setup_session_slot_tables(session);
8018         dprintk("slot table setup returned %d\n", status);
8019         if (status)
8020                 goto out;
8021
8022         ptr = (unsigned *)&session->sess_id.data[0];
8023         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
8024                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
8025 out:
8026         dprintk("<-- %s\n", __func__);
8027         return status;
8028 }
8029
8030 /*
8031  * Issue the over-the-wire RPC DESTROY_SESSION.
8032  * The caller must serialize access to this routine.
8033  */
8034 int nfs4_proc_destroy_session(struct nfs4_session *session,
8035                 struct rpc_cred *cred)
8036 {
8037         struct rpc_message msg = {
8038                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
8039                 .rpc_argp = session,
8040                 .rpc_cred = cred,
8041         };
8042         int status = 0;
8043
8044         dprintk("--> nfs4_proc_destroy_session\n");
8045
8046         /* session is still being setup */
8047         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
8048                 return 0;
8049
8050         status = rpc_call_sync(session->clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
8051         trace_nfs4_destroy_session(session->clp, status);
8052
8053         if (status)
8054                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
8055                         "Session has been destroyed regardless...\n", status);
8056
8057         dprintk("<-- nfs4_proc_destroy_session\n");
8058         return status;
8059 }
8060
8061 /*
8062  * Renew the cl_session lease.
8063  */
8064 struct nfs4_sequence_data {
8065         struct nfs_client *clp;
8066         struct nfs4_sequence_args args;
8067         struct nfs4_sequence_res res;
8068 };
8069
8070 static void nfs41_sequence_release(void *data)
8071 {
8072         struct nfs4_sequence_data *calldata = data;
8073         struct nfs_client *clp = calldata->clp;
8074
8075         if (atomic_read(&clp->cl_count) > 1)
8076                 nfs4_schedule_state_renewal(clp);
8077         nfs_put_client(clp);
8078         kfree(calldata);
8079 }
8080
8081 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8082 {
8083         switch(task->tk_status) {
8084         case -NFS4ERR_DELAY:
8085                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8086                 return -EAGAIN;
8087         default:
8088                 nfs4_schedule_lease_recovery(clp);
8089         }
8090         return 0;
8091 }
8092
8093 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
8094 {
8095         struct nfs4_sequence_data *calldata = data;
8096         struct nfs_client *clp = calldata->clp;
8097
8098         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
8099                 return;
8100
8101         trace_nfs4_sequence(clp, task->tk_status);
8102         if (task->tk_status < 0) {
8103                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
8104                 if (atomic_read(&clp->cl_count) == 1)
8105                         goto out;
8106
8107                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
8108                         rpc_restart_call_prepare(task);
8109                         return;
8110                 }
8111         }
8112         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
8113 out:
8114         dprintk("<-- %s\n", __func__);
8115 }
8116
8117 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
8118 {
8119         struct nfs4_sequence_data *calldata = data;
8120         struct nfs_client *clp = calldata->clp;
8121         struct nfs4_sequence_args *args;
8122         struct nfs4_sequence_res *res;
8123
8124         args = task->tk_msg.rpc_argp;
8125         res = task->tk_msg.rpc_resp;
8126
8127         nfs41_setup_sequence(clp->cl_session, args, res, task);
8128 }
8129
8130 static const struct rpc_call_ops nfs41_sequence_ops = {
8131         .rpc_call_done = nfs41_sequence_call_done,
8132         .rpc_call_prepare = nfs41_sequence_prepare,
8133         .rpc_release = nfs41_sequence_release,
8134 };
8135
8136 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
8137                 struct rpc_cred *cred,
8138                 bool is_privileged)
8139 {
8140         struct nfs4_sequence_data *calldata;
8141         struct rpc_message msg = {
8142                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
8143                 .rpc_cred = cred,
8144         };
8145         struct rpc_task_setup task_setup_data = {
8146                 .rpc_client = clp->cl_rpcclient,
8147                 .rpc_message = &msg,
8148                 .callback_ops = &nfs41_sequence_ops,
8149                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
8150         };
8151
8152         if (!atomic_inc_not_zero(&clp->cl_count))
8153                 return ERR_PTR(-EIO);
8154         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8155         if (calldata == NULL) {
8156                 nfs_put_client(clp);
8157                 return ERR_PTR(-ENOMEM);
8158         }
8159         nfs4_init_sequence(&calldata->args, &calldata->res, 0);
8160         if (is_privileged)
8161                 nfs4_set_sequence_privileged(&calldata->args);
8162         msg.rpc_argp = &calldata->args;
8163         msg.rpc_resp = &calldata->res;
8164         calldata->clp = clp;
8165         task_setup_data.callback_data = calldata;
8166
8167         return rpc_run_task(&task_setup_data);
8168 }
8169
8170 static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
8171 {
8172         struct rpc_task *task;
8173         int ret = 0;
8174
8175         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
8176                 return -EAGAIN;
8177         task = _nfs41_proc_sequence(clp, cred, false);
8178         if (IS_ERR(task))
8179                 ret = PTR_ERR(task);
8180         else
8181                 rpc_put_task_async(task);
8182         dprintk("<-- %s status=%d\n", __func__, ret);
8183         return ret;
8184 }
8185
8186 static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred)
8187 {
8188         struct rpc_task *task;
8189         int ret;
8190
8191         task = _nfs41_proc_sequence(clp, cred, true);
8192         if (IS_ERR(task)) {
8193                 ret = PTR_ERR(task);
8194                 goto out;
8195         }
8196         ret = rpc_wait_for_completion_task(task);
8197         if (!ret)
8198                 ret = task->tk_status;
8199         rpc_put_task(task);
8200 out:
8201         dprintk("<-- %s status=%d\n", __func__, ret);
8202         return ret;
8203 }
8204
