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 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "sysfs.h"
  67 #include "nfs4idmap.h"
  68 #include "nfs4session.h"
  69 #include "fscache.h"
  70 #include "nfs42.h"
  71
  72 #include "nfs4trace.h"
  73
  74 #ifdef CONFIG_NFS_V4_2
  75 #include "nfs42.h"
  76 #endif /* CONFIG_NFS_V4_2 */
  77
  78 #define NFSDBG_FACILITY         NFSDBG_PROC
  79
  80 #define NFS4_BITMASK_SZ         3
  81
  82 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  83 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  84
  85 /* file attributes which can be mapped to nfs attributes */
  86 #define NFS4_VALID_ATTRS (ATTR_MODE \
  87         | ATTR_UID \
  88         | ATTR_GID \
  89         | ATTR_SIZE \
  90         | ATTR_ATIME \
  91         | ATTR_MTIME \
  92         | ATTR_CTIME \
  93         | ATTR_ATIME_SET \
  94         | ATTR_MTIME_SET)
  95
  96 struct nfs4_opendata;
  97 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  98 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  99 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
 100 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
 101 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
 102                             struct nfs_fattr *fattr, struct iattr *sattr,
 103                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
 104                             struct nfs4_label *olabel);
 105 #ifdef CONFIG_NFS_V4_1
 106 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 107                 const struct cred *cred,
 108                 struct nfs4_slot *slot,
 109                 bool is_privileged);
 110 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 111                 const struct cred *);
 112 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 113                 const struct cred *, bool);
 114 #endif
 115 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ],
 116                              const __u32 *src, struct inode *inode,
 117                              struct nfs_server *server,
 118                              struct nfs4_label *label);
 119
 120 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 121 static inline struct nfs4_label *
 122 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 123         struct iattr *sattr, struct nfs4_label *label)
 124 {
 125         int err;
 126
 127         if (label == NULL)
 128                 return NULL;
 129
 130         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 131                 return NULL;
 132
 133         label->lfs = 0;
 134         label->pi = 0;
 135         label->len = 0;
 136         label->label = NULL;
 137
 138         err = security_dentry_init_security(dentry, sattr->ia_mode,
 139                                 &dentry->d_name, (void **)&label->label, &label->len);
 140         if (err == 0)
 141                 return label;
 142
 143         return NULL;
 144 }
 145 static inline void
 146 nfs4_label_release_security(struct nfs4_label *label)
 147 {
 148         if (label)
 149                 security_release_secctx(label->label, label->len);
 150 }
 151 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 152 {
 153         if (label)
 154                 return server->attr_bitmask;
 155
 156         return server->attr_bitmask_nl;
 157 }
 158 #else
 159 static inline struct nfs4_label *
 160 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 161         struct iattr *sattr, struct nfs4_label *l)
 162 { return NULL; }
 163 static inline void
 164 nfs4_label_release_security(struct nfs4_label *label)
 165 { return; }
 166 static inline u32 *
 167 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 168 { return server->attr_bitmask; }
 169 #endif
 170
 171 /* Prevent leaks of NFSv4 errors into userland */
 172 static int nfs4_map_errors(int err)
 173 {
 174         if (err >= -1000)
 175                 return err;
 176         switch (err) {
 177         case -NFS4ERR_RESOURCE:
 178         case -NFS4ERR_LAYOUTTRYLATER:
 179         case -NFS4ERR_RECALLCONFLICT:
 180         case -NFS4ERR_RETURNCONFLICT:
 181                 return -EREMOTEIO;
 182         case -NFS4ERR_WRONGSEC:
 183         case -NFS4ERR_WRONG_CRED:
 184                 return -EPERM;
 185         case -NFS4ERR_BADOWNER:
 186         case -NFS4ERR_BADNAME:
 187                 return -EINVAL;
 188         case -NFS4ERR_SHARE_DENIED:
 189                 return -EACCES;
 190         case -NFS4ERR_MINOR_VERS_MISMATCH:
 191                 return -EPROTONOSUPPORT;
 192         case -NFS4ERR_FILE_OPEN:
 193                 return -EBUSY;
 194         default:
 195                 dprintk("%s could not handle NFSv4 error %d\n",
 196                                 __func__, -err);
 197                 break;
 198         }
 199         return -EIO;
 200 }
 201
 202 /*
 203  * This is our standard bitmap for GETATTR requests.
 204  */
 205 const u32 nfs4_fattr_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_WORD1_MOUNTED_ON_FILEID,
 221 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 222         FATTR4_WORD2_SECURITY_LABEL
 223 #endif
 224 };
 225
 226 static const u32 nfs4_pnfs_open_bitmap[3] = {
 227         FATTR4_WORD0_TYPE
 228         | FATTR4_WORD0_CHANGE
 229         | FATTR4_WORD0_SIZE
 230         | FATTR4_WORD0_FSID
 231         | FATTR4_WORD0_FILEID,
 232         FATTR4_WORD1_MODE
 233         | FATTR4_WORD1_NUMLINKS
 234         | FATTR4_WORD1_OWNER
 235         | FATTR4_WORD1_OWNER_GROUP
 236         | FATTR4_WORD1_RAWDEV
 237         | FATTR4_WORD1_SPACE_USED
 238         | FATTR4_WORD1_TIME_ACCESS
 239         | FATTR4_WORD1_TIME_METADATA
 240         | FATTR4_WORD1_TIME_MODIFY,
 241         FATTR4_WORD2_MDSTHRESHOLD
 242 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 243         | FATTR4_WORD2_SECURITY_LABEL
 244 #endif
 245 };
 246
 247 static const u32 nfs4_open_noattr_bitmap[3] = {
 248         FATTR4_WORD0_TYPE
 249         | FATTR4_WORD0_FILEID,
 250 };
 251
 252 const u32 nfs4_statfs_bitmap[3] = {
 253         FATTR4_WORD0_FILES_AVAIL
 254         | FATTR4_WORD0_FILES_FREE
 255         | FATTR4_WORD0_FILES_TOTAL,
 256         FATTR4_WORD1_SPACE_AVAIL
 257         | FATTR4_WORD1_SPACE_FREE
 258         | FATTR4_WORD1_SPACE_TOTAL
 259 };
 260
 261 const u32 nfs4_pathconf_bitmap[3] = {
 262         FATTR4_WORD0_MAXLINK
 263         | FATTR4_WORD0_MAXNAME,
 264         0
 265 };
 266
 267 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 268                         | FATTR4_WORD0_MAXREAD
 269                         | FATTR4_WORD0_MAXWRITE
 270                         | FATTR4_WORD0_LEASE_TIME,
 271                         FATTR4_WORD1_TIME_DELTA
 272                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 273                         FATTR4_WORD2_LAYOUT_BLKSIZE
 274                         | FATTR4_WORD2_CLONE_BLKSIZE
 275                         | FATTR4_WORD2_XATTR_SUPPORT
 276 };
 277
 278 const u32 nfs4_fs_locations_bitmap[3] = {
 279         FATTR4_WORD0_CHANGE
 280         | FATTR4_WORD0_SIZE
 281         | FATTR4_WORD0_FSID
 282         | FATTR4_WORD0_FILEID
 283         | FATTR4_WORD0_FS_LOCATIONS,
 284         FATTR4_WORD1_OWNER
 285         | FATTR4_WORD1_OWNER_GROUP
 286         | FATTR4_WORD1_RAWDEV
 287         | FATTR4_WORD1_SPACE_USED
 288         | FATTR4_WORD1_TIME_ACCESS
 289         | FATTR4_WORD1_TIME_METADATA
 290         | FATTR4_WORD1_TIME_MODIFY
 291         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 292 };
 293
 294 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 295                 struct inode *inode)
 296 {
 297         unsigned long cache_validity;
 298
 299         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 300         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 301                 return;
 302
 303         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 304         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 305                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 306                                 | NFS_INO_INVALID_SIZE);
 307
 308         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 309                 dst[0] &= ~FATTR4_WORD0_SIZE;
 310
 311         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 312                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 313 }
 314
 315 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 316                 const __u32 *src, struct inode *inode)
 317 {
 318         nfs4_bitmap_copy_adjust(dst, src, inode);
 319 }
 320
 321 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 322                 struct nfs4_readdir_arg *readdir)
 323 {
 324         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 325         __be32 *start, *p;
 326
 327         if (cookie > 2) {
 328                 readdir->cookie = cookie;
 329                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 330                 return;
 331         }
 332
 333         readdir->cookie = 0;
 334         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 335         if (cookie == 2)
 336                 return;
 337
 338         /*
 339          * NFSv4 servers do not return entries for '.' and '..'
 340          * Therefore, we fake these entries here.  We let '.'
 341          * have cookie 0 and '..' have cookie 1.  Note that
 342          * when talking to the server, we always send cookie 0
 343          * instead of 1 or 2.
 344          */
 345         start = p = kmap_atomic(*readdir->pages);
 346
 347         if (cookie == 0) {
 348                 *p++ = xdr_one;                                  /* next */
 349                 *p++ = xdr_zero;                   /* cookie, first word */
 350                 *p++ = xdr_one;                   /* cookie, second word */
 351                 *p++ = xdr_one;                             /* entry len */
 352                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 353                 p++;
 354                 *p++ = xdr_one;                         /* bitmap length */
 355                 *p++ = htonl(attrs);                           /* bitmap */
 356                 *p++ = htonl(12);             /* attribute buffer length */
 357                 *p++ = htonl(NF4DIR);
 358                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 359         }
 360
 361         *p++ = xdr_one;                                  /* next */
 362         *p++ = xdr_zero;                   /* cookie, first word */
 363         *p++ = xdr_two;                   /* cookie, second word */
 364         *p++ = xdr_two;                             /* entry len */
 365         memcpy(p, "..\0\0", 4);                         /* entry */
 366         p++;
 367         *p++ = xdr_one;                         /* bitmap length */
 368         *p++ = htonl(attrs);                           /* bitmap */
 369         *p++ = htonl(12);             /* attribute buffer length */
 370         *p++ = htonl(NF4DIR);
 371         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 372
 373         readdir->pgbase = (char *)p - (char *)start;
 374         readdir->count -= readdir->pgbase;
 375         kunmap_atomic(start);
 376 }
 377
 378 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
 379 {
 380         if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
 381                 fattr->pre_change_attr = version;
 382                 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
 383         }
 384 }
 385
 386 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 387                 nfs4_stateid *stateid,
 388                 const struct cred *cred)
 389 {
 390         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 391
 392         ops->test_and_free_expired(server, stateid, cred);
 393 }
 394
 395 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 396                 nfs4_stateid *stateid,
 397                 const struct cred *cred)
 398 {
 399         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 400         nfs4_test_and_free_stateid(server, stateid, cred);
 401 }
 402
 403 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 404                 const nfs4_stateid *stateid,
 405                 const struct cred *cred)
 406 {
 407         nfs4_stateid tmp;
 408
 409         nfs4_stateid_copy(&tmp, stateid);
 410         __nfs4_free_revoked_stateid(server, &tmp, cred);
 411 }
 412
 413 static long nfs4_update_delay(long *timeout)
 414 {
 415         long ret;
 416         if (!timeout)
 417                 return NFS4_POLL_RETRY_MAX;
 418         if (*timeout <= 0)
 419                 *timeout = NFS4_POLL_RETRY_MIN;
 420         if (*timeout > NFS4_POLL_RETRY_MAX)
 421                 *timeout = NFS4_POLL_RETRY_MAX;
 422         ret = *timeout;
 423         *timeout <<= 1;
 424         return ret;
 425 }
 426
 427 static int nfs4_delay_killable(long *timeout)
 428 {
 429         might_sleep();
 430
 431         freezable_schedule_timeout_killable_unsafe(
 432                 nfs4_update_delay(timeout));
 433         if (!__fatal_signal_pending(current))
 434                 return 0;
 435         return -EINTR;
 436 }
 437
 438 static int nfs4_delay_interruptible(long *timeout)
 439 {
 440         might_sleep();
 441
 442         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 443         if (!signal_pending(current))
 444                 return 0;
 445         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 446 }
 447
 448 static int nfs4_delay(long *timeout, bool interruptible)
 449 {
 450         if (interruptible)
 451                 return nfs4_delay_interruptible(timeout);
 452         return nfs4_delay_killable(timeout);
 453 }
 454
 455 static const nfs4_stateid *
 456 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 457 {
 458         if (!stateid)
 459                 return NULL;
 460         switch (stateid->type) {
 461         case NFS4_OPEN_STATEID_TYPE:
 462         case NFS4_LOCK_STATEID_TYPE:
 463         case NFS4_DELEGATION_STATEID_TYPE:
 464                 return stateid;
 465         default:
 466                 break;
 467         }
 468         return NULL;
 469 }
 470
 471 /* This is the error handling routine for processes that are allowed
 472  * to sleep.
 473  */
 474 static int nfs4_do_handle_exception(struct nfs_server *server,
 475                 int errorcode, struct nfs4_exception *exception)
 476 {
 477         struct nfs_client *clp = server->nfs_client;
 478         struct nfs4_state *state = exception->state;
 479         const nfs4_stateid *stateid;
 480         struct inode *inode = exception->inode;
 481         int ret = errorcode;
 482
 483         exception->delay = 0;
 484         exception->recovering = 0;
 485         exception->retry = 0;
 486
 487         stateid = nfs4_recoverable_stateid(exception->stateid);
 488         if (stateid == NULL && state != NULL)
 489                 stateid = nfs4_recoverable_stateid(&state->stateid);
 490
 491         switch(errorcode) {
 492                 case 0:
 493                         return 0;
 494                 case -NFS4ERR_BADHANDLE:
 495                 case -ESTALE:
 496                         if (inode != NULL && S_ISREG(inode->i_mode))
 497                                 pnfs_destroy_layout(NFS_I(inode));
 498                         break;
 499                 case -NFS4ERR_DELEG_REVOKED:
 500                 case -NFS4ERR_ADMIN_REVOKED:
 501                 case -NFS4ERR_EXPIRED:
 502                 case -NFS4ERR_BAD_STATEID:
 503                 case -NFS4ERR_PARTNER_NO_AUTH:
 504                         if (inode != NULL && stateid != NULL) {
 505                                 nfs_inode_find_state_and_recover(inode,
 506                                                 stateid);
 507                                 goto wait_on_recovery;
 508                         }
 509                         fallthrough;
 510                 case -NFS4ERR_OPENMODE:
 511                         if (inode) {
 512                                 int err;
 513
 514                                 err = nfs_async_inode_return_delegation(inode,
 515                                                 stateid);
 516                                 if (err == 0)
 517                                         goto wait_on_recovery;
 518                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 519                                         exception->retry = 1;
 520                                         break;
 521                                 }
 522                         }
 523                         if (state == NULL)
 524                                 break;
 525                         ret = nfs4_schedule_stateid_recovery(server, state);
 526                         if (ret < 0)
 527                                 break;
 528                         goto wait_on_recovery;
 529                 case -NFS4ERR_STALE_STATEID:
 530                 case -NFS4ERR_STALE_CLIENTID:
 531                         nfs4_schedule_lease_recovery(clp);
 532                         goto wait_on_recovery;
 533                 case -NFS4ERR_MOVED:
 534                         ret = nfs4_schedule_migration_recovery(server);
 535                         if (ret < 0)
 536                                 break;
 537                         goto wait_on_recovery;
 538                 case -NFS4ERR_LEASE_MOVED:
 539                         nfs4_schedule_lease_moved_recovery(clp);
 540                         goto wait_on_recovery;
 541 #if defined(CONFIG_NFS_V4_1)
 542                 case -NFS4ERR_BADSESSION:
 543                 case -NFS4ERR_BADSLOT:
 544                 case -NFS4ERR_BAD_HIGH_SLOT:
 545                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 546                 case -NFS4ERR_DEADSESSION:
 547                 case -NFS4ERR_SEQ_FALSE_RETRY:
 548                 case -NFS4ERR_SEQ_MISORDERED:
 549                         /* Handled in nfs41_sequence_process() */
 550                         goto wait_on_recovery;
 551 #endif /* defined(CONFIG_NFS_V4_1) */
 552                 case -NFS4ERR_FILE_OPEN:
 553                         if (exception->timeout > HZ) {
 554                                 /* We have retried a decent amount, time to
 555                                  * fail
 556                                  */
 557                                 ret = -EBUSY;
 558                                 break;
 559                         }
 560                         fallthrough;
 561                 case -NFS4ERR_DELAY:
 562                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 563                         fallthrough;
 564                 case -NFS4ERR_GRACE:
 565                 case -NFS4ERR_LAYOUTTRYLATER:
 566                 case -NFS4ERR_RECALLCONFLICT:
 567                 case -NFS4ERR_RETURNCONFLICT:
 568                         exception->delay = 1;
 569                         return 0;
 570
 571                 case -NFS4ERR_RETRY_UNCACHED_REP:
 572                 case -NFS4ERR_OLD_STATEID:
 573                         exception->retry = 1;
 574                         break;
 575                 case -NFS4ERR_BADOWNER:
 576                         /* The following works around a Linux server bug! */
 577                 case -NFS4ERR_BADNAME:
 578                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 579                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 580                                 exception->retry = 1;
 581                                 printk(KERN_WARNING "NFS: v4 server %s "
 582                                                 "does not accept raw "
 583                                                 "uid/gids. "
 584                                                 "Reenabling the idmapper.\n",
 585                                                 server->nfs_client->cl_hostname);
 586                         }
 587         }
 588         /* We failed to handle the error */
 589         return nfs4_map_errors(ret);
 590 wait_on_recovery:
 591         exception->recovering = 1;
 592         return 0;
 593 }
 594
 595 /* This is the error handling routine for processes that are allowed
 596  * to sleep.
 597  */
 598 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 599 {
 600         struct nfs_client *clp = server->nfs_client;
 601         int ret;
 602
 603         ret = nfs4_do_handle_exception(server, errorcode, exception);
 604         if (exception->delay) {
 605                 ret = nfs4_delay(&exception->timeout,
 606                                 exception->interruptible);
 607                 goto out_retry;
 608         }
 609         if (exception->recovering) {
 610                 if (exception->task_is_privileged)
 611                         return -EDEADLOCK;
 612                 ret = nfs4_wait_clnt_recover(clp);
 613                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 614                         return -EIO;
 615                 goto out_retry;
 616         }
 617         return ret;
 618 out_retry:
 619         if (ret == 0)
 620                 exception->retry = 1;
 621         return ret;
 622 }
 623
 624 static int
 625 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 626                 int errorcode, struct nfs4_exception *exception)
 627 {
 628         struct nfs_client *clp = server->nfs_client;
 629         int ret;
 630
 631         ret = nfs4_do_handle_exception(server, errorcode, exception);
 632         if (exception->delay) {
 633                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 634                 goto out_retry;
 635         }
 636         if (exception->recovering) {
 637                 if (exception->task_is_privileged)
 638                         return -EDEADLOCK;
 639                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 640                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 641                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 642                 goto out_retry;
 643         }
 644         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 645                 ret = -EIO;
 646         return ret;
 647 out_retry:
 648         if (ret == 0) {
 649                 exception->retry = 1;
 650                 /*
 651                  * For NFS4ERR_MOVED, the client transport will need to
 652                  * be recomputed after migration recovery has completed.
 653                  */
 654                 if (errorcode == -NFS4ERR_MOVED)
 655                         rpc_task_release_transport(task);
 656         }
 657         return ret;
 658 }
 659
 660 int
 661 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 662                         struct nfs4_state *state, long *timeout)
 663 {
 664         struct nfs4_exception exception = {
 665                 .state = state,
 666         };
 667
 668         if (task->tk_status >= 0)
 669                 return 0;
 670         if (timeout)
 671                 exception.timeout = *timeout;
 672         task->tk_status = nfs4_async_handle_exception(task, server,
 673                         task->tk_status,
 674                         &exception);
 675         if (exception.delay && timeout)
 676                 *timeout = exception.timeout;
 677         if (exception.retry)
 678                 return -EAGAIN;
 679         return 0;
 680 }
 681
 682 /*
 683  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 684  * or 'false' otherwise.
 685  */
 686 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 687 {
 688         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 689         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 690 }
 691
 692 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 693 {
 694         spin_lock(&clp->cl_lock);
 695         if (time_before(clp->cl_last_renewal,timestamp))
 696                 clp->cl_last_renewal = timestamp;
 697         spin_unlock(&clp->cl_lock);
 698 }
 699
 700 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 701 {
 702         struct nfs_client *clp = server->nfs_client;
 703
 704         if (!nfs4_has_session(clp))
 705                 do_renew_lease(clp, timestamp);
 706 }
 707
 708 struct nfs4_call_sync_data {
 709         const struct nfs_server *seq_server;
 710         struct nfs4_sequence_args *seq_args;
 711         struct nfs4_sequence_res *seq_res;
 712 };
 713
 714 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 715                         struct nfs4_sequence_res *res, int cache_reply,
 716                         int privileged)
 717 {
 718         args->sa_slot = NULL;
 719         args->sa_cache_this = cache_reply;
 720         args->sa_privileged = privileged;
 721
 722         res->sr_slot = NULL;
 723 }
 724
 725 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 726 {
 727         struct nfs4_slot *slot = res->sr_slot;
 728         struct nfs4_slot_table *tbl;
 729
 730         tbl = slot->table;
 731         spin_lock(&tbl->slot_tbl_lock);
 732         if (!nfs41_wake_and_assign_slot(tbl, slot))
 733                 nfs4_free_slot(tbl, slot);
 734         spin_unlock(&tbl->slot_tbl_lock);
 735
 736         res->sr_slot = NULL;
 737 }
 738
 739 static int nfs40_sequence_done(struct rpc_task *task,
 740                                struct nfs4_sequence_res *res)
 741 {
 742         if (res->sr_slot != NULL)
 743                 nfs40_sequence_free_slot(res);
 744         return 1;
 745 }
 746
 747 #if defined(CONFIG_NFS_V4_1)
 748
 749 static void nfs41_release_slot(struct nfs4_slot *slot)
 750 {
 751         struct nfs4_session *session;
 752         struct nfs4_slot_table *tbl;
 753         bool send_new_highest_used_slotid = false;
 754
 755         if (!slot)
 756                 return;
 757         tbl = slot->table;
 758         session = tbl->session;
 759
 760         /* Bump the slot sequence number */
 761         if (slot->seq_done)
 762                 slot->seq_nr++;
 763         slot->seq_done = 0;
 764
 765         spin_lock(&tbl->slot_tbl_lock);
 766         /* Be nice to the server: try to ensure that the last transmitted
 767          * value for highest_user_slotid <= target_highest_slotid
 768          */
 769         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 770                 send_new_highest_used_slotid = true;
 771
 772         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 773                 send_new_highest_used_slotid = false;
 774                 goto out_unlock;
 775         }
 776         nfs4_free_slot(tbl, slot);
 777
 778         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 779                 send_new_highest_used_slotid = false;
 780 out_unlock:
 781         spin_unlock(&tbl->slot_tbl_lock);
 782         if (send_new_highest_used_slotid)
 783                 nfs41_notify_server(session->clp);
 784         if (waitqueue_active(&tbl->slot_waitq))
 785                 wake_up_all(&tbl->slot_waitq);
 786 }
 787
 788 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 789 {
 790         nfs41_release_slot(res->sr_slot);
 791         res->sr_slot = NULL;
 792 }
 793
 794 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 795                 u32 seqnr)
 796 {
 797         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 798                 slot->seq_nr_highest_sent = seqnr;
 799 }
 800 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
 801 {
 802         nfs4_slot_sequence_record_sent(slot, seqnr);
 803         slot->seq_nr_last_acked = seqnr;
 804 }
 805
 806 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 807                                 struct nfs4_slot *slot)
 808 {
 809         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 810         if (!IS_ERR(task))
 811                 rpc_put_task_async(task);
 812 }
 813
 814 static int nfs41_sequence_process(struct rpc_task *task,
 815                 struct nfs4_sequence_res *res)
 816 {
 817         struct nfs4_session *session;
 818         struct nfs4_slot *slot = res->sr_slot;
 819         struct nfs_client *clp;
 820         int status;
 821         int ret = 1;
 822
 823         if (slot == NULL)
 824                 goto out_noaction;
 825         /* don't increment the sequence number if the task wasn't sent */
 826         if (!RPC_WAS_SENT(task) || slot->seq_done)
 827                 goto out;
 828
 829         session = slot->table->session;
 830         clp = session->clp;
 831
 832         trace_nfs4_sequence_done(session, res);
 833
 834         status = res->sr_status;
 835         if (task->tk_status == -NFS4ERR_DEADSESSION)
 836                 status = -NFS4ERR_DEADSESSION;
 837
 838         /* Check the SEQUENCE operation status */
 839         switch (status) {
 840         case 0:
 841                 /* Mark this sequence number as having been acked */
 842                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 843                 /* Update the slot's sequence and clientid lease timer */
 844                 slot->seq_done = 1;
 845                 do_renew_lease(clp, res->sr_timestamp);
 846                 /* Check sequence flags */
 847                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 848                                 !!slot->privileged);
 849                 nfs41_update_target_slotid(slot->table, slot, res);
 850                 break;
 851         case 1:
 852                 /*
 853                  * sr_status remains 1 if an RPC level error occurred.
 854                  * The server may or may not have processed the sequence
 855                  * operation..
 856                  */
 857                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 858                 slot->seq_done = 1;
 859                 goto out;
 860         case -NFS4ERR_DELAY:
 861                 /* The server detected a resend of the RPC call and
 862                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 863                  * of RFC5661.
 864                  */
 865                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 866                         __func__,
 867                         slot->slot_nr,
 868                         slot->seq_nr);
 869                 goto out_retry;
 870         case -NFS4ERR_RETRY_UNCACHED_REP:
 871         case -NFS4ERR_SEQ_FALSE_RETRY:
 872                 /*
 873                  * The server thinks we tried to replay a request.
 874                  * Retry the call after bumping the sequence ID.
 875                  */
 876                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 877                 goto retry_new_seq;
 878         case -NFS4ERR_BADSLOT:
 879                 /*
 880                  * The slot id we used was probably retired. Try again
 881                  * using a different slot id.
 882                  */
 883                 if (slot->slot_nr < slot->table->target_highest_slotid)
 884                         goto session_recover;
 885                 goto retry_nowait;
 886         case -NFS4ERR_SEQ_MISORDERED:
 887                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 888                 /*
 889                  * Were one or more calls using this slot interrupted?
 890                  * If the server never received the request, then our
 891                  * transmitted slot sequence number may be too high. However,
 892                  * if the server did receive the request then it might
 893                  * accidentally give us a reply with a mismatched operation.
 894                  * We can sort this out by sending a lone sequence operation
 895                  * to the server on the same slot.
 896                  */
 897                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 898                         slot->seq_nr--;
 899                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 900                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 901                                 res->sr_slot = NULL;
 902                         }
 903                         goto retry_nowait;
 904                 }
 905                 /*
 906                  * RFC5661:
 907                  * A retry might be sent while the original request is
 908                  * still in progress on the replier. The replier SHOULD
 909                  * deal with the issue by returning NFS4ERR_DELAY as the
 910                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 911                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 912                  *
 913                  * Restart the search after a delay.