8205 struct nfs4_reclaim_complete_data {
8206         struct nfs_client *clp;
8207         struct nfs41_reclaim_complete_args arg;
8208         struct nfs41_reclaim_complete_res res;
8209 };
8210
8211 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
8212 {
8213         struct nfs4_reclaim_complete_data *calldata = data;
8214
8215         nfs41_setup_sequence(calldata->clp->cl_session,
8216                         &calldata->arg.seq_args,
8217                         &calldata->res.seq_res,
8218                         task);
8219 }
8220
8221 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
8222 {
8223         switch(task->tk_status) {
8224         case 0:
8225         case -NFS4ERR_COMPLETE_ALREADY:
8226         case -NFS4ERR_WRONG_CRED: /* What to do here? */
8227                 break;
8228         case -NFS4ERR_DELAY:
8229                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
8230                 /* fall through */
8231         case -NFS4ERR_RETRY_UNCACHED_REP:
8232                 return -EAGAIN;
8233         case -NFS4ERR_BADSESSION:
8234         case -NFS4ERR_DEADSESSION:
8235         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
8236                 nfs4_schedule_session_recovery(clp->cl_session,
8237                                 task->tk_status);
8238                 break;
8239         default:
8240                 nfs4_schedule_lease_recovery(clp);
8241         }
8242         return 0;
8243 }
8244
8245 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
8246 {
8247         struct nfs4_reclaim_complete_data *calldata = data;
8248         struct nfs_client *clp = calldata->clp;
8249         struct nfs4_sequence_res *res = &calldata->res.seq_res;
8250
8251         dprintk("--> %s\n", __func__);
8252         if (!nfs41_sequence_done(task, res))
8253                 return;
8254
8255         trace_nfs4_reclaim_complete(clp, task->tk_status);
8256         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
8257                 rpc_restart_call_prepare(task);
8258                 return;
8259         }
8260         dprintk("<-- %s\n", __func__);
8261 }
8262
8263 static void nfs4_free_reclaim_complete_data(void *data)
8264 {
8265         struct nfs4_reclaim_complete_data *calldata = data;
8266
8267         kfree(calldata);
8268 }
8269
8270 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
8271         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
8272         .rpc_call_done = nfs4_reclaim_complete_done,
8273         .rpc_release = nfs4_free_reclaim_complete_data,
8274 };
8275
8276 /*
8277  * Issue a global reclaim complete.
8278  */
8279 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
8280                 struct rpc_cred *cred)
8281 {
8282         struct nfs4_reclaim_complete_data *calldata;
8283         struct rpc_task *task;
8284         struct rpc_message msg = {
8285                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
8286                 .rpc_cred = cred,
8287         };
8288         struct rpc_task_setup task_setup_data = {
8289                 .rpc_client = clp->cl_rpcclient,
8290                 .rpc_message = &msg,
8291                 .callback_ops = &nfs4_reclaim_complete_call_ops,
8292                 .flags = RPC_TASK_ASYNC,
8293         };
8294         int status = -ENOMEM;
8295
8296         dprintk("--> %s\n", __func__);
8297         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8298         if (calldata == NULL)
8299                 goto out;
8300         calldata->clp = clp;
8301         calldata->arg.one_fs = 0;
8302
8303         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0);
8304         nfs4_set_sequence_privileged(&calldata->arg.seq_args);
8305         msg.rpc_argp = &calldata->arg;
8306         msg.rpc_resp = &calldata->res;
8307         task_setup_data.callback_data = calldata;
8308         task = rpc_run_task(&task_setup_data);
8309         if (IS_ERR(task)) {
8310                 status = PTR_ERR(task);
8311                 goto out;
8312         }
8313         status = nfs4_wait_for_completion_rpc_task(task);
8314         if (status == 0)
8315                 status = task->tk_status;
8316         rpc_put_task(task);
8317 out:
8318         dprintk("<-- %s status=%d\n", __func__, status);
8319         return status;
8320 }
8321
8322 static void
8323 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
8324 {
8325         struct nfs4_layoutget *lgp = calldata;
8326         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
8327         struct nfs4_session *session = nfs4_get_session(server);
8328
8329         dprintk("--> %s\n", __func__);
8330         nfs41_setup_sequence(session, &lgp->args.seq_args,
8331                                 &lgp->res.seq_res, task);
8332         dprintk("<-- %s\n", __func__);
8333 }
8334
8335 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
8336 {
8337         struct nfs4_layoutget *lgp = calldata;
8338
8339         dprintk("--> %s\n", __func__);
8340         nfs41_sequence_process(task, &lgp->res.seq_res);
8341         dprintk("<-- %s\n", __func__);
8342 }
8343
8344 static int
8345 nfs4_layoutget_handle_exception(struct rpc_task *task,
8346                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
8347 {
8348         struct inode *inode = lgp->args.inode;
8349         struct nfs_server *server = NFS_SERVER(inode);
8350         struct pnfs_layout_hdr *lo;
8351         int nfs4err = task->tk_status;
8352         int err, status = 0;
8353         LIST_HEAD(head);
8354
8355         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
8356
8357         switch (nfs4err) {
8358         case 0:
8359                 goto out;
8360
8361         /*
8362          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
8363          * on the file. set tk_status to -ENODATA to tell upper layer to
8364          * retry go inband.
8365          */
8366         case -NFS4ERR_LAYOUTUNAVAILABLE:
8367                 status = -ENODATA;
8368                 goto out;
8369         /*
8370          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
8371          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
8372          */
8373         case -NFS4ERR_BADLAYOUT:
8374                 status = -EOVERFLOW;
8375                 goto out;
8376         /*
8377          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
8378          * (or clients) writing to the same RAID stripe except when
8379          * the minlength argument is 0 (see RFC5661 section 18.43.3).
8380          *
8381          * Treat it like we would RECALLCONFLICT -- we retry for a little
8382          * while, and then eventually give up.
8383          */
8384         case -NFS4ERR_LAYOUTTRYLATER:
8385                 if (lgp->args.minlength == 0) {
8386                         status = -EOVERFLOW;
8387                         goto out;
8388                 }
8389                 status = -EBUSY;
8390                 break;
8391         case -NFS4ERR_RECALLCONFLICT:
8392                 status = -ERECALLCONFLICT;
8393                 break;
8394         case -NFS4ERR_DELEG_REVOKED:
8395         case -NFS4ERR_ADMIN_REVOKED:
8396         case -NFS4ERR_EXPIRED:
8397         case -NFS4ERR_BAD_STATEID:
8398                 exception->timeout = 0;
8399                 spin_lock(&inode->i_lock);
8400                 lo = NFS_I(inode)->layout;
8401                 /* If the open stateid was bad, then recover it. */
8402                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
8403                     nfs4_stateid_match_other(&lgp->args.stateid,
8404                                         &lgp->args.ctx->state->stateid)) {
8405                         spin_unlock(&inode->i_lock);
8406                         exception->state = lgp->args.ctx->state;
8407                         exception->stateid = &lgp->args.stateid;
8408                         break;
8409                 }
8410
8411                 /*
8412                  * Mark the bad layout state as invalid, then retry
8413                  */
8414                 pnfs_mark_layout_stateid_invalid(lo, &head);
8415                 spin_unlock(&inode->i_lock);
8416                 pnfs_free_lseg_list(&head);
8417                 status = -EAGAIN;
8418                 goto out;
8419         }
8420
8421         nfs4_sequence_free_slot(&lgp->res.seq_res);
8422         err = nfs4_handle_exception(server, nfs4err, exception);
8423         if (!status) {
8424                 if (exception->retry)
8425                         status = -EAGAIN;
8426                 else
8427                         status = err;
8428         }
8429 out:
8430         dprintk("<-- %s\n", __func__);
8431         return status;
8432 }
8433
8434 static size_t max_response_pages(struct nfs_server *server)
8435 {
8436         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
8437         return nfs_page_array_len(0, max_resp_sz);
8438 }
8439
8440 static void nfs4_free_pages(struct page **pages, size_t size)
8441 {
8442         int i;
8443
8444         if (!pages)
8445                 return;
8446
8447         for (i = 0; i < size; i++) {
8448                 if (!pages[i])
8449                         break;
8450                 __free_page(pages[i]);
8451         }
8452         kfree(pages);
8453 }
8454
8455 static struct page **nfs4_alloc_pages(size_t size, gfp_t gfp_flags)
8456 {
8457         struct page **pages;
8458         int i;
8459
8460         pages = kcalloc(size, sizeof(struct page *), gfp_flags);
8461         if (!pages) {
8462                 dprintk("%s: can't alloc array of %zu pages\n", __func__, size);
8463                 return NULL;
8464         }
8465
8466         for (i = 0; i < size; i++) {
8467                 pages[i] = alloc_page(gfp_flags);
8468                 if (!pages[i]) {
8469                         dprintk("%s: failed to allocate page\n", __func__);
8470                         nfs4_free_pages(pages, size);
8471                         return NULL;
8472                 }
8473         }
8474
8475         return pages;
8476 }
8477
8478 static void nfs4_layoutget_release(void *calldata)
8479 {
8480         struct nfs4_layoutget *lgp = calldata;
8481         struct inode *inode = lgp->args.inode;
8482         struct nfs_server *server = NFS_SERVER(inode);
8483         size_t max_pages = max_response_pages(server);
8484
8485         dprintk("--> %s\n", __func__);
8486         nfs4_sequence_free_slot(&lgp->res.seq_res);
8487         nfs4_free_pages(lgp->args.layout.pages, max_pages);
8488         pnfs_put_layout_hdr(NFS_I(inode)->layout);
8489         put_nfs_open_context(lgp->args.ctx);
8490         kfree(calldata);
8491         dprintk("<-- %s\n", __func__);
8492 }
8493
8494 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
8495         .rpc_call_prepare = nfs4_layoutget_prepare,
8496         .rpc_call_done = nfs4_layoutget_done,
8497         .rpc_release = nfs4_layoutget_release,
8498 };
8499
8500 struct pnfs_layout_segment *
8501 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags)
8502 {
8503         struct inode *inode = lgp->args.inode;
8504         struct nfs_server *server = NFS_SERVER(inode);
8505         size_t max_pages = max_response_pages(server);
8506         struct rpc_task *task;
8507         struct rpc_message msg = {
8508                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
8509                 .rpc_argp = &lgp->args,
8510                 .rpc_resp = &lgp->res,
8511                 .rpc_cred = lgp->cred,
8512         };
8513         struct rpc_task_setup task_setup_data = {
8514                 .rpc_client = server->client,
8515                 .rpc_message = &msg,
8516                 .callback_ops = &nfs4_layoutget_call_ops,
8517                 .callback_data = lgp,
8518                 .flags = RPC_TASK_ASYNC,
8519         };
8520         struct pnfs_layout_segment *lseg = NULL;
8521         struct nfs4_exception exception = {
8522                 .inode = inode,
8523                 .timeout = *timeout,
8524         };
8525         int status = 0;
8526
8527         dprintk("--> %s\n", __func__);
8528
8529         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
8530         pnfs_get_layout_hdr(NFS_I(inode)->layout);
8531
8532         lgp->args.layout.pages = nfs4_alloc_pages(max_pages, gfp_flags);
8533         if (!lgp->args.layout.pages) {
8534                 nfs4_layoutget_release(lgp);
8535                 return ERR_PTR(-ENOMEM);
8536         }
8537         lgp->args.layout.pglen = max_pages * PAGE_SIZE;
8538
8539         lgp->res.layoutp = &lgp->args.layout;
8540         lgp->res.seq_res.sr_slot = NULL;
8541         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0);
8542
8543         task = rpc_run_task(&task_setup_data);
8544         if (IS_ERR(task))
8545                 return ERR_CAST(task);
8546         status = nfs4_wait_for_completion_rpc_task(task);
8547         if (status == 0) {
8548                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
8549                 *timeout = exception.timeout;
8550         }
8551
8552         trace_nfs4_layoutget(lgp->args.ctx,
8553                         &lgp->args.range,
8554                         &lgp->res.range,
8555                         &lgp->res.stateid,
8556                         status);
8557
8558         /* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
8559         if (status == 0 && lgp->res.layoutp->len)
8560                 lseg = pnfs_layout_process(lgp);
8561         rpc_put_task(task);
8562         dprintk("<-- %s status=%d\n", __func__, status);
8563         if (status)
8564                 return ERR_PTR(status);
8565         return lseg;
8566 }
8567
8568 static void
8569 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
8570 {
8571         struct nfs4_layoutreturn *lrp = calldata;