 914                  */
 915                 slot->seq_nr = slot->seq_nr_highest_sent;
 916                 goto out_retry;
 917         case -NFS4ERR_BADSESSION:
 918         case -NFS4ERR_DEADSESSION:
 919         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 920                 goto session_recover;
 921         default:
 922                 /* Just update the slot sequence no. */
 923                 slot->seq_done = 1;
 924         }
 925 out:
 926         /* The session may be reset by one of the error handlers. */
 927         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 928 out_noaction:
 929         return ret;
 930 session_recover:
 931         set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
 932         nfs4_schedule_session_recovery(session, status);
 933         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 934         nfs41_sequence_free_slot(res);
 935         goto out;
 936 retry_new_seq:
 937         ++slot->seq_nr;
 938 retry_nowait:
 939         if (rpc_restart_call_prepare(task)) {
 940                 nfs41_sequence_free_slot(res);
 941                 task->tk_status = 0;
 942                 ret = 0;
 943         }
 944         goto out;
 945 out_retry:
 946         if (!rpc_restart_call(task))
 947                 goto out;
 948         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 949         return 0;
 950 }
 951
 952 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 953 {
 954         if (!nfs41_sequence_process(task, res))
 955                 return 0;
 956         if (res->sr_slot != NULL)
 957                 nfs41_sequence_free_slot(res);
 958         return 1;
 959
 960 }
 961 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 962
 963 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 964 {
 965         if (res->sr_slot == NULL)
 966                 return 1;
 967         if (res->sr_slot->table->session != NULL)
 968                 return nfs41_sequence_process(task, res);
 969         return nfs40_sequence_done(task, res);
 970 }
 971
 972 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 973 {
 974         if (res->sr_slot != NULL) {
 975                 if (res->sr_slot->table->session != NULL)
 976                         nfs41_sequence_free_slot(res);
 977                 else
 978                         nfs40_sequence_free_slot(res);
 979         }
 980 }
 981
 982 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 983 {
 984         if (res->sr_slot == NULL)
 985                 return 1;
 986         if (!res->sr_slot->table->session)
 987                 return nfs40_sequence_done(task, res);
 988         return nfs41_sequence_done(task, res);
 989 }
 990 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 991
 992 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 993 {
 994         struct nfs4_call_sync_data *data = calldata;
 995
 996         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 997
 998         nfs4_setup_sequence(data->seq_server->nfs_client,
 999                             data->seq_args, data->seq_res, task);
1000 }
1001
1002 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
1003 {
1004         struct nfs4_call_sync_data *data = calldata;
1005
1006         nfs41_sequence_done(task, data->seq_res);
1007 }
1008
1009 static const struct rpc_call_ops nfs41_call_sync_ops = {
1010         .rpc_call_prepare = nfs41_call_sync_prepare,
1011         .rpc_call_done = nfs41_call_sync_done,
1012 };
1013
1014 #else   /* !CONFIG_NFS_V4_1 */
1015
1016 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1017 {
1018         return nfs40_sequence_done(task, res);
1019 }
1020
1021 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1022 {
1023         if (res->sr_slot != NULL)
1024                 nfs40_sequence_free_slot(res);
1025 }
1026
1027 int nfs4_sequence_done(struct rpc_task *task,
1028                        struct nfs4_sequence_res *res)
1029 {
1030         return nfs40_sequence_done(task, res);
1031 }
1032 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1033
1034 #endif  /* !CONFIG_NFS_V4_1 */
1035
1036 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1037 {
1038         res->sr_timestamp = jiffies;
1039         res->sr_status_flags = 0;
1040         res->sr_status = 1;
1041 }
1042
1043 static
1044 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1045                 struct nfs4_sequence_res *res,
1046                 struct nfs4_slot *slot)
1047 {
1048         if (!slot)
1049                 return;
1050         slot->privileged = args->sa_privileged ? 1 : 0;
1051         args->sa_slot = slot;
1052
1053         res->sr_slot = slot;
1054 }
1055
1056 int nfs4_setup_sequence(struct nfs_client *client,
1057                         struct nfs4_sequence_args *args,
1058                         struct nfs4_sequence_res *res,
1059                         struct rpc_task *task)
1060 {
1061         struct nfs4_session *session = nfs4_get_session(client);
1062         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1063         struct nfs4_slot *slot;
1064
1065         /* slot already allocated? */
1066         if (res->sr_slot != NULL)
1067                 goto out_start;
1068
1069         if (session)
1070                 tbl = &session->fc_slot_table;
1071
1072         spin_lock(&tbl->slot_tbl_lock);
1073         /* The state manager will wait until the slot table is empty */
1074         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1075                 goto out_sleep;
1076
1077         slot = nfs4_alloc_slot(tbl);
1078         if (IS_ERR(slot)) {
1079                 if (slot == ERR_PTR(-ENOMEM))
1080                         goto out_sleep_timeout;
1081                 goto out_sleep;
1082         }
1083         spin_unlock(&tbl->slot_tbl_lock);
1084
1085         nfs4_sequence_attach_slot(args, res, slot);
1086
1087         trace_nfs4_setup_sequence(session, args);
1088 out_start:
1089         nfs41_sequence_res_init(res);
1090         rpc_call_start(task);
1091         return 0;
1092 out_sleep_timeout:
1093         /* Try again in 1/4 second */
1094         if (args->sa_privileged)
1095                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1096                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1097         else
1098                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1099                                 NULL, jiffies + (HZ >> 2));
1100         spin_unlock(&tbl->slot_tbl_lock);
1101         return -EAGAIN;
1102 out_sleep:
1103         if (args->sa_privileged)
1104                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1105                                 RPC_PRIORITY_PRIVILEGED);
1106         else
1107                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1108         spin_unlock(&tbl->slot_tbl_lock);
1109         return -EAGAIN;
1110 }
1111 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1112
1113 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1114 {
1115         struct nfs4_call_sync_data *data = calldata;
1116         nfs4_setup_sequence(data->seq_server->nfs_client,
1117                                 data->seq_args, data->seq_res, task);
1118 }
1119
1120 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1121 {
1122         struct nfs4_call_sync_data *data = calldata;
1123         nfs4_sequence_done(task, data->seq_res);
1124 }
1125
1126 static const struct rpc_call_ops nfs40_call_sync_ops = {
1127         .rpc_call_prepare = nfs40_call_sync_prepare,
1128         .rpc_call_done = nfs40_call_sync_done,
1129 };
1130
1131 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1132 {
1133         int ret;
1134         struct rpc_task *task;
1135
1136         task = rpc_run_task(task_setup);
1137         if (IS_ERR(task))
1138                 return PTR_ERR(task);
1139
1140         ret = task->tk_status;
1141         rpc_put_task(task);
1142         return ret;
1143 }
1144
1145 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1146                              struct nfs_server *server,
1147                              struct rpc_message *msg,
1148                              struct nfs4_sequence_args *args,
1149                              struct nfs4_sequence_res *res,
1150                              unsigned short task_flags)
1151 {
1152         struct nfs_client *clp = server->nfs_client;
1153         struct nfs4_call_sync_data data = {
1154                 .seq_server = server,
1155                 .seq_args = args,
1156                 .seq_res = res,
1157         };
1158         struct rpc_task_setup task_setup = {
1159                 .rpc_client = clnt,
1160                 .rpc_message = msg,
1161                 .callback_ops = clp->cl_mvops->call_sync_ops,
1162                 .callback_data = &data,
1163                 .flags = task_flags,
1164         };
1165
1166         return nfs4_call_sync_custom(&task_setup);
1167 }
1168
1169 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1170                                    struct nfs_server *server,
1171                                    struct rpc_message *msg,
1172                                    struct nfs4_sequence_args *args,
1173                                    struct nfs4_sequence_res *res)
1174 {
1175         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1176 }
1177
1178
1179 int nfs4_call_sync(struct rpc_clnt *clnt,
1180                    struct nfs_server *server,
1181                    struct rpc_message *msg,
1182                    struct nfs4_sequence_args *args,
1183                    struct nfs4_sequence_res *res,
1184                    int cache_reply)
1185 {
1186         nfs4_init_sequence(args, res, cache_reply, 0);
1187         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1188 }
1189
1190 static void
1191 nfs4_inc_nlink_locked(struct inode *inode)
1192 {
1193         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1194         inc_nlink(inode);
1195 }
1196
1197 static void
1198 nfs4_dec_nlink_locked(struct inode *inode)
1199 {
1200         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1201         drop_nlink(inode);
1202 }
1203
1204 static void
1205 nfs4_update_changeattr_locked(struct inode *inode,
1206                 struct nfs4_change_info *cinfo,
1207                 unsigned long timestamp, unsigned long cache_validity)
1208 {
1209         struct nfs_inode *nfsi = NFS_I(inode);
1210
1211         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1212                 | NFS_INO_INVALID_MTIME
1213                 | cache_validity;
1214
1215         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) {
1216                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1217                 nfsi->attrtimeo_timestamp = jiffies;
1218         } else {
1219                 if (S_ISDIR(inode->i_mode)) {
1220                         nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1221                         nfs_force_lookup_revalidate(inode);
1222                 } else {
1223                         if (!NFS_PROTO(inode)->have_delegation(inode,
1224                                                                FMODE_READ))
1225                                 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1226                 }
1227
1228                 if (cinfo->before != inode_peek_iversion_raw(inode))
1229                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1230                                                 NFS_INO_INVALID_ACL |
1231                                                 NFS_INO_INVALID_XATTR;
1232         }
1233         inode_set_iversion_raw(inode, cinfo->after);
1234         nfsi->read_cache_jiffies = timestamp;
1235         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1236         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1237
1238         if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
1239                 nfs_fscache_invalidate(inode);
1240 }
1241
1242 void
1243 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1244                 unsigned long timestamp, unsigned long cache_validity)
1245 {
1246         spin_lock(&dir->i_lock);
1247         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1248         spin_unlock(&dir->i_lock);
1249 }
1250
1251 struct nfs4_open_createattrs {
1252         struct nfs4_label *label;
1253         struct iattr *sattr;
1254         const __u32 verf[2];
1255 };
1256
1257 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1258                 int err, struct nfs4_exception *exception)
1259 {
1260         if (err != -EINVAL)
1261                 return false;
1262         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1263                 return false;
1264         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1265         exception->retry = 1;
1266         return true;
1267 }
1268
1269 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1270 {
1271          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1272 }
1273
1274 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1275 {
1276         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1277
1278         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1279 }
1280
1281 static u32
1282 nfs4_map_atomic_open_share(struct nfs_server *server,
1283                 fmode_t fmode, int openflags)
1284 {
1285         u32 res = 0;
1286
1287         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1288         case FMODE_READ:
1289                 res = NFS4_SHARE_ACCESS_READ;
1290                 break;
1291         case FMODE_WRITE:
1292                 res = NFS4_SHARE_ACCESS_WRITE;
1293                 break;
1294         case FMODE_READ|FMODE_WRITE:
1295                 res = NFS4_SHARE_ACCESS_BOTH;
1296         }
1297         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1298                 goto out;
1299         /* Want no delegation if we're using O_DIRECT */
1300         if (openflags & O_DIRECT)
1301                 res |= NFS4_SHARE_WANT_NO_DELEG;
1302 out:
1303         return res;
1304 }
1305
1306 static enum open_claim_type4
1307 nfs4_map_atomic_open_claim(struct nfs_server *server,
1308                 enum open_claim_type4 claim)
1309 {
1310         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1311                 return claim;
1312         switch (claim) {
1313         default:
1314                 return claim;
1315         case NFS4_OPEN_CLAIM_FH:
1316                 return NFS4_OPEN_CLAIM_NULL;
1317         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1318                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1319         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1320                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1321         }
1322 }
1323
1324 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1325 {
1326         p->o_res.f_attr = &p->f_attr;
1327         p->o_res.f_label = p->f_label;
1328         p->o_res.seqid = p->o_arg.seqid;
1329         p->c_res.seqid = p->c_arg.seqid;
1330         p->o_res.server = p->o_arg.server;
1331         p->o_res.access_request = p->o_arg.access;
1332         nfs_fattr_init(&p->f_attr);
1333         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1334 }
1335
1336 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1337                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1338                 const struct nfs4_open_createattrs *c,
1339                 enum open_claim_type4 claim,
1340                 gfp_t gfp_mask)
1341 {
1342         struct dentry *parent = dget_parent(dentry);
1343         struct inode *dir = d_inode(parent);
1344         struct nfs_server *server = NFS_SERVER(dir);
1345         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1346         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1347         struct nfs4_opendata *p;
1348
1349         p = kzalloc(sizeof(*p), gfp_mask);
1350         if (p == NULL)
1351                 goto err;
1352
1353         p->f_label = nfs4_label_alloc(server, gfp_mask);
1354         if (IS_ERR(p->f_label))
1355                 goto err_free_p;
1356
1357         p->a_label = nfs4_label_alloc(server, gfp_mask);
1358         if (IS_ERR(p->a_label))
1359                 goto err_free_f;
1360
1361         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1362         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1363         if (IS_ERR(p->o_arg.seqid))
1364                 goto err_free_label;
1365         nfs_sb_active(dentry->d_sb);
1366         p->dentry = dget(dentry);
1367         p->dir = parent;
1368         p->owner = sp;
1369         atomic_inc(&sp->so_count);
1370         p->o_arg.open_flags = flags;
1371         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1372         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1373         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1374                         fmode, flags);
1375         if (flags & O_CREAT) {
1376                 p->o_arg.umask = current_umask();
1377                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1378                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1379                         p->o_arg.u.attrs = &p->attrs;
1380                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1381
1382                         memcpy(p->o_arg.u.verifier.data, c->verf,
1383                                         sizeof(p->o_arg.u.verifier.data));
1384                 }
1385         }
1386         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1387          * will return permission denied for all bits until close */
1388         if (!(flags & O_EXCL)) {
1389                 /* ask server to check for all possible rights as results
1390                  * are cached */
1391                 switch (p->o_arg.claim) {
1392                 default:
1393                         break;
1394                 case NFS4_OPEN_CLAIM_NULL:
1395                 case NFS4_OPEN_CLAIM_FH:
1396                         p->o_arg.access = NFS4_ACCESS_READ |
1397                                 NFS4_ACCESS_MODIFY |
1398                                 NFS4_ACCESS_EXTEND |
1399                                 NFS4_ACCESS_EXECUTE;
1400 #ifdef CONFIG_NFS_V4_2
1401                         if (server->caps & NFS_CAP_XATTR)
1402                                 p->o_arg.access |= NFS4_ACCESS_XAREAD |
1403                                     NFS4_ACCESS_XAWRITE |
1404                                     NFS4_ACCESS_XALIST;
1405 #endif
1406                 }
1407         }
1408         p->o_arg.clientid = server->nfs_client->cl_clientid;
1409         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1410         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1411         p->o_arg.name = &dentry->d_name;
1412         p->o_arg.server = server;
1413         p->o_arg.bitmask = nfs4_bitmask(server, label);
1414         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1415         switch (p->o_arg.claim) {
1416         case NFS4_OPEN_CLAIM_NULL:
1417         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1418         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1419                 p->o_arg.fh = NFS_FH(dir);
1420                 break;
1421         case NFS4_OPEN_CLAIM_PREVIOUS:
1422         case NFS4_OPEN_CLAIM_FH:
1423         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1424         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1425                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1426         }
1427         p->c_arg.fh = &p->o_res.fh;
1428         p->c_arg.stateid = &p->o_res.stateid;
1429         p->c_arg.seqid = p->o_arg.seqid;
1430         nfs4_init_opendata_res(p);
1431         kref_init(&p->kref);
1432         return p;
1433
1434 err_free_label:
1435         nfs4_label_free(p->a_label);
1436 err_free_f:
1437         nfs4_label_free(p->f_label);
1438 err_free_p:
1439         kfree(p);
1440 err:
1441         dput(parent);
1442         return NULL;
1443 }
1444
1445 static void nfs4_opendata_free(struct kref *kref)
1446 {
1447         struct nfs4_opendata *p = container_of(kref,
1448                         struct nfs4_opendata, kref);
1449         struct super_block *sb = p->dentry->d_sb;
1450
1451         nfs4_lgopen_release(p->lgp);
1452         nfs_free_seqid(p->o_arg.seqid);
1453         nfs4_sequence_free_slot(&p->o_res.seq_res);
1454         if (p->state != NULL)
1455                 nfs4_put_open_state(p->state);
1456         nfs4_put_state_owner(p->owner);
1457
1458         nfs4_label_free(p->a_label);
1459         nfs4_label_free(p->f_label);
1460
1461         dput(p->dir);
1462         dput(p->dentry);
1463         nfs_sb_deactive(sb);
1464         nfs_fattr_free_names(&p->f_attr);
1465         kfree(p->f_attr.mdsthreshold);
1466         kfree(p);
1467 }
1468
1469 static void nfs4_opendata_put(struct nfs4_opendata *p)
1470 {
1471         if (p != NULL)
1472                 kref_put(&p->kref, nfs4_opendata_free);
1473 }
1474
1475 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1476                 fmode_t fmode)
1477 {
1478         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1479         case FMODE_READ|FMODE_WRITE:
1480                 return state->n_rdwr != 0;
1481         case FMODE_WRITE:
1482                 return state->n_wronly != 0;
1483         case FMODE_READ:
1484                 return state->n_rdonly != 0;
1485         }
1486         WARN_ON_ONCE(1);
1487         return false;
1488 }
1489
1490 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1491                 int open_mode, enum open_claim_type4 claim)
1492 {
1493         int ret = 0;
1494
1495         if (open_mode & (O_EXCL|O_TRUNC))
1496                 goto out;
1497         switch (claim) {
1498         case NFS4_OPEN_CLAIM_NULL:
1499         case NFS4_OPEN_CLAIM_FH:
1500                 goto out;
1501         default:
1502                 break;
1503         }
1504         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1505                 case FMODE_READ:
1506                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1507                                 && state->n_rdonly != 0;
1508                         break;
1509                 case FMODE_WRITE:
1510                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1511                                 && state->n_wronly != 0;
1512                         break;
1513                 case FMODE_READ|FMODE_WRITE:
1514                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1515                                 && state->n_rdwr != 0;
1516         }
1517 out:
1518         return ret;
1519 }
1520
1521 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1522                 enum open_claim_type4 claim)
1523 {
1524         if (delegation == NULL)
1525                 return 0;
1526         if ((delegation->type & fmode) != fmode)
1527                 return 0;
1528         switch (claim) {
1529         case NFS4_OPEN_CLAIM_NULL:
1530         case NFS4_OPEN_CLAIM_FH:
1531                 break;
1532         case NFS4_OPEN_CLAIM_PREVIOUS:
1533                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1534                         break;
1535                 fallthrough;
1536         default:
1537                 return 0;
1538         }
1539         nfs_mark_delegation_referenced(delegation);
1540         return 1;
1541 }
1542
1543 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1544 {
1545         switch (fmode) {
1546                 case FMODE_WRITE:
1547                         state->n_wronly++;
1548                         break;
1549                 case FMODE_READ:
1550                         state->n_rdonly++;
1551                         break;
1552                 case FMODE_READ|FMODE_WRITE:
1553                         state->n_rdwr++;
1554         }
1555         nfs4_state_set_mode_locked(state, state->state | fmode);
1556 }
1557
1558 #ifdef CONFIG_NFS_V4_1
1559 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1560 {
1561         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1562                 return true;
1563         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1564                 return true;
1565         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1566                 return true;
1567         return false;
1568 }
1569 #endif /* CONFIG_NFS_V4_1 */
1570
1571 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1572 {
1573         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1574                 wake_up_all(&state->waitq);
1575 }
1576
1577 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1578 {
1579         struct nfs_client *clp = state->owner->so_server->nfs_client;
1580         bool need_recover = false;
1581
1582         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1583                 need_recover = true;
1584         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1585                 need_recover = true;
1586         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1587                 need_recover = true;
1588         if (need_recover)
1589                 nfs4_state_mark_reclaim_nograce(clp, state);
1590 }
1591
1592 /*
1593  * Check for whether or not the caller may update the open stateid
1594  * to the value passed in by stateid.
1595  *
1596  * Note: This function relies heavily on the server implementing
1597  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1598  * correctly.
1599  * i.e. The stateid seqids have to be initialised to 1, and
1600  * are then incremented on every state transition.
1601  */
1602 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1603                 const nfs4_stateid *stateid)
1604 {
1605         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1606                 /* The common case - we're updating to a new sequence number */
1607                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1608                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1609                                 return true;
1610                         return false;
1611                 }
1612                 /* The server returned a new stateid */
1613         }
1614         /* This is the first OPEN in this generation */
1615         if (stateid->seqid == cpu_to_be32(1))
1616                 return true;
1617         return false;
1618 }
1619
1620 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1621 {
1622         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1623                 return;
1624         if (state->n_wronly)
1625                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1626         if (state->n_rdonly)
1627                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1628         if (state->n_rdwr)
1629                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1630         set_bit(NFS_OPEN_STATE, &state->flags);
1631 }
1632
1633 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1634                 nfs4_stateid *stateid, fmode_t fmode)
1635 {
1636         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1637         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1638         case FMODE_WRITE:
1639                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1640                 break;
1641         case FMODE_READ:
1642                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1643                 break;
1644         case 0:
1645                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1646                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1647                 clear_bit(NFS_OPEN_STATE, &state->flags);
1648         }
1649         if (stateid == NULL)
1650                 return;
1651         /* Handle OPEN+OPEN_DOWNGRADE races */
1652         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1653             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1654                 nfs_resync_open_stateid_locked(state);
1655                 goto out;
1656         }
1657         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1658                 nfs4_stateid_copy(&state->stateid, stateid);
1659         nfs4_stateid_copy(&state->open_stateid, stateid);
1660         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1661 out:
1662         nfs_state_log_update_open_stateid(state);
1663 }
1664
1665 static void nfs_clear_open_stateid(struct nfs4_state *state,
1666         nfs4_stateid *arg_stateid,
1667         nfs4_stateid *stateid, fmode_t fmode)
1668 {
1669         write_seqlock(&state->seqlock);
1670         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1671         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1672                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1673         write_sequnlock(&state->seqlock);
1674         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1675                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1676 }
1677
1678 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1679                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1680         __must_hold(&state->owner->so_lock)
1681         __must_hold(&state->seqlock)
1682         __must_hold(RCU)
1683
1684 {
1685         DEFINE_WAIT(wait);
1686         int status = 0;
1687         for (;;) {
1688
1689                 if (nfs_stateid_is_sequential(state, stateid))
1690                         break;
1691
1692                 if (status)
1693                         break;
1694                 /* Rely on seqids for serialisation with NFSv4.0 */
1695                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1696                         break;
1697
1698                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1699                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1700                 /*
1701                  * Ensure we process the state changes in the same order
1702                  * in which the server processed them by delaying the
1703                  * update of the stateid until we are in sequence.
1704                  */
1705                 write_sequnlock(&state->seqlock);
1706                 spin_unlock(&state->owner->so_lock);
1707                 rcu_read_unlock();
1708                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1709
1710                 if (!fatal_signal_pending(current)) {
1711                         if (schedule_timeout(5*HZ) == 0)
1712                                 status = -EAGAIN;
1713                         else
1714                                 status = 0;
1715                 } else
1716                         status = -EINTR;
1717                 finish_wait(&state->waitq, &wait);
1718                 rcu_read_lock();
1719                 spin_lock(&state->owner->so_lock);
1720                 write_seqlock(&state->seqlock);
1721         }
1722
1723         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1724             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1725                 nfs4_stateid_copy(freeme, &state->open_stateid);
1726                 nfs_test_and_clear_all_open_stateid(state);
1727         }
1728
1729         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1730                 nfs4_stateid_copy(&state->stateid, stateid);
1731         nfs4_stateid_copy(&state->open_stateid, stateid);
1732         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1733         nfs_state_log_update_open_stateid(state);
1734 }
1735
1736 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1737                 const nfs4_stateid *open_stateid,
1738                 fmode_t fmode,
1739                 nfs4_stateid *freeme)
1740 {
1741         /*
1742          * Protect the call to nfs4_state_set_mode_locked and
1743          * serialise the stateid update
1744          */
1745         write_seqlock(&state->seqlock);
1746         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1747         switch (fmode) {
1748         case FMODE_READ:
1749                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1750                 break;
1751         case FMODE_WRITE:
1752                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1753                 break;
1754         case FMODE_READ|FMODE_WRITE:
1755                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1756         }
1757         set_bit(NFS_OPEN_STATE, &state->flags);
1758         write_sequnlock(&state->seqlock);
1759 }
1760
1761 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1762 {
1763         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1764         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1765         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1766         clear_bit(NFS_OPEN_STATE, &state->flags);
1767 }
1768
1769 static void nfs_state_set_delegation(struct nfs4_state *state,
1770                 const nfs4_stateid *deleg_stateid,
1771                 fmode_t fmode)
1772 {
1773         /*
1774          * Protect the call to nfs4_state_set_mode_locked and
1775          * serialise the stateid update
1776          */
1777         write_seqlock(&state->seqlock);
1778         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1779         set_bit(NFS_DELEGATED_STATE, &state->flags);
1780         write_sequnlock(&state->seqlock);
1781 }
1782
1783 static void nfs_state_clear_delegation(struct nfs4_state *state)
1784 {
1785         write_seqlock(&state->seqlock);
1786         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1787         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1788         write_sequnlock(&state->seqlock);
1789 }
1790
1791 int update_open_stateid(struct nfs4_state *state,
1792                 const nfs4_stateid *open_stateid,
1793                 const nfs4_stateid *delegation,
1794                 fmode_t fmode)
1795 {
1796         struct nfs_server *server = NFS_SERVER(state->inode);
1797         struct nfs_client *clp = server->nfs_client;
1798         struct nfs_inode *nfsi = NFS_I(state->inode);
1799         struct nfs_delegation *deleg_cur;
1800         nfs4_stateid freeme = { };
1801         int ret = 0;
1802
1803         fmode &= (FMODE_READ|FMODE_WRITE);
1804
1805         rcu_read_lock();
1806         spin_lock(&state->owner->so_lock);
1807         if (open_stateid != NULL) {
1808                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1809                 ret = 1;
1810         }
1811
1812         deleg_cur = nfs4_get_valid_delegation(state->inode);
1813         if (deleg_cur == NULL)
1814                 goto no_delegation;
1815
1816         spin_lock(&deleg_cur->lock);
1817         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1818            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1819             (deleg_cur->type & fmode) != fmode)
1820                 goto no_delegation_unlock;
1821
1822         if (delegation == NULL)
1823                 delegation = &deleg_cur->stateid;
1824         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1825                 goto no_delegation_unlock;
1826
1827         nfs_mark_delegation_referenced(deleg_cur);
1828         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1829         ret = 1;
1830 no_delegation_unlock:
1831         spin_unlock(&deleg_cur->lock);
1832 no_delegation:
1833         if (ret)
1834                 update_open_stateflags(state, fmode);
1835         spin_unlock(&state->owner->so_lock);
1836         rcu_read_unlock();
1837
1838         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1839                 nfs4_schedule_state_manager(clp);
1840         if (freeme.type != 0)
1841                 nfs4_test_and_free_stateid(server, &freeme,
1842                                 state->owner->so_cred);
1843
1844         return ret;
1845 }
1846
1847 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1848                 const nfs4_stateid *stateid)
1849 {
1850         struct nfs4_state *state = lsp->ls_state;
1851         bool ret = false;
1852
1853         spin_lock(&state->state_lock);
1854         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1855                 goto out_noupdate;
1856         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1857                 goto out_noupdate;
1858         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1859         ret = true;
1860 out_noupdate:
1861         spin_unlock(&state->state_lock);
1862         return ret;
1863 }
1864
1865 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1866 {
1867         struct nfs_delegation *delegation;
1868
1869         fmode &= FMODE_READ|FMODE_WRITE;
1870         rcu_read_lock();
1871         delegation = nfs4_get_valid_delegation(inode);
1872         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1873                 rcu_read_unlock();
1874                 return;
1875         }
1876         rcu_read_unlock();
1877         nfs4_inode_return_delegation(inode);
1878 }
1879
1880 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1881 {
1882         struct nfs4_state *state = opendata->state;
1883         struct nfs_delegation *delegation;
1884         int open_mode = opendata->o_arg.open_flags;
1885         fmode_t fmode = opendata->o_arg.fmode;
1886         enum open_claim_type4 claim = opendata->o_arg.claim;
1887         nfs4_stateid stateid;
1888         int ret = -EAGAIN;
1889
1890         for (;;) {
1891                 spin_lock(&state->owner->so_lock);
1892                 if (can_open_cached(state, fmode, open_mode, claim)) {
1893                         update_open_stateflags(state, fmode);
1894                         spin_unlock(&state->owner->so_lock);
1895                         goto out_return_state;
1896                 }
1897                 spin_unlock(&state->owner->so_lock);
1898                 rcu_read_lock();
1899                 delegation = nfs4_get_valid_delegation(state->inode);
1900                 if (!can_open_delegated(delegation, fmode, claim)) {
1901                         rcu_read_unlock();
1902                         break;
1903                 }
1904                 /* Save the delegation */
1905                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1906                 rcu_read_unlock();
1907                 nfs_release_seqid(opendata->o_arg.seqid);
1908                 if (!opendata->is_recover) {
1909                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1910                         if (ret != 0)
1911                                 goto out;
1912                 }
1913                 ret = -EAGAIN;
1914
1915                 /* Try to update the stateid using the delegation */
1916                 if (update_open_stateid(state, NULL, &stateid, fmode))
1917                         goto out_return_state;
1918         }
1919 out:
1920         return ERR_PTR(ret);
1921 out_return_state:
1922         refcount_inc(&state->count);
1923         return state;
1924 }
1925
1926 static void
1927 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1928 {
1929         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1930         struct nfs_delegation *delegation;
1931         int delegation_flags = 0;
1932
1933         rcu_read_lock();
1934         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1935         if (delegation)
1936                 delegation_flags = delegation->flags;
1937         rcu_read_unlock();
1938         switch (data->o_arg.claim) {
1939         default:
1940                 break;
1941         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1942         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1943                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1944                                    "returning a delegation for "
1945                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1946                                    clp->cl_hostname);
1947                 return;
1948         }
1949         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1950                 nfs_inode_set_delegation(state->inode,
1951                                 data->owner->so_cred,
1952                                 data->o_res.delegation_type,
1953                                 &data->o_res.delegation,
1954                                 data->o_res.pagemod_limit);
1955         else
1956                 nfs_inode_reclaim_delegation(state->inode,
1957                                 data->owner->so_cred,
1958                                 data->o_res.delegation_type,
1959                                 &data->o_res.delegation,
1960                                 data->o_res.pagemod_limit);
1961
1962         if (data->o_res.do_recall)
1963                 nfs_async_inode_return_delegation(state->inode,
1964                                                   &data->o_res.delegation);
1965 }
1966
1967 /*
1968  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1969  * and update the nfs4_state.
1970  */
1971 static struct nfs4_state *
1972 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1973 {
1974         struct inode *inode = data->state->inode;
1975         struct nfs4_state *state = data->state;
1976         int ret;
1977
1978         if (!data->rpc_done) {
1979                 if (data->rpc_status)
1980                         return ERR_PTR(data->rpc_status);
1981                 return nfs4_try_open_cached(data);
1982         }
1983
1984         ret = nfs_refresh_inode(inode, &data->f_attr);
1985         if (ret)
1986                 return ERR_PTR(ret);
1987
1988         if (data->o_res.delegation_type != 0)
1989                 nfs4_opendata_check_deleg(data, state);
1990
1991         if (!update_open_stateid(state, &data->o_res.stateid,
1992                                 NULL, data->o_arg.fmode))
1993                 return ERR_PTR(-EAGAIN);
1994         refcount_inc(&state->count);
1995
1996         return state;
1997 }
1998
1999 static struct inode *
2000 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2001 {
2002         struct inode *inode;
2003
2004         switch (data->o_arg.claim) {
2005         case NFS4_OPEN_CLAIM_NULL:
2006         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2007         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2008                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2009                         return ERR_PTR(-EAGAIN);
2010                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2011                                 &data->f_attr, data->f_label);
2012                 break;
2013         default:
2014                 inode = d_inode(data->dentry);
2015                 ihold(inode);
2016                 nfs_refresh_inode(inode, &data->f_attr);
2017         }
2018         return inode;
2019 }
2020
2021 static struct nfs4_state *
2022 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2023 {
2024         struct nfs4_state *state;
2025         struct inode *inode;
2026
2027         inode = nfs4_opendata_get_inode(data);
2028         if (IS_ERR(inode))
2029                 return ERR_CAST(inode);
2030         if (data->state != NULL && data->state->inode == inode) {
2031                 state = data->state;
2032                 refcount_inc(&state->count);
2033         } else
2034                 state = nfs4_get_open_state(inode, data->owner);
2035         iput(inode);
2036         if (state == NULL)
2037                 state = ERR_PTR(-ENOMEM);
2038         return state;
2039 }
2040
2041 static struct nfs4_state *
2042 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2043 {
2044         struct nfs4_state *state;
2045
2046         if (!data->rpc_done) {
2047                 state = nfs4_try_open_cached(data);
2048                 trace_nfs4_cached_open(data->state);
2049                 goto out;
2050         }
2051
2052         state = nfs4_opendata_find_nfs4_state(data);
2053         if (IS_ERR(state))
2054                 goto out;
2055
2056         if (data->o_res.delegation_type != 0)
2057                 nfs4_opendata_check_deleg(data, state);
2058         if (!update_open_stateid(state, &data->o_res.stateid,
2059                                 NULL, data->o_arg.fmode)) {
2060                 nfs4_put_open_state(state);
2061                 state = ERR_PTR(-EAGAIN);
2062         }
2063 out:
2064         nfs_release_seqid(data->o_arg.seqid);
2065         return state;
2066 }
2067
2068 static struct nfs4_state *
2069 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2070 {
2071         struct nfs4_state *ret;
2072
2073         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2074                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2075         else
2076                 ret = _nfs4_opendata_to_nfs4_state(data);
2077         nfs4_sequence_free_slot(&data->o_res.seq_res);
2078         return ret;
2079 }
2080
2081 static struct nfs_open_context *
2082 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2083 {
2084         struct nfs_inode *nfsi = NFS_I(state->inode);
2085         struct nfs_open_context *ctx;
2086
2087         rcu_read_lock();
2088         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2089                 if (ctx->state != state)
2090                         continue;
2091                 if ((ctx->mode & mode) != mode)
2092                         continue;
2093                 if (!get_nfs_open_context(ctx))
2094                         continue;
2095                 rcu_read_unlock();
2096                 return ctx;
2097         }
2098         rcu_read_unlock();
2099         return ERR_PTR(-ENOENT);
2100 }
2101
2102 static struct nfs_open_context *
2103 nfs4_state_find_open_context(struct nfs4_state *state)
2104 {
2105         struct nfs_open_context *ctx;
2106
2107         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2108         if (!IS_ERR(ctx))
2109                 return ctx;
2110         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2111         if (!IS_ERR(ctx))
2112                 return ctx;
2113         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2114 }
2115
2116 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2117                 struct nfs4_state *state, enum open_claim_type4 claim)
2118 {
2119         struct nfs4_opendata *opendata;
2120
2121         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2122                         NULL, claim, GFP_NOFS);
2123         if (opendata == NULL)
2124                 return ERR_PTR(-ENOMEM);
2125         opendata->state = state;
2126         refcount_inc(&state->count);
2127         return opendata;
2128 }
2129
2130 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2131                                     fmode_t fmode)
2132 {
2133         struct nfs4_state *newstate;
2134         struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2135         int openflags = opendata->o_arg.open_flags;
2136         int ret;
2137
2138         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2139                 return 0;
2140         opendata->o_arg.fmode = fmode;
2141         opendata->o_arg.share_access =
2142                 nfs4_map_atomic_open_share(server, fmode, openflags);
2143         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2144         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2145         nfs4_init_opendata_res(opendata);
2146         ret = _nfs4_recover_proc_open(opendata);
2147         if (ret != 0)
2148                 return ret;
2149         newstate = nfs4_opendata_to_nfs4_state(opendata);
2150         if (IS_ERR(newstate))
2151                 return PTR_ERR(newstate);
2152         if (newstate != opendata->state)
2153                 ret = -ESTALE;
2154         nfs4_close_state(newstate, fmode);
2155         return ret;
2156 }
2157
2158 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2159 {
2160         int ret;
2161
2162         /* memory barrier prior to reading state->n_* */
2163         smp_rmb();
2164         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2165         if (ret != 0)
2166                 return ret;
2167         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2168         if (ret != 0)
2169                 return ret;
2170         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2171         if (ret != 0)
2172                 return ret;
2173         /*
2174          * We may have performed cached opens for all three recoveries.
2175          * Check if we need to update the current stateid.
2176          */
2177         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2178             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2179                 write_seqlock(&state->seqlock);
2180                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2181                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2182                 write_sequnlock(&state->seqlock);
2183         }
2184         return 0;
2185 }
2186
2187 /*
2188  * OPEN_RECLAIM:
2189  *      reclaim state on the server after a reboot.