8572
8573         dprintk("--> %s\n", __func__);
8574         nfs41_setup_sequence(lrp->clp->cl_session,
8575                         &lrp->args.seq_args,
8576                         &lrp->res.seq_res,
8577                         task);
8578 }
8579
8580 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
8581 {
8582         struct nfs4_layoutreturn *lrp = calldata;
8583         struct nfs_server *server;
8584
8585         dprintk("--> %s\n", __func__);
8586
8587         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
8588                 return;
8589
8590         server = NFS_SERVER(lrp->args.inode);
8591         switch (task->tk_status) {
8592         default:
8593                 task->tk_status = 0;
8594         case 0:
8595                 break;
8596         case -NFS4ERR_DELAY:
8597                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
8598                         break;
8599                 nfs4_sequence_free_slot(&lrp->res.seq_res);
8600                 rpc_restart_call_prepare(task);
8601                 return;
8602         }
8603         dprintk("<-- %s\n", __func__);
8604 }
8605
8606 static void nfs4_layoutreturn_release(void *calldata)
8607 {
8608         struct nfs4_layoutreturn *lrp = calldata;
8609         struct pnfs_layout_hdr *lo = lrp->args.layout;
8610         LIST_HEAD(freeme);
8611
8612         dprintk("--> %s\n", __func__);
8613         spin_lock(&lo->plh_inode->i_lock);
8614         if (lrp->res.lrs_present) {
8615                 pnfs_mark_matching_lsegs_invalid(lo, &freeme,
8616                                 &lrp->args.range,
8617                                 be32_to_cpu(lrp->args.stateid.seqid));
8618                 pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
8619         } else
8620                 pnfs_mark_layout_stateid_invalid(lo, &freeme);
8621         pnfs_clear_layoutreturn_waitbit(lo);
8622         spin_unlock(&lo->plh_inode->i_lock);
8623         nfs4_sequence_free_slot(&lrp->res.seq_res);
8624         pnfs_free_lseg_list(&freeme);
8625         pnfs_put_layout_hdr(lrp->args.layout);
8626         nfs_iput_and_deactive(lrp->inode);
8627         kfree(calldata);
8628         dprintk("<-- %s\n", __func__);
8629 }
8630
8631 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
8632         .rpc_call_prepare = nfs4_layoutreturn_prepare,
8633         .rpc_call_done = nfs4_layoutreturn_done,
8634         .rpc_release = nfs4_layoutreturn_release,
8635 };
8636
8637 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
8638 {
8639         struct rpc_task *task;
8640         struct rpc_message msg = {
8641                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
8642                 .rpc_argp = &lrp->args,
8643                 .rpc_resp = &lrp->res,
8644                 .rpc_cred = lrp->cred,
8645         };
8646         struct rpc_task_setup task_setup_data = {
8647                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
8648                 .rpc_message = &msg,
8649                 .callback_ops = &nfs4_layoutreturn_call_ops,
8650                 .callback_data = lrp,
8651         };
8652         int status = 0;
8653
8654         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
8655                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
8656                         &task_setup_data.rpc_client, &msg);
8657
8658         dprintk("--> %s\n", __func__);
8659         if (!sync) {
8660                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
8661                 if (!lrp->inode) {
8662                         nfs4_layoutreturn_release(lrp);
8663                         return -EAGAIN;
8664                 }
8665                 task_setup_data.flags |= RPC_TASK_ASYNC;
8666         }
8667         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1);
8668         task = rpc_run_task(&task_setup_data);
8669         if (IS_ERR(task))
8670                 return PTR_ERR(task);
8671         if (sync)
8672                 status = task->tk_status;
8673         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
8674         dprintk("<-- %s status=%d\n", __func__, status);
8675         rpc_put_task(task);
8676         return status;
8677 }
8678
8679 static int
8680 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
8681                 struct pnfs_device *pdev,
8682                 struct rpc_cred *cred)
8683 {
8684         struct nfs4_getdeviceinfo_args args = {
8685                 .pdev = pdev,
8686                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
8687                         NOTIFY_DEVICEID4_DELETE,
8688         };
8689         struct nfs4_getdeviceinfo_res res = {
8690                 .pdev = pdev,
8691         };
8692         struct rpc_message msg = {
8693                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
8694                 .rpc_argp = &args,
8695                 .rpc_resp = &res,
8696                 .rpc_cred = cred,
8697         };
8698         int status;
8699
8700         dprintk("--> %s\n", __func__);
8701         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
8702         if (res.notification & ~args.notify_types)
8703                 dprintk("%s: unsupported notification\n", __func__);
8704         if (res.notification != args.notify_types)
8705                 pdev->nocache = 1;
8706
8707         dprintk("<-- %s status=%d\n", __func__, status);
8708
8709         return status;
8710 }
8711
8712 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
8713                 struct pnfs_device *pdev,
8714                 struct rpc_cred *cred)
8715 {
8716         struct nfs4_exception exception = { };
8717         int err;
8718
8719         do {
8720                 err = nfs4_handle_exception(server,
8721                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
8722                                         &exception);
8723         } while (exception.retry);
8724         return err;
8725 }
8726 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
8727
8728 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
8729 {
8730         struct nfs4_layoutcommit_data *data = calldata;
8731         struct nfs_server *server = NFS_SERVER(data->args.inode);
8732         struct nfs4_session *session = nfs4_get_session(server);
8733
8734         nfs41_setup_sequence(session,
8735                         &data->args.seq_args,
8736                         &data->res.seq_res,
8737                         task);
8738 }
8739
8740 static void
8741 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
8742 {
8743         struct nfs4_layoutcommit_data *data = calldata;
8744         struct nfs_server *server = NFS_SERVER(data->args.inode);
8745
8746         if (!nfs41_sequence_done(task, &data->res.seq_res))
8747                 return;
8748
8749         switch (task->tk_status) { /* Just ignore these failures */
8750         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
8751         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
8752         case -NFS4ERR_BADLAYOUT:     /* no layout */
8753         case -NFS4ERR_GRACE:        /* loca_recalim always false */