2190  */
2191 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2192 {
2193         struct nfs_delegation *delegation;
2194         struct nfs4_opendata *opendata;
2195         fmode_t delegation_type = 0;
2196         int status;
2197
2198         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2199                         NFS4_OPEN_CLAIM_PREVIOUS);
2200         if (IS_ERR(opendata))
2201                 return PTR_ERR(opendata);
2202         rcu_read_lock();
2203         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2204         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2205                 delegation_type = delegation->type;
2206         rcu_read_unlock();
2207         opendata->o_arg.u.delegation_type = delegation_type;
2208         status = nfs4_open_recover(opendata, state);
2209         nfs4_opendata_put(opendata);
2210         return status;
2211 }
2212
2213 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2214 {
2215         struct nfs_server *server = NFS_SERVER(state->inode);
2216         struct nfs4_exception exception = { };
2217         int err;
2218         do {
2219                 err = _nfs4_do_open_reclaim(ctx, state);
2220                 trace_nfs4_open_reclaim(ctx, 0, err);
2221                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2222                         continue;
2223                 if (err != -NFS4ERR_DELAY)
2224                         break;
2225                 nfs4_handle_exception(server, err, &exception);
2226         } while (exception.retry);
2227         return err;
2228 }
2229
2230 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2231 {
2232         struct nfs_open_context *ctx;
2233         int ret;
2234
2235         ctx = nfs4_state_find_open_context(state);
2236         if (IS_ERR(ctx))
2237                 return -EAGAIN;
2238         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2239         nfs_state_clear_open_state_flags(state);
2240         ret = nfs4_do_open_reclaim(ctx, state);
2241         put_nfs_open_context(ctx);
2242         return ret;
2243 }
2244
2245 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)
2246 {
2247         switch (err) {
2248                 default:
2249                         printk(KERN_ERR "NFS: %s: unhandled error "
2250                                         "%d.\n", __func__, err);
2251                 case 0:
2252                 case -ENOENT:
2253                 case -EAGAIN:
2254                 case -ESTALE:
2255                 case -ETIMEDOUT:
2256                         break;
2257                 case -NFS4ERR_BADSESSION:
2258                 case -NFS4ERR_BADSLOT:
2259                 case -NFS4ERR_BAD_HIGH_SLOT:
2260                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2261                 case -NFS4ERR_DEADSESSION:
2262                         return -EAGAIN;
2263                 case -NFS4ERR_STALE_CLIENTID:
2264                 case -NFS4ERR_STALE_STATEID:
2265                         /* Don't recall a delegation if it was lost */
2266                         nfs4_schedule_lease_recovery(server->nfs_client);
2267                         return -EAGAIN;
2268                 case -NFS4ERR_MOVED:
2269                         nfs4_schedule_migration_recovery(server);
2270                         return -EAGAIN;
2271                 case -NFS4ERR_LEASE_MOVED:
2272                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2273                         return -EAGAIN;
2274                 case -NFS4ERR_DELEG_REVOKED:
2275                 case -NFS4ERR_ADMIN_REVOKED:
2276                 case -NFS4ERR_EXPIRED:
2277                 case -NFS4ERR_BAD_STATEID:
2278                 case -NFS4ERR_OPENMODE:
2279                         nfs_inode_find_state_and_recover(state->inode,
2280                                         stateid);
2281                         nfs4_schedule_stateid_recovery(server, state);
2282                         return -EAGAIN;
2283                 case -NFS4ERR_DELAY:
2284                 case -NFS4ERR_GRACE:
2285                         ssleep(1);
2286                         return -EAGAIN;
2287                 case -ENOMEM:
2288                 case -NFS4ERR_DENIED:
2289                         if (fl) {
2290                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2291                                 if (lsp)
2292                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2293                         }
2294                         return 0;
2295         }
2296         return err;
2297 }
2298
2299 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2300                 struct nfs4_state *state, const nfs4_stateid *stateid)
2301 {
2302         struct nfs_server *server = NFS_SERVER(state->inode);
2303         struct nfs4_opendata *opendata;
2304         int err = 0;
2305
2306         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2307                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2308         if (IS_ERR(opendata))
2309                 return PTR_ERR(opendata);
2310         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2311         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2312                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2313                 if (err)
2314                         goto out;
2315         }
2316         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2317                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2318                 if (err)
2319                         goto out;
2320         }
2321         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2322                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2323                 if (err)
2324                         goto out;
2325         }
2326         nfs_state_clear_delegation(state);
2327 out:
2328         nfs4_opendata_put(opendata);
2329         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2330 }
2331
2332 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2333 {
2334         struct nfs4_opendata *data = calldata;
2335
2336         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2337                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2338 }
2339
2340 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2341 {
2342         struct nfs4_opendata *data = calldata;
2343
2344         nfs40_sequence_done(task, &data->c_res.seq_res);
2345
2346         data->rpc_status = task->tk_status;
2347         if (data->rpc_status == 0) {
2348                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2349                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2350                 renew_lease(data->o_res.server, data->timestamp);
2351                 data->rpc_done = true;
2352         }
2353 }
2354
2355 static void nfs4_open_confirm_release(void *calldata)
2356 {
2357         struct nfs4_opendata *data = calldata;
2358         struct nfs4_state *state = NULL;
2359
2360         /* If this request hasn't been cancelled, do nothing */
2361         if (!data->cancelled)
2362                 goto out_free;
2363         /* In case of error, no cleanup! */
2364         if (!data->rpc_done)
2365                 goto out_free;
2366         state = nfs4_opendata_to_nfs4_state(data);
2367         if (!IS_ERR(state))
2368                 nfs4_close_state(state, data->o_arg.fmode);
2369 out_free:
2370         nfs4_opendata_put(data);
2371 }
2372
2373 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2374         .rpc_call_prepare = nfs4_open_confirm_prepare,
2375         .rpc_call_done = nfs4_open_confirm_done,
2376         .rpc_release = nfs4_open_confirm_release,
2377 };
2378
2379 /*
2380  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2381  */
2382 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2383 {
2384         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2385         struct rpc_task *task;
2386         struct  rpc_message msg = {
2387                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2388                 .rpc_argp = &data->c_arg,
2389                 .rpc_resp = &data->c_res,
2390                 .rpc_cred = data->owner->so_cred,
2391         };
2392         struct rpc_task_setup task_setup_data = {
2393                 .rpc_client = server->client,
2394                 .rpc_message = &msg,
2395                 .callback_ops = &nfs4_open_confirm_ops,
2396                 .callback_data = data,
2397                 .workqueue = nfsiod_workqueue,
2398                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2399         };
2400         int status;
2401
2402         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2403                                 data->is_recover);
2404         kref_get(&data->kref);
2405         data->rpc_done = false;
2406         data->rpc_status = 0;
2407         data->timestamp = jiffies;
2408         task = rpc_run_task(&task_setup_data);
2409         if (IS_ERR(task))
2410                 return PTR_ERR(task);
2411         status = rpc_wait_for_completion_task(task);
2412         if (status != 0) {
2413                 data->cancelled = true;
2414                 smp_wmb();
2415         } else
2416                 status = data->rpc_status;
2417         rpc_put_task(task);
2418         return status;
2419 }
2420
2421 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2422 {
2423         struct nfs4_opendata *data = calldata;
2424         struct nfs4_state_owner *sp = data->owner;
2425         struct nfs_client *clp = sp->so_server->nfs_client;
2426         enum open_claim_type4 claim = data->o_arg.claim;
2427
2428         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2429                 goto out_wait;
2430         /*
2431          * Check if we still need to send an OPEN call, or if we can use
2432          * a delegation instead.
2433          */
2434         if (data->state != NULL) {
2435                 struct nfs_delegation *delegation;
2436
2437                 if (can_open_cached(data->state, data->o_arg.fmode,
2438                                         data->o_arg.open_flags, claim))
2439                         goto out_no_action;
2440                 rcu_read_lock();
2441                 delegation = nfs4_get_valid_delegation(data->state->inode);
2442                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2443                         goto unlock_no_action;
2444                 rcu_read_unlock();
2445         }
2446         /* Update client id. */
2447         data->o_arg.clientid = clp->cl_clientid;
2448         switch (claim) {
2449         default:
2450                 break;
2451         case NFS4_OPEN_CLAIM_PREVIOUS:
2452         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2453         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2454                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2455                 fallthrough;
2456         case NFS4_OPEN_CLAIM_FH:
2457                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2458         }
2459         data->timestamp = jiffies;
2460         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2461                                 &data->o_arg.seq_args,
2462                                 &data->o_res.seq_res,
2463                                 task) != 0)
2464                 nfs_release_seqid(data->o_arg.seqid);
2465
2466         /* Set the create mode (note dependency on the session type) */
2467         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2468         if (data->o_arg.open_flags & O_EXCL) {
2469                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2470                 if (nfs4_has_persistent_session(clp))
2471                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2472                 else if (clp->cl_mvops->minor_version > 0)
2473                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2474         }
2475         return;
2476 unlock_no_action:
2477         trace_nfs4_cached_open(data->state);
2478         rcu_read_unlock();
2479 out_no_action:
2480         task->tk_action = NULL;
2481 out_wait:
2482         nfs4_sequence_done(task, &data->o_res.seq_res);
2483 }
2484
2485 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2486 {
2487         struct nfs4_opendata *data = calldata;
2488
2489         data->rpc_status = task->tk_status;
2490
2491         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2492                 return;
2493
2494         if (task->tk_status == 0) {
2495                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2496                         switch (data->o_res.f_attr->mode & S_IFMT) {
2497                         case S_IFREG:
2498                                 break;
2499                         case S_IFLNK:
2500                                 data->rpc_status = -ELOOP;
2501                                 break;
2502                         case S_IFDIR:
2503                                 data->rpc_status = -EISDIR;
2504                                 break;
2505                         default:
2506                                 data->rpc_status = -ENOTDIR;
2507                         }
2508                 }
2509                 renew_lease(data->o_res.server, data->timestamp);
2510                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2511                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2512         }
2513         data->rpc_done = true;
2514 }
2515
2516 static void nfs4_open_release(void *calldata)
2517 {
2518         struct nfs4_opendata *data = calldata;
2519         struct nfs4_state *state = NULL;
2520
2521         /* If this request hasn't been cancelled, do nothing */
2522         if (!data->cancelled)
2523                 goto out_free;
2524         /* In case of error, no cleanup! */
2525         if (data->rpc_status != 0 || !data->rpc_done)
2526                 goto out_free;
2527         /* In case we need an open_confirm, no cleanup! */
2528         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2529                 goto out_free;
2530         state = nfs4_opendata_to_nfs4_state(data);
2531         if (!IS_ERR(state))
2532                 nfs4_close_state(state, data->o_arg.fmode);
2533 out_free:
2534         nfs4_opendata_put(data);
2535 }
2536
2537 static const struct rpc_call_ops nfs4_open_ops = {
2538         .rpc_call_prepare = nfs4_open_prepare,
2539         .rpc_call_done = nfs4_open_done,
2540         .rpc_release = nfs4_open_release,
2541 };
2542
2543 static int nfs4_run_open_task(struct nfs4_opendata *data,
2544                               struct nfs_open_context *ctx)
2545 {
2546         struct inode *dir = d_inode(data->dir);
2547         struct nfs_server *server = NFS_SERVER(dir);
2548         struct nfs_openargs *o_arg = &data->o_arg;
2549         struct nfs_openres *o_res = &data->o_res;
2550         struct rpc_task *task;
2551         struct rpc_message msg = {
2552                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2553                 .rpc_argp = o_arg,
2554                 .rpc_resp = o_res,
2555                 .rpc_cred = data->owner->so_cred,
2556         };
2557         struct rpc_task_setup task_setup_data = {
2558                 .rpc_client = server->client,
2559                 .rpc_message = &msg,
2560                 .callback_ops = &nfs4_open_ops,
2561                 .callback_data = data,
2562                 .workqueue = nfsiod_workqueue,
2563                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2564         };
2565         int status;
2566
2567         kref_get(&data->kref);
2568         data->rpc_done = false;
2569         data->rpc_status = 0;
2570         data->cancelled = false;
2571         data->is_recover = false;
2572         if (!ctx) {
2573                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2574                 data->is_recover = true;
2575                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2576         } else {
2577                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2578                 pnfs_lgopen_prepare(data, ctx);
2579         }
2580         task = rpc_run_task(&task_setup_data);
2581         if (IS_ERR(task))
2582                 return PTR_ERR(task);
2583         status = rpc_wait_for_completion_task(task);
2584         if (status != 0) {
2585                 data->cancelled = true;
2586                 smp_wmb();
2587         } else
2588                 status = data->rpc_status;
2589         rpc_put_task(task);
2590
2591         return status;
2592 }
2593
2594 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2595 {
2596         struct inode *dir = d_inode(data->dir);
2597         struct nfs_openres *o_res = &data->o_res;
2598         int status;
2599
2600         status = nfs4_run_open_task(data, NULL);
2601         if (status != 0 || !data->rpc_done)
2602                 return status;
2603
2604         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2605
2606         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2607                 status = _nfs4_proc_open_confirm(data);
2608
2609         return status;
2610 }
2611
2612 /*
2613  * Additional permission checks in order to distinguish between an
2614  * open for read, and an open for execute. This works around the
2615  * fact that NFSv4 OPEN treats read and execute permissions as being
2616  * the same.
2617  * Note that in the non-execute case, we want to turn off permission
2618  * checking if we just created a new file (POSIX open() semantics).
2619  */
2620 static int nfs4_opendata_access(const struct cred *cred,
2621                                 struct nfs4_opendata *opendata,
2622                                 struct nfs4_state *state, fmode_t fmode,
2623                                 int openflags)
2624 {
2625         struct nfs_access_entry cache;
2626         u32 mask, flags;
2627
2628         /* access call failed or for some reason the server doesn't
2629          * support any access modes -- defer access call until later */
2630         if (opendata->o_res.access_supported == 0)
2631                 return 0;
2632
2633         mask = 0;
2634         /*
2635          * Use openflags to check for exec, because fmode won't
2636          * always have FMODE_EXEC set when file open for exec.
2637          */
2638         if (openflags & __FMODE_EXEC) {
2639                 /* ONLY check for exec rights */
2640                 if (S_ISDIR(state->inode->i_mode))
2641                         mask = NFS4_ACCESS_LOOKUP;
2642                 else
2643                         mask = NFS4_ACCESS_EXECUTE;
2644         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2645                 mask = NFS4_ACCESS_READ;
2646
2647         cache.cred = cred;
2648         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2649         nfs_access_add_cache(state->inode, &cache);
2650
2651         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2652         if ((mask & ~cache.mask & flags) == 0)
2653                 return 0;
2654
2655         return -EACCES;
2656 }
2657
2658 /*
2659  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2660  */
2661 static int _nfs4_proc_open(struct nfs4_opendata *data,
2662                            struct nfs_open_context *ctx)
2663 {
2664         struct inode *dir = d_inode(data->dir);
2665         struct nfs_server *server = NFS_SERVER(dir);
2666         struct nfs_openargs *o_arg = &data->o_arg;
2667         struct nfs_openres *o_res = &data->o_res;
2668         int status;
2669
2670         status = nfs4_run_open_task(data, ctx);
2671         if (!data->rpc_done)
2672                 return status;
2673         if (status != 0) {
2674                 if (status == -NFS4ERR_BADNAME &&
2675                                 !(o_arg->open_flags & O_CREAT))
2676                         return -ENOENT;
2677                 return status;
2678         }
2679
2680         nfs_fattr_map_and_free_names(server, &data->f_attr);
2681
2682         if (o_arg->open_flags & O_CREAT) {
2683                 if (o_arg->open_flags & O_EXCL)
2684                         data->file_created = true;
2685                 else if (o_res->cinfo.before != o_res->cinfo.after)
2686                         data->file_created = true;
2687                 if (data->file_created ||
2688                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2689                         nfs4_update_changeattr(dir, &o_res->cinfo,
2690                                         o_res->f_attr->time_start,
2691                                         NFS_INO_INVALID_DATA);
2692         }
2693         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2694                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2695         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2696                 status = _nfs4_proc_open_confirm(data);
2697                 if (status != 0)
2698                         return status;
2699         }
2700         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2701                 nfs4_sequence_free_slot(&o_res->seq_res);
2702                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2703                                 o_res->f_label, NULL);
2704         }
2705         return 0;
2706 }
2707
2708 /*
2709  * OPEN_EXPIRED:
2710  *      reclaim state on the server after a network partition.
2711  *      Assumes caller holds the appropriate lock
2712  */
2713 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2714 {
2715         struct nfs4_opendata *opendata;
2716         int ret;
2717
2718         opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2719         if (IS_ERR(opendata))
2720                 return PTR_ERR(opendata);
2721         /*
2722          * We're not recovering a delegation, so ask for no delegation.
2723          * Otherwise the recovery thread could deadlock with an outstanding
2724          * delegation return.
2725          */
2726         opendata->o_arg.open_flags = O_DIRECT;
2727         ret = nfs4_open_recover(opendata, state);
2728         if (ret == -ESTALE)
2729                 d_drop(ctx->dentry);
2730         nfs4_opendata_put(opendata);
2731         return ret;
2732 }
2733
2734 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2735 {
2736         struct nfs_server *server = NFS_SERVER(state->inode);
2737         struct nfs4_exception exception = { };
2738         int err;
2739
2740         do {
2741                 err = _nfs4_open_expired(ctx, state);
2742                 trace_nfs4_open_expired(ctx, 0, err);
2743                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2744                         continue;
2745                 switch (err) {
2746                 default:
2747                         goto out;
2748                 case -NFS4ERR_GRACE:
2749                 case -NFS4ERR_DELAY:
2750                         nfs4_handle_exception(server, err, &exception);
2751                         err = 0;
2752                 }
2753         } while (exception.retry);
2754 out:
2755         return err;
2756 }
2757
2758 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2759 {
2760         struct nfs_open_context *ctx;
2761         int ret;
2762
2763         ctx = nfs4_state_find_open_context(state);
2764         if (IS_ERR(ctx))
2765                 return -EAGAIN;
2766         ret = nfs4_do_open_expired(ctx, state);
2767         put_nfs_open_context(ctx);
2768         return ret;
2769 }
2770
2771 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2772                 const nfs4_stateid *stateid)
2773 {
2774         nfs_remove_bad_delegation(state->inode, stateid);
2775         nfs_state_clear_delegation(state);
2776 }
2777
2778 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2779 {
2780         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2781                 nfs_finish_clear_delegation_stateid(state, NULL);
2782 }
2783
2784 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2785 {
2786         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2787         nfs40_clear_delegation_stateid(state);
2788         nfs_state_clear_open_state_flags(state);
2789         return nfs4_open_expired(sp, state);
2790 }
2791
2792 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2793                 nfs4_stateid *stateid,
2794                 const struct cred *cred)
2795 {
2796         return -NFS4ERR_BAD_STATEID;
2797 }
2798
2799 #if defined(CONFIG_NFS_V4_1)
2800 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2801                 nfs4_stateid *stateid,
2802                 const struct cred *cred)
2803 {
2804         int status;
2805
2806         switch (stateid->type) {
2807         default:
2808                 break;
2809         case NFS4_INVALID_STATEID_TYPE:
2810         case NFS4_SPECIAL_STATEID_TYPE:
2811                 return -NFS4ERR_BAD_STATEID;
2812         case NFS4_REVOKED_STATEID_TYPE:
2813                 goto out_free;
2814         }
2815
2816         status = nfs41_test_stateid(server, stateid, cred);
2817         switch (status) {
2818         case -NFS4ERR_EXPIRED:
2819         case -NFS4ERR_ADMIN_REVOKED:
2820         case -NFS4ERR_DELEG_REVOKED:
2821                 break;
2822         default:
2823                 return status;
2824         }
2825 out_free:
2826         /* Ack the revoked state to the server */
2827         nfs41_free_stateid(server, stateid, cred, true);
2828         return -NFS4ERR_EXPIRED;
2829 }
2830
2831 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2832 {
2833         struct nfs_server *server = NFS_SERVER(state->inode);
2834         nfs4_stateid stateid;
2835         struct nfs_delegation *delegation;
2836         const struct cred *cred = NULL;
2837         int status, ret = NFS_OK;
2838
2839         /* Get the delegation credential for use by test/free_stateid */
2840         rcu_read_lock();
2841         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2842         if (delegation == NULL) {
2843                 rcu_read_unlock();
2844                 nfs_state_clear_delegation(state);
2845                 return NFS_OK;
2846         }
2847
2848         spin_lock(&delegation->lock);
2849         nfs4_stateid_copy(&stateid, &delegation->stateid);
2850
2851         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2852                                 &delegation->flags)) {
2853                 spin_unlock(&delegation->lock);
2854                 rcu_read_unlock();
2855                 return NFS_OK;
2856         }
2857
2858         if (delegation->cred)
2859                 cred = get_cred(delegation->cred);
2860         spin_unlock(&delegation->lock);
2861         rcu_read_unlock();
2862         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2863         trace_nfs4_test_delegation_stateid(state, NULL, status);
2864         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2865                 nfs_finish_clear_delegation_stateid(state, &stateid);
2866         else
2867                 ret = status;
2868
2869         put_cred(cred);
2870         return ret;
2871 }
2872
2873 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2874 {
2875         nfs4_stateid tmp;
2876
2877         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2878             nfs4_copy_delegation_stateid(state->inode, state->state,
2879                                 &tmp, NULL) &&
2880             nfs4_stateid_match_other(&state->stateid, &tmp))
2881                 nfs_state_set_delegation(state, &tmp, state->state);
2882         else
2883                 nfs_state_clear_delegation(state);
2884 }
2885
2886 /**
2887  * nfs41_check_expired_locks - possibly free a lock stateid
2888  *
2889  * @state: NFSv4 state for an inode
2890  *
2891  * Returns NFS_OK if recovery for this stateid is now finished.
2892  * Otherwise a negative NFS4ERR value is returned.
2893  */
2894 static int nfs41_check_expired_locks(struct nfs4_state *state)
2895 {
2896         int status, ret = NFS_OK;
2897         struct nfs4_lock_state *lsp, *prev = NULL;
2898         struct nfs_server *server = NFS_SERVER(state->inode);
2899
2900         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2901                 goto out;
2902
2903         spin_lock(&state->state_lock);
2904         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2905                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2906                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2907
2908                         refcount_inc(&lsp->ls_count);
2909                         spin_unlock(&state->state_lock);
2910
2911                         nfs4_put_lock_state(prev);
2912                         prev = lsp;
2913
2914                         status = nfs41_test_and_free_expired_stateid(server,
2915                                         &lsp->ls_stateid,
2916                                         cred);
2917                         trace_nfs4_test_lock_stateid(state, lsp, status);
2918                         if (status == -NFS4ERR_EXPIRED ||
2919                             status == -NFS4ERR_BAD_STATEID) {
2920                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2921                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2922                                 if (!recover_lost_locks)
2923                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2924                         } else if (status != NFS_OK) {
2925                                 ret = status;
2926                                 nfs4_put_lock_state(prev);
2927                                 goto out;
2928                         }
2929                         spin_lock(&state->state_lock);
2930                 }
2931         }
2932         spin_unlock(&state->state_lock);
2933         nfs4_put_lock_state(prev);
2934 out:
2935         return ret;
2936 }
2937
2938 /**
2939  * nfs41_check_open_stateid - possibly free an open stateid
2940  *
2941  * @state: NFSv4 state for an inode
2942  *
2943  * Returns NFS_OK if recovery for this stateid is now finished.
2944  * Otherwise a negative NFS4ERR value is returned.
2945  */
2946 static int nfs41_check_open_stateid(struct nfs4_state *state)
2947 {
2948         struct nfs_server *server = NFS_SERVER(state->inode);
2949         nfs4_stateid *stateid = &state->open_stateid;
2950         const struct cred *cred = state->owner->so_cred;
2951         int status;
2952
2953         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2954                 return -NFS4ERR_BAD_STATEID;
2955         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2956         trace_nfs4_test_open_stateid(state, NULL, status);
2957         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2958                 nfs_state_clear_open_state_flags(state);
2959                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2960                 return status;
2961         }
2962         if (nfs_open_stateid_recover_openmode(state))
2963                 return -NFS4ERR_OPENMODE;
2964         return NFS_OK;
2965 }
2966
2967 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2968 {
2969         int status;
2970
2971         status = nfs41_check_delegation_stateid(state);
2972         if (status != NFS_OK)
2973                 return status;
2974         nfs41_delegation_recover_stateid(state);
2975
2976         status = nfs41_check_expired_locks(state);
2977         if (status != NFS_OK)
2978                 return status;
2979         status = nfs41_check_open_stateid(state);
2980         if (status != NFS_OK)
2981                 status = nfs4_open_expired(sp, state);
2982         return status;
2983 }
2984 #endif
2985
2986 /*
2987  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2988  * fields corresponding to attributes that were used to store the verifier.
2989  * Make sure we clobber those fields in the later setattr call
2990  */
2991 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2992                                 struct iattr *sattr, struct nfs4_label **label)
2993 {
2994         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2995         __u32 attrset[3];
2996         unsigned ret;
2997         unsigned i;
2998
2999         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3000                 attrset[i] = opendata->o_res.attrset[i];
3001                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3002                         attrset[i] &= ~bitmask[i];
3003         }
3004
3005         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3006                 sattr->ia_valid : 0;
3007
3008         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3009                 if (sattr->ia_valid & ATTR_ATIME_SET)
3010                         ret |= ATTR_ATIME_SET;
3011                 else
3012                         ret |= ATTR_ATIME;
3013         }
3014
3015         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3016                 if (sattr->ia_valid & ATTR_MTIME_SET)
3017                         ret |= ATTR_MTIME_SET;
3018                 else
3019                         ret |= ATTR_MTIME;
3020         }
3021
3022         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3023                 *label = NULL;
3024         return ret;
3025 }
3026
3027 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3028                 int flags, struct nfs_open_context *ctx)
3029 {
3030         struct nfs4_state_owner *sp = opendata->owner;
3031         struct nfs_server *server = sp->so_server;
3032         struct dentry *dentry;
3033         struct nfs4_state *state;
3034         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3035         struct inode *dir = d_inode(opendata->dir);
3036         unsigned long dir_verifier;
3037         unsigned int seq;
3038         int ret;
3039
3040         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3041         dir_verifier = nfs_save_change_attribute(dir);
3042
3043         ret = _nfs4_proc_open(opendata, ctx);
3044         if (ret != 0)
3045                 goto out;
3046
3047         state = _nfs4_opendata_to_nfs4_state(opendata);
3048         ret = PTR_ERR(state);
3049         if (IS_ERR(state))
3050                 goto out;
3051         ctx->state = state;
3052         if (server->caps & NFS_CAP_POSIX_LOCK)
3053                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3054         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3055                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3056
3057         dentry = opendata->dentry;
3058         if (d_really_is_negative(dentry)) {
3059                 struct dentry *alias;
3060                 d_drop(dentry);
3061                 alias = d_exact_alias(dentry, state->inode);
3062                 if (!alias)
3063                         alias = d_splice_alias(igrab(state->inode), dentry);
3064                 /* d_splice_alias() can't fail here - it's a non-directory */
3065                 if (alias) {
3066                         dput(ctx->dentry);
3067                         ctx->dentry = dentry = alias;
3068                 }
3069         }
3070
3071         switch(opendata->o_arg.claim) {
3072         default:
3073                 break;
3074         case NFS4_OPEN_CLAIM_NULL:
3075         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3076         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3077                 if (!opendata->rpc_done)
3078                         break;
3079                 if (opendata->o_res.delegation_type != 0)
3080                         dir_verifier = nfs_save_change_attribute(dir);
3081                 nfs_set_verifier(dentry, dir_verifier);
3082         }
3083
3084         /* Parse layoutget results before we check for access */
3085         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3086
3087         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3088                         acc_mode, flags);
3089         if (ret != 0)
3090                 goto out;
3091
3092         if (d_inode(dentry) == state->inode) {
3093                 nfs_inode_attach_open_context(ctx);
3094                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3095                         nfs4_schedule_stateid_recovery(server, state);
3096         }
3097
3098 out:
3099         if (!opendata->cancelled) {
3100                 if (opendata->lgp) {
3101                         nfs4_lgopen_release(opendata->lgp);
3102                         opendata->lgp = NULL;
3103                 }
3104                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3105         }
3106         return ret;
3107 }
3108
3109 /*
3110  * Returns a referenced nfs4_state
3111  */
3112 static int _nfs4_do_open(struct inode *dir,
3113                         struct nfs_open_context *ctx,
3114                         int flags,
3115                         const struct nfs4_open_createattrs *c,
3116                         int *opened)
3117 {
3118         struct nfs4_state_owner  *sp;
3119         struct nfs4_state     *state = NULL;
3120         struct nfs_server       *server = NFS_SERVER(dir);
3121         struct nfs4_opendata *opendata;
3122         struct dentry *dentry = ctx->dentry;
3123         const struct cred *cred = ctx->cred;
3124         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3125         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3126         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3127         struct iattr *sattr = c->sattr;
3128         struct nfs4_label *label = c->label;
3129         struct nfs4_label *olabel = NULL;
3130         int status;
3131
3132         /* Protect against reboot recovery conflicts */
3133         status = -ENOMEM;
3134         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3135         if (sp == NULL) {
3136                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3137                 goto out_err;
3138         }
3139         status = nfs4_client_recover_expired_lease(server->nfs_client);
3140         if (status != 0)
3141                 goto err_put_state_owner;
3142         if (d_really_is_positive(dentry))
3143                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3144         status = -ENOMEM;
3145         if (d_really_is_positive(dentry))
3146                 claim = NFS4_OPEN_CLAIM_FH;
3147         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3148                         c, claim, GFP_KERNEL);
3149         if (opendata == NULL)
3150                 goto err_put_state_owner;
3151
3152         if (label) {
3153                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3154                 if (IS_ERR(olabel)) {
3155                         status = PTR_ERR(olabel);
3156                         goto err_opendata_put;
3157                 }
3158         }
3159
3160         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3161                 if (!opendata->f_attr.mdsthreshold) {
3162                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3163                         if (!opendata->f_attr.mdsthreshold)
3164                                 goto err_free_label;
3165                 }
3166                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3167         }
3168         if (d_really_is_positive(dentry))
3169                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3170
3171         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3172         if (status != 0)
3173                 goto err_free_label;
3174         state = ctx->state;
3175
3176         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3177             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3178                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3179                 /*
3180                  * send create attributes which was not set by open
3181                  * with an extra setattr.
3182                  */
3183                 if (attrs || label) {
3184                         unsigned ia_old = sattr->ia_valid;
3185
3186                         sattr->ia_valid = attrs;
3187                         nfs_fattr_init(opendata->o_res.f_attr);
3188                         status = nfs4_do_setattr(state->inode, cred,
3189                                         opendata->o_res.f_attr, sattr,
3190                                         ctx, label, olabel);
3191                         if (status == 0) {
3192                                 nfs_setattr_update_inode(state->inode, sattr,
3193                                                 opendata->o_res.f_attr);
3194                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3195                         }
3196                         sattr->ia_valid = ia_old;
3197                 }
3198         }
3199         if (opened && opendata->file_created)
3200                 *opened = 1;
3201
3202         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3203                 *ctx_th = opendata->f_attr.mdsthreshold;
3204                 opendata->f_attr.mdsthreshold = NULL;
3205         }
3206
3207         nfs4_label_free(olabel);
3208
3209         nfs4_opendata_put(opendata);
3210         nfs4_put_state_owner(sp);
3211         return 0;
3212 err_free_label:
3213         nfs4_label_free(olabel);
3214 err_opendata_put:
3215         nfs4_opendata_put(opendata);
3216 err_put_state_owner:
3217         nfs4_put_state_owner(sp);
3218 out_err:
3219         return status;
3220 }
3221
3222
3223 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3224                                         struct nfs_open_context *ctx,
3225                                         int flags,
3226                                         struct iattr *sattr,
3227                                         struct nfs4_label *label,
3228                                         int *opened)
3229 {
3230         struct nfs_server *server = NFS_SERVER(dir);
3231         struct nfs4_exception exception = {
3232                 .interruptible = true,
3233         };
3234         struct nfs4_state *res;
3235         struct nfs4_open_createattrs c = {
3236                 .label = label,
3237                 .sattr = sattr,
3238                 .verf = {
3239                         [0] = (__u32)jiffies,
3240                         [1] = (__u32)current->pid,
3241                 },
3242         };
3243         int status;
3244
3245         do {
3246                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3247                 res = ctx->state;
3248                 trace_nfs4_open_file(ctx, flags, status);
3249                 if (status == 0)
3250                         break;
3251                 /* NOTE: BAD_SEQID means the server and client disagree about the
3252                  * book-keeping w.r.t. state-changing operations
3253                  * (OPEN/CLOSE/LOCK/LOCKU...)
3254                  * It is actually a sign of a bug on the client or on the server.
3255                  *
3256                  * If we receive a BAD_SEQID error in the particular case of
3257                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3258                  * have unhashed the old state_owner for us, and that we can
3259                  * therefore safely retry using a new one. We should still warn
3260                  * the user though...
3261                  */
3262                 if (status == -NFS4ERR_BAD_SEQID) {
3263                         pr_warn_ratelimited("NFS: v4 server %s "
3264                                         " returned a bad sequence-id error!\n",
3265                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3266                         exception.retry = 1;
3267                         continue;
3268                 }
3269                 /*
3270                  * BAD_STATEID on OPEN means that the server cancelled our
3271                  * state before it received the OPEN_CONFIRM.
3272                  * Recover by retrying the request as per the discussion
3273                  * on Page 181 of RFC3530.
3274                  */
3275                 if (status == -NFS4ERR_BAD_STATEID) {
3276                         exception.retry = 1;
3277                         continue;
3278                 }
3279                 if (status == -NFS4ERR_EXPIRED) {
3280                         nfs4_schedule_lease_recovery(server->nfs_client);
3281                         exception.retry = 1;
3282                         continue;
3283                 }
3284                 if (status == -EAGAIN) {
3285                         /* We must have found a delegation */
3286                         exception.retry = 1;
3287                         continue;
3288                 }
3289                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3290                         continue;
3291                 res = ERR_PTR(nfs4_handle_exception(server,
3292                                         status, &exception));
3293         } while (exception.retry);
3294         return res;
3295 }
3296
3297 static int _nfs4_do_setattr(struct inode *inode,
3298                             struct nfs_setattrargs *arg,
3299                             struct nfs_setattrres *res,
3300                             const struct cred *cred,
3301                             struct nfs_open_context *ctx)
3302 {
3303         struct nfs_server *server = NFS_SERVER(inode);
3304         struct rpc_message msg = {
3305                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3306                 .rpc_argp       = arg,
3307                 .rpc_resp       = res,
3308                 .rpc_cred       = cred,
3309         };
3310         const struct cred *delegation_cred = NULL;
3311         unsigned long timestamp = jiffies;
3312         bool truncate;
3313         int status;
3314
3315         nfs_fattr_init(res->fattr);
3316
3317         /* Servers should only apply open mode checks for file size changes */
3318         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3319         if (!truncate) {
3320                 nfs4_inode_make_writeable(inode);
3321                 goto zero_stateid;
3322         }
3323
3324         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3325                 /* Use that stateid */
3326         } else if (ctx != NULL && ctx->state) {
3327                 struct nfs_lock_context *l_ctx;
3328                 if (!nfs4_valid_open_stateid(ctx->state))
3329                         return -EBADF;
3330                 l_ctx = nfs_get_lock_context(ctx);
3331                 if (IS_ERR(l_ctx))
3332                         return PTR_ERR(l_ctx);
3333                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3334                                                 &arg->stateid, &delegation_cred);
3335                 nfs_put_lock_context(l_ctx);
3336                 if (status == -EIO)
3337                         return -EBADF;
3338                 else if (status == -EAGAIN)
3339                         goto zero_stateid;
3340         } else {
3341 zero_stateid:
3342                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3343         }
3344         if (delegation_cred)
3345                 msg.rpc_cred = delegation_cred;
3346
3347         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3348
3349         put_cred(delegation_cred);
3350         if (status == 0 && ctx != NULL)
3351                 renew_lease(server, timestamp);
3352         trace_nfs4_setattr(inode, &arg->stateid, status);
3353         return status;
3354 }
3355
3356 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3357                            struct nfs_fattr *fattr, struct iattr *sattr,
3358                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3359                            struct nfs4_label *olabel)
3360 {
3361         struct nfs_server *server = NFS_SERVER(inode);
3362         __u32 bitmask[NFS4_BITMASK_SZ];
3363         struct nfs4_state *state = ctx ? ctx->state : NULL;
3364         struct nfs_setattrargs  arg = {
3365                 .fh             = NFS_FH(inode),
3366                 .iap            = sattr,
3367                 .server         = server,
3368                 .bitmask = bitmask,
3369                 .label          = ilabel,
3370         };
3371         struct nfs_setattrres  res = {
3372                 .fattr          = fattr,
3373                 .label          = olabel,
3374                 .server         = server,
3375         };
3376         struct nfs4_exception exception = {
3377                 .state = state,
3378                 .inode = inode,
3379                 .stateid = &arg.stateid,
3380         };
3381         int err;
3382
3383         do {
3384                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3385                                 nfs4_bitmask(server, olabel),
3386                                 inode);
3387
3388                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3389                 switch (err) {
3390                 case -NFS4ERR_OPENMODE:
3391                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3392                                 pr_warn_once("NFSv4: server %s is incorrectly "
3393                                                 "applying open mode checks to "
3394                                                 "a SETATTR that is not "
3395                                                 "changing file size.\n",
3396                                                 server->nfs_client->cl_hostname);
3397                         }
3398                         if (state && !(state->state & FMODE_WRITE)) {
3399                                 err = -EBADF;
3400                                 if (sattr->ia_valid & ATTR_OPEN)
3401                                         err = -EACCES;
3402                                 goto out;
3403                         }
3404                 }
3405                 err = nfs4_handle_exception(server, err, &exception);
3406         } while (exception.retry);
3407 out:
3408         return err;
3409 }
3410
3411 static bool
3412 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3413 {
3414         if (inode == NULL || !nfs_have_layout(inode))
3415                 return false;
3416
3417         return pnfs_wait_on_layoutreturn(inode, task);
3418 }
3419
3420 /*
3421  * Update the seqid of an open stateid
3422  */
3423 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3424                 struct nfs4_state *state)
3425 {
3426         __be32 seqid_open;
3427         u32 dst_seqid;
3428         int seq;
3429
3430         for (;;) {
3431                 if (!nfs4_valid_open_stateid(state))
3432                         break;
3433                 seq = read_seqbegin(&state->seqlock);
3434                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3435                         nfs4_stateid_copy(dst, &state->open_stateid);
3436                         if (read_seqretry(&state->seqlock, seq))
3437                                 continue;
3438                         break;
3439                 }
3440                 seqid_open = state->open_stateid.seqid;
3441                 if (read_seqretry(&state->seqlock, seq))
3442                         continue;
3443
3444                 dst_seqid = be32_to_cpu(dst->seqid);
3445                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3446                         dst->seqid = seqid_open;
3447                 break;
3448         }
3449 }
3450
3451 /*
3452  * Update the seqid of an open stateid after receiving
3453  * NFS4ERR_OLD_STATEID
3454  */
3455 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3456                 struct nfs4_state *state)
3457 {
3458         __be32 seqid_open;
3459         u32 dst_seqid;
3460         bool ret;
3461         int seq, status = -EAGAIN;
3462         DEFINE_WAIT(wait);
3463
3464         for (;;) {
3465                 ret = false;
3466                 if (!nfs4_valid_open_stateid(state))
3467                         break;
3468                 seq = read_seqbegin(&state->seqlock);
3469                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3470                         if (read_seqretry(&state->seqlock, seq))
3471                                 continue;
3472                         break;
3473                 }
3474
3475                 write_seqlock(&state->seqlock);
3476                 seqid_open = state->open_stateid.seqid;
3477
3478                 dst_seqid = be32_to_cpu(dst->seqid);
3479
3480                 /* Did another OPEN bump the state's seqid?  try again: */
3481                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3482                         dst->seqid = seqid_open;
3483                         write_sequnlock(&state->seqlock);
3484                         ret = true;
3485                         break;
3486                 }
3487
3488                 /* server says we're behind but we haven't seen the update yet */
3489                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3490                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3491                 write_sequnlock(&state->seqlock);
3492                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3493
3494                 if (fatal_signal_pending(current))
3495                         status = -EINTR;
3496                 else
3497                         if (schedule_timeout(5*HZ) != 0)
3498                                 status = 0;
3499
3500                 finish_wait(&state->waitq, &wait);
3501
3502                 if (!status)
3503                         continue;
3504                 if (status == -EINTR)
3505                         break;
3506
3507                 /* we slept the whole 5 seconds, we must have lost a seqid */
3508                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3509                 ret = true;
3510                 break;
3511         }
3512
3513         return ret;
3514 }
3515
3516 struct nfs4_closedata {
3517         struct inode *inode;
3518         struct nfs4_state *state;
3519         struct nfs_closeargs arg;
3520         struct nfs_closeres res;
3521         struct {
3522                 struct nfs4_layoutreturn_args arg;
3523                 struct nfs4_layoutreturn_res res;
3524                 struct nfs4_xdr_opaque_data ld_private;
3525                 u32 roc_barrier;
3526                 bool roc;
3527         } lr;
3528         struct nfs_fattr fattr;
3529         unsigned long timestamp;
3530 };
3531
3532 static void nfs4_free_closedata(void *data)
3533 {
3534         struct nfs4_closedata *calldata = data;
3535         struct nfs4_state_owner *sp = calldata->state->owner;
3536         struct super_block *sb = calldata->state->inode->i_sb;
3537
3538         if (calldata->lr.roc)
3539                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3540                                 calldata->res.lr_ret);
3541         nfs4_put_open_state(calldata->state);
3542         nfs_free_seqid(calldata->arg.seqid);
3543         nfs4_put_state_owner(sp);
3544         nfs_sb_deactive(sb);
3545         kfree(calldata);
3546 }
3547
3548 static void nfs4_close_done(struct rpc_task *task, void *data)
3549 {
3550         struct nfs4_closedata *calldata = data;
3551         struct nfs4_state *state = calldata->state;
3552         struct nfs_server *server = NFS_SERVER(calldata->inode);
3553         nfs4_stateid *res_stateid = NULL;
3554         struct nfs4_exception exception = {
3555                 .state = state,
3556                 .inode = calldata->inode,
3557                 .stateid = &calldata->arg.stateid,
3558         };
3559
3560         dprintk("%s: begin!\n", __func__);
3561         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3562                 return;
3563         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3564
3565         /* Handle Layoutreturn errors */
3566         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3567                           &calldata->res.lr_ret) == -EAGAIN)
3568                 goto out_restart;
3569
3570         /* hmm. we are done with the inode, and in the process of freeing
3571          * the state_owner. we keep this around to process errors
3572          */
3573         switch (task->tk_status) {
3574                 case 0:
3575                         res_stateid = &calldata->res.stateid;
3576                         renew_lease(server, calldata->timestamp);
3577                         break;
3578                 case -NFS4ERR_ACCESS:
3579                         if (calldata->arg.bitmask != NULL) {
3580                                 calldata->arg.bitmask = NULL;
3581                                 calldata->res.fattr = NULL;
3582                                 goto out_restart;
3583
3584                         }
3585                         break;
3586                 case -NFS4ERR_OLD_STATEID:
3587                         /* Did we race with OPEN? */
3588                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3589                                                 state))
3590                                 goto out_restart;
3591                         goto out_release;
3592                 case -NFS4ERR_ADMIN_REVOKED:
3593                 case -NFS4ERR_STALE_STATEID:
3594                 case -NFS4ERR_EXPIRED:
3595                         nfs4_free_revoked_stateid(server,
3596                                         &calldata->arg.stateid,
3597                                         task->tk_msg.rpc_cred);
3598                         fallthrough;
3599                 case -NFS4ERR_BAD_STATEID:
3600                         if (calldata->arg.fmode == 0)
3601                                 break;
3602                         fallthrough;
3603                 default:
3604                         task->tk_status = nfs4_async_handle_exception(task,
3605                                         server, task->tk_status, &exception);
3606                         if (exception.retry)
3607                                 goto out_restart;
3608         }
3609         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3610                         res_stateid, calldata->arg.fmode);
3611 out_release:
3612         task->tk_status = 0;
3613         nfs_release_seqid(calldata->arg.seqid);
3614         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3615         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3616         return;
3617 out_restart:
3618         task->tk_status = 0;
3619         rpc_restart_call_prepare(task);
3620         goto out_release;
3621 }
3622
3623 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3624 {
3625         struct nfs4_closedata *calldata = data;
3626         struct nfs4_state *state = calldata->state;
3627         struct inode *inode = calldata->inode;
3628         struct nfs_server *server = NFS_SERVER(inode);
3629         struct pnfs_layout_hdr *lo;
3630         bool is_rdonly, is_wronly, is_rdwr;
3631         int call_close = 0;
3632
3633         dprintk("%s: begin!\n", __func__);
3634         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3635                 goto out_wait;
3636
3637         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3638         spin_lock(&state->owner->so_lock);
3639         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3640         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3641         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3642         /* Calculate the change in open mode */
3643         calldata->arg.fmode = 0;
3644         if (state->n_rdwr == 0) {
3645                 if (state->n_rdonly == 0)
3646                         call_close |= is_rdonly;
3647                 else if (is_rdonly)
3648                         calldata->arg.fmode |= FMODE_READ;
3649                 if (state->n_wronly == 0)
3650                         call_close |= is_wronly;
3651                 else if (is_wronly)
3652                         calldata->arg.fmode |= FMODE_WRITE;
3653                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3654                         call_close |= is_rdwr;
3655         } else if (is_rdwr)
3656                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3657
3658         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3659         if (!nfs4_valid_open_stateid(state))
3660                 call_close = 0;
3661         spin_unlock(&state->owner->so_lock);
3662
3663         if (!call_close) {
3664                 /* Note: exit _without_ calling nfs4_close_done */
3665                 goto out_no_action;
3666         }
3667
3668         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3669                 nfs_release_seqid(calldata->arg.seqid);
3670                 goto out_wait;
3671         }
3672
3673         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3674         if (lo && !pnfs_layout_is_valid(lo)) {
3675                 calldata->arg.lr_args = NULL;
3676                 calldata->res.lr_res = NULL;
3677         }
3678
3679         if (calldata->arg.fmode == 0)
3680                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3681
3682         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3683                 /* Close-to-open cache consistency revalidation */
3684                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3685                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3686                                          server->cache_consistency_bitmask,
3687                                          inode, server, NULL);
3688                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3689                 } else
3690                         calldata->arg.bitmask = NULL;
3691         }
3692
3693         calldata->arg.share_access =
3694                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3695                                 calldata->arg.fmode, 0);
3696
3697         if (calldata->res.fattr == NULL)
3698                 calldata->arg.bitmask = NULL;
3699         else if (calldata->arg.bitmask == NULL)
3700                 calldata->res.fattr = NULL;
3701         calldata->timestamp = jiffies;
3702         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3703                                 &calldata->arg.seq_args,
3704                                 &calldata->res.seq_res,
3705                                 task) != 0)
3706                 nfs_release_seqid(calldata->arg.seqid);
3707         dprintk("%s: done!\n", __func__);
3708         return;
3709 out_no_action:
3710         task->tk_action = NULL;
3711 out_wait:
3712         nfs4_sequence_done(task, &calldata->res.seq_res);
3713 }
3714
3715 static const struct rpc_call_ops nfs4_close_ops = {
3716         .rpc_call_prepare = nfs4_close_prepare,
3717         .rpc_call_done = nfs4_close_done,
3718         .rpc_release = nfs4_free_closedata,
3719 };
3720
3721 /*
3722  * It is possible for data to be read/written from a mem-mapped file
3723  * after the sys_close call (which hits the vfs layer as a flush).
3724  * This means that we can't safely call nfsv4 close on a file until
3725  * the inode is cleared. This in turn means that we are not good
3726  * NFSv4 citizens - we do not indicate to the server to update the file's
3727  * share state even when we are done with one of the three share
3728  * stateid's in the inode.
3729  *
3730  * NOTE: Caller must be holding the sp->so_owner semaphore!
3731  */
3732 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3733 {
3734         struct nfs_server *server = NFS_SERVER(state->inode);
3735         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3736         struct nfs4_closedata *calldata;
3737         struct nfs4_state_owner *sp = state->owner;
3738         struct rpc_task *task;
3739         struct rpc_message msg = {
3740                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3741                 .rpc_cred = state->owner->so_cred,
3742         };
3743         struct rpc_task_setup task_setup_data = {
3744                 .rpc_client = server->client,
3745                 .rpc_message = &msg,
3746                 .callback_ops = &nfs4_close_ops,
3747                 .workqueue = nfsiod_workqueue,
3748                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3749         };
3750         int status = -ENOMEM;
3751
3752         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3753                 &task_setup_data.rpc_client, &msg);
3754
3755         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3756         if (calldata == NULL)
3757                 goto out;
3758         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3759         calldata->inode = state->inode;
3760         calldata->state = state;
3761         calldata->arg.fh = NFS_FH(state->inode);
3762         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3763                 goto out_free_calldata;
3764         /* Serialization for the sequence id */
3765         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3766         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3767         if (IS_ERR(calldata->arg.seqid))
3768                 goto out_free_calldata;
3769         nfs_fattr_init(&calldata->fattr);
3770         calldata->arg.fmode = 0;
3771         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3772         calldata->res.fattr = &calldata->fattr;
3773         calldata->res.seqid = calldata->arg.seqid;
3774         calldata->res.server = server;
3775         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3776         calldata->lr.roc = pnfs_roc(state->inode,
3777                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3778         if (calldata->lr.roc) {
3779                 calldata->arg.lr_args = &calldata->lr.arg;
3780                 calldata->res.lr_res = &calldata->lr.res;
3781         }
3782         nfs_sb_active(calldata->inode->i_sb);
3783
3784         msg.rpc_argp = &calldata->arg;
3785         msg.rpc_resp = &calldata->res;
3786         task_setup_data.callback_data = calldata;
3787         task = rpc_run_task(&task_setup_data);
3788         if (IS_ERR(task))
3789                 return PTR_ERR(task);
3790         status = 0;
3791         if (wait)
3792                 status = rpc_wait_for_completion_task(task);
3793         rpc_put_task(task);
3794         return status;
3795 out_free_calldata:
3796         kfree(calldata);
3797 out:
3798         nfs4_put_open_state(state);
3799         nfs4_put_state_owner(sp);
3800         return status;
3801 }
3802
3803 static struct inode *
3804 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3805                 int open_flags, struct iattr *attr, int *opened)
3806 {
3807         struct nfs4_state *state;
3808         struct nfs4_label l, *label;
3809
3810         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3811
3812         /* Protect against concurrent sillydeletes */
3813         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3814
3815         nfs4_label_release_security(label);
3816
3817         if (IS_ERR(state))
3818                 return ERR_CAST(state);
3819         return state->inode;
3820 }
3821
3822 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3823 {
3824         if (ctx->state == NULL)
3825                 return;
3826         if (is_sync)
3827                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3828         else
3829                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3830 }
3831
3832 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3833 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3834 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3835
3836 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3837 {
3838         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3839         struct nfs4_server_caps_arg args = {
3840                 .fhandle = fhandle,
3841                 .bitmask = bitmask,
3842         };
3843         struct nfs4_server_caps_res res = {};
3844         struct rpc_message msg = {
3845                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3846                 .rpc_argp = &args,
3847                 .rpc_resp = &res,
3848         };
3849         int status;
3850         int i;
3851
3852         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3853                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3854                      FATTR4_WORD0_LINK_SUPPORT |
3855                      FATTR4_WORD0_SYMLINK_SUPPORT |
3856                      FATTR4_WORD0_ACLSUPPORT;
3857         if (minorversion)
3858                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3859
3860         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3861         if (status == 0) {
3862                 /* Sanity check the server answers */
3863                 switch (minorversion) {
3864                 case 0:
3865                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3866                         res.attr_bitmask[2] = 0;
3867                         break;
3868                 case 1:
3869                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3870                         break;
3871                 case 2:
3872                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3873                 }
3874                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3875                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3876                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3877                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3878                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3879                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3880                                 NFS_CAP_SECURITY_LABEL);
3881                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3882                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3883                         server->caps |= NFS_CAP_ACLS;
3884                 if (res.has_links != 0)
3885                         server->caps |= NFS_CAP_HARDLINKS;
3886                 if (res.has_symlinks != 0)
3887                         server->caps |= NFS_CAP_SYMLINKS;
3888                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3889                         server->caps |= NFS_CAP_FILEID;
3890                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3891                         server->caps |= NFS_CAP_MODE;
3892                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3893                         server->caps |= NFS_CAP_NLINK;
3894                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3895                         server->caps |= NFS_CAP_OWNER;
3896                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3897                         server->caps |= NFS_CAP_OWNER_GROUP;
3898                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3899                         server->caps |= NFS_CAP_ATIME;
3900                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3901                         server->caps |= NFS_CAP_CTIME;
3902                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3903                         server->caps |= NFS_CAP_MTIME;
3904 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3905                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3906                         server->caps |= NFS_CAP_SECURITY_LABEL;
3907 #endif
3908                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3909                                 sizeof(server->attr_bitmask));
3910                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3911
3912                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3913                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3914                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3915                 server->cache_consistency_bitmask[2] = 0;
3916
3917                 /* Avoid a regression due to buggy server */
3918                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3919                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3920                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3921                         sizeof(server->exclcreat_bitmask));
3922
3923                 server->acl_bitmask = res.acl_bitmask;
3924                 server->fh_expire_type = res.fh_expire_type;
3925         }
3926
3927         return status;
3928 }
3929
3930 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3931 {
3932         struct nfs4_exception exception = {
3933                 .interruptible = true,
3934         };
3935         int err;
3936         do {
3937                 err = nfs4_handle_exception(server,
3938                                 _nfs4_server_capabilities(server, fhandle),
3939                                 &exception);
3940         } while (exception.retry);
3941         return err;
3942 }
3943
3944 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3945                 struct nfs_fsinfo *info)
3946 {
3947         u32 bitmask[3];
3948         struct nfs4_lookup_root_arg args = {
3949                 .bitmask = bitmask,
3950         };
3951         struct nfs4_lookup_res res = {
3952                 .server = server,
3953                 .fattr = info->fattr,
3954                 .fh = fhandle,
3955         };
3956         struct rpc_message msg = {
3957                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3958                 .rpc_argp = &args,
3959                 .rpc_resp = &res,
3960         };
3961
3962         bitmask[0] = nfs4_fattr_bitmap[0];
3963         bitmask[1] = nfs4_fattr_bitmap[1];
3964         /*
3965          * Process the label in the upcoming getfattr
3966          */
3967         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3968
3969         nfs_fattr_init(info->fattr);
3970         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3971 }
3972
3973 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3974                 struct nfs_fsinfo *info)
3975 {
3976         struct nfs4_exception exception = {
3977                 .interruptible = true,
3978         };
3979         int err;
3980         do {
3981                 err = _nfs4_lookup_root(server, fhandle, info);
3982                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3983                 switch (err) {
3984                 case 0:
3985                 case -NFS4ERR_WRONGSEC:
3986                         goto out;
3987                 default:
3988                         err = nfs4_handle_exception(server, err, &exception);
3989                 }
3990         } while (exception.retry);
3991 out:
3992         return err;
3993 }
3994
3995 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3996                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3997 {
3998         struct rpc_auth_create_args auth_args = {
3999                 .pseudoflavor = flavor,
4000         };
4001         struct rpc_auth *auth;
4002
4003         auth = rpcauth_create(&auth_args, server->client);
4004         if (IS_ERR(auth))
4005                 return -EACCES;
4006         return nfs4_lookup_root(server, fhandle, info);
4007 }
4008
4009 /*
4010  * Retry pseudoroot lookup with various security flavors.  We do this when:
4011  *
4012  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4013  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4014  *
4015  * Returns zero on success, or a negative NFS4ERR value, or a
4016  * negative errno value.
4017  */
4018 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4019                               struct nfs_fsinfo *info)
4020 {
4021         /* Per 3530bis 15.33.5 */
4022         static const rpc_authflavor_t flav_array[] = {
4023                 RPC_AUTH_GSS_KRB5P,
4024                 RPC_AUTH_GSS_KRB5I,
4025                 RPC_AUTH_GSS_KRB5,
4026                 RPC_AUTH_UNIX,                  /* courtesy */
4027                 RPC_AUTH_NULL,
4028         };
4029         int status = -EPERM;
4030         size_t i;
4031
4032         if (server->auth_info.flavor_len > 0) {
4033                 /* try each flavor specified by user */
4034                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4035                         status = nfs4_lookup_root_sec(server, fhandle, info,
4036                                                 server->auth_info.flavors[i]);
4037                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4038                                 continue;
4039                         break;
4040                 }
4041         } else {
4042                 /* no flavors specified by user, try default list */
4043                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4044                         status = nfs4_lookup_root_sec(server, fhandle, info,
4045                                                       flav_array[i]);
4046                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4047                                 continue;
4048                         break;
4049                 }
4050         }
4051
4052         /*
4053          * -EACCES could mean that the user doesn't have correct permissions
4054          * to access the mount.  It could also mean that we tried to mount
4055          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4056          * existing mount programs don't handle -EACCES very well so it should
4057          * be mapped to -EPERM instead.
4058          */
4059         if (status == -EACCES)
4060                 status = -EPERM;
4061         return status;
4062 }
4063
4064 /**
4065  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4066  * @server: initialized nfs_server handle
4067  * @fhandle: we fill in the pseudo-fs root file handle
4068  * @info: we fill in an FSINFO struct
4069  * @auth_probe: probe the auth flavours
4070  *
4071  * Returns zero on success, or a negative errno.
4072  */
4073 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4074                          struct nfs_fsinfo *info,
4075                          bool auth_probe)
4076 {
4077         int status = 0;
4078
4079         if (!auth_probe)
4080                 status = nfs4_lookup_root(server, fhandle, info);
4081
4082         if (auth_probe || status == NFS4ERR_WRONGSEC)
4083                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4084                                 fhandle, info);
4085
4086         if (status == 0)
4087                 status = nfs4_server_capabilities(server, fhandle);
4088         if (status == 0)
4089                 status = nfs4_do_fsinfo(server, fhandle, info);
4090
4091         return nfs4_map_errors(status);
4092 }
4093
4094 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4095                               struct nfs_fsinfo *info)
4096 {
4097         int error;
4098         struct nfs_fattr *fattr = info->fattr;
4099         struct nfs4_label *label = fattr->label;
4100
4101         error = nfs4_server_capabilities(server, mntfh);
4102         if (error < 0) {
4103                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4104                 return error;
4105         }
4106
4107         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4108         if (error < 0) {
4109                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4110                 goto out;
4111         }
4112
4113         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4114             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4115                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4116
4117 out:
4118         return error;
4119 }
4120
4121 /*
4122  * Get locations and (maybe) other attributes of a referral.
4123  * Note that we'll actually follow the referral later when
4124  * we detect fsid mismatch in inode revalidation
4125  */
4126 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4127                              const struct qstr *name, struct nfs_fattr *fattr,
4128                              struct nfs_fh *fhandle)
4129 {
4130         int status = -ENOMEM;
4131         struct page *page = NULL;
4132         struct nfs4_fs_locations *locations = NULL;
4133
4134         page = alloc_page(GFP_KERNEL);
4135         if (page == NULL)
4136                 goto out;
4137         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4138         if (locations == NULL)
4139                 goto out;
4140
4141         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4142         if (status != 0)
4143                 goto out;
4144
4145         /*
4146          * If the fsid didn't change, this is a migration event, not a
4147          * referral.  Cause us to drop into the exception handler, which
4148          * will kick off migration recovery.
4149          */
4150         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4151                 dprintk("%s: server did not return a different fsid for"
4152                         " a referral at %s\n", __func__, name->name);
4153                 status = -NFS4ERR_MOVED;
4154                 goto out;
4155         }
4156         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4157         nfs_fixup_referral_attributes(&locations->fattr);
4158
4159         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4160         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4161         memset(fhandle, 0, sizeof(struct nfs_fh));
4162 out:
4163         if (page)
4164                 __free_page(page);
4165         kfree(locations);
4166         return status;
4167 }
4168
4169 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4170                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4171                                 struct inode *inode)
4172 {
4173         __u32 bitmask[NFS4_BITMASK_SZ];
4174         struct nfs4_getattr_arg args = {
4175                 .fh = fhandle,
4176                 .bitmask = bitmask,
4177         };
4178         struct nfs4_getattr_res res = {
4179                 .fattr = fattr,
4180                 .label = label,
4181                 .server = server,
4182         };
4183         struct rpc_message msg = {
4184                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4185                 .rpc_argp = &args,
4186                 .rpc_resp = &res,
4187         };
4188         unsigned short task_flags = 0;
4189
4190         /* Is this is an attribute revalidation, subject to softreval? */
4191         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4192                 task_flags |= RPC_TASK_TIMEOUT;
4193
4194         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4195
4196         nfs_fattr_init(fattr);
4197         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4198         return nfs4_do_call_sync(server->client, server, &msg,
4199                         &args.seq_args, &res.seq_res, task_flags);
4200 }
4201
4202 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4203                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4204                                 struct inode *inode)
4205 {
4206         struct nfs4_exception exception = {
4207                 .interruptible = true,
4208         };
4209         int err;
4210         do {
4211                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4212                 trace_nfs4_getattr(server, fhandle, fattr, err);
4213                 err = nfs4_handle_exception(server, err,
4214                                 &exception);
4215         } while (exception.retry);
4216         return err;
4217 }
4218
4219 /*
4220  * The file is not closed if it is opened due to the a request to change
4221  * the size of the file. The open call will not be needed once the
4222  * VFS layer lookup-intents are implemented.
4223  *
4224  * Close is called when the inode is destroyed.
4225  * If we haven't opened the file for O_WRONLY, we
4226  * need to in the size_change case to obtain a stateid.
4227  *
4228  * Got race?
4229  * Because OPEN is always done by name in nfsv4, it is
4230  * possible that we opened a different file by the same
4231  * name.  We can recognize this race condition, but we
4232  * can't do anything about it besides returning an error.
4233  *
4234  * This will be fixed with VFS changes (lookup-intent).
4235  */
4236 static int
4237 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4238                   struct iattr *sattr)
4239 {
4240         struct inode *inode = d_inode(dentry);
4241         const struct cred *cred = NULL;
4242         struct nfs_open_context *ctx = NULL;
4243         struct nfs4_label *label = NULL;
4244         int status;
4245
4246         if (pnfs_ld_layoutret_on_setattr(inode) &&
4247             sattr->ia_valid & ATTR_SIZE &&
4248             sattr->ia_size < i_size_read(inode))
4249                 pnfs_commit_and_return_layout(inode);
4250
4251         nfs_fattr_init(fattr);
4252
4253         /* Deal with open(O_TRUNC) */
4254         if (sattr->ia_valid & ATTR_OPEN)
4255                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4256
4257         /* Optimization: if the end result is no change, don't RPC */
4258         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4259                 return 0;
4260
4261         /* Search for an existing open(O_WRITE) file */
4262         if (sattr->ia_valid & ATTR_FILE) {
4263
4264                 ctx = nfs_file_open_context(sattr->ia_file);
4265                 if (ctx)
4266                         cred = ctx->cred;
4267         }
4268
4269         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4270         if (IS_ERR(label))
4271                 return PTR_ERR(label);
4272
4273         /* Return any delegations if we're going to change ACLs */
4274         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4275                 nfs4_inode_make_writeable(inode);
4276
4277         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4278         if (status == 0) {
4279                 nfs_setattr_update_inode(inode, sattr, fattr);
4280                 nfs_setsecurity(inode, fattr, label);
4281         }
4282         nfs4_label_free(label);
4283         return status;
4284 }
4285
4286 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4287                 struct dentry *dentry, struct nfs_fh *fhandle,
4288                 struct nfs_fattr *fattr, struct nfs4_label *label)
4289 {
4290         struct nfs_server *server = NFS_SERVER(dir);
4291         int                    status;
4292         struct nfs4_lookup_arg args = {
4293                 .bitmask = server->attr_bitmask,
4294                 .dir_fh = NFS_FH(dir),
4295                 .name = &dentry->d_name,
4296         };
4297         struct nfs4_lookup_res res = {
4298                 .server = server,
4299                 .fattr = fattr,
4300                 .label = label,
4301                 .fh = fhandle,
4302         };
4303         struct rpc_message msg = {
4304                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4305                 .rpc_argp = &args,
4306                 .rpc_resp = &res,
4307         };
4308         unsigned short task_flags = 0;
4309
4310         /* Is this is an attribute revalidation, subject to softreval? */
4311         if (nfs_lookup_is_soft_revalidate(dentry))
4312                 task_flags |= RPC_TASK_TIMEOUT;
4313
4314         args.bitmask = nfs4_bitmask(server, label);
4315
4316         nfs_fattr_init(fattr);
4317
4318         dprintk("NFS call  lookup %pd2\n", dentry);
4319         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4320         status = nfs4_do_call_sync(clnt, server, &msg,
4321                         &args.seq_args, &res.seq_res, task_flags);
4322         dprintk("NFS reply lookup: %d\n", status);
4323         return status;
4324 }
4325
4326 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4327 {
4328         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4329                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4330         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4331         fattr->nlink = 2;
4332 }
4333
4334 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4335                                    struct dentry *dentry, struct nfs_fh *fhandle,
4336                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4337 {
4338         struct nfs4_exception exception = {
4339                 .interruptible = true,
4340         };
4341         struct rpc_clnt *client = *clnt;
4342         const struct qstr *name = &dentry->d_name;
4343         int err;
4344         do {
4345                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4346                 trace_nfs4_lookup(dir, name, err);
4347                 switch (err) {
4348                 case -NFS4ERR_BADNAME:
4349                         err = -ENOENT;
4350                         goto out;
4351                 case -NFS4ERR_MOVED:
4352                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4353                         if (err == -NFS4ERR_MOVED)
4354                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4355                         goto out;
4356                 case -NFS4ERR_WRONGSEC:
4357                         err = -EPERM;
4358                         if (client != *clnt)
4359                                 goto out;
4360                         client = nfs4_negotiate_security(client, dir, name);
4361                         if (IS_ERR(client))
4362                                 return PTR_ERR(client);
4363
4364                         exception.retry = 1;
4365                         break;
4366                 default:
4367                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4368                 }
4369         } while (exception.retry);
4370
4371 out:
4372         if (err == 0)
4373                 *clnt = client;
4374         else if (client != *clnt)
4375                 rpc_shutdown_client(client);
4376
4377         return err;
4378 }
4379
4380 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4381                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4382                             struct nfs4_label *label)
4383 {
4384         int status;
4385         struct rpc_clnt *client = NFS_CLIENT(dir);
4386
4387         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4388         if (client != NFS_CLIENT(dir)) {
4389                 rpc_shutdown_client(client);
4390                 nfs_fixup_secinfo_attributes(fattr);
4391         }
4392         return status;
4393 }
4394
4395 struct rpc_clnt *
4396 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4397                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4398 {
4399         struct rpc_clnt *client = NFS_CLIENT(dir);
4400         int status;
4401
4402         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4403         if (status < 0)
4404                 return ERR_PTR(status);
4405         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4406 }
4407
4408 static int _nfs4_proc_lookupp(struct inode *inode,
4409                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4410                 struct nfs4_label *label)
4411 {
4412         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4413         struct nfs_server *server = NFS_SERVER(inode);
4414         int                    status;
4415         struct nfs4_lookupp_arg args = {
4416                 .bitmask = server->attr_bitmask,
4417                 .fh = NFS_FH(inode),
4418         };
4419         struct nfs4_lookupp_res res = {
4420                 .server = server,
4421                 .fattr = fattr,
4422                 .label = label,
4423                 .fh = fhandle,
4424         };
4425         struct rpc_message msg = {
4426                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4427                 .rpc_argp = &args,
4428                 .rpc_resp = &res,
4429         };
4430
4431         args.bitmask = nfs4_bitmask(server, label);
4432
4433         nfs_fattr_init(fattr);
4434
4435         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4436         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4437                                 &res.seq_res, 0);
4438         dprintk("NFS reply lookupp: %d\n", status);
4439         return status;
4440 }
4441
4442 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4443                              struct nfs_fattr *fattr, struct nfs4_label *label)
4444 {
4445         struct nfs4_exception exception = {
4446                 .interruptible = true,
4447         };
4448         int err;
4449         do {
4450                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4451                 trace_nfs4_lookupp(inode, err);
4452                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4453                                 &exception);
4454         } while (exception.retry);
4455         return err;
4456 }
4457
4458 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4459 {
4460         struct nfs_server *server = NFS_SERVER(inode);
4461         struct nfs4_accessargs args = {
4462                 .fh = NFS_FH(inode),
4463                 .access = entry->mask,
4464         };
4465         struct nfs4_accessres res = {
4466                 .server = server,
4467         };
4468         struct rpc_message msg = {
4469                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4470                 .rpc_argp = &args,
4471                 .rpc_resp = &res,
4472                 .rpc_cred = entry->cred,
4473         };
4474         int status = 0;
4475
4476         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4477                 res.fattr = nfs_alloc_fattr();
4478                 if (res.fattr == NULL)
4479                         return -ENOMEM;
4480                 args.bitmask = server->cache_consistency_bitmask;
4481         }
4482         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4483         if (!status) {
4484                 nfs_access_set_mask(entry, res.access);
4485                 if (res.fattr)
4486                         nfs_refresh_inode(inode, res.fattr);
4487         }
4488         nfs_free_fattr(res.fattr);
4489         return status;
4490 }
4491
4492 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4493 {
4494         struct nfs4_exception exception = {
4495                 .interruptible = true,
4496         };
4497         int err;
4498         do {
4499                 err = _nfs4_proc_access(inode, entry);
4500                 trace_nfs4_access(inode, err);
4501                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4502                                 &exception);
4503         } while (exception.retry);
4504         return err;
4505 }
4506
4507 /*
4508  * TODO: For the time being, we don't try to get any attributes
4509  * along with any of the zero-copy operations READ, READDIR,
4510  * READLINK, WRITE.
4511  *
4512  * In the case of the first three, we want to put the GETATTR
4513  * after the read-type operation -- this is because it is hard
4514  * to predict the length of a GETATTR response in v4, and thus
4515  * align the READ data correctly.  This means that the GETATTR
4516  * may end up partially falling into the page cache, and we should
4517  * shift it into the 'tail' of the xdr_buf before processing.
4518  * To do this efficiently, we need to know the total length
4519  * of data received, which doesn't seem to be available outside
4520  * of the RPC layer.