8754                 task->tk_status = 0;
8755         case 0:
8756                 break;
8757         default:
8758                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
8759                         rpc_restart_call_prepare(task);
8760                         return;
8761                 }
8762         }
8763 }
8764
8765 static void nfs4_layoutcommit_release(void *calldata)
8766 {
8767         struct nfs4_layoutcommit_data *data = calldata;
8768
8769         pnfs_cleanup_layoutcommit(data);
8770         nfs_post_op_update_inode_force_wcc(data->args.inode,
8771                                            data->res.fattr);
8772         put_rpccred(data->cred);
8773         nfs_iput_and_deactive(data->inode);
8774         kfree(data);
8775 }
8776
8777 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
8778         .rpc_call_prepare = nfs4_layoutcommit_prepare,
8779         .rpc_call_done = nfs4_layoutcommit_done,
8780         .rpc_release = nfs4_layoutcommit_release,
8781 };
8782
8783 int
8784 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
8785 {
8786         struct rpc_message msg = {
8787                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
8788                 .rpc_argp = &data->args,
8789                 .rpc_resp = &data->res,
8790                 .rpc_cred = data->cred,
8791         };
8792         struct rpc_task_setup task_setup_data = {
8793                 .task = &data->task,
8794                 .rpc_client = NFS_CLIENT(data->args.inode),
8795                 .rpc_message = &msg,
8796                 .callback_ops = &nfs4_layoutcommit_ops,
8797                 .callback_data = data,
8798         };
8799         struct rpc_task *task;
8800         int status = 0;
8801
8802         dprintk("NFS: initiating layoutcommit call. sync %d "
8803                 "lbw: %llu inode %lu\n", sync,
8804                 data->args.lastbytewritten,
8805                 data->args.inode->i_ino);
8806
8807         if (!sync) {
8808                 data->inode = nfs_igrab_and_active(data->args.inode);
8809                 if (data->inode == NULL) {
8810                         nfs4_layoutcommit_release(data);
8811                         return -EAGAIN;
8812                 }
8813                 task_setup_data.flags = RPC_TASK_ASYNC;
8814         }
8815         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
8816         task = rpc_run_task(&task_setup_data);
8817         if (IS_ERR(task))
8818                 return PTR_ERR(task);
8819         if (sync)
8820                 status = task->tk_status;
8821         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
8822         dprintk("%s: status %d\n", __func__, status);
8823         rpc_put_task(task);
8824         return status;
8825 }
8826
8827 /**
8828  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
8829  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
8830  */
8831 static int
8832 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8833                     struct nfs_fsinfo *info,
8834                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8835 {
8836         struct nfs41_secinfo_no_name_args args = {
8837                 .style = SECINFO_STYLE_CURRENT_FH,
8838         };
8839         struct nfs4_secinfo_res res = {
8840                 .flavors = flavors,
8841         };
8842         struct rpc_message msg = {
8843                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
8844                 .rpc_argp = &args,
8845                 .rpc_resp = &res,
8846         };
8847         struct rpc_clnt *clnt = server->client;
8848         struct rpc_cred *cred = NULL;
8849         int status;
8850
8851         if (use_integrity) {
8852                 clnt = server->nfs_client->cl_rpcclient;
8853                 cred = nfs4_get_clid_cred(server->nfs_client);
8854                 msg.rpc_cred = cred;
8855         }
8856
8857         dprintk("--> %s\n", __func__);
8858         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
8859                                 &res.seq_res, 0);
8860         dprintk("<-- %s status=%d\n", __func__, status);
8861
8862         if (cred)
8863                 put_rpccred(cred);
8864
8865         return status;
8866 }
8867
8868 static int
8869 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
8870                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
8871 {
8872         struct nfs4_exception exception = { };
8873         int err;
8874         do {
8875                 /* first try using integrity protection */
8876                 err = -NFS4ERR_WRONGSEC;
8877
8878                 /* try to use integrity protection with machine cred */
8879                 if (_nfs4_is_integrity_protected(server->nfs_client))
8880                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8881                                                           flavors, true);
8882
8883                 /*
8884                  * if unable to use integrity protection, or SECINFO with
8885                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8886                  * disallowed by spec, but exists in deployed servers) use
8887                  * the current filesystem's rpc_client and the user cred.
8888                  */
8889                 if (err == -NFS4ERR_WRONGSEC)
8890                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
8891                                                           flavors, false);
8892
8893                 switch (err) {
8894                 case 0:
8895                 case -NFS4ERR_WRONGSEC:
8896                 case -ENOTSUPP:
8897                         goto out;
8898                 default:
8899                         err = nfs4_handle_exception(server, err, &exception);
8900                 }
8901         } while (exception.retry);
8902 out:
8903         return err;
8904 }
8905
8906 static int
8907 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
8908                     struct nfs_fsinfo *info)
8909 {
8910         int err;
8911         struct page *page;
8912         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
8913         struct nfs4_secinfo_flavors *flavors;
8914         struct nfs4_secinfo4 *secinfo;
8915         int i;
8916
8917         page = alloc_page(GFP_KERNEL);
8918         if (!page) {
8919                 err = -ENOMEM;
8920                 goto out;
8921         }
8922
8923         flavors = page_address(page);
8924         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
8925
8926         /*
8927          * Fall back on "guess and check" method if
8928          * the server doesn't support SECINFO_NO_NAME
8929          */
8930         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
8931                 err = nfs4_find_root_sec(server, fhandle, info);
8932                 goto out_freepage;
8933         }
8934         if (err)
8935                 goto out_freepage;
8936
8937         for (i = 0; i < flavors->num_flavors; i++) {
8938                 secinfo = &flavors->flavors[i];
8939
8940                 switch (secinfo->flavor) {
8941                 case RPC_AUTH_NULL:
8942                 case RPC_AUTH_UNIX:
8943                 case RPC_AUTH_GSS:
8944                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
8945                                         &secinfo->flavor_info);
8946                         break;
8947                 default:
8948                         flavor = RPC_AUTH_MAXFLAVOR;
8949                         break;
8950                 }
8951
8952                 if (!nfs_auth_info_match(&server->auth_info, flavor))
8953                         flavor = RPC_AUTH_MAXFLAVOR;
8954
8955                 if (flavor != RPC_AUTH_MAXFLAVOR) {
8956                         err = nfs4_lookup_root_sec(server, fhandle,
8957                                                    info, flavor);
8958                         if (!err)
8959                                 break;
8960                 }
8961         }
8962
8963         if (flavor == RPC_AUTH_MAXFLAVOR)
8964                 err = -EPERM;
8965
8966 out_freepage:
8967         put_page(page);
8968         if (err == -EACCES)
8969                 return -EPERM;
8970 out:
8971         return err;
8972 }
8973
8974 static int _nfs41_test_stateid(struct nfs_server *server,
8975                 nfs4_stateid *stateid,
8976                 struct rpc_cred *cred)
8977 {
8978         int status;
8979         struct nfs41_test_stateid_args args = {
8980                 .stateid = stateid,
8981         };
8982         struct nfs41_test_stateid_res res;
8983         struct rpc_message msg = {
8984                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
8985                 .rpc_argp = &args,
8986                 .rpc_resp = &res,
8987                 .rpc_cred = cred,
8988         };
8989         struct rpc_clnt *rpc_client = server->client;
8990
8991         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
8992                 &rpc_client, &msg);
8993
8994         dprintk("NFS call  test_stateid %p\n", stateid);
8995         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0);
8996         nfs4_set_sequence_privileged(&args.seq_args);
8997         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
8998                         &args.seq_args, &res.seq_res);
8999         if (status != NFS_OK) {
9000                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9001                 return status;
9002         }
9003         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9004         return -res.status;
9005 }
9006
9007 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9008                 int err, struct nfs4_exception *exception)
9009 {
9010         exception->retry = 0;
9011         switch(err) {
9012         case -NFS4ERR_DELAY:
9013         case -NFS4ERR_RETRY_UNCACHED_REP:
9014                 nfs4_handle_exception(server, err, exception);
9015                 break;
9016         case -NFS4ERR_BADSESSION:
9017         case -NFS4ERR_BADSLOT:
9018         case -NFS4ERR_BAD_HIGH_SLOT:
9019         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9020         case -NFS4ERR_DEADSESSION:
9021                 nfs4_do_handle_exception(server, err, exception);
9022         }
9023 }
9024
9025 /**
9026  * nfs41_test_stateid - perform a TEST_STATEID operation
9027  *
9028  * @server: server / transport on which to perform the operation
9029  * @stateid: state ID to test
9030  * @cred: credential
9031  *
9032  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9033  * Otherwise a negative NFS4ERR value is returned if the operation
9034  * failed or the state ID is not currently valid.
9035  */
9036 static int nfs41_test_stateid(struct nfs_server *server,
9037                 nfs4_stateid *stateid,
9038                 struct rpc_cred *cred)
9039 {
9040         struct nfs4_exception exception = { };
9041         int err;
9042         do {
9043                 err = _nfs41_test_stateid(server, stateid, cred);
9044                 nfs4_handle_delay_or_session_error(server, err, &exception);
9045         } while (exception.retry);
9046         return err;
9047 }
9048
9049 struct nfs_free_stateid_data {
9050         struct nfs_server *server;
9051         struct nfs41_free_stateid_args args;
9052         struct nfs41_free_stateid_res res;
9053 };
9054
9055 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
9056 {
9057         struct nfs_free_stateid_data *data = calldata;
9058         nfs41_setup_sequence(nfs4_get_session(data->server),
9059                         &data->args.seq_args,
9060                         &data->res.seq_res,
9061                         task);
9062 }
9063
9064 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
9065 {
9066         struct nfs_free_stateid_data *data = calldata;
9067
9068         nfs41_sequence_done(task, &data->res.seq_res);
9069
9070         switch (task->tk_status) {
9071         case -NFS4ERR_DELAY:
9072                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
9073                         rpc_restart_call_prepare(task);
9074         }
9075 }
9076
9077 static void nfs41_free_stateid_release(void *calldata)
9078 {
9079         kfree(calldata);
9080 }
9081
9082 static const struct rpc_call_ops nfs41_free_stateid_ops = {
9083         .rpc_call_prepare = nfs41_free_stateid_prepare,
9084         .rpc_call_done = nfs41_free_stateid_done,
9085         .rpc_release = nfs41_free_stateid_release,
9086 };
9087
9088 static struct rpc_task *_nfs41_free_stateid(struct nfs_server *server,
9089                 const nfs4_stateid *stateid,
9090                 struct rpc_cred *cred,
9091                 bool privileged)
9092 {
9093         struct rpc_message msg = {
9094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
9095                 .rpc_cred = cred,
9096         };
9097         struct rpc_task_setup task_setup = {
9098                 .rpc_client = server->client,
9099                 .rpc_message = &msg,
9100                 .callback_ops = &nfs41_free_stateid_ops,
9101                 .flags = RPC_TASK_ASYNC,
9102         };
9103         struct nfs_free_stateid_data *data;
9104
9105         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9106                 &task_setup.rpc_client, &msg);
9107
9108         dprintk("NFS call  free_stateid %p\n", stateid);
9109         data = kmalloc(sizeof(*data), GFP_NOFS);
9110         if (!data)
9111                 return ERR_PTR(-ENOMEM);
9112         data->server = server;
9113         nfs4_stateid_copy(&data->args.stateid, stateid);
9114
9115         task_setup.callback_data = data;
9116
9117         msg.rpc_argp = &data->args;
9118         msg.rpc_resp = &data->res;
9119         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
9120         if (privileged)
9121                 nfs4_set_sequence_privileged(&data->args.seq_args);
9122
9123         return rpc_run_task(&task_setup);
9124 }
9125
9126 /**
9127  * nfs41_free_stateid - perform a FREE_STATEID operation
9128  *
9129  * @server: server / transport on which to perform the operation
9130  * @stateid: state ID to release
9131  * @cred: credential
9132  * @is_recovery: set to true if this call needs to be privileged
9133  *
9134  * Note: this function is always asynchronous.