4521  *
4522  * In the case of WRITE, we also want to put the GETATTR after
4523  * the operation -- in this case because we want to make sure
4524  * we get the post-operation mtime and size.
4525  *
4526  * Both of these changes to the XDR layer would in fact be quite
4527  * minor, but I decided to leave them for a subsequent patch.
4528  */
4529 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4530                 unsigned int pgbase, unsigned int pglen)
4531 {
4532         struct nfs4_readlink args = {
4533                 .fh       = NFS_FH(inode),
4534                 .pgbase   = pgbase,
4535                 .pglen    = pglen,
4536                 .pages    = &page,
4537         };
4538         struct nfs4_readlink_res res;
4539         struct rpc_message msg = {
4540                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4541                 .rpc_argp = &args,
4542                 .rpc_resp = &res,
4543         };
4544
4545         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4546 }
4547
4548 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4549                 unsigned int pgbase, unsigned int pglen)
4550 {
4551         struct nfs4_exception exception = {
4552                 .interruptible = true,
4553         };
4554         int err;
4555         do {
4556                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4557                 trace_nfs4_readlink(inode, err);
4558                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4559                                 &exception);
4560         } while (exception.retry);
4561         return err;
4562 }
4563
4564 /*
4565  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4566  */
4567 static int
4568 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4569                  int flags)
4570 {
4571         struct nfs_server *server = NFS_SERVER(dir);
4572         struct nfs4_label l, *ilabel;
4573         struct nfs_open_context *ctx;
4574         struct nfs4_state *state;
4575         int status = 0;
4576
4577         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4578         if (IS_ERR(ctx))
4579                 return PTR_ERR(ctx);
4580
4581         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4582
4583         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4584                 sattr->ia_mode &= ~current_umask();
4585         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4586         if (IS_ERR(state)) {
4587                 status = PTR_ERR(state);
4588                 goto out;
4589         }
4590 out:
4591         nfs4_label_release_security(ilabel);
4592         put_nfs_open_context(ctx);
4593         return status;
4594 }
4595
4596 static int
4597 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4598 {
4599         struct nfs_server *server = NFS_SERVER(dir);
4600         struct nfs_removeargs args = {
4601                 .fh = NFS_FH(dir),
4602                 .name = *name,
4603         };
4604         struct nfs_removeres res = {
4605                 .server = server,
4606         };
4607         struct rpc_message msg = {
4608                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4609                 .rpc_argp = &args,
4610                 .rpc_resp = &res,
4611         };
4612         unsigned long timestamp = jiffies;
4613         int status;
4614
4615         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4616         if (status == 0) {
4617                 spin_lock(&dir->i_lock);
4618                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4619                                               NFS_INO_INVALID_DATA);
4620                 /* Removing a directory decrements nlink in the parent */
4621                 if (ftype == NF4DIR && dir->i_nlink > 2)
4622                         nfs4_dec_nlink_locked(dir);
4623                 spin_unlock(&dir->i_lock);
4624         }
4625         return status;
4626 }
4627
4628 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4629 {
4630         struct nfs4_exception exception = {
4631                 .interruptible = true,
4632         };
4633         struct inode *inode = d_inode(dentry);
4634         int err;
4635
4636         if (inode) {
4637                 if (inode->i_nlink == 1)
4638                         nfs4_inode_return_delegation(inode);
4639                 else
4640                         nfs4_inode_make_writeable(inode);
4641         }
4642         do {
4643                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4644                 trace_nfs4_remove(dir, &dentry->d_name, err);
4645                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4646                                 &exception);
4647         } while (exception.retry);
4648         return err;
4649 }
4650
4651 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4652 {
4653         struct nfs4_exception exception = {
4654                 .interruptible = true,
4655         };
4656         int err;
4657
4658         do {
4659                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4660                 trace_nfs4_remove(dir, name, err);
4661                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4662                                 &exception);
4663         } while (exception.retry);
4664         return err;
4665 }
4666
4667 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4668                 struct dentry *dentry,
4669                 struct inode *inode)
4670 {
4671         struct nfs_removeargs *args = msg->rpc_argp;
4672         struct nfs_removeres *res = msg->rpc_resp;
4673
4674         res->server = NFS_SB(dentry->d_sb);
4675         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4676         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4677
4678         nfs_fattr_init(res->dir_attr);
4679
4680         if (inode)
4681                 nfs4_inode_return_delegation(inode);
4682 }
4683
4684 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4685 {
4686         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4687                         &data->args.seq_args,
4688                         &data->res.seq_res,
4689                         task);
4690 }
4691
4692 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4693 {
4694         struct nfs_unlinkdata *data = task->tk_calldata;
4695         struct nfs_removeres *res = &data->res;
4696
4697         if (!nfs4_sequence_done(task, &res->seq_res))
4698                 return 0;
4699         if (nfs4_async_handle_error(task, res->server, NULL,
4700                                     &data->timeout) == -EAGAIN)
4701                 return 0;
4702         if (task->tk_status == 0)
4703                 nfs4_update_changeattr(dir, &res->cinfo,
4704                                 res->dir_attr->time_start,
4705                                 NFS_INO_INVALID_DATA);
4706         return 1;
4707 }
4708
4709 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4710                 struct dentry *old_dentry,
4711                 struct dentry *new_dentry)
4712 {
4713         struct nfs_renameargs *arg = msg->rpc_argp;
4714         struct nfs_renameres *res = msg->rpc_resp;
4715         struct inode *old_inode = d_inode(old_dentry);
4716         struct inode *new_inode = d_inode(new_dentry);
4717
4718         if (old_inode)
4719                 nfs4_inode_make_writeable(old_inode);
4720         if (new_inode)
4721                 nfs4_inode_return_delegation(new_inode);
4722         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4723         res->server = NFS_SB(old_dentry->d_sb);
4724         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4725 }
4726
4727 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4728 {
4729         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4730                         &data->args.seq_args,
4731                         &data->res.seq_res,
4732                         task);
4733 }
4734
4735 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4736                                  struct inode *new_dir)
4737 {
4738         struct nfs_renamedata *data = task->tk_calldata;
4739         struct nfs_renameres *res = &data->res;
4740
4741         if (!nfs4_sequence_done(task, &res->seq_res))
4742                 return 0;
4743         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4744                 return 0;
4745
4746         if (task->tk_status == 0) {
4747                 if (new_dir != old_dir) {
4748                         /* Note: If we moved a directory, nlink will change */
4749                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4750                                         res->old_fattr->time_start,
4751                                         NFS_INO_INVALID_OTHER |
4752                                             NFS_INO_INVALID_DATA);
4753                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4754                                         res->new_fattr->time_start,
4755                                         NFS_INO_INVALID_OTHER |
4756                                             NFS_INO_INVALID_DATA);
4757                 } else
4758                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4759                                         res->old_fattr->time_start,
4760                                         NFS_INO_INVALID_DATA);
4761         }
4762         return 1;
4763 }
4764
4765 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4766 {
4767         struct nfs_server *server = NFS_SERVER(inode);
4768         __u32 bitmask[NFS4_BITMASK_SZ];
4769         struct nfs4_link_arg arg = {
4770                 .fh     = NFS_FH(inode),
4771                 .dir_fh = NFS_FH(dir),
4772                 .name   = name,
4773                 .bitmask = bitmask,
4774         };
4775         struct nfs4_link_res res = {
4776                 .server = server,
4777                 .label = NULL,
4778         };
4779         struct rpc_message msg = {
4780                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4781                 .rpc_argp = &arg,
4782                 .rpc_resp = &res,
4783         };
4784         int status = -ENOMEM;
4785
4786         res.fattr = nfs_alloc_fattr();
4787         if (res.fattr == NULL)
4788                 goto out;
4789
4790         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4791         if (IS_ERR(res.label)) {
4792                 status = PTR_ERR(res.label);
4793                 goto out;
4794         }
4795
4796         nfs4_inode_make_writeable(inode);
4797         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4798
4799         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4800         if (!status) {
4801                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4802                                        NFS_INO_INVALID_DATA);
4803                 status = nfs_post_op_update_inode(inode, res.fattr);
4804                 if (!status)
4805                         nfs_setsecurity(inode, res.fattr, res.label);
4806         }
4807
4808
4809         nfs4_label_free(res.label);
4810
4811 out:
4812         nfs_free_fattr(res.fattr);
4813         return status;
4814 }
4815
4816 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4817 {
4818         struct nfs4_exception exception = {
4819                 .interruptible = true,
4820         };
4821         int err;
4822         do {
4823                 err = nfs4_handle_exception(NFS_SERVER(inode),
4824                                 _nfs4_proc_link(inode, dir, name),
4825                                 &exception);
4826         } while (exception.retry);
4827         return err;
4828 }
4829
4830 struct nfs4_createdata {
4831         struct rpc_message msg;
4832         struct nfs4_create_arg arg;
4833         struct nfs4_create_res res;
4834         struct nfs_fh fh;
4835         struct nfs_fattr fattr;
4836         struct nfs4_label *label;
4837 };
4838
4839 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4840                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4841 {
4842         struct nfs4_createdata *data;
4843
4844         data = kzalloc(sizeof(*data), GFP_KERNEL);
4845         if (data != NULL) {
4846                 struct nfs_server *server = NFS_SERVER(dir);
4847
4848                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4849                 if (IS_ERR(data->label))
4850                         goto out_free;
4851
4852                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4853                 data->msg.rpc_argp = &data->arg;
4854                 data->msg.rpc_resp = &data->res;
4855                 data->arg.dir_fh = NFS_FH(dir);
4856                 data->arg.server = server;
4857                 data->arg.name = name;
4858                 data->arg.attrs = sattr;
4859                 data->arg.ftype = ftype;
4860                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4861                 data->arg.umask = current_umask();
4862                 data->res.server = server;
4863                 data->res.fh = &data->fh;
4864                 data->res.fattr = &data->fattr;
4865                 data->res.label = data->label;
4866                 nfs_fattr_init(data->res.fattr);
4867         }
4868         return data;
4869 out_free:
4870         kfree(data);
4871         return NULL;
4872 }
4873
4874 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4875 {
4876         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4877                                     &data->arg.seq_args, &data->res.seq_res, 1);
4878         if (status == 0) {
4879                 spin_lock(&dir->i_lock);
4880                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4881                                 data->res.fattr->time_start,
4882                                 NFS_INO_INVALID_DATA);
4883                 /* Creating a directory bumps nlink in the parent */
4884                 if (data->arg.ftype == NF4DIR)
4885                         nfs4_inc_nlink_locked(dir);
4886                 spin_unlock(&dir->i_lock);
4887                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4888         }
4889         return status;
4890 }
4891
4892 static void nfs4_free_createdata(struct nfs4_createdata *data)
4893 {
4894         nfs4_label_free(data->label);
4895         kfree(data);
4896 }
4897
4898 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4899                 struct page *page, unsigned int len, struct iattr *sattr,
4900                 struct nfs4_label *label)
4901 {
4902         struct nfs4_createdata *data;
4903         int status = -ENAMETOOLONG;
4904
4905         if (len > NFS4_MAXPATHLEN)
4906                 goto out;
4907
4908         status = -ENOMEM;
4909         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4910         if (data == NULL)
4911                 goto out;
4912
4913         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4914         data->arg.u.symlink.pages = &page;
4915         data->arg.u.symlink.len = len;
4916         data->arg.label = label;
4917
4918         status = nfs4_do_create(dir, dentry, data);
4919
4920         nfs4_free_createdata(data);
4921 out:
4922         return status;
4923 }
4924
4925 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4926                 struct page *page, unsigned int len, struct iattr *sattr)
4927 {
4928         struct nfs4_exception exception = {
4929                 .interruptible = true,
4930         };
4931         struct nfs4_label l, *label;
4932         int err;
4933
4934         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4935
4936         do {
4937                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4938                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4939                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4940                                 &exception);
4941         } while (exception.retry);
4942
4943         nfs4_label_release_security(label);
4944         return err;
4945 }
4946
4947 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4948                 struct iattr *sattr, struct nfs4_label *label)
4949 {
4950         struct nfs4_createdata *data;
4951         int status = -ENOMEM;
4952
4953         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4954         if (data == NULL)
4955                 goto out;
4956
4957         data->arg.label = label;
4958         status = nfs4_do_create(dir, dentry, data);
4959
4960         nfs4_free_createdata(data);
4961 out:
4962         return status;
4963 }
4964
4965 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4966                 struct iattr *sattr)
4967 {
4968         struct nfs_server *server = NFS_SERVER(dir);
4969         struct nfs4_exception exception = {
4970                 .interruptible = true,
4971         };
4972         struct nfs4_label l, *label;
4973         int err;
4974
4975         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4976
4977         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4978                 sattr->ia_mode &= ~current_umask();
4979         do {
4980                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4981                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4982                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4983                                 &exception);
4984         } while (exception.retry);
4985         nfs4_label_release_security(label);
4986
4987         return err;
4988 }
4989
4990 static int _nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
4991                 u64 cookie, struct page **pages, unsigned int count, bool plus)
4992 {
4993         struct inode            *dir = d_inode(dentry);
4994         struct nfs_server       *server = NFS_SERVER(dir);
4995         struct nfs4_readdir_arg args = {
4996                 .fh = NFS_FH(dir),
4997                 .pages = pages,
4998                 .pgbase = 0,
4999                 .count = count,
5000                 .plus = plus,
5001         };
5002         struct nfs4_readdir_res res;
5003         struct rpc_message msg = {
5004                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5005                 .rpc_argp = &args,
5006                 .rpc_resp = &res,
5007                 .rpc_cred = cred,
5008         };
5009         int                     status;
5010
5011         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
5012                         dentry,
5013                         (unsigned long long)cookie);
5014         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5015                 args.bitmask = server->attr_bitmask_nl;
5016         else
5017                 args.bitmask = server->attr_bitmask;
5018
5019         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
5020         res.pgbase = args.pgbase;
5021         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5022                         &res.seq_res, 0);
5023         if (status >= 0) {
5024                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
5025                 status += args.pgbase;
5026         }
5027
5028         nfs_invalidate_atime(dir);
5029
5030         dprintk("%s: returns %d\n", __func__, status);
5031         return status;
5032 }
5033
5034 static int nfs4_proc_readdir(struct dentry *dentry, const struct cred *cred,
5035                 u64 cookie, struct page **pages, unsigned int count, bool plus)
5036 {
5037         struct nfs4_exception exception = {
5038                 .interruptible = true,
5039         };
5040         int err;
5041         do {
5042                 err = _nfs4_proc_readdir(dentry, cred, cookie,
5043                                 pages, count, plus);
5044                 trace_nfs4_readdir(d_inode(dentry), err);
5045                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
5046                                 &exception);
5047         } while (exception.retry);
5048         return err;
5049 }
5050
5051 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5052                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5053 {
5054         struct nfs4_createdata *data;
5055         int mode = sattr->ia_mode;
5056         int status = -ENOMEM;
5057
5058         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5059         if (data == NULL)
5060                 goto out;
5061
5062         if (S_ISFIFO(mode))
5063                 data->arg.ftype = NF4FIFO;
5064         else if (S_ISBLK(mode)) {
5065                 data->arg.ftype = NF4BLK;
5066                 data->arg.u.device.specdata1 = MAJOR(rdev);
5067                 data->arg.u.device.specdata2 = MINOR(rdev);
5068         }
5069         else if (S_ISCHR(mode)) {
5070                 data->arg.ftype = NF4CHR;
5071                 data->arg.u.device.specdata1 = MAJOR(rdev);
5072                 data->arg.u.device.specdata2 = MINOR(rdev);
5073         } else if (!S_ISSOCK(mode)) {
5074                 status = -EINVAL;
5075                 goto out_free;
5076         }
5077
5078         data->arg.label = label;
5079         status = nfs4_do_create(dir, dentry, data);
5080 out_free:
5081         nfs4_free_createdata(data);
5082 out:
5083         return status;
5084 }
5085
5086 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5087                 struct iattr *sattr, dev_t rdev)
5088 {
5089         struct nfs_server *server = NFS_SERVER(dir);
5090         struct nfs4_exception exception = {
5091                 .interruptible = true,
5092         };
5093         struct nfs4_label l, *label;
5094         int err;
5095
5096         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5097
5098         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5099                 sattr->ia_mode &= ~current_umask();
5100         do {
5101                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5102                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5103                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5104                                 &exception);
5105         } while (exception.retry);
5106
5107         nfs4_label_release_security(label);
5108
5109         return err;
5110 }
5111
5112 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5113                  struct nfs_fsstat *fsstat)
5114 {
5115         struct nfs4_statfs_arg args = {
5116                 .fh = fhandle,
5117                 .bitmask = server->attr_bitmask,
5118         };
5119         struct nfs4_statfs_res res = {
5120                 .fsstat = fsstat,
5121         };
5122         struct rpc_message msg = {
5123                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5124                 .rpc_argp = &args,
5125                 .rpc_resp = &res,
5126         };
5127
5128         nfs_fattr_init(fsstat->fattr);
5129         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5130 }
5131
5132 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5133 {
5134         struct nfs4_exception exception = {
5135                 .interruptible = true,
5136         };
5137         int err;
5138         do {
5139                 err = nfs4_handle_exception(server,
5140                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5141                                 &exception);
5142         } while (exception.retry);
5143         return err;
5144 }
5145
5146 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5147                 struct nfs_fsinfo *fsinfo)
5148 {
5149         struct nfs4_fsinfo_arg args = {
5150                 .fh = fhandle,
5151                 .bitmask = server->attr_bitmask,
5152         };
5153         struct nfs4_fsinfo_res res = {
5154                 .fsinfo = fsinfo,
5155         };
5156         struct rpc_message msg = {
5157                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5158                 .rpc_argp = &args,
5159                 .rpc_resp = &res,
5160         };
5161
5162         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5163 }
5164
5165 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5166 {
5167         struct nfs4_exception exception = {
5168                 .interruptible = true,
5169         };
5170         int err;
5171
5172         do {
5173                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5174                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5175                 if (err == 0) {
5176                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5177                         break;
5178                 }
5179                 err = nfs4_handle_exception(server, err, &exception);
5180         } while (exception.retry);
5181         return err;
5182 }
5183
5184 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5185 {
5186         int error;
5187
5188         nfs_fattr_init(fsinfo->fattr);
5189         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5190         if (error == 0) {
5191                 /* block layout checks this! */
5192                 server->pnfs_blksize = fsinfo->blksize;
5193                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5194         }
5195
5196         return error;
5197 }
5198
5199 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5200                 struct nfs_pathconf *pathconf)
5201 {
5202         struct nfs4_pathconf_arg args = {
5203                 .fh = fhandle,
5204                 .bitmask = server->attr_bitmask,
5205         };
5206         struct nfs4_pathconf_res res = {
5207                 .pathconf = pathconf,
5208         };
5209         struct rpc_message msg = {
5210                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5211                 .rpc_argp = &args,
5212                 .rpc_resp = &res,
5213         };
5214
5215         /* None of the pathconf attributes are mandatory to implement */
5216         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5217                 memset(pathconf, 0, sizeof(*pathconf));
5218                 return 0;
5219         }
5220
5221         nfs_fattr_init(pathconf->fattr);
5222         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5223 }
5224
5225 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5226                 struct nfs_pathconf *pathconf)
5227 {
5228         struct nfs4_exception exception = {
5229                 .interruptible = true,
5230         };
5231         int err;
5232
5233         do {
5234                 err = nfs4_handle_exception(server,
5235                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5236                                 &exception);
5237         } while (exception.retry);
5238         return err;
5239 }
5240
5241 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5242                 const struct nfs_open_context *ctx,
5243                 const struct nfs_lock_context *l_ctx,
5244                 fmode_t fmode)
5245 {
5246         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5247 }
5248 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5249
5250 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5251                 const struct nfs_open_context *ctx,
5252                 const struct nfs_lock_context *l_ctx,
5253                 fmode_t fmode)
5254 {
5255         nfs4_stateid _current_stateid;
5256
5257         /* If the current stateid represents a lost lock, then exit */
5258         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5259                 return true;
5260         return nfs4_stateid_match(stateid, &_current_stateid);
5261 }
5262
5263 static bool nfs4_error_stateid_expired(int err)
5264 {
5265         switch (err) {
5266         case -NFS4ERR_DELEG_REVOKED:
5267         case -NFS4ERR_ADMIN_REVOKED:
5268         case -NFS4ERR_BAD_STATEID:
5269         case -NFS4ERR_STALE_STATEID:
5270         case -NFS4ERR_OLD_STATEID:
5271         case -NFS4ERR_OPENMODE:
5272         case -NFS4ERR_EXPIRED:
5273                 return true;
5274         }
5275         return false;
5276 }
5277
5278 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5279 {
5280         struct nfs_server *server = NFS_SERVER(hdr->inode);
5281
5282         trace_nfs4_read(hdr, task->tk_status);
5283         if (task->tk_status < 0) {
5284                 struct nfs4_exception exception = {
5285                         .inode = hdr->inode,
5286                         .state = hdr->args.context->state,
5287                         .stateid = &hdr->args.stateid,
5288                 };
5289                 task->tk_status = nfs4_async_handle_exception(task,
5290                                 server, task->tk_status, &exception);
5291                 if (exception.retry) {
5292                         rpc_restart_call_prepare(task);
5293                         return -EAGAIN;
5294                 }
5295         }
5296
5297         if (task->tk_status > 0)
5298                 renew_lease(server, hdr->timestamp);
5299         return 0;
5300 }
5301
5302 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5303                 struct nfs_pgio_args *args)
5304 {
5305
5306         if (!nfs4_error_stateid_expired(task->tk_status) ||
5307                 nfs4_stateid_is_current(&args->stateid,
5308                                 args->context,
5309                                 args->lock_context,
5310                                 FMODE_READ))
5311                 return false;
5312         rpc_restart_call_prepare(task);
5313         return true;
5314 }
5315
5316 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5317                                          struct nfs_pgio_header *hdr)
5318 {
5319         struct nfs_server *server = NFS_SERVER(hdr->inode);
5320         struct rpc_message *msg = &task->tk_msg;
5321
5322         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5323             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5324                 server->caps &= ~NFS_CAP_READ_PLUS;
5325                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5326                 rpc_restart_call_prepare(task);
5327                 return true;
5328         }
5329         return false;
5330 }
5331
5332 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5333 {
5334         dprintk("--> %s\n", __func__);
5335
5336         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5337                 return -EAGAIN;
5338         if (nfs4_read_stateid_changed(task, &hdr->args))
5339                 return -EAGAIN;
5340         if (nfs4_read_plus_not_supported(task, hdr))
5341                 return -EAGAIN;
5342         if (task->tk_status > 0)
5343                 nfs_invalidate_atime(hdr->inode);
5344         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5345                                     nfs4_read_done_cb(task, hdr);
5346 }
5347
5348 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5349 static void nfs42_read_plus_support(struct nfs_server *server, struct rpc_message *msg)
5350 {
5351         if (server->caps & NFS_CAP_READ_PLUS)
5352                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5353         else
5354                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5355 }
5356 #else
5357 static void nfs42_read_plus_support(struct nfs_server *server, struct rpc_message *msg)
5358 {
5359         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5360 }
5361 #endif /* CONFIG_NFS_V4_2 */
5362
5363 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5364                                  struct rpc_message *msg)
5365 {
5366         hdr->timestamp   = jiffies;
5367         if (!hdr->pgio_done_cb)
5368                 hdr->pgio_done_cb = nfs4_read_done_cb;
5369         nfs42_read_plus_support(NFS_SERVER(hdr->inode), msg);
5370         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5371 }
5372
5373 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5374                                       struct nfs_pgio_header *hdr)
5375 {
5376         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5377                         &hdr->args.seq_args,
5378                         &hdr->res.seq_res,
5379                         task))
5380                 return 0;
5381         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5382                                 hdr->args.lock_context,
5383                                 hdr->rw_mode) == -EIO)
5384                 return -EIO;
5385         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5386                 return -EIO;
5387         return 0;
5388 }
5389
5390 static int nfs4_write_done_cb(struct rpc_task *task,
5391                               struct nfs_pgio_header *hdr)
5392 {
5393         struct inode *inode = hdr->inode;
5394
5395         trace_nfs4_write(hdr, task->tk_status);
5396         if (task->tk_status < 0) {
5397                 struct nfs4_exception exception = {
5398                         .inode = hdr->inode,
5399                         .state = hdr->args.context->state,
5400                         .stateid = &hdr->args.stateid,
5401                 };
5402                 task->tk_status = nfs4_async_handle_exception(task,
5403                                 NFS_SERVER(inode), task->tk_status,
5404                                 &exception);
5405                 if (exception.retry) {
5406                         rpc_restart_call_prepare(task);
5407                         return -EAGAIN;
5408                 }
5409         }
5410         if (task->tk_status >= 0) {
5411                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5412                 nfs_writeback_update_inode(hdr);
5413         }
5414         return 0;
5415 }
5416
5417 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5418                 struct nfs_pgio_args *args)
5419 {
5420
5421         if (!nfs4_error_stateid_expired(task->tk_status) ||
5422                 nfs4_stateid_is_current(&args->stateid,
5423                                 args->context,
5424                                 args->lock_context,
5425                                 FMODE_WRITE))
5426                 return false;
5427         rpc_restart_call_prepare(task);
5428         return true;
5429 }
5430
5431 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5432 {
5433         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5434                 return -EAGAIN;
5435         if (nfs4_write_stateid_changed(task, &hdr->args))
5436                 return -EAGAIN;
5437         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5438                 nfs4_write_done_cb(task, hdr);
5439 }
5440
5441 static
5442 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5443 {
5444         /* Don't request attributes for pNFS or O_DIRECT writes */
5445         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5446                 return false;
5447         /* Otherwise, request attributes if and only if we don't hold
5448          * a delegation
5449          */
5450         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5451 }
5452
5453 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src,
5454                              struct inode *inode, struct nfs_server *server,
5455                              struct nfs4_label *label)
5456 {
5457         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5458         unsigned int i;
5459
5460         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5461
5462         if (cache_validity & (NFS_INO_INVALID_CHANGE | NFS_INO_REVAL_PAGECACHE))
5463                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5464         if (cache_validity & NFS_INO_INVALID_ATIME)
5465                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5466         if (cache_validity & NFS_INO_INVALID_OTHER)
5467                 bitmask[1] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER |
5468                                 FATTR4_WORD1_OWNER_GROUP |
5469                                 FATTR4_WORD1_NUMLINKS;
5470         if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL)
5471                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5472         if (cache_validity & NFS_INO_INVALID_CTIME)
5473                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5474         if (cache_validity & NFS_INO_INVALID_MTIME)
5475                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5476         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5477                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5478
5479         if (nfs4_have_delegation(inode, FMODE_READ) &&
5480             !(cache_validity & NFS_INO_REVAL_FORCED))
5481                 bitmask[0] &= ~FATTR4_WORD0_SIZE;
5482         else if (cache_validity &
5483                  (NFS_INO_INVALID_SIZE | NFS_INO_REVAL_PAGECACHE))
5484                 bitmask[0] |= FATTR4_WORD0_SIZE;
5485
5486         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5487                 bitmask[i] &= server->attr_bitmask[i];
5488 }
5489
5490 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5491                                   struct rpc_message *msg,
5492                                   struct rpc_clnt **clnt)
5493 {
5494         struct nfs_server *server = NFS_SERVER(hdr->inode);
5495
5496         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5497                 hdr->args.bitmask = NULL;
5498                 hdr->res.fattr = NULL;
5499         } else {
5500                 nfs4_bitmask_set(hdr->args.bitmask_store,
5501                                  server->cache_consistency_bitmask,
5502                                  hdr->inode, server, NULL);
5503                 hdr->args.bitmask = hdr->args.bitmask_store;
5504         }
5505
5506         if (!hdr->pgio_done_cb)
5507                 hdr->pgio_done_cb = nfs4_write_done_cb;
5508         hdr->res.server = server;
5509         hdr->timestamp   = jiffies;
5510
5511         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5512         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5513         nfs4_state_protect_write(hdr->ds_clp ? hdr->ds_clp : server->nfs_client, clnt, msg, hdr);
5514 }
5515
5516 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5517 {
5518         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5519                         &data->args.seq_args,
5520                         &data->res.seq_res,
5521                         task);
5522 }
5523
5524 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5525 {
5526         struct inode *inode = data->inode;
5527
5528         trace_nfs4_commit(data, task->tk_status);
5529         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5530                                     NULL, NULL) == -EAGAIN) {
5531                 rpc_restart_call_prepare(task);
5532                 return -EAGAIN;
5533         }
5534         return 0;
5535 }
5536
5537 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5538 {
5539         if (!nfs4_sequence_done(task, &data->res.seq_res))
5540                 return -EAGAIN;
5541         return data->commit_done_cb(task, data);
5542 }
5543
5544 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5545                                    struct rpc_clnt **clnt)
5546 {
5547         struct nfs_server *server = NFS_SERVER(data->inode);
5548
5549         if (data->commit_done_cb == NULL)
5550                 data->commit_done_cb = nfs4_commit_done_cb;
5551         data->res.server = server;
5552         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5553         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5554         nfs4_state_protect(data->ds_clp ? data->ds_clp : server->nfs_client,
5555                         NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5556 }
5557
5558 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5559                                 struct nfs_commitres *res)
5560 {
5561         struct inode *dst_inode = file_inode(dst);
5562         struct nfs_server *server = NFS_SERVER(dst_inode);
5563         struct rpc_message msg = {
5564                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5565                 .rpc_argp = args,
5566                 .rpc_resp = res,
5567         };
5568
5569         args->fh = NFS_FH(dst_inode);
5570         return nfs4_call_sync(server->client, server, &msg,
5571                         &args->seq_args, &res->seq_res, 1);
5572 }
5573
5574 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5575 {
5576         struct nfs_commitargs args = {
5577                 .offset = offset,
5578                 .count = count,
5579         };
5580         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5581         struct nfs4_exception exception = { };
5582         int status;
5583
5584         do {
5585                 status = _nfs4_proc_commit(dst, &args, res);
5586                 status = nfs4_handle_exception(dst_server, status, &exception);
5587         } while (exception.retry);
5588
5589         return status;
5590 }
5591
5592 struct nfs4_renewdata {
5593         struct nfs_client       *client;
5594         unsigned long           timestamp;
5595 };
5596
5597 /*
5598  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5599  * standalone procedure for queueing an asynchronous RENEW.
5600  */
5601 static void nfs4_renew_release(void *calldata)
5602 {
5603         struct nfs4_renewdata *data = calldata;
5604         struct nfs_client *clp = data->client;
5605
5606         if (refcount_read(&clp->cl_count) > 1)
5607                 nfs4_schedule_state_renewal(clp);
5608         nfs_put_client(clp);
5609         kfree(data);
5610 }
5611
5612 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5613 {
5614         struct nfs4_renewdata *data = calldata;
5615         struct nfs_client *clp = data->client;
5616         unsigned long timestamp = data->timestamp;
5617
5618         trace_nfs4_renew_async(clp, task->tk_status);
5619         switch (task->tk_status) {
5620         case 0:
5621                 break;
5622         case -NFS4ERR_LEASE_MOVED:
5623                 nfs4_schedule_lease_moved_recovery(clp);
5624                 break;
5625         default:
5626                 /* Unless we're shutting down, schedule state recovery! */
5627                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5628                         return;
5629                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5630                         nfs4_schedule_lease_recovery(clp);
5631                         return;
5632                 }
5633                 nfs4_schedule_path_down_recovery(clp);
5634         }
5635         do_renew_lease(clp, timestamp);
5636 }
5637
5638 static const struct rpc_call_ops nfs4_renew_ops = {
5639         .rpc_call_done = nfs4_renew_done,
5640         .rpc_release = nfs4_renew_release,
5641 };
5642
5643 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5644 {
5645         struct rpc_message msg = {
5646                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5647                 .rpc_argp       = clp,
5648                 .rpc_cred       = cred,
5649         };
5650         struct nfs4_renewdata *data;
5651
5652         if (renew_flags == 0)
5653                 return 0;
5654         if (!refcount_inc_not_zero(&clp->cl_count))
5655                 return -EIO;
5656         data = kmalloc(sizeof(*data), GFP_NOFS);
5657         if (data == NULL) {
5658                 nfs_put_client(clp);
5659                 return -ENOMEM;
5660         }
5661         data->client = clp;
5662         data->timestamp = jiffies;
5663         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5664                         &nfs4_renew_ops, data);
5665 }
5666
5667 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5668 {
5669         struct rpc_message msg = {
5670                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5671                 .rpc_argp       = clp,
5672                 .rpc_cred       = cred,
5673         };
5674         unsigned long now = jiffies;
5675         int status;
5676
5677         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5678         if (status < 0)
5679                 return status;
5680         do_renew_lease(clp, now);
5681         return 0;
5682 }
5683
5684 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5685 {
5686         return server->caps & NFS_CAP_ACLS;
5687 }
5688
5689 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5690  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5691  * the stack.
5692  */
5693 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5694
5695 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5696                 struct page **pages)
5697 {
5698         struct page *newpage, **spages;
5699         int rc = 0;
5700         size_t len;
5701         spages = pages;
5702
5703         do {
5704                 len = min_t(size_t, PAGE_SIZE, buflen);
5705                 newpage = alloc_page(GFP_KERNEL);
5706
5707                 if (newpage == NULL)
5708                         goto unwind;
5709                 memcpy(page_address(newpage), buf, len);
5710                 buf += len;
5711                 buflen -= len;
5712                 *pages++ = newpage;
5713                 rc++;
5714         } while (buflen != 0);
5715
5716         return rc;
5717
5718 unwind:
5719         for(; rc > 0; rc--)
5720                 __free_page(spages[rc-1]);
5721         return -ENOMEM;
5722 }
5723
5724 struct nfs4_cached_acl {
5725         int cached;
5726         size_t len;
5727         char data[];
5728 };
5729
5730 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5731 {
5732         struct nfs_inode *nfsi = NFS_I(inode);
5733
5734         spin_lock(&inode->i_lock);
5735         kfree(nfsi->nfs4_acl);
5736         nfsi->nfs4_acl = acl;
5737         spin_unlock(&inode->i_lock);
5738 }
5739
5740 static void nfs4_zap_acl_attr(struct inode *inode)
5741 {
5742         nfs4_set_cached_acl(inode, NULL);
5743 }
5744
5745 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5746 {
5747         struct nfs_inode *nfsi = NFS_I(inode);
5748         struct nfs4_cached_acl *acl;
5749         int ret = -ENOENT;
5750
5751         spin_lock(&inode->i_lock);
5752         acl = nfsi->nfs4_acl;
5753         if (acl == NULL)
5754                 goto out;
5755         if (buf == NULL) /* user is just asking for length */
5756                 goto out_len;
5757         if (acl->cached == 0)
5758                 goto out;
5759         ret = -ERANGE; /* see getxattr(2) man page */
5760         if (acl->len > buflen)
5761                 goto out;
5762         memcpy(buf, acl->data, acl->len);
5763 out_len:
5764         ret = acl->len;
5765 out:
5766         spin_unlock(&inode->i_lock);
5767         return ret;
5768 }
5769
5770 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5771 {
5772         struct nfs4_cached_acl *acl;
5773         size_t buflen = sizeof(*acl) + acl_len;
5774
5775         if (buflen <= PAGE_SIZE) {
5776                 acl = kmalloc(buflen, GFP_KERNEL);
5777                 if (acl == NULL)
5778                         goto out;
5779                 acl->cached = 1;
5780                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5781         } else {
5782                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5783                 if (acl == NULL)
5784                         goto out;
5785                 acl->cached = 0;
5786         }
5787         acl->len = acl_len;
5788 out:
5789         nfs4_set_cached_acl(inode, acl);
5790 }
5791
5792 /*
5793  * The getxattr API returns the required buffer length when called with a
5794  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5795  * the required buf.  On a NULL buf, we send a page of data to the server
5796  * guessing that the ACL request can be serviced by a page. If so, we cache
5797  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5798  * the cache. If not so, we throw away the page, and cache the required
5799  * length. The next getxattr call will then produce another round trip to
5800  * the server, this time with the input buf of the required size.
5801  */
5802 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5803 {
5804         struct page **pages;
5805         struct nfs_getaclargs args = {
5806                 .fh = NFS_FH(inode),
5807                 .acl_len = buflen,
5808         };
5809         struct nfs_getaclres res = {
5810                 .acl_len = buflen,
5811         };
5812         struct rpc_message msg = {
5813                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5814                 .rpc_argp = &args,
5815                 .rpc_resp = &res,
5816         };
5817         unsigned int npages;
5818         int ret = -ENOMEM, i;
5819         struct nfs_server *server = NFS_SERVER(inode);
5820
5821         if (buflen == 0)
5822                 buflen = server->rsize;
5823
5824         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5825         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5826         if (!pages)
5827                 return -ENOMEM;
5828
5829         args.acl_pages = pages;
5830
5831         for (i = 0; i < npages; i++) {
5832                 pages[i] = alloc_page(GFP_KERNEL);
5833                 if (!pages[i])
5834                         goto out_free;
5835         }
5836
5837         /* for decoding across pages */
5838         res.acl_scratch = alloc_page(GFP_KERNEL);
5839         if (!res.acl_scratch)
5840                 goto out_free;
5841
5842         args.acl_len = npages * PAGE_SIZE;
5843
5844         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5845                 __func__, buf, buflen, npages, args.acl_len);
5846         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5847                              &msg, &args.seq_args, &res.seq_res, 0);
5848         if (ret)
5849                 goto out_free;
5850
5851         /* Handle the case where the passed-in buffer is too short */
5852         if (res.acl_flags & NFS4_ACL_TRUNC) {
5853                 /* Did the user only issue a request for the acl length? */
5854                 if (buf == NULL)
5855                         goto out_ok;
5856                 ret = -ERANGE;
5857                 goto out_free;
5858         }
5859         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5860         if (buf) {
5861                 if (res.acl_len > buflen) {
5862                         ret = -ERANGE;
5863                         goto out_free;
5864                 }
5865                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5866         }
5867 out_ok:
5868         ret = res.acl_len;
5869 out_free:
5870         while (--i >= 0)
5871                 __free_page(pages[i]);
5872         if (res.acl_scratch)
5873                 __free_page(res.acl_scratch);
5874         kfree(pages);
5875         return ret;
5876 }
5877
5878 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5879 {
5880         struct nfs4_exception exception = {
5881                 .interruptible = true,
5882         };
5883         ssize_t ret;
5884         do {
5885                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5886                 trace_nfs4_get_acl(inode, ret);
5887                 if (ret >= 0)
5888                         break;
5889                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5890         } while (exception.retry);
5891         return ret;
5892 }
5893
5894 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5895 {
5896         struct nfs_server *server = NFS_SERVER(inode);
5897         int ret;
5898
5899         if (!nfs4_server_supports_acls(server))
5900                 return -EOPNOTSUPP;
5901         ret = nfs_revalidate_inode(server, inode);
5902         if (ret < 0)
5903                 return ret;
5904         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5905                 nfs_zap_acl_cache(inode);
5906         ret = nfs4_read_cached_acl(inode, buf, buflen);
5907         if (ret != -ENOENT)
5908                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5909                  * but no cached acl data, just the acl length */
5910                 return ret;
5911         return nfs4_get_acl_uncached(inode, buf, buflen);
5912 }
5913
5914 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5915 {
5916         struct nfs_server *server = NFS_SERVER(inode);
5917         struct page *pages[NFS4ACL_MAXPAGES];
5918         struct nfs_setaclargs arg = {
5919                 .fh             = NFS_FH(inode),
5920                 .acl_pages      = pages,
5921                 .acl_len        = buflen,
5922         };
5923         struct nfs_setaclres res;
5924         struct rpc_message msg = {
5925                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5926                 .rpc_argp       = &arg,
5927                 .rpc_resp       = &res,
5928         };
5929         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5930         int ret, i;
5931
5932         /* You can't remove system.nfs4_acl: */
5933         if (buflen == 0)
5934                 return -EINVAL;
5935         if (!nfs4_server_supports_acls(server))
5936                 return -EOPNOTSUPP;
5937         if (npages > ARRAY_SIZE(pages))
5938                 return -ERANGE;
5939         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5940         if (i < 0)
5941                 return i;
5942         nfs4_inode_make_writeable(inode);
5943         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5944
5945         /*
5946          * Free each page after tx, so the only ref left is
5947          * held by the network stack
5948          */
5949         for (; i > 0; i--)
5950                 put_page(pages[i-1]);
5951
5952         /*
5953          * Acl update can result in inode attribute update.
5954          * so mark the attribute cache invalid.
5955          */
5956         spin_lock(&inode->i_lock);
5957         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5958                 | NFS_INO_INVALID_CTIME
5959                 | NFS_INO_REVAL_FORCED;
5960         spin_unlock(&inode->i_lock);
5961         nfs_access_zap_cache(inode);
5962         nfs_zap_acl_cache(inode);
5963         return ret;
5964 }
5965
5966 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5967 {
5968         struct nfs4_exception exception = { };
5969         int err;
5970         do {
5971                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5972                 trace_nfs4_set_acl(inode, err);
5973                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
5974                         /*
5975                          * no need to retry since the kernel
5976                          * isn't involved in encoding the ACEs.
5977                          */
5978                         err = -EINVAL;
5979                         break;
5980                 }
5981                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5982                                 &exception);
5983         } while (exception.retry);
5984         return err;
5985 }
5986
5987 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5988 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5989                                         size_t buflen)
5990 {
5991         struct nfs_server *server = NFS_SERVER(inode);
5992         struct nfs_fattr fattr;
5993         struct nfs4_label label = {0, 0, buflen, buf};
5994
5995         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5996         struct nfs4_getattr_arg arg = {
5997                 .fh             = NFS_FH(inode),
5998                 .bitmask        = bitmask,
5999         };
6000         struct nfs4_getattr_res res = {
6001                 .fattr          = &fattr,
6002                 .label          = &label,
6003                 .server         = server,
6004         };
6005         struct rpc_message msg = {
6006                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6007                 .rpc_argp       = &arg,
6008                 .rpc_resp       = &res,
6009         };
6010         int ret;
6011
6012         nfs_fattr_init(&fattr);
6013
6014         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6015         if (ret)
6016                 return ret;
6017         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6018                 return -ENOENT;
6019         return label.len;
6020 }
6021
6022 static int nfs4_get_security_label(struct inode *inode, void *buf,
6023                                         size_t buflen)
6024 {
6025         struct nfs4_exception exception = {
6026                 .interruptible = true,
6027         };
6028         int err;
6029
6030         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6031                 return -EOPNOTSUPP;
6032
6033         do {
6034                 err = _nfs4_get_security_label(inode, buf, buflen);
6035                 trace_nfs4_get_security_label(inode, err);
6036                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6037                                 &exception);
6038         } while (exception.retry);
6039         return err;
6040 }
6041
6042 static int _nfs4_do_set_security_label(struct inode *inode,
6043                 struct nfs4_label *ilabel,
6044                 struct nfs_fattr *fattr,
6045                 struct nfs4_label *olabel)
6046 {
6047
6048         struct iattr sattr = {0};
6049         struct nfs_server *server = NFS_SERVER(inode);
6050         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6051         struct nfs_setattrargs arg = {
6052                 .fh             = NFS_FH(inode),
6053                 .iap            = &sattr,
6054                 .server         = server,
6055                 .bitmask        = bitmask,
6056                 .label          = ilabel,
6057         };
6058         struct nfs_setattrres res = {
6059                 .fattr          = fattr,
6060                 .label          = olabel,
6061                 .server         = server,
6062         };
6063         struct rpc_message msg = {
6064                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6065                 .rpc_argp       = &arg,
6066                 .rpc_resp       = &res,
6067         };
6068         int status;
6069
6070         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6071
6072         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6073         if (status)
6074                 dprintk("%s failed: %d\n", __func__, status);
6075
6076         return status;
6077 }
6078
6079 static int nfs4_do_set_security_label(struct inode *inode,
6080                 struct nfs4_label *ilabel,
6081                 struct nfs_fattr *fattr,
6082                 struct nfs4_label *olabel)
6083 {
6084         struct nfs4_exception exception = { };
6085         int err;
6086
6087         do {
6088                 err = _nfs4_do_set_security_label(inode, ilabel,
6089                                 fattr, olabel);
6090                 trace_nfs4_set_security_label(inode, err);
6091                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6092                                 &exception);
6093         } while (exception.retry);
6094         return err;
6095 }
6096
6097 static int
6098 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6099 {
6100         struct nfs4_label ilabel, *olabel = NULL;
6101         struct nfs_fattr fattr;
6102         int status;
6103
6104         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6105                 return -EOPNOTSUPP;
6106
6107         nfs_fattr_init(&fattr);
6108
6109         ilabel.pi = 0;
6110         ilabel.lfs = 0;
6111         ilabel.label = (char *)buf;
6112         ilabel.len = buflen;
6113
6114         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6115         if (IS_ERR(olabel)) {
6116                 status = -PTR_ERR(olabel);
6117                 goto out;
6118         }
6119
6120         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6121         if (status == 0)
6122                 nfs_setsecurity(inode, &fattr, olabel);
6123
6124         nfs4_label_free(olabel);
6125 out:
6126         return status;
6127 }
6128 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6129
6130
6131 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6132                                     nfs4_verifier *bootverf)
6133 {
6134         __be32 verf[2];
6135
6136         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6137                 /* An impossible timestamp guarantees this value
6138                  * will never match a generated boot time. */
6139                 verf[0] = cpu_to_be32(U32_MAX);
6140                 verf[1] = cpu_to_be32(U32_MAX);
6141         } else {
6142                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6143                 u64 ns = ktime_to_ns(nn->boot_time);
6144
6145                 verf[0] = cpu_to_be32(ns >> 32);
6146                 verf[1] = cpu_to_be32(ns);
6147         }
6148         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6149 }
6150
6151 static size_t
6152 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6153 {
6154         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6155         struct nfs_netns_client *nn_clp = nn->nfs_client;
6156         const char *id;
6157
6158         buf[0] = '\0';
6159
6160         if (nn_clp) {
6161                 rcu_read_lock();
6162                 id = rcu_dereference(nn_clp->identifier);
6163                 if (id)
6164                         strscpy(buf, id, buflen);
6165                 rcu_read_unlock();
6166         }
6167
6168         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6169                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6170
6171         return strlen(buf);
6172 }
6173
6174 static int
6175 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6176 {
6177         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6178         size_t buflen;
6179         size_t len;
6180         char *str;
6181
6182         if (clp->cl_owner_id != NULL)
6183                 return 0;
6184
6185         rcu_read_lock();
6186         len = 14 +
6187                 strlen(clp->cl_rpcclient->cl_nodename) +
6188                 1 +
6189                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6190                 1;
6191         rcu_read_unlock();
6192
6193         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6194         if (buflen)
6195                 len += buflen + 1;
6196
6197         if (len > NFS4_OPAQUE_LIMIT + 1)
6198                 return -EINVAL;
6199
6200         /*
6201          * Since this string is allocated at mount time, and held until the
6202          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6203          * about a memory-reclaim deadlock.
6204          */
6205         str = kmalloc(len, GFP_KERNEL);
6206         if (!str)
6207                 return -ENOMEM;
6208
6209         rcu_read_lock();
6210         if (buflen)
6211                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6212                           clp->cl_rpcclient->cl_nodename, buf,
6213                           rpc_peeraddr2str(clp->cl_rpcclient,
6214                                            RPC_DISPLAY_ADDR));
6215         else
6216                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6217                           clp->cl_rpcclient->cl_nodename,
6218                           rpc_peeraddr2str(clp->cl_rpcclient,
6219                                            RPC_DISPLAY_ADDR));
6220         rcu_read_unlock();
6221
6222         clp->cl_owner_id = str;
6223         return 0;
6224 }
6225
6226 static int
6227 nfs4_init_uniform_client_string(struct nfs_client *clp)
6228 {
6229         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6230         size_t buflen;
6231         size_t len;
6232         char *str;
6233
6234         if (clp->cl_owner_id != NULL)
6235                 return 0;
6236
6237         len = 10 + 10 + 1 + 10 + 1 +
6238                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6239
6240         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6241         if (buflen)
6242                 len += buflen + 1;
6243
6244         if (len > NFS4_OPAQUE_LIMIT + 1)
6245                 return -EINVAL;
6246
6247         /*
6248          * Since this string is allocated at mount time, and held until the
6249          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6250          * about a memory-reclaim deadlock.
6251          */
6252         str = kmalloc(len, GFP_KERNEL);
6253         if (!str)
6254                 return -ENOMEM;
6255
6256         if (buflen)
6257                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6258                           clp->rpc_ops->version, clp->cl_minorversion,
6259                           buf, clp->cl_rpcclient->cl_nodename);
6260         else
6261                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6262                           clp->rpc_ops->version, clp->cl_minorversion,
6263                           clp->cl_rpcclient->cl_nodename);
6264         clp->cl_owner_id = str;
6265         return 0;
6266 }
6267
6268 /*
6269  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6270  * services.  Advertise one based on the address family of the
6271  * clientaddr.
6272  */
6273 static unsigned int
6274 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6275 {
6276         if (strchr(clp->cl_ipaddr, ':') != NULL)
6277                 return scnprintf(buf, len, "tcp6");
6278         else
6279                 return scnprintf(buf, len, "tcp");
6280 }
6281
6282 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6283 {
6284         struct nfs4_setclientid *sc = calldata;
6285
6286         if (task->tk_status == 0)
6287                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6288 }
6289
6290 static const struct rpc_call_ops nfs4_setclientid_ops = {
6291         .rpc_call_done = nfs4_setclientid_done,
6292 };
6293
6294 /**
6295  * nfs4_proc_setclientid - Negotiate client ID
6296  * @clp: state data structure
6297  * @program: RPC program for NFSv4 callback service
6298  * @port: IP port number for NFS4 callback service
6299  * @cred: credential to use for this call
6300  * @res: where to place the result
6301  *
6302  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6303  */
6304 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6305                 unsigned short port, const struct cred *cred,
6306                 struct nfs4_setclientid_res *res)
6307 {
6308         nfs4_verifier sc_verifier;
6309         struct nfs4_setclientid setclientid = {
6310                 .sc_verifier = &sc_verifier,
6311                 .sc_prog = program,
6312                 .sc_clnt = clp,
6313         };
6314         struct rpc_message msg = {
6315                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6316                 .rpc_argp = &setclientid,
6317                 .rpc_resp = res,
6318                 .rpc_cred = cred,
6319         };
6320         struct rpc_task_setup task_setup_data = {
6321                 .rpc_client = clp->cl_rpcclient,
6322                 .rpc_message = &msg,
6323                 .callback_ops = &nfs4_setclientid_ops,
6324                 .callback_data = &setclientid,
6325                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6326         };
6327         unsigned long now = jiffies;
6328         int status;
6329
6330         /* nfs_client_id4 */
6331         nfs4_init_boot_verifier(clp, &sc_verifier);
6332
6333         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6334                 status = nfs4_init_uniform_client_string(clp);
6335         else
6336                 status = nfs4_init_nonuniform_client_string(clp);
6337
6338         if (status)
6339                 goto out;
6340
6341         /* cb_client4 */
6342         setclientid.sc_netid_len =
6343                                 nfs4_init_callback_netid(clp,
6344                                                 setclientid.sc_netid,
6345                                                 sizeof(setclientid.sc_netid));
6346         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6347                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6348                                 clp->cl_ipaddr, port >> 8, port & 255);
6349
6350         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6351                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6352                 clp->cl_owner_id);
6353
6354         status = nfs4_call_sync_custom(&task_setup_data);
6355         if (setclientid.sc_cred) {
6356                 kfree(clp->cl_acceptor);
6357                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6358                 put_rpccred(setclientid.sc_cred);
6359         }
6360
6361         if (status == 0)
6362                 do_renew_lease(clp, now);
6363 out:
6364         trace_nfs4_setclientid(clp, status);
6365         dprintk("NFS reply setclientid: %d\n", status);
6366         return status;
6367 }
6368
6369 /**
6370  * nfs4_proc_setclientid_confirm - Confirm client ID
6371  * @clp: state data structure
6372  * @arg: result of a previous SETCLIENTID
6373  * @cred: credential to use for this call
6374  *
6375  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6376  */
6377 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6378                 struct nfs4_setclientid_res *arg,
6379                 const struct cred *cred)
6380 {
6381         struct rpc_message msg = {
6382                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6383                 .rpc_argp = arg,
6384                 .rpc_cred = cred,
6385         };
6386         int status;
6387
6388         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6389                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6390                 clp->cl_clientid);
6391         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6392                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6393         trace_nfs4_setclientid_confirm(clp, status);
6394         dprintk("NFS reply setclientid_confirm: %d\n", status);
6395         return status;
6396 }
6397
6398 struct nfs4_delegreturndata {
6399         struct nfs4_delegreturnargs args;
6400         struct nfs4_delegreturnres res;
6401         struct nfs_fh fh;
6402         nfs4_stateid stateid;
6403         unsigned long timestamp;
6404         struct {
6405                 struct nfs4_layoutreturn_args arg;
6406                 struct nfs4_layoutreturn_res res;
6407                 struct nfs4_xdr_opaque_data ld_private;
6408                 u32 roc_barrier;
6409                 bool roc;
6410         } lr;
6411         struct nfs_fattr fattr;
6412         int rpc_status;
6413         struct inode *inode;
6414 };
6415
6416 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6417 {
6418         struct nfs4_delegreturndata *data = calldata;
6419         struct nfs4_exception exception = {
6420                 .inode = data->inode,
6421                 .stateid = &data->stateid,
6422                 .task_is_privileged = data->args.seq_args.sa_privileged,
6423         };
6424
6425         if (!nfs4_sequence_done(task, &data->res.seq_res))
6426                 return;
6427
6428         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6429
6430         /* Handle Layoutreturn errors */
6431         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6432                           &data->res.lr_ret) == -EAGAIN)
6433                 goto out_restart;
6434
6435         switch (task->tk_status) {
6436         case 0:
6437                 renew_lease(data->res.server, data->timestamp);
6438                 break;
6439         case -NFS4ERR_ADMIN_REVOKED:
6440         case -NFS4ERR_DELEG_REVOKED:
6441         case -NFS4ERR_EXPIRED:
6442                 nfs4_free_revoked_stateid(data->res.server,
6443                                 data->args.stateid,
6444                                 task->tk_msg.rpc_cred);
6445                 fallthrough;
6446         case -NFS4ERR_BAD_STATEID:
6447         case -NFS4ERR_STALE_STATEID:
6448         case -ETIMEDOUT:
6449                 task->tk_status = 0;
6450                 break;
6451         case -NFS4ERR_OLD_STATEID:
6452                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6453                         nfs4_stateid_seqid_inc(&data->stateid);
6454                 if (data->args.bitmask) {
6455                         data->args.bitmask = NULL;
6456                         data->res.fattr = NULL;
6457                 }
6458                 goto out_restart;
6459         case -NFS4ERR_ACCESS:
6460                 if (data->args.bitmask) {
6461                         data->args.bitmask = NULL;
6462                         data->res.fattr = NULL;
6463                         goto out_restart;
6464                 }
6465                 fallthrough;
6466         default:
6467                 task->tk_status = nfs4_async_handle_exception(task,
6468                                 data->res.server, task->tk_status,
6469                                 &exception);
6470                 if (exception.retry)
6471                         goto out_restart;
6472         }
6473         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6474         data->rpc_status = task->tk_status;
6475         return;
6476 out_restart:
6477         task->tk_status = 0;
6478         rpc_restart_call_prepare(task);
6479 }
6480
6481 static void nfs4_delegreturn_release(void *calldata)
6482 {
6483         struct nfs4_delegreturndata *data = calldata;
6484         struct inode *inode = data->inode;
6485
6486         if (data->lr.roc)
6487                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6488                                  data->res.lr_ret);
6489         if (inode) {
6490                 nfs4_fattr_set_prechange(&data->fattr,
6491                                          inode_peek_iversion_raw(inode));
6492                 nfs_refresh_inode(inode, &data->fattr);
6493                 nfs_iput_and_deactive(inode);
6494         }
6495         kfree(calldata);
6496 }
6497
6498 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6499 {
6500         struct nfs4_delegreturndata *d_data;
6501         struct pnfs_layout_hdr *lo;
6502
6503         d_data = (struct nfs4_delegreturndata *)data;
6504
6505         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6506                 nfs4_sequence_done(task, &d_data->res.seq_res);
6507                 return;
6508         }
6509
6510         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6511         if (lo && !pnfs_layout_is_valid(lo)) {
6512                 d_data->args.lr_args = NULL;
6513                 d_data->res.lr_res = NULL;
6514         }
6515
6516         nfs4_setup_sequence(d_data->res.server->nfs_client,
6517                         &d_data->args.seq_args,
6518                         &d_data->res.seq_res,
6519                         task);
6520 }
6521
6522 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6523         .rpc_call_prepare = nfs4_delegreturn_prepare,
6524         .rpc_call_done = nfs4_delegreturn_done,
6525         .rpc_release = nfs4_delegreturn_release,
6526 };
6527
6528 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6529 {
6530         struct nfs4_delegreturndata *data;
6531         struct nfs_server *server = NFS_SERVER(inode);
6532         struct rpc_task *task;
6533         struct rpc_message msg = {
6534                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6535                 .rpc_cred = cred,
6536         };
6537         struct rpc_task_setup task_setup_data = {
6538                 .rpc_client = server->client,
6539                 .rpc_message = &msg,
6540                 .callback_ops = &nfs4_delegreturn_ops,
6541                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6542         };
6543         int status = 0;
6544
6545         data = kzalloc(sizeof(*data), GFP_NOFS);
6546         if (data == NULL)
6547                 return -ENOMEM;
6548
6549         nfs4_state_protect(server->nfs_client,
6550                         NFS_SP4_MACH_CRED_CLEANUP,
6551                         &task_setup_data.rpc_client, &msg);
6552
6553         data->args.fhandle = &data->fh;
6554         data->args.stateid = &data->stateid;
6555         nfs4_bitmask_set(data->args.bitmask_store,
6556                          server->cache_consistency_bitmask, inode, server,
6557                          NULL);
6558         data->args.bitmask = data->args.bitmask_store;
6559         nfs_copy_fh(&data->fh, NFS_FH(inode));
6560         nfs4_stateid_copy(&data->stateid, stateid);
6561         data->res.fattr = &data->fattr;
6562         data->res.server = server;
6563         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6564         data->lr.arg.ld_private = &data->lr.ld_private;
6565         nfs_fattr_init(data->res.fattr);
6566         data->timestamp = jiffies;
6567         data->rpc_status = 0;
6568         data->inode = nfs_igrab_and_active(inode);
6569         if (data->inode || issync) {
6570                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6571                                         cred);
6572                 if (data->lr.roc) {
6573                         data->args.lr_args = &data->lr.arg;
6574                         data->res.lr_res = &data->lr.res;
6575                 }
6576         }
6577
6578         if (!data->inode)
6579                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6580                                    1);
6581         else
6582                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6583                                    0);
6584         task_setup_data.callback_data = data;
6585         msg.rpc_argp = &data->args;
6586         msg.rpc_resp = &data->res;
6587         task = rpc_run_task(&task_setup_data);
6588         if (IS_ERR(task))
6589                 return PTR_ERR(task);
6590         if (!issync)
6591                 goto out;
6592         status = rpc_wait_for_completion_task(task);
6593         if (status != 0)
6594                 goto out;
6595         status = data->rpc_status;
6596 out:
6597         rpc_put_task(task);
6598         return status;
6599 }
6600
6601 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6602 {
6603         struct nfs_server *server = NFS_SERVER(inode);
6604         struct nfs4_exception exception = { };
6605         int err;
6606         do {
6607                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6608                 trace_nfs4_delegreturn(inode, stateid, err);
6609                 switch (err) {
6610                         case -NFS4ERR_STALE_STATEID:
6611                         case -NFS4ERR_EXPIRED:
6612                         case 0:
6613                                 return 0;
6614                 }
6615                 err = nfs4_handle_exception(server, err, &exception);
6616         } while (exception.retry);
6617         return err;
6618 }
6619
6620 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6621 {
6622         struct inode *inode = state->inode;
6623         struct nfs_server *server = NFS_SERVER(inode);
6624         struct nfs_client *clp = server->nfs_client;
6625         struct nfs_lockt_args arg = {
6626                 .fh = NFS_FH(inode),
6627                 .fl = request,
6628         };
6629         struct nfs_lockt_res res = {
6630                 .denied = request,
6631         };
6632         struct rpc_message msg = {
6633                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6634                 .rpc_argp       = &arg,
6635                 .rpc_resp       = &res,
6636                 .rpc_cred       = state->owner->so_cred,
6637         };
6638         struct nfs4_lock_state *lsp;
6639         int status;
6640
6641         arg.lock_owner.clientid = clp->cl_clientid;
6642         status = nfs4_set_lock_state(state, request);
6643         if (status != 0)
6644                 goto out;
6645         lsp = request->fl_u.nfs4_fl.owner;
6646         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6647         arg.lock_owner.s_dev = server->s_dev;
6648         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6649         switch (status) {
6650                 case 0:
6651                         request->fl_type = F_UNLCK;
6652                         break;
6653                 case -NFS4ERR_DENIED:
6654                         status = 0;
6655         }
6656         request->fl_ops->fl_release_private(request);
6657         request->fl_ops = NULL;
6658 out:
6659         return status;
6660 }
6661
6662 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6663 {
6664         struct nfs4_exception exception = {
6665                 .interruptible = true,
6666         };
6667         int err;
6668
6669         do {
6670                 err = _nfs4_proc_getlk(state, cmd, request);
6671                 trace_nfs4_get_lock(request, state, cmd, err);
6672                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6673                                 &exception);
6674         } while (exception.retry);
6675         return err;
6676 }
6677
6678 /*
6679  * Update the seqid of a lock stateid after receiving
6680  * NFS4ERR_OLD_STATEID
6681  */
6682 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6683                 struct nfs4_lock_state *lsp)
6684 {
6685         struct nfs4_state *state = lsp->ls_state;
6686         bool ret = false;
6687
6688         spin_lock(&state->state_lock);
6689         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6690                 goto out;
6691         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6692                 nfs4_stateid_seqid_inc(dst);
6693         else
6694                 dst->seqid = lsp->ls_stateid.seqid;
6695         ret = true;
6696 out:
6697         spin_unlock(&state->state_lock);
6698         return ret;
6699 }
6700
6701 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6702                 struct nfs4_lock_state *lsp)
6703 {
6704         struct nfs4_state *state = lsp->ls_state;
6705         bool ret;
6706
6707         spin_lock(&state->state_lock);
6708         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6709         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6710         spin_unlock(&state->state_lock);
6711         return ret;
6712 }
6713
6714 struct nfs4_unlockdata {
6715         struct nfs_locku_args arg;
6716         struct nfs_locku_res res;
6717         struct nfs4_lock_state *lsp;
6718         struct nfs_open_context *ctx;
6719         struct nfs_lock_context *l_ctx;
6720         struct file_lock fl;
6721         struct nfs_server *server;
6722         unsigned long timestamp;
6723 };
6724
6725 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6726                 struct nfs_open_context *ctx,
6727                 struct nfs4_lock_state *lsp,
6728                 struct nfs_seqid *seqid)
6729 {
6730         struct nfs4_unlockdata *p;
6731         struct nfs4_state *state = lsp->ls_state;
6732         struct inode *inode = state->inode;
6733
6734         p = kzalloc(sizeof(*p), GFP_NOFS);
6735         if (p == NULL)
6736                 return NULL;
6737         p->arg.fh = NFS_FH(inode);
6738         p->arg.fl = &p->fl;
6739         p->arg.seqid = seqid;
6740         p->res.seqid = seqid;
6741         p->lsp = lsp;
6742         /* Ensure we don't close file until we're done freeing locks! */
6743         p->ctx = get_nfs_open_context(ctx);
6744         p->l_ctx = nfs_get_lock_context(ctx);
6745         locks_init_lock(&p->fl);
6746         locks_copy_lock(&p->fl, fl);
6747         p->server = NFS_SERVER(inode);
6748         spin_lock(&state->state_lock);
6749         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6750         spin_unlock(&state->state_lock);
6751         return p;
6752 }
6753
6754 static void nfs4_locku_release_calldata(void *data)
6755 {
6756         struct nfs4_unlockdata *calldata = data;
6757         nfs_free_seqid(calldata->arg.seqid);
6758         nfs4_put_lock_state(calldata->lsp);
6759         nfs_put_lock_context(calldata->l_ctx);
6760         put_nfs_open_context(calldata->ctx);
6761         kfree(calldata);
6762 }
6763
6764 static void nfs4_locku_done(struct rpc_task *task, void *data)
6765 {
6766         struct nfs4_unlockdata *calldata = data;
6767         struct nfs4_exception exception = {
6768                 .inode = calldata->lsp->ls_state->inode,
6769                 .stateid = &calldata->arg.stateid,
6770         };
6771
6772         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6773                 return;
6774         switch (task->tk_status) {
6775                 case 0:
6776                         renew_lease(calldata->server, calldata->timestamp);
6777                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6778                         if (nfs4_update_lock_stateid(calldata->lsp,
6779                                         &calldata->res.stateid))
6780                                 break;
6781                         fallthrough;
6782                 case -NFS4ERR_ADMIN_REVOKED:
6783                 case -NFS4ERR_EXPIRED:
6784                         nfs4_free_revoked_stateid(calldata->server,
6785                                         &calldata->arg.stateid,
6786                                         task->tk_msg.rpc_cred);
6787                         fallthrough;
6788                 case -NFS4ERR_BAD_STATEID:
6789                 case -NFS4ERR_STALE_STATEID:
6790                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6791                                                 calldata->lsp))
6792                                 rpc_restart_call_prepare(task);
6793                         break;
6794                 case -NFS4ERR_OLD_STATEID:
6795                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6796                                                 calldata->lsp))
6797                                 rpc_restart_call_prepare(task);
6798                         break;
6799                 default:
6800                         task->tk_status = nfs4_async_handle_exception(task,
6801                                         calldata->server, task->tk_status,
6802                                         &exception);
6803                         if (exception.retry)
6804                                 rpc_restart_call_prepare(task);
6805         }
6806         nfs_release_seqid(calldata->arg.seqid);
6807 }
6808
6809 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6810 {
6811         struct nfs4_unlockdata *calldata = data;
6812
6813         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6814                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6815                 return;
6816
6817         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6818                 goto out_wait;
6819         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6820                 /* Note: exit _without_ running nfs4_locku_done */
6821                 goto out_no_action;
6822         }
6823         calldata->timestamp = jiffies;
6824         if (nfs4_setup_sequence(calldata->server->nfs_client,
6825                                 &calldata->arg.seq_args,
6826                                 &calldata->res.seq_res,
6827                                 task) != 0)
6828                 nfs_release_seqid(calldata->arg.seqid);
6829         return;
6830 out_no_action:
6831         task->tk_action = NULL;
6832 out_wait:
6833         nfs4_sequence_done(task, &calldata->res.seq_res);
6834 }
6835
6836 static const struct rpc_call_ops nfs4_locku_ops = {
6837         .rpc_call_prepare = nfs4_locku_prepare,
6838         .rpc_call_done = nfs4_locku_done,
6839         .rpc_release = nfs4_locku_release_calldata,
6840 };
6841
6842 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6843                 struct nfs_open_context *ctx,
6844                 struct nfs4_lock_state *lsp,
6845                 struct nfs_seqid *seqid)
6846 {
6847         struct nfs4_unlockdata *data;
6848         struct rpc_message msg = {
6849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6850                 .rpc_cred = ctx->cred,
6851         };
6852         struct rpc_task_setup task_setup_data = {
6853                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6854                 .rpc_message = &msg,
6855                 .callback_ops = &nfs4_locku_ops,
6856                 .workqueue = nfsiod_workqueue,
6857                 .flags = RPC_TASK_ASYNC,
6858         };
6859
6860         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6861                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6862
6863         /* Ensure this is an unlock - when canceling a lock, the
6864          * canceled lock is passed in, and it won't be an unlock.