9135  */
9136 static int nfs41_free_stateid(struct nfs_server *server,
9137                 const nfs4_stateid *stateid,
9138                 struct rpc_cred *cred,
9139                 bool is_recovery)
9140 {
9141         struct rpc_task *task;
9142
9143         task = _nfs41_free_stateid(server, stateid, cred, is_recovery);
9144         if (IS_ERR(task))
9145                 return PTR_ERR(task);
9146         rpc_put_task(task);
9147         return 0;
9148 }
9149
9150 static void
9151 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
9152 {
9153         struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
9154
9155         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
9156         nfs4_free_lock_state(server, lsp);
9157 }
9158
9159 static bool nfs41_match_stateid(const nfs4_stateid *s1,
9160                 const nfs4_stateid *s2)
9161 {
9162         if (s1->type != s2->type)
9163                 return false;
9164
9165         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
9166                 return false;
9167
9168         if (s1->seqid == s2->seqid)
9169                 return true;
9170         if (s1->seqid == 0 || s2->seqid == 0)
9171                 return true;
9172
9173         return false;
9174 }
9175
9176 #endif /* CONFIG_NFS_V4_1 */
9177
9178 static bool nfs4_match_stateid(const nfs4_stateid *s1,
9179                 const nfs4_stateid *s2)
9180 {
9181         return nfs4_stateid_match(s1, s2);
9182 }
9183
9184
9185 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
9186         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9187         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9188         .recover_open   = nfs4_open_reclaim,
9189         .recover_lock   = nfs4_lock_reclaim,
9190         .establish_clid = nfs4_init_clientid,
9191         .detect_trunking = nfs40_discover_server_trunking,
9192 };
9193
9194 #if defined(CONFIG_NFS_V4_1)
9195 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
9196         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
9197         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
9198         .recover_open   = nfs4_open_reclaim,
9199         .recover_lock   = nfs4_lock_reclaim,
9200         .establish_clid = nfs41_init_clientid,
9201         .reclaim_complete = nfs41_proc_reclaim_complete,
9202         .detect_trunking = nfs41_discover_server_trunking,
9203 };
9204 #endif /* CONFIG_NFS_V4_1 */
9205
9206 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
9207         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9208         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9209         .recover_open   = nfs40_open_expired,
9210         .recover_lock   = nfs4_lock_expired,
9211         .establish_clid = nfs4_init_clientid,
9212 };
9213
9214 #if defined(CONFIG_NFS_V4_1)
9215 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
9216         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
9217         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
9218         .recover_open   = nfs41_open_expired,
9219         .recover_lock   = nfs41_lock_expired,
9220         .establish_clid = nfs41_init_clientid,
9221 };
9222 #endif /* CONFIG_NFS_V4_1 */
9223
9224 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
9225         .sched_state_renewal = nfs4_proc_async_renew,
9226         .get_state_renewal_cred_locked = nfs4_get_renew_cred_locked,
9227         .renew_lease = nfs4_proc_renew,
9228 };
9229
9230 #if defined(CONFIG_NFS_V4_1)
9231 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
9232         .sched_state_renewal = nfs41_proc_async_sequence,
9233         .get_state_renewal_cred_locked = nfs4_get_machine_cred_locked,
9234         .renew_lease = nfs4_proc_sequence,
9235 };
9236 #endif
9237
9238 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
9239         .get_locations = _nfs40_proc_get_locations,
9240         .fsid_present = _nfs40_proc_fsid_present,
9241 };
9242
9243 #if defined(CONFIG_NFS_V4_1)
9244 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
9245         .get_locations = _nfs41_proc_get_locations,
9246         .fsid_present = _nfs41_proc_fsid_present,
9247 };
9248 #endif  /* CONFIG_NFS_V4_1 */
9249
9250 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
9251         .minor_version = 0,
9252         .init_caps = NFS_CAP_READDIRPLUS
9253                 | NFS_CAP_ATOMIC_OPEN
9254                 | NFS_CAP_POSIX_LOCK,
9255         .init_client = nfs40_init_client,
9256         .shutdown_client = nfs40_shutdown_client,
9257         .match_stateid = nfs4_match_stateid,
9258         .find_root_sec = nfs4_find_root_sec,
9259         .free_lock_state = nfs4_release_lockowner,
9260         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
9261         .alloc_seqid = nfs_alloc_seqid,
9262         .call_sync_ops = &nfs40_call_sync_ops,
9263         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
9264         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
9265         .state_renewal_ops = &nfs40_state_renewal_ops,
9266         .mig_recovery_ops = &nfs40_mig_recovery_ops,
9267 };
9268
9269 #if defined(CONFIG_NFS_V4_1)
9270 static struct nfs_seqid *
9271 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
9272 {
9273         return NULL;
9274 }
9275
9276 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
9277         .minor_version = 1,
9278         .init_caps = NFS_CAP_READDIRPLUS
9279                 | NFS_CAP_ATOMIC_OPEN
9280                 | NFS_CAP_POSIX_LOCK
9281                 | NFS_CAP_STATEID_NFSV41
9282                 | NFS_CAP_ATOMIC_OPEN_V1,
9283         .init_client = nfs41_init_client,
9284         .shutdown_client = nfs41_shutdown_client,
9285         .match_stateid = nfs41_match_stateid,
9286         .find_root_sec = nfs41_find_root_sec,
9287         .free_lock_state = nfs41_free_lock_state,
9288         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9289         .alloc_seqid = nfs_alloc_no_seqid,
9290         .session_trunk = nfs4_test_session_trunk,
9291         .call_sync_ops = &nfs41_call_sync_ops,
9292         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9293         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9294         .state_renewal_ops = &nfs41_state_renewal_ops,
9295         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9296 };
9297 #endif
9298
9299 #if defined(CONFIG_NFS_V4_2)
9300 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
9301         .minor_version = 2,
9302         .init_caps = NFS_CAP_READDIRPLUS
9303                 | NFS_CAP_ATOMIC_OPEN
9304                 | NFS_CAP_POSIX_LOCK
9305                 | NFS_CAP_STATEID_NFSV41
9306                 | NFS_CAP_ATOMIC_OPEN_V1
9307                 | NFS_CAP_ALLOCATE
9308                 | NFS_CAP_COPY
9309                 | NFS_CAP_DEALLOCATE
9310                 | NFS_CAP_SEEK
9311                 | NFS_CAP_LAYOUTSTATS
9312                 | NFS_CAP_CLONE,
9313         .init_client = nfs41_init_client,
9314         .shutdown_client = nfs41_shutdown_client,
9315         .match_stateid = nfs41_match_stateid,
9316         .find_root_sec = nfs41_find_root_sec,
9317         .free_lock_state = nfs41_free_lock_state,
9318         .call_sync_ops = &nfs41_call_sync_ops,
9319         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
9320         .alloc_seqid = nfs_alloc_no_seqid,
9321         .session_trunk = nfs4_test_session_trunk,
9322         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
9323         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
9324         .state_renewal_ops = &nfs41_state_renewal_ops,
9325         .mig_recovery_ops = &nfs41_mig_recovery_ops,
9326 };
9327 #endif
9328
9329 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
9330         [0] = &nfs_v4_0_minor_ops,
9331 #if defined(CONFIG_NFS_V4_1)
9332         [1] = &nfs_v4_1_minor_ops,
9333 #endif
9334 #if defined(CONFIG_NFS_V4_2)
9335         [2] = &nfs_v4_2_minor_ops,
9336 #endif
9337 };
9338
9339 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
9340 {
9341         ssize_t error, error2;
9342
9343         error = generic_listxattr(dentry, list, size);
9344         if (error < 0)
9345                 return error;
9346         if (list) {
9347                 list += error;
9348                 size -= error;
9349         }
9350
9351         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
9352         if (error2 < 0)
9353                 return error2;
9354         return error + error2;
9355 }
9356
9357 static const struct inode_operations nfs4_dir_inode_operations = {
9358         .create         = nfs_create,
9359         .lookup         = nfs_lookup,
9360         .atomic_open    = nfs_atomic_open,
9361         .link           = nfs_link,
9362         .unlink         = nfs_unlink,
9363         .symlink        = nfs_symlink,
9364         .mkdir          = nfs_mkdir,
9365         .rmdir          = nfs_rmdir,
9366         .mknod          = nfs_mknod,
9367         .rename         = nfs_rename,
9368         .permission     = nfs_permission,
9369         .getattr        = nfs_getattr,
9370         .setattr        = nfs_setattr,
9371         .listxattr      = nfs4_listxattr,
9372 };
9373
9374 static const struct inode_operations nfs4_file_inode_operations = {
9375         .permission     = nfs_permission,
9376         .getattr        = nfs_getattr,
9377         .setattr        = nfs_setattr,
9378         .listxattr      = nfs4_listxattr,
9379 };
9380
9381 const struct nfs_rpc_ops nfs_v4_clientops = {
9382         .version        = 4,                    /* protocol version */
9383         .dentry_ops     = &nfs4_dentry_operations,
9384         .dir_inode_ops  = &nfs4_dir_inode_operations,
9385         .file_inode_ops = &nfs4_file_inode_operations,
9386         .file_ops       = &nfs4_file_operations,
9387         .getroot        = nfs4_proc_get_root,
9388         .submount       = nfs4_submount,
9389         .try_mount      = nfs4_try_mount,
9390         .getattr        = nfs4_proc_getattr,
9391         .setattr        = nfs4_proc_setattr,
9392         .lookup         = nfs4_proc_lookup,
9393         .access         = nfs4_proc_access,
9394         .readlink       = nfs4_proc_readlink,
9395         .create         = nfs4_proc_create,
9396         .remove         = nfs4_proc_remove,
9397         .unlink_setup   = nfs4_proc_unlink_setup,
9398         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
9399         .unlink_done    = nfs4_proc_unlink_done,
9400         .rename_setup   = nfs4_proc_rename_setup,
9401         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
9402         .rename_done    = nfs4_proc_rename_done,
9403         .link           = nfs4_proc_link,
9404         .symlink        = nfs4_proc_symlink,
9405         .mkdir          = nfs4_proc_mkdir,
9406         .rmdir          = nfs4_proc_remove,
9407         .readdir        = nfs4_proc_readdir,
9408         .mknod          = nfs4_proc_mknod,
9409         .statfs         = nfs4_proc_statfs,
9410         .fsinfo         = nfs4_proc_fsinfo,
9411         .pathconf       = nfs4_proc_pathconf,
9412         .set_capabilities = nfs4_server_capabilities,
9413         .decode_dirent  = nfs4_decode_dirent,
9414         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
9415         .read_setup     = nfs4_proc_read_setup,
9416         .read_done      = nfs4_read_done,
9417         .write_setup    = nfs4_proc_write_setup,
9418         .write_done     = nfs4_write_done,
9419         .commit_setup   = nfs4_proc_commit_setup,
9420         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
9421         .commit_done    = nfs4_commit_done,
9422         .lock           = nfs4_proc_lock,
9423         .clear_acl_cache = nfs4_zap_acl_attr,
9424         .close_context  = nfs4_close_context,
9425         .open_context   = nfs4_atomic_open,
9426         .have_delegation = nfs4_have_delegation,
9427         .return_delegation = nfs4_inode_return_delegation,
9428         .alloc_client   = nfs4_alloc_client,
9429         .init_client    = nfs4_init_client,
9430         .free_client    = nfs4_free_client,
9431         .create_server  = nfs4_create_server,
9432         .clone_server   = nfs_clone_server,
9433 };
9434
9435 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
9436         .name   = XATTR_NAME_NFSV4_ACL,
9437         .list   = nfs4_xattr_list_nfs4_acl,
9438         .get    = nfs4_xattr_get_nfs4_acl,
9439         .set    = nfs4_xattr_set_nfs4_acl,
9440 };
9441
9442 const struct xattr_handler *nfs4_xattr_handlers[] = {
9443         &nfs4_xattr_nfs4_acl_handler,
9444 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
9445         &nfs4_xattr_nfs4_label_handler,
9446 #endif
9447         NULL
9448 };
9449
9450 /*
9451  * Local variables:
9452  *  c-basic-offset: 8
9453  * End:
9454  */