6865          */
6866         fl->fl_type = F_UNLCK;
6867         if (fl->fl_flags & FL_CLOSE)
6868                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6869
6870         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6871         if (data == NULL) {
6872                 nfs_free_seqid(seqid);
6873                 return ERR_PTR(-ENOMEM);
6874         }
6875
6876         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6877         msg.rpc_argp = &data->arg;
6878         msg.rpc_resp = &data->res;
6879         task_setup_data.callback_data = data;
6880         return rpc_run_task(&task_setup_data);
6881 }
6882
6883 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6884 {
6885         struct inode *inode = state->inode;
6886         struct nfs4_state_owner *sp = state->owner;
6887         struct nfs_inode *nfsi = NFS_I(inode);
6888         struct nfs_seqid *seqid;
6889         struct nfs4_lock_state *lsp;
6890         struct rpc_task *task;
6891         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6892         int status = 0;
6893         unsigned char fl_flags = request->fl_flags;
6894
6895         status = nfs4_set_lock_state(state, request);
6896         /* Unlock _before_ we do the RPC call */
6897         request->fl_flags |= FL_EXISTS;
6898         /* Exclude nfs_delegation_claim_locks() */
6899         mutex_lock(&sp->so_delegreturn_mutex);
6900         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6901         down_read(&nfsi->rwsem);
6902         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6903                 up_read(&nfsi->rwsem);
6904                 mutex_unlock(&sp->so_delegreturn_mutex);
6905                 goto out;
6906         }
6907         up_read(&nfsi->rwsem);
6908         mutex_unlock(&sp->so_delegreturn_mutex);
6909         if (status != 0)
6910                 goto out;
6911         /* Is this a delegated lock? */
6912         lsp = request->fl_u.nfs4_fl.owner;
6913         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6914                 goto out;
6915         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6916         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6917         status = -ENOMEM;
6918         if (IS_ERR(seqid))
6919                 goto out;
6920         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6921         status = PTR_ERR(task);
6922         if (IS_ERR(task))
6923                 goto out;
6924         status = rpc_wait_for_completion_task(task);
6925         rpc_put_task(task);
6926 out:
6927         request->fl_flags = fl_flags;
6928         trace_nfs4_unlock(request, state, F_SETLK, status);
6929         return status;
6930 }
6931
6932 struct nfs4_lockdata {
6933         struct nfs_lock_args arg;
6934         struct nfs_lock_res res;
6935         struct nfs4_lock_state *lsp;
6936         struct nfs_open_context *ctx;
6937         struct file_lock fl;
6938         unsigned long timestamp;
6939         int rpc_status;
6940         int cancelled;
6941         struct nfs_server *server;
6942 };
6943
6944 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6945                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6946                 gfp_t gfp_mask)
6947 {
6948         struct nfs4_lockdata *p;
6949         struct inode *inode = lsp->ls_state->inode;
6950         struct nfs_server *server = NFS_SERVER(inode);
6951         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6952
6953         p = kzalloc(sizeof(*p), gfp_mask);
6954         if (p == NULL)
6955                 return NULL;
6956
6957         p->arg.fh = NFS_FH(inode);
6958         p->arg.fl = &p->fl;
6959         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6960         if (IS_ERR(p->arg.open_seqid))
6961                 goto out_free;
6962         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6963         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6964         if (IS_ERR(p->arg.lock_seqid))
6965                 goto out_free_seqid;
6966         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6967         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6968         p->arg.lock_owner.s_dev = server->s_dev;
6969         p->res.lock_seqid = p->arg.lock_seqid;
6970         p->lsp = lsp;
6971         p->server = server;
6972         p->ctx = get_nfs_open_context(ctx);
6973         locks_init_lock(&p->fl);
6974         locks_copy_lock(&p->fl, fl);
6975         return p;
6976 out_free_seqid:
6977         nfs_free_seqid(p->arg.open_seqid);
6978 out_free:
6979         kfree(p);
6980         return NULL;
6981 }
6982
6983 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6984 {
6985         struct nfs4_lockdata *data = calldata;
6986         struct nfs4_state *state = data->lsp->ls_state;
6987
6988         dprintk("%s: begin!\n", __func__);
6989         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6990                 goto out_wait;
6991         /* Do we need to do an open_to_lock_owner? */
6992         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6993                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6994                         goto out_release_lock_seqid;
6995                 }
6996                 nfs4_stateid_copy(&data->arg.open_stateid,
6997                                 &state->open_stateid);
6998                 data->arg.new_lock_owner = 1;
6999                 data->res.open_seqid = data->arg.open_seqid;
7000         } else {
7001                 data->arg.new_lock_owner = 0;
7002                 nfs4_stateid_copy(&data->arg.lock_stateid,
7003                                 &data->lsp->ls_stateid);
7004         }
7005         if (!nfs4_valid_open_stateid(state)) {
7006                 data->rpc_status = -EBADF;
7007                 task->tk_action = NULL;
7008                 goto out_release_open_seqid;
7009         }
7010         data->timestamp = jiffies;
7011         if (nfs4_setup_sequence(data->server->nfs_client,
7012                                 &data->arg.seq_args,
7013                                 &data->res.seq_res,
7014                                 task) == 0)
7015                 return;
7016 out_release_open_seqid:
7017         nfs_release_seqid(data->arg.open_seqid);
7018 out_release_lock_seqid:
7019         nfs_release_seqid(data->arg.lock_seqid);
7020 out_wait:
7021         nfs4_sequence_done(task, &data->res.seq_res);
7022         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
7023 }
7024
7025 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7026 {
7027         struct nfs4_lockdata *data = calldata;
7028         struct nfs4_lock_state *lsp = data->lsp;
7029         struct nfs_server *server = NFS_SERVER(d_inode(data->ctx->dentry));
7030
7031         dprintk("%s: begin!\n", __func__);
7032
7033         if (!nfs4_sequence_done(task, &data->res.seq_res))
7034                 return;
7035
7036         data->rpc_status = task->tk_status;
7037         switch (task->tk_status) {
7038         case 0:
7039                 renew_lease(server, data->timestamp);
7040                 if (data->arg.new_lock && !data->cancelled) {
7041                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7042                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7043                                 goto out_restart;
7044                 }
7045                 if (data->arg.new_lock_owner != 0) {
7046                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7047                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7048                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7049                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7050                         goto out_restart;
7051                 break;
7052         case -NFS4ERR_OLD_STATEID:
7053                 if (data->arg.new_lock_owner != 0 &&
7054                         nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7055                                         lsp->ls_state))
7056                         goto out_restart;
7057                 if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7058                         goto out_restart;
7059                 fallthrough;
7060         case -NFS4ERR_BAD_STATEID:
7061         case -NFS4ERR_STALE_STATEID:
7062         case -NFS4ERR_EXPIRED:
7063                 if (data->arg.new_lock_owner != 0) {
7064                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7065                                                 &lsp->ls_state->open_stateid))
7066                                 goto out_restart;
7067                         else if (nfs4_async_handle_error(task, server, lsp->ls_state, NULL) == -EAGAIN)
7068                                 goto out_restart;
7069                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7070                                                 &lsp->ls_stateid))
7071                                 goto out_restart;
7072         }
7073 out_done:
7074         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7075         return;
7076 out_restart:
7077         if (!data->cancelled)
7078                 rpc_restart_call_prepare(task);
7079         goto out_done;
7080 }
7081
7082 static void nfs4_lock_release(void *calldata)
7083 {
7084         struct nfs4_lockdata *data = calldata;
7085
7086         dprintk("%s: begin!\n", __func__);
7087         nfs_free_seqid(data->arg.open_seqid);
7088         if (data->cancelled && data->rpc_status == 0) {
7089                 struct rpc_task *task;
7090                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7091                                 data->arg.lock_seqid);
7092                 if (!IS_ERR(task))
7093                         rpc_put_task_async(task);
7094                 dprintk("%s: cancelling lock!\n", __func__);
7095         } else
7096                 nfs_free_seqid(data->arg.lock_seqid);
7097         nfs4_put_lock_state(data->lsp);
7098         put_nfs_open_context(data->ctx);
7099         kfree(data);
7100         dprintk("%s: done!\n", __func__);
7101 }
7102
7103 static const struct rpc_call_ops nfs4_lock_ops = {
7104         .rpc_call_prepare = nfs4_lock_prepare,
7105         .rpc_call_done = nfs4_lock_done,
7106         .rpc_release = nfs4_lock_release,
7107 };
7108
7109 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7110 {
7111         switch (error) {
7112         case -NFS4ERR_ADMIN_REVOKED:
7113         case -NFS4ERR_EXPIRED:
7114         case -NFS4ERR_BAD_STATEID:
7115                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7116                 if (new_lock_owner != 0 ||
7117                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7118                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7119                 break;
7120         case -NFS4ERR_STALE_STATEID:
7121                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7122                 nfs4_schedule_lease_recovery(server->nfs_client);
7123         }
7124 }
7125
7126 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7127 {
7128         struct nfs4_lockdata *data;
7129         struct rpc_task *task;
7130         struct rpc_message msg = {
7131                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7132                 .rpc_cred = state->owner->so_cred,
7133         };
7134         struct rpc_task_setup task_setup_data = {
7135                 .rpc_client = NFS_CLIENT(state->inode),
7136                 .rpc_message = &msg,
7137                 .callback_ops = &nfs4_lock_ops,
7138                 .workqueue = nfsiod_workqueue,
7139                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7140         };
7141         int ret;
7142
7143         dprintk("%s: begin!\n", __func__);
7144         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7145                         fl->fl_u.nfs4_fl.owner,
7146                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7147         if (data == NULL)
7148                 return -ENOMEM;
7149         if (IS_SETLKW(cmd))
7150                 data->arg.block = 1;
7151         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7152                                 recovery_type > NFS_LOCK_NEW);
7153         msg.rpc_argp = &data->arg;
7154         msg.rpc_resp = &data->res;
7155         task_setup_data.callback_data = data;
7156         if (recovery_type > NFS_LOCK_NEW) {
7157                 if (recovery_type == NFS_LOCK_RECLAIM)
7158                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7159         } else
7160                 data->arg.new_lock = 1;
7161         task = rpc_run_task(&task_setup_data);
7162         if (IS_ERR(task))
7163                 return PTR_ERR(task);
7164         ret = rpc_wait_for_completion_task(task);
7165         if (ret == 0) {
7166                 ret = data->rpc_status;
7167                 if (ret)
7168                         nfs4_handle_setlk_error(data->server, data->lsp,
7169                                         data->arg.new_lock_owner, ret);
7170         } else
7171                 data->cancelled = true;
7172         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7173         rpc_put_task(task);
7174         dprintk("%s: done, ret = %d!\n", __func__, ret);
7175         return ret;
7176 }
7177
7178 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7179 {
7180         struct nfs_server *server = NFS_SERVER(state->inode);
7181         struct nfs4_exception exception = {
7182                 .inode = state->inode,
7183         };
7184         int err;
7185
7186         do {
7187                 /* Cache the lock if possible... */
7188                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7189                         return 0;
7190                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7191                 if (err != -NFS4ERR_DELAY)
7192                         break;
7193                 nfs4_handle_exception(server, err, &exception);
7194         } while (exception.retry);
7195         return err;
7196 }
7197
7198 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7199 {
7200         struct nfs_server *server = NFS_SERVER(state->inode);
7201         struct nfs4_exception exception = {
7202                 .inode = state->inode,
7203         };
7204         int err;
7205
7206         err = nfs4_set_lock_state(state, request);
7207         if (err != 0)
7208                 return err;
7209         if (!recover_lost_locks) {
7210                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7211                 return 0;
7212         }
7213         do {
7214                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7215                         return 0;
7216                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7217                 switch (err) {
7218                 default:
7219                         goto out;
7220                 case -NFS4ERR_GRACE:
7221                 case -NFS4ERR_DELAY:
7222                         nfs4_handle_exception(server, err, &exception);
7223                         err = 0;
7224                 }
7225         } while (exception.retry);
7226 out:
7227         return err;
7228 }
7229
7230 #if defined(CONFIG_NFS_V4_1)
7231 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7232 {
7233         struct nfs4_lock_state *lsp;
7234         int status;
7235
7236         status = nfs4_set_lock_state(state, request);
7237         if (status != 0)
7238                 return status;
7239         lsp = request->fl_u.nfs4_fl.owner;
7240         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7241             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7242                 return 0;
7243         return nfs4_lock_expired(state, request);
7244 }
7245 #endif
7246
7247 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7248 {
7249         struct nfs_inode *nfsi = NFS_I(state->inode);
7250         struct nfs4_state_owner *sp = state->owner;
7251         unsigned char fl_flags = request->fl_flags;
7252         int status;
7253
7254         request->fl_flags |= FL_ACCESS;
7255         status = locks_lock_inode_wait(state->inode, request);
7256         if (status < 0)
7257                 goto out;
7258         mutex_lock(&sp->so_delegreturn_mutex);
7259         down_read(&nfsi->rwsem);
7260         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7261                 /* Yes: cache locks! */
7262                 /* ...but avoid races with delegation recall... */
7263                 request->fl_flags = fl_flags & ~FL_SLEEP;
7264                 status = locks_lock_inode_wait(state->inode, request);
7265                 up_read(&nfsi->rwsem);
7266                 mutex_unlock(&sp->so_delegreturn_mutex);
7267                 goto out;
7268         }
7269         up_read(&nfsi->rwsem);
7270         mutex_unlock(&sp->so_delegreturn_mutex);
7271         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7272 out:
7273         request->fl_flags = fl_flags;
7274         return status;
7275 }
7276
7277 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7278 {
7279         struct nfs4_exception exception = {
7280                 .state = state,
7281                 .inode = state->inode,
7282                 .interruptible = true,
7283         };
7284         int err;
7285
7286         do {
7287                 err = _nfs4_proc_setlk(state, cmd, request);
7288                 if (err == -NFS4ERR_DENIED)
7289                         err = -EAGAIN;
7290                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7291                                 err, &exception);
7292         } while (exception.retry);
7293         return err;
7294 }
7295
7296 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7297 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7298
7299 static int
7300 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7301                         struct file_lock *request)
7302 {
7303         int             status = -ERESTARTSYS;
7304         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7305
7306         while(!signalled()) {
7307                 status = nfs4_proc_setlk(state, cmd, request);
7308                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7309                         break;
7310                 freezable_schedule_timeout_interruptible(timeout);
7311                 timeout *= 2;
7312                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7313                 status = -ERESTARTSYS;
7314         }
7315         return status;
7316 }
7317
7318 #ifdef CONFIG_NFS_V4_1
7319 struct nfs4_lock_waiter {
7320         struct task_struct      *task;
7321         struct inode            *inode;
7322         struct nfs_lowner       *owner;
7323 };
7324
7325 static int
7326 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7327 {
7328         int ret;
7329         struct nfs4_lock_waiter *waiter = wait->private;
7330
7331         /* NULL key means to wake up everyone */
7332         if (key) {
7333                 struct cb_notify_lock_args      *cbnl = key;
7334                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7335                                                 *wowner = waiter->owner;
7336
7337                 /* Only wake if the callback was for the same owner. */
7338                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7339                         return 0;
7340
7341                 /* Make sure it's for the right inode */
7342                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7343                         return 0;
7344         }
7345
7346         /* override "private" so we can use default_wake_function */
7347         wait->private = waiter->task;
7348         ret = woken_wake_function(wait, mode, flags, key);
7349         if (ret)
7350                 list_del_init(&wait->entry);
7351         wait->private = waiter;
7352         return ret;
7353 }
7354
7355 static int
7356 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7357 {
7358         int status = -ERESTARTSYS;
7359         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7360         struct nfs_server *server = NFS_SERVER(state->inode);
7361         struct nfs_client *clp = server->nfs_client;
7362         wait_queue_head_t *q = &clp->cl_lock_waitq;
7363         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7364                                     .id = lsp->ls_seqid.owner_id,
7365                                     .s_dev = server->s_dev };
7366         struct nfs4_lock_waiter waiter = { .task  = current,
7367                                            .inode = state->inode,
7368                                            .owner = &owner};
7369         wait_queue_entry_t wait;
7370
7371         /* Don't bother with waitqueue if we don't expect a callback */
7372         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7373                 return nfs4_retry_setlk_simple(state, cmd, request);
7374
7375         init_wait(&wait);
7376         wait.private = &waiter;
7377         wait.func = nfs4_wake_lock_waiter;
7378
7379         while(!signalled()) {
7380                 add_wait_queue(q, &wait);
7381                 status = nfs4_proc_setlk(state, cmd, request);
7382                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7383                         finish_wait(q, &wait);
7384                         break;
7385                 }
7386
7387                 status = -ERESTARTSYS;
7388                 freezer_do_not_count();
7389                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7390                 freezer_count();
7391                 finish_wait(q, &wait);
7392         }
7393
7394         return status;
7395 }
7396 #else /* !CONFIG_NFS_V4_1 */
7397 static inline int
7398 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7399 {
7400         return nfs4_retry_setlk_simple(state, cmd, request);
7401 }
7402 #endif
7403
7404 static int
7405 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7406 {
7407         struct nfs_open_context *ctx;
7408         struct nfs4_state *state;
7409         int status;
7410
7411         /* verify open state */
7412         ctx = nfs_file_open_context(filp);
7413         state = ctx->state;
7414
7415         if (IS_GETLK(cmd)) {
7416                 if (state != NULL)
7417                         return nfs4_proc_getlk(state, F_GETLK, request);
7418                 return 0;
7419         }
7420
7421         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7422                 return -EINVAL;
7423
7424         if (request->fl_type == F_UNLCK) {
7425                 if (state != NULL)
7426                         return nfs4_proc_unlck(state, cmd, request);
7427                 return 0;
7428         }
7429
7430         if (state == NULL)
7431                 return -ENOLCK;
7432
7433         if ((request->fl_flags & FL_POSIX) &&
7434             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7435                 return -ENOLCK;
7436
7437         /*
7438          * Don't rely on the VFS having checked the file open mode,
7439          * since it won't do this for flock() locks.
7440          */
7441         switch (request->fl_type) {
7442         case F_RDLCK:
7443                 if (!(filp->f_mode & FMODE_READ))
7444                         return -EBADF;
7445                 break;
7446         case F_WRLCK:
7447                 if (!(filp->f_mode & FMODE_WRITE))
7448                         return -EBADF;
7449         }
7450
7451         status = nfs4_set_lock_state(state, request);
7452         if (status != 0)
7453                 return status;
7454
7455         return nfs4_retry_setlk(state, cmd, request);
7456 }
7457
7458 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7459 {
7460         struct nfs_server *server = NFS_SERVER(state->inode);
7461         int err;
7462
7463         err = nfs4_set_lock_state(state, fl);
7464         if (err != 0)
7465                 return err;
7466         do {
7467                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7468                 if (err != -NFS4ERR_DELAY)
7469                         break;
7470                 ssleep(1);
7471         } while (err == -NFS4ERR_DELAY);
7472         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7473 }
7474
7475 struct nfs_release_lockowner_data {
7476         struct nfs4_lock_state *lsp;
7477         struct nfs_server *server;
7478         struct nfs_release_lockowner_args args;
7479         struct nfs_release_lockowner_res res;
7480         unsigned long timestamp;
7481 };
7482
7483 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7484 {
7485         struct nfs_release_lockowner_data *data = calldata;
7486         struct nfs_server *server = data->server;
7487         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7488                            &data->res.seq_res, task);
7489         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7490         data->timestamp = jiffies;
7491 }
7492
7493 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7494 {
7495         struct nfs_release_lockowner_data *data = calldata;
7496         struct nfs_server *server = data->server;
7497
7498         nfs40_sequence_done(task, &data->res.seq_res);
7499
7500         switch (task->tk_status) {
7501         case 0:
7502                 renew_lease(server, data->timestamp);
7503                 break;
7504         case -NFS4ERR_STALE_CLIENTID:
7505         case -NFS4ERR_EXPIRED:
7506                 nfs4_schedule_lease_recovery(server->nfs_client);
7507                 break;
7508         case -NFS4ERR_LEASE_MOVED:
7509         case -NFS4ERR_DELAY:
7510                 if (nfs4_async_handle_error(task, server,
7511                                             NULL, NULL) == -EAGAIN)
7512                         rpc_restart_call_prepare(task);
7513         }
7514 }
7515
7516 static void nfs4_release_lockowner_release(void *calldata)
7517 {
7518         struct nfs_release_lockowner_data *data = calldata;
7519         nfs4_free_lock_state(data->server, data->lsp);
7520         kfree(calldata);
7521 }
7522
7523 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7524         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7525         .rpc_call_done = nfs4_release_lockowner_done,
7526         .rpc_release = nfs4_release_lockowner_release,
7527 };
7528
7529 static void
7530 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7531 {
7532         struct nfs_release_lockowner_data *data;
7533         struct rpc_message msg = {
7534                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7535         };
7536
7537         if (server->nfs_client->cl_mvops->minor_version != 0)
7538                 return;
7539
7540         data = kmalloc(sizeof(*data), GFP_NOFS);
7541         if (!data)
7542                 return;
7543         data->lsp = lsp;
7544         data->server = server;
7545         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7546         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7547         data->args.lock_owner.s_dev = server->s_dev;
7548
7549         msg.rpc_argp = &data->args;
7550         msg.rpc_resp = &data->res;
7551         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7552         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7553 }
7554
7555 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7556
7557 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7558                                    struct dentry *unused, struct inode *inode,
7559                                    const char *key, const void *buf,
7560                                    size_t buflen, int flags)
7561 {
7562         return nfs4_proc_set_acl(inode, buf, buflen);
7563 }
7564
7565 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7566                                    struct dentry *unused, struct inode *inode,
7567                                    const char *key, void *buf, size_t buflen)
7568 {
7569         return nfs4_proc_get_acl(inode, buf, buflen);
7570 }
7571
7572 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7573 {
7574         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7575 }
7576
7577 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7578
7579 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7580                                      struct dentry *unused, struct inode *inode,
7581                                      const char *key, const void *buf,
7582                                      size_t buflen, int flags)
7583 {
7584         if (security_ismaclabel(key))
7585                 return nfs4_set_security_label(inode, buf, buflen);
7586
7587         return -EOPNOTSUPP;
7588 }
7589
7590 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7591                                      struct dentry *unused, struct inode *inode,
7592                                      const char *key, void *buf, size_t buflen)
7593 {
7594         if (security_ismaclabel(key))
7595                 return nfs4_get_security_label(inode, buf, buflen);
7596         return -EOPNOTSUPP;
7597 }
7598
7599 static ssize_t
7600 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7601 {
7602         int len = 0;
7603
7604         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7605                 len = security_inode_listsecurity(inode, list, list_len);
7606                 if (len >= 0 && list_len && len > list_len)
7607                         return -ERANGE;
7608         }
7609         return len;
7610 }
7611
7612 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7613         .prefix = XATTR_SECURITY_PREFIX,
7614         .get    = nfs4_xattr_get_nfs4_label,
7615         .set    = nfs4_xattr_set_nfs4_label,
7616 };
7617
7618 #else
7619
7620 static ssize_t
7621 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7622 {
7623         return 0;
7624 }
7625
7626 #endif
7627
7628 #ifdef CONFIG_NFS_V4_2
7629 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7630                                     struct dentry *unused, struct inode *inode,
7631                                     const char *key, const void *buf,
7632                                     size_t buflen, int flags)
7633 {
7634         u32 mask;
7635         int ret;
7636
7637         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7638                 return -EOPNOTSUPP;
7639
7640         /*
7641          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7642          * flags right now. Handling of xattr operations use the normal
7643          * file read/write permissions.
7644          *
7645          * Just in case the server has other ideas (which RFC 8276 allows),
7646          * do a cached access check for the XA* flags to possibly avoid
7647          * doing an RPC and getting EACCES back.
7648          */
7649         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7650                 if (!(mask & NFS_ACCESS_XAWRITE))
7651                         return -EACCES;
7652         }
7653
7654         if (buf == NULL) {
7655                 ret = nfs42_proc_removexattr(inode, key);
7656                 if (!ret)
7657                         nfs4_xattr_cache_remove(inode, key);
7658         } else {
7659                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7660                 if (!ret)
7661                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7662         }
7663
7664         return ret;
7665 }
7666
7667 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7668                                     struct dentry *unused, struct inode *inode,
7669                                     const char *key, void *buf, size_t buflen)
7670 {
7671         u32 mask;
7672         ssize_t ret;
7673
7674         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7675                 return -EOPNOTSUPP;
7676
7677         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7678                 if (!(mask & NFS_ACCESS_XAREAD))
7679                         return -EACCES;
7680         }
7681
7682         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7683         if (ret)
7684                 return ret;
7685
7686         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7687         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7688                 return ret;
7689
7690         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7691
7692         return ret;
7693 }
7694
7695 static ssize_t
7696 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7697 {
7698         u64 cookie;
7699         bool eof;
7700         ssize_t ret, size;
7701         char *buf;
7702         size_t buflen;
7703         u32 mask;
7704
7705         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7706                 return 0;
7707
7708         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7709                 if (!(mask & NFS_ACCESS_XALIST))
7710                         return 0;
7711         }
7712
7713         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7714         if (ret)
7715                 return ret;
7716
7717         ret = nfs4_xattr_cache_list(inode, list, list_len);
7718         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7719                 return ret;
7720
7721         cookie = 0;
7722         eof = false;
7723         buflen = list_len ? list_len : XATTR_LIST_MAX;
7724         buf = list_len ? list : NULL;
7725         size = 0;
7726
7727         while (!eof) {
7728                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7729                     &cookie, &eof);
7730                 if (ret < 0)
7731                         return ret;
7732
7733                 if (list_len) {
7734                         buf += ret;
7735                         buflen -= ret;
7736                 }
7737                 size += ret;
7738         }
7739
7740         if (list_len)
7741                 nfs4_xattr_cache_set_list(inode, list, size);
7742
7743         return size;
7744 }
7745
7746 #else
7747
7748 static ssize_t
7749 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7750 {
7751         return 0;
7752 }
7753 #endif /* CONFIG_NFS_V4_2 */
7754
7755 /*
7756  * nfs_fhget will use either the mounted_on_fileid or the fileid
7757  */
7758 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7759 {
7760         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7761                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7762               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7763               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7764                 return;
7765
7766         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7767                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7768         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7769         fattr->nlink = 2;
7770 }
7771
7772 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7773                                    const struct qstr *name,
7774                                    struct nfs4_fs_locations *fs_locations,
7775                                    struct page *page)
7776 {
7777         struct nfs_server *server = NFS_SERVER(dir);
7778         u32 bitmask[3];
7779         struct nfs4_fs_locations_arg args = {
7780                 .dir_fh = NFS_FH(dir),
7781                 .name = name,
7782                 .page = page,
7783                 .bitmask = bitmask,
7784         };
7785         struct nfs4_fs_locations_res res = {
7786                 .fs_locations = fs_locations,
7787         };
7788         struct rpc_message msg = {
7789                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7790                 .rpc_argp = &args,
7791                 .rpc_resp = &res,
7792         };
7793         int status;
7794
7795         dprintk("%s: start\n", __func__);
7796
7797         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7798         bitmask[1] = nfs4_fattr_bitmap[1];
7799
7800         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7801          * is not supported */
7802         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7803                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7804         else
7805                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7806
7807         nfs_fattr_init(&fs_locations->fattr);
7808         fs_locations->server = server;
7809         fs_locations->nlocations = 0;
7810         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7811         dprintk("%s: returned status = %d\n", __func__, status);
7812         return status;
7813 }
7814
7815 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7816                            const struct qstr *name,
7817                            struct nfs4_fs_locations *fs_locations,
7818                            struct page *page)
7819 {
7820         struct nfs4_exception exception = {
7821                 .interruptible = true,
7822         };
7823         int err;
7824         do {
7825                 err = _nfs4_proc_fs_locations(client, dir, name,
7826                                 fs_locations, page);
7827                 trace_nfs4_get_fs_locations(dir, name, err);
7828                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7829                                 &exception);
7830         } while (exception.retry);
7831         return err;
7832 }
7833
7834 /*
7835  * This operation also signals the server that this client is
7836  * performing migration recovery.  The server can stop returning
7837  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7838  * appended to this compound to identify the client ID which is
7839  * performing recovery.
7840  */
7841 static int _nfs40_proc_get_locations(struct inode *inode,
7842                                      struct nfs4_fs_locations *locations,
7843                                      struct page *page, const struct cred *cred)
7844 {
7845         struct nfs_server *server = NFS_SERVER(inode);
7846         struct rpc_clnt *clnt = server->client;
7847         u32 bitmask[2] = {
7848                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7849         };
7850         struct nfs4_fs_locations_arg args = {
7851                 .clientid       = server->nfs_client->cl_clientid,
7852                 .fh             = NFS_FH(inode),
7853                 .page           = page,
7854                 .bitmask        = bitmask,
7855                 .migration      = 1,            /* skip LOOKUP */
7856                 .renew          = 1,            /* append RENEW */
7857         };
7858         struct nfs4_fs_locations_res res = {
7859                 .fs_locations   = locations,
7860                 .migration      = 1,
7861                 .renew          = 1,
7862         };
7863         struct rpc_message msg = {
7864                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7865                 .rpc_argp       = &args,
7866                 .rpc_resp       = &res,
7867                 .rpc_cred       = cred,
7868         };
7869         unsigned long now = jiffies;
7870         int status;
7871
7872         nfs_fattr_init(&locations->fattr);
7873         locations->server = server;
7874         locations->nlocations = 0;
7875
7876         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7877         status = nfs4_call_sync_sequence(clnt, server, &msg,
7878                                         &args.seq_args, &res.seq_res);
7879         if (status)
7880                 return status;
7881
7882         renew_lease(server, now);
7883         return 0;
7884 }
7885
7886 #ifdef CONFIG_NFS_V4_1
7887
7888 /*
7889  * This operation also signals the server that this client is
7890  * performing migration recovery.  The server can stop asserting
7891  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7892  * performing this operation is identified in the SEQUENCE
7893  * operation in this compound.
7894  *
7895  * When the client supports GETATTR(fs_locations_info), it can
7896  * be plumbed in here.
7897  */
7898 static int _nfs41_proc_get_locations(struct inode *inode,
7899                                      struct nfs4_fs_locations *locations,
7900                                      struct page *page, const struct cred *cred)
7901 {
7902         struct nfs_server *server = NFS_SERVER(inode);
7903         struct rpc_clnt *clnt = server->client;
7904         u32 bitmask[2] = {
7905                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7906         };
7907         struct nfs4_fs_locations_arg args = {
7908                 .fh             = NFS_FH(inode),
7909                 .page           = page,
7910                 .bitmask        = bitmask,
7911                 .migration      = 1,            /* skip LOOKUP */
7912         };
7913         struct nfs4_fs_locations_res res = {
7914                 .fs_locations   = locations,
7915                 .migration      = 1,
7916         };
7917         struct rpc_message msg = {
7918                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7919                 .rpc_argp       = &args,
7920                 .rpc_resp       = &res,
7921                 .rpc_cred       = cred,
7922         };
7923         int status;
7924
7925         nfs_fattr_init(&locations->fattr);
7926         locations->server = server;
7927         locations->nlocations = 0;
7928
7929         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7930         status = nfs4_call_sync_sequence(clnt, server, &msg,
7931                                         &args.seq_args, &res.seq_res);
7932         if (status == NFS4_OK &&
7933             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7934                 status = -NFS4ERR_LEASE_MOVED;
7935         return status;
7936 }
7937
7938 #endif  /* CONFIG_NFS_V4_1 */
7939
7940 /**
7941  * nfs4_proc_get_locations - discover locations for a migrated FSID
7942  * @inode: inode on FSID that is migrating
7943  * @locations: result of query
7944  * @page: buffer
7945  * @cred: credential to use for this operation
7946  *
7947  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7948  * operation failed, or a negative errno if a local error occurred.
7949  *
7950  * On success, "locations" is filled in, but if the server has
7951  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7952  * asserted.
7953  *
7954  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7955  * from this client that require migration recovery.
7956  */
7957 int nfs4_proc_get_locations(struct inode *inode,
7958                             struct nfs4_fs_locations *locations,
7959                             struct page *page, const struct cred *cred)
7960 {
7961         struct nfs_server *server = NFS_SERVER(inode);
7962         struct nfs_client *clp = server->nfs_client;
7963         const struct nfs4_mig_recovery_ops *ops =
7964                                         clp->cl_mvops->mig_recovery_ops;
7965         struct nfs4_exception exception = {
7966                 .interruptible = true,
7967         };
7968         int status;
7969
7970         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7971                 (unsigned long long)server->fsid.major,
7972                 (unsigned long long)server->fsid.minor,
7973                 clp->cl_hostname);
7974         nfs_display_fhandle(NFS_FH(inode), __func__);
7975
7976         do {
7977                 status = ops->get_locations(inode, locations, page, cred);
7978                 if (status != -NFS4ERR_DELAY)
7979                         break;
7980                 nfs4_handle_exception(server, status, &exception);
7981         } while (exception.retry);
7982         return status;
7983 }
7984
7985 /*
7986  * This operation also signals the server that this client is
7987  * performing "lease moved" recovery.  The server can stop
7988  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7989  * is appended to this compound to identify the client ID which is
7990  * performing recovery.
7991  */
7992 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7993 {
7994         struct nfs_server *server = NFS_SERVER(inode);
7995         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7996         struct rpc_clnt *clnt = server->client;
7997         struct nfs4_fsid_present_arg args = {
7998                 .fh             = NFS_FH(inode),
7999                 .clientid       = clp->cl_clientid,
8000                 .renew          = 1,            /* append RENEW */
8001         };
8002         struct nfs4_fsid_present_res res = {
8003                 .renew          = 1,
8004         };
8005         struct rpc_message msg = {
8006                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8007                 .rpc_argp       = &args,
8008                 .rpc_resp       = &res,
8009                 .rpc_cred       = cred,
8010         };
8011         unsigned long now = jiffies;
8012         int status;
8013
8014         res.fh = nfs_alloc_fhandle();
8015         if (res.fh == NULL)
8016                 return -ENOMEM;
8017
8018         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8019         status = nfs4_call_sync_sequence(clnt, server, &msg,
8020                                                 &args.seq_args, &res.seq_res);
8021         nfs_free_fhandle(res.fh);
8022         if (status)
8023                 return status;
8024
8025         do_renew_lease(clp, now);
8026         return 0;
8027 }
8028
8029 #ifdef CONFIG_NFS_V4_1
8030
8031 /*
8032  * This operation also signals the server that this client is
8033  * performing "lease moved" recovery.  The server can stop asserting
8034  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8035  * this operation is identified in the SEQUENCE operation in this
8036  * compound.
8037  */
8038 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8039 {
8040         struct nfs_server *server = NFS_SERVER(inode);
8041         struct rpc_clnt *clnt = server->client;
8042         struct nfs4_fsid_present_arg args = {
8043                 .fh             = NFS_FH(inode),
8044         };
8045         struct nfs4_fsid_present_res res = {
8046         };
8047         struct rpc_message msg = {
8048                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8049                 .rpc_argp       = &args,
8050                 .rpc_resp       = &res,
8051                 .rpc_cred       = cred,
8052         };
8053         int status;
8054
8055         res.fh = nfs_alloc_fhandle();
8056         if (res.fh == NULL)
8057                 return -ENOMEM;
8058
8059         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8060         status = nfs4_call_sync_sequence(clnt, server, &msg,
8061                                                 &args.seq_args, &res.seq_res);
8062         nfs_free_fhandle(res.fh);
8063         if (status == NFS4_OK &&
8064             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8065                 status = -NFS4ERR_LEASE_MOVED;
8066         return status;
8067 }
8068
8069 #endif  /* CONFIG_NFS_V4_1 */
8070
8071 /**
8072  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8073  * @inode: inode on FSID to check
8074  * @cred: credential to use for this operation
8075  *
8076  * Server indicates whether the FSID is present, moved, or not
8077  * recognized.  This operation is necessary to clear a LEASE_MOVED
8078  * condition for this client ID.
8079  *
8080  * Returns NFS4_OK if the FSID is present on this server,
8081  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8082  *  NFS4ERR code if some error occurred on the server, or a
8083  *  negative errno if a local failure occurred.
8084  */
8085 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8086 {
8087         struct nfs_server *server = NFS_SERVER(inode);
8088         struct nfs_client *clp = server->nfs_client;
8089         const struct nfs4_mig_recovery_ops *ops =
8090                                         clp->cl_mvops->mig_recovery_ops;
8091         struct nfs4_exception exception = {
8092                 .interruptible = true,
8093         };
8094         int status;
8095
8096         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8097                 (unsigned long long)server->fsid.major,
8098                 (unsigned long long)server->fsid.minor,
8099                 clp->cl_hostname);
8100         nfs_display_fhandle(NFS_FH(inode), __func__);
8101
8102         do {
8103                 status = ops->fsid_present(inode, cred);
8104                 if (status != -NFS4ERR_DELAY)
8105                         break;
8106                 nfs4_handle_exception(server, status, &exception);
8107         } while (exception.retry);
8108         return status;
8109 }
8110
8111 /*
8112  * If 'use_integrity' is true and the state managment nfs_client
8113  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8114  * and the machine credential as per RFC3530bis and RFC5661 Security
8115  * Considerations sections. Otherwise, just use the user cred with the
8116  * filesystem's rpc_client.
8117  */
8118 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8119 {
8120         int status;
8121         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8122         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8123         struct nfs4_secinfo_arg args = {
8124                 .dir_fh = NFS_FH(dir),
8125                 .name   = name,
8126         };
8127         struct nfs4_secinfo_res res = {
8128                 .flavors     = flavors,
8129         };
8130         struct rpc_message msg = {
8131                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8132                 .rpc_argp = &args,
8133                 .rpc_resp = &res,
8134         };
8135         struct nfs4_call_sync_data data = {
8136                 .seq_server = NFS_SERVER(dir),
8137                 .seq_args = &args.seq_args,
8138                 .seq_res = &res.seq_res,
8139         };
8140         struct rpc_task_setup task_setup = {
8141                 .rpc_client = clnt,
8142                 .rpc_message = &msg,
8143                 .callback_ops = clp->cl_mvops->call_sync_ops,
8144                 .callback_data = &data,
8145                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8146         };
8147         const struct cred *cred = NULL;
8148
8149         if (use_integrity) {
8150                 clnt = clp->cl_rpcclient;
8151                 task_setup.rpc_client = clnt;
8152
8153                 cred = nfs4_get_clid_cred(clp);
8154                 msg.rpc_cred = cred;
8155         }
8156
8157         dprintk("NFS call  secinfo %s\n", name->name);
8158
8159         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8160         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8161         status = nfs4_call_sync_custom(&task_setup);
8162
8163         dprintk("NFS reply  secinfo: %d\n", status);
8164
8165         put_cred(cred);
8166         return status;
8167 }
8168
8169 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8170                       struct nfs4_secinfo_flavors *flavors)
8171 {
8172         struct nfs4_exception exception = {
8173                 .interruptible = true,
8174         };
8175         int err;
8176         do {
8177                 err = -NFS4ERR_WRONGSEC;
8178
8179                 /* try to use integrity protection with machine cred */
8180                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8181                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8182
8183                 /*
8184                  * if unable to use integrity protection, or SECINFO with
8185                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8186                  * disallowed by spec, but exists in deployed servers) use
8187                  * the current filesystem's rpc_client and the user cred.
8188                  */
8189                 if (err == -NFS4ERR_WRONGSEC)
8190                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8191
8192                 trace_nfs4_secinfo(dir, name, err);
8193                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8194                                 &exception);
8195         } while (exception.retry);
8196         return err;
8197 }
8198
8199 #ifdef CONFIG_NFS_V4_1
8200 /*
8201  * Check the exchange flags returned by the server for invalid flags, having
8202  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8203  * DS flags set.
8204  */
8205 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8206 {
8207         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8208                 goto out_inval;
8209         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8210                 goto out_inval;
8211         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8212             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8213                 goto out_inval;
8214         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8215                 goto out_inval;
8216         return NFS_OK;
8217 out_inval:
8218         return -NFS4ERR_INVAL;
8219 }
8220
8221 static bool
8222 nfs41_same_server_scope(struct nfs41_server_scope *a,
8223                         struct nfs41_server_scope *b)
8224 {
8225         if (a->server_scope_sz != b->server_scope_sz)
8226                 return false;
8227         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8228 }
8229
8230 static void
8231 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8232 {
8233         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8234         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8235         struct nfs_client *clp = args->client;
8236
8237         switch (task->tk_status) {
8238         case -NFS4ERR_BADSESSION:
8239         case -NFS4ERR_DEADSESSION:
8240                 nfs4_schedule_session_recovery(clp->cl_session,
8241                                 task->tk_status);
8242                 return;
8243         }
8244         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8245                         res->dir != NFS4_CDFS4_BOTH) {
8246                 rpc_task_close_connection(task);
8247                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8248                         rpc_restart_call(task);
8249         }
8250 }
8251
8252 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8253         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8254 };
8255
8256 /*
8257  * nfs4_proc_bind_one_conn_to_session()
8258  *
8259  * The 4.1 client currently uses the same TCP connection for the
8260  * fore and backchannel.
8261  */
8262 static
8263 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8264                 struct rpc_xprt *xprt,
8265                 struct nfs_client *clp,
8266                 const struct cred *cred)
8267 {
8268         int status;
8269         struct nfs41_bind_conn_to_session_args args = {
8270                 .client = clp,
8271                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8272                 .retries = 0,
8273         };
8274         struct nfs41_bind_conn_to_session_res res;
8275         struct rpc_message msg = {
8276                 .rpc_proc =
8277                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8278                 .rpc_argp = &args,
8279                 .rpc_resp = &res,
8280                 .rpc_cred = cred,
8281         };
8282         struct rpc_task_setup task_setup_data = {
8283                 .rpc_client = clnt,
8284                 .rpc_xprt = xprt,
8285                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8286                 .rpc_message = &msg,
8287                 .flags = RPC_TASK_TIMEOUT,
8288         };
8289         struct rpc_task *task;
8290
8291         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8292         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8293                 args.dir = NFS4_CDFC4_FORE;
8294
8295         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8296         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8297                 args.dir = NFS4_CDFC4_FORE;
8298
8299         task = rpc_run_task(&task_setup_data);
8300         if (!IS_ERR(task)) {
8301                 status = task->tk_status;
8302                 rpc_put_task(task);
8303         } else
8304                 status = PTR_ERR(task);
8305         trace_nfs4_bind_conn_to_session(clp, status);
8306         if (status == 0) {
8307                 if (memcmp(res.sessionid.data,
8308                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8309                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8310                         return -EIO;
8311                 }
8312                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8313                         dprintk("NFS: %s: Unexpected direction from server\n",
8314                                 __func__);
8315                         return -EIO;
8316                 }
8317                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8318                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8319                                 __func__);
8320                         return -EIO;
8321                 }
8322         }
8323
8324         return status;
8325 }
8326
8327 struct rpc_bind_conn_calldata {
8328         struct nfs_client *clp;
8329         const struct cred *cred;
8330 };
8331
8332 static int
8333 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8334                 struct rpc_xprt *xprt,
8335                 void *calldata)
8336 {
8337         struct rpc_bind_conn_calldata *p = calldata;
8338
8339         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8340 }
8341
8342 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8343 {
8344         struct rpc_bind_conn_calldata data = {
8345                 .clp = clp,
8346                 .cred = cred,
8347         };
8348         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8349                         nfs4_proc_bind_conn_to_session_callback, &data);
8350 }
8351
8352 /*
8353  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8354  * and operations we'd like to see to enable certain features in the allow map
8355  */
8356 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8357         .how = SP4_MACH_CRED,
8358         .enforce.u.words = {
8359                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8360                       1 << (OP_EXCHANGE_ID - 32) |
8361                       1 << (OP_CREATE_SESSION - 32) |
8362                       1 << (OP_DESTROY_SESSION - 32) |
8363                       1 << (OP_DESTROY_CLIENTID - 32)
8364         },
8365         .allow.u.words = {
8366                 [0] = 1 << (OP_CLOSE) |
8367                       1 << (OP_OPEN_DOWNGRADE) |
8368                       1 << (OP_LOCKU) |
8369                       1 << (OP_DELEGRETURN) |
8370                       1 << (OP_COMMIT),
8371                 [1] = 1 << (OP_SECINFO - 32) |
8372                       1 << (OP_SECINFO_NO_NAME - 32) |
8373                       1 << (OP_LAYOUTRETURN - 32) |
8374                       1 << (OP_TEST_STATEID - 32) |
8375                       1 << (OP_FREE_STATEID - 32) |
8376                       1 << (OP_WRITE - 32)
8377         }
8378 };
8379
8380 /*
8381  * Select the state protection mode for client `clp' given the server results
8382  * from exchange_id in `sp'.
8383  *
8384  * Returns 0 on success, negative errno otherwise.
8385  */
8386 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8387                                  struct nfs41_state_protection *sp)
8388 {
8389         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8390                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8391                       1 << (OP_EXCHANGE_ID - 32) |
8392                       1 << (OP_CREATE_SESSION - 32) |
8393                       1 << (OP_DESTROY_SESSION - 32) |
8394                       1 << (OP_DESTROY_CLIENTID - 32)
8395         };
8396         unsigned long flags = 0;
8397         unsigned int i;
8398         int ret = 0;
8399
8400         if (sp->how == SP4_MACH_CRED) {
8401                 /* Print state protect result */
8402                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8403                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8404                         if (test_bit(i, sp->enforce.u.longs))
8405                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8406                         if (test_bit(i, sp->allow.u.longs))
8407                                 dfprintk(MOUNT, "  allow op %d\n", i);
8408                 }
8409
8410                 /* make sure nothing is on enforce list that isn't supported */
8411                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8412                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8413                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8414                                 ret = -EINVAL;
8415                                 goto out;
8416                         }
8417                 }
8418
8419                 /*
8420                  * Minimal mode - state operations are allowed to use machine
8421                  * credential.  Note this already happens by default, so the
8422                  * client doesn't have to do anything more than the negotiation.
8423                  *
8424                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8425                  *       we're already using the machine cred for exchange_id
8426                  *       and will never use a different cred.
8427                  */
8428                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8429                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8430                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8431                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8432                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8433                         dfprintk(MOUNT, "  minimal mode enabled\n");
8434                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8435                 } else {
8436                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8437                         ret = -EINVAL;
8438                         goto out;
8439                 }
8440
8441                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8442                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8443                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8444                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8445                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8446                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8447                 }
8448
8449                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8450                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8451                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8452                 }
8453
8454                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8455                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8456                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8457                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8458                 }
8459
8460                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8461                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8462                         dfprintk(MOUNT, "  stateid mode enabled\n");
8463                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8464                 }
8465
8466                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8467                         dfprintk(MOUNT, "  write mode enabled\n");
8468                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8469                 }
8470
8471                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8472                         dfprintk(MOUNT, "  commit mode enabled\n");
8473                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8474                 }
8475         }
8476 out:
8477         clp->cl_sp4_flags = flags;
8478         return ret;
8479 }
8480
8481 struct nfs41_exchange_id_data {
8482         struct nfs41_exchange_id_res res;
8483         struct nfs41_exchange_id_args args;
8484 };
8485
8486 static void nfs4_exchange_id_release(void *data)
8487 {
8488         struct nfs41_exchange_id_data *cdata =
8489                                         (struct nfs41_exchange_id_data *)data;
8490
8491         nfs_put_client(cdata->args.client);
8492         kfree(cdata->res.impl_id);
8493         kfree(cdata->res.server_scope);
8494         kfree(cdata->res.server_owner);
8495         kfree(cdata);
8496 }
8497
8498 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8499         .rpc_release = nfs4_exchange_id_release,
8500 };
8501
8502 /*
8503  * _nfs4_proc_exchange_id()
8504  *
8505  * Wrapper for EXCHANGE_ID operation.
8506  */
8507 static struct rpc_task *
8508 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8509                         u32 sp4_how, struct rpc_xprt *xprt)
8510 {
8511         struct rpc_message msg = {
8512                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8513                 .rpc_cred = cred,
8514         };
8515         struct rpc_task_setup task_setup_data = {
8516                 .rpc_client = clp->cl_rpcclient,
8517                 .callback_ops = &nfs4_exchange_id_call_ops,
8518                 .rpc_message = &msg,
8519                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8520         };
8521         struct nfs41_exchange_id_data *calldata;
8522         int status;
8523
8524         if (!refcount_inc_not_zero(&clp->cl_count))
8525                 return ERR_PTR(-EIO);
8526
8527         status = -ENOMEM;
8528         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8529         if (!calldata)
8530                 goto out;
8531
8532         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8533
8534         status = nfs4_init_uniform_client_string(clp);
8535         if (status)
8536                 goto out_calldata;
8537
8538         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8539                                                 GFP_NOFS);
8540         status = -ENOMEM;
8541         if (unlikely(calldata->res.server_owner == NULL))
8542                 goto out_calldata;
8543
8544         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8545                                         GFP_NOFS);
8546         if (unlikely(calldata->res.server_scope == NULL))
8547                 goto out_server_owner;
8548
8549         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8550         if (unlikely(calldata->res.impl_id == NULL))
8551                 goto out_server_scope;
8552
8553         switch (sp4_how) {
8554         case SP4_NONE:
8555                 calldata->args.state_protect.how = SP4_NONE;
8556                 break;
8557
8558         case SP4_MACH_CRED:
8559                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8560                 break;
8561
8562         default:
8563                 /* unsupported! */
8564                 WARN_ON_ONCE(1);
8565                 status = -EINVAL;
8566                 goto out_impl_id;
8567         }
8568         if (xprt) {
8569                 task_setup_data.rpc_xprt = xprt;
8570                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8571                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8572                                 sizeof(calldata->args.verifier.data));
8573         }
8574         calldata->args.client = clp;
8575         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8576         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8577 #ifdef CONFIG_NFS_V4_1_MIGRATION
8578         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8579 #endif
8580         msg.rpc_argp = &calldata->args;
8581         msg.rpc_resp = &calldata->res;
8582         task_setup_data.callback_data = calldata;
8583
8584         return rpc_run_task(&task_setup_data);
8585
8586 out_impl_id:
8587         kfree(calldata->res.impl_id);
8588 out_server_scope:
8589         kfree(calldata->res.server_scope);
8590 out_server_owner:
8591         kfree(calldata->res.server_owner);
8592 out_calldata:
8593         kfree(calldata);
8594 out:
8595         nfs_put_client(clp);
8596         return ERR_PTR(status);
8597 }
8598
8599 /*
8600  * _nfs4_proc_exchange_id()
8601  *
8602  * Wrapper for EXCHANGE_ID operation.
8603  */
8604 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8605                         u32 sp4_how)
8606 {
8607         struct rpc_task *task;
8608         struct nfs41_exchange_id_args *argp;
8609         struct nfs41_exchange_id_res *resp;
8610         unsigned long now = jiffies;
8611         int status;
8612
8613         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8614         if (IS_ERR(task))
8615                 return PTR_ERR(task);
8616
8617         argp = task->tk_msg.rpc_argp;
8618         resp = task->tk_msg.rpc_resp;
8619         status = task->tk_status;
8620         if (status  != 0)
8621                 goto out;
8622
8623         status = nfs4_check_cl_exchange_flags(resp->flags,
8624                         clp->cl_mvops->minor_version);
8625         if (status  != 0)
8626                 goto out;
8627
8628         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8629         if (status != 0)
8630                 goto out;
8631
8632         do_renew_lease(clp, now);
8633
8634         clp->cl_clientid = resp->clientid;
8635         clp->cl_exchange_flags = resp->flags;
8636         clp->cl_seqid = resp->seqid;
8637         /* Client ID is not confirmed */
8638         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8639                 clear_bit(NFS4_SESSION_ESTABLISHED,
8640                           &clp->cl_session->session_state);
8641
8642         if (clp->cl_serverscope != NULL &&
8643             !nfs41_same_server_scope(clp->cl_serverscope,
8644                                 resp->server_scope)) {
8645                 dprintk("%s: server_scope mismatch detected\n",
8646                         __func__);
8647                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8648         }
8649
8650         swap(clp->cl_serverowner, resp->server_owner);
8651         swap(clp->cl_serverscope, resp->server_scope);
8652         swap(clp->cl_implid, resp->impl_id);
8653
8654         /* Save the EXCHANGE_ID verifier session trunk tests */
8655         memcpy(clp->cl_confirm.data, argp->verifier.data,
8656                sizeof(clp->cl_confirm.data));
8657 out:
8658         trace_nfs4_exchange_id(clp, status);
8659         rpc_put_task(task);
8660         return status;
8661 }
8662
8663 /*
8664  * nfs4_proc_exchange_id()
8665  *
8666  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8667  *
8668  * Since the clientid has expired, all compounds using sessions
8669  * associated with the stale clientid will be returning
8670  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8671  * be in some phase of session reset.
8672  *
8673  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8674  */
8675 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8676 {
8677         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8678         int status;
8679
8680         /* try SP4_MACH_CRED if krb5i/p */
8681         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8682             authflavor == RPC_AUTH_GSS_KRB5P) {
8683                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8684                 if (!status)
8685                         return 0;
8686         }
8687
8688         /* try SP4_NONE */
8689         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8690 }
8691
8692 /**
8693  * nfs4_test_session_trunk
8694  *
8695  * This is an add_xprt_test() test function called from
8696  * rpc_clnt_setup_test_and_add_xprt.
8697  *
8698  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8699  * and is dereferrenced in nfs4_exchange_id_release
8700  *
8701  * Upon success, add the new transport to the rpc_clnt
8702  *
8703  * @clnt: struct rpc_clnt to get new transport
8704  * @xprt: the rpc_xprt to test
8705  * @data: call data for _nfs4_proc_exchange_id.
8706  */
8707 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8708                             void *data)
8709 {
8710         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8711         struct rpc_task *task;
8712         int status;
8713
8714         u32 sp4_how;
8715
8716         dprintk("--> %s try %s\n", __func__,
8717                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8718
8719         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8720
8721         /* Test connection for session trunking. Async exchange_id call */
8722         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8723         if (IS_ERR(task))
8724                 return;
8725
8726         status = task->tk_status;
8727         if (status == 0)
8728                 status = nfs4_detect_session_trunking(adata->clp,
8729                                 task->tk_msg.rpc_resp, xprt);
8730
8731         if (status == 0)
8732                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8733
8734         rpc_put_task(task);
8735 }
8736 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8737
8738 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8739                 const struct cred *cred)
8740 {
8741         struct rpc_message msg = {
8742                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8743                 .rpc_argp = clp,
8744                 .rpc_cred = cred,
8745         };
8746         int status;
8747
8748         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8749                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8750         trace_nfs4_destroy_clientid(clp, status);
8751         if (status)
8752                 dprintk("NFS: Got error %d from the server %s on "
8753                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8754         return status;
8755 }
8756
8757 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8758                 const struct cred *cred)
8759 {
8760         unsigned int loop;
8761         int ret;
8762
8763         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8764                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8765                 switch (ret) {
8766                 case -NFS4ERR_DELAY:
8767                 case -NFS4ERR_CLIENTID_BUSY:
8768                         ssleep(1);
8769                         break;
8770                 default:
8771                         return ret;
8772                 }
8773         }
8774         return 0;
8775 }
8776
8777 int nfs4_destroy_clientid(struct nfs_client *clp)
8778 {
8779         const struct cred *cred;
8780         int ret = 0;
8781
8782         if (clp->cl_mvops->minor_version < 1)
8783                 goto out;
8784         if (clp->cl_exchange_flags == 0)
8785                 goto out;
8786         if (clp->cl_preserve_clid)
8787                 goto out;
8788         cred = nfs4_get_clid_cred(clp);
8789         ret = nfs4_proc_destroy_clientid(clp, cred);
8790         put_cred(cred);
8791         switch (ret) {
8792         case 0:
8793         case -NFS4ERR_STALE_CLIENTID:
8794                 clp->cl_exchange_flags = 0;
8795         }
8796 out:
8797         return ret;
8798 }
8799
8800 #endif /* CONFIG_NFS_V4_1 */
8801
8802 struct nfs4_get_lease_time_data {
8803         struct nfs4_get_lease_time_args *args;
8804         struct nfs4_get_lease_time_res *res;
8805         struct nfs_client *clp;
8806 };
8807
8808 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8809                                         void *calldata)
8810 {
8811         struct nfs4_get_lease_time_data *data =
8812                         (struct nfs4_get_lease_time_data *)calldata;
8813
8814         dprintk("--> %s\n", __func__);
8815         /* just setup sequence, do not trigger session recovery
8816            since we're invoked within one */
8817         nfs4_setup_sequence(data->clp,
8818                         &data->args->la_seq_args,
8819                         &data->res->lr_seq_res,
8820                         task);
8821         dprintk("<-- %s\n", __func__);
8822 }
8823
8824 /*
8825  * Called from nfs4_state_manager thread for session setup, so don't recover
8826  * from sequence operation or clientid errors.
8827  */
8828 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8829 {
8830         struct nfs4_get_lease_time_data *data =
8831                         (struct nfs4_get_lease_time_data *)calldata;
8832
8833         dprintk("--> %s\n", __func__);
8834         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8835                 return;
8836         switch (task->tk_status) {
8837         case -NFS4ERR_DELAY:
8838         case -NFS4ERR_GRACE:
8839                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8840                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8841                 task->tk_status = 0;
8842                 fallthrough;
8843         case -NFS4ERR_RETRY_UNCACHED_REP:
8844                 rpc_restart_call_prepare(task);
8845                 return;
8846         }
8847         dprintk("<-- %s\n", __func__);
8848 }
8849
8850 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8851         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8852         .rpc_call_done = nfs4_get_lease_time_done,
8853 };
8854
8855 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8856 {
8857         struct nfs4_get_lease_time_args args;
8858         struct nfs4_get_lease_time_res res = {
8859                 .lr_fsinfo = fsinfo,
8860         };
8861         struct nfs4_get_lease_time_data data = {
8862                 .args = &args,
8863                 .res = &res,
8864                 .clp = clp,
8865         };
8866         struct rpc_message msg = {
8867                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8868                 .rpc_argp = &args,
8869                 .rpc_resp = &res,
8870         };
8871         struct rpc_task_setup task_setup = {
8872                 .rpc_client = clp->cl_rpcclient,
8873                 .rpc_message = &msg,
8874                 .callback_ops = &nfs4_get_lease_time_ops,
8875                 .callback_data = &data,
8876                 .flags = RPC_TASK_TIMEOUT,
8877         };
8878
8879         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8880         return nfs4_call_sync_custom(&task_setup);
8881 }
8882
8883 #ifdef CONFIG_NFS_V4_1
8884
8885 /*
8886  * Initialize the values to be used by the client in CREATE_SESSION
8887  * If nfs4_init_session set the fore channel request and response sizes,
8888  * use them.
8889  *
8890  * Set the back channel max_resp_sz_cached to zero to force the client to
8891  * always set csa_cachethis to FALSE because the current implementation
8892  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8893  */
8894 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8895                                     struct rpc_clnt *clnt)
8896 {
8897         unsigned int max_rqst_sz, max_resp_sz;
8898         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8899         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8900
8901         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8902         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8903
8904         /* Fore channel attributes */
8905         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8906         args->fc_attrs.max_resp_sz = max_resp_sz;
8907         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8908         args->fc_attrs.max_reqs = max_session_slots;
8909
8910         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8911                 "max_ops=%u max_reqs=%u\n",
8912                 __func__,
8913                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8914                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8915
8916         /* Back channel attributes */
8917         args->bc_attrs.max_rqst_sz = max_bc_payload;
8918         args->bc_attrs.max_resp_sz = max_bc_payload;
8919         args->bc_attrs.max_resp_sz_cached = 0;
8920         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8921         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8922         if (args->bc_attrs.max_reqs > max_bc_slots)
8923                 args->bc_attrs.max_reqs = max_bc_slots;
8924
8925         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8926                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8927                 __func__,
8928                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8929                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8930                 args->bc_attrs.max_reqs);
8931 }
8932
8933 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8934                 struct nfs41_create_session_res *res)
8935 {
8936         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8937         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8938
8939         if (rcvd->max_resp_sz > sent->max_resp_sz)
8940                 return -EINVAL;
8941         /*
8942          * Our requested max_ops is the minimum we need; we're not
8943          * prepared to break up compounds into smaller pieces than that.
8944          * So, no point even trying to continue if the server won't
8945          * cooperate:
8946          */
8947         if (rcvd->max_ops < sent->max_ops)
8948                 return -EINVAL;
8949         if (rcvd->max_reqs == 0)
8950                 return -EINVAL;
8951         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8952                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8953         return 0;
8954 }
8955
8956 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8957                 struct nfs41_create_session_res *res)
8958 {
8959         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8960         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8961
8962         if (!(res->flags & SESSION4_BACK_CHAN))
8963                 goto out;
8964         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8965                 return -EINVAL;
8966         if (rcvd->max_resp_sz < sent->max_resp_sz)
8967                 return -EINVAL;
8968         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8969                 return -EINVAL;
8970         if (rcvd->max_ops > sent->max_ops)
8971                 return -EINVAL;
8972         if (rcvd->max_reqs > sent->max_reqs)
8973                 return -EINVAL;
8974 out:
8975         return 0;
8976 }
8977
8978 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8979                                      struct nfs41_create_session_res *res)
8980 {
8981         int ret;
8982
8983         ret = nfs4_verify_fore_channel_attrs(args, res);
8984         if (ret)
8985                 return ret;
8986         return nfs4_verify_back_channel_attrs(args, res);
8987 }
8988
8989 static void nfs4_update_session(struct nfs4_session *session,
8990                 struct nfs41_create_session_res *res)
8991 {
8992         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8993         /* Mark client id and session as being confirmed */
8994         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8995         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8996         session->flags = res->flags;
8997         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8998         if (res->flags & SESSION4_BACK_CHAN)
8999                 memcpy(&session->bc_attrs, &res->bc_attrs,
9000                                 sizeof(session->bc_attrs));
9001 }
9002
9003 static int _nfs4_proc_create_session(struct nfs_client *clp,
9004                 const struct cred *cred)
9005 {
9006         struct nfs4_session *session = clp->cl_session;
9007         struct nfs41_create_session_args args = {
9008                 .client = clp,
9009                 .clientid = clp->cl_clientid,
9010                 .seqid = clp->cl_seqid,
9011                 .cb_program = NFS4_CALLBACK,
9012         };
9013         struct nfs41_create_session_res res;
9014
9015         struct rpc_message msg = {
9016                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9017                 .rpc_argp = &args,
9018                 .rpc_resp = &res,
9019                 .rpc_cred = cred,
9020         };
9021         int status;
9022
9023         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9024         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9025
9026         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9027                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9028         trace_nfs4_create_session(clp, status);
9029
9030         switch (status) {
9031         case -NFS4ERR_STALE_CLIENTID:
9032         case -NFS4ERR_DELAY:
9033         case -ETIMEDOUT:
9034         case -EACCES:
9035         case -EAGAIN:
9036                 goto out;
9037         }
9038
9039         clp->cl_seqid++;
9040         if (!status) {
9041                 /* Verify the session's negotiated channel_attrs values */
9042                 status = nfs4_verify_channel_attrs(&args, &res);
9043                 /* Increment the clientid slot sequence id */
9044                 if (status)
9045                         goto out;
9046                 nfs4_update_session(session, &res);
9047         }
9048 out:
9049         return status;
9050 }
9051
9052 /*
9053  * Issues a CREATE_SESSION operation to the server.
9054  * It is the responsibility of the caller to verify the session is
9055  * expired before calling this routine.
9056  */
9057 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9058 {
9059         int status;
9060         unsigned *ptr;
9061         struct nfs4_session *session = clp->cl_session;
9062
9063         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9064
9065         status = _nfs4_proc_create_session(clp, cred);
9066         if (status)
9067                 goto out;
9068
9069         /* Init or reset the session slot tables */
9070         status = nfs4_setup_session_slot_tables(session);
9071         dprintk("slot table setup returned %d\n", status);
9072         if (status)
9073                 goto out;
9074
9075         ptr = (unsigned *)&session->sess_id.data[0];
9076         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9077                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9078 out:
9079         dprintk("<-- %s\n", __func__);
9080         return status;
9081 }
9082
9083 /*
9084  * Issue the over-the-wire RPC DESTROY_SESSION.
9085  * The caller must serialize access to this routine.
9086  */
9087 int nfs4_proc_destroy_session(struct nfs4_session *session,
9088                 const struct cred *cred)
9089 {
9090         struct rpc_message msg = {
9091                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9092                 .rpc_argp = session,
9093                 .rpc_cred = cred,
9094         };
9095         int status = 0;
9096
9097         dprintk("--> nfs4_proc_destroy_session\n");
9098
9099         /* session is still being setup */
9100         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9101                 return 0;
9102
9103         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9104                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9105         trace_nfs4_destroy_session(session->clp, status);
9106
9107         if (status)
9108                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9109                         "Session has been destroyed regardless...\n", status);
9110
9111         dprintk("<-- nfs4_proc_destroy_session\n");
9112         return status;
9113 }
9114
9115 /*
9116  * Renew the cl_session lease.
9117  */
9118 struct nfs4_sequence_data {
9119         struct nfs_client *clp;
9120         struct nfs4_sequence_args args;
9121         struct nfs4_sequence_res res;
9122 };
9123
9124 static void nfs41_sequence_release(void *data)
9125 {
9126         struct nfs4_sequence_data *calldata = data;
9127         struct nfs_client *clp = calldata->clp;
9128
9129         if (refcount_read(&clp->cl_count) > 1)
9130                 nfs4_schedule_state_renewal(clp);
9131         nfs_put_client(clp);
9132         kfree(calldata);
9133 }
9134
9135 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9136 {
9137         switch(task->tk_status) {
9138         case -NFS4ERR_DELAY:
9139                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9140                 return -EAGAIN;
9141         default:
9142                 nfs4_schedule_lease_recovery(clp);
9143         }
9144         return 0;
9145 }
9146
9147 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9148 {
9149         struct nfs4_sequence_data *calldata = data;
9150         struct nfs_client *clp = calldata->clp;
9151
9152         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9153                 return;
9154
9155         trace_nfs4_sequence(clp, task->tk_status);
9156         if (task->tk_status < 0) {
9157                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9158                 if (refcount_read(&clp->cl_count) == 1)
9159                         goto out;
9160
9161                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9162                         rpc_restart_call_prepare(task);
9163                         return;
9164                 }
9165         }
9166         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9167 out:
9168         dprintk("<-- %s\n", __func__);
9169 }
9170
9171 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9172 {
9173         struct nfs4_sequence_data *calldata = data;
9174         struct nfs_client *clp = calldata->clp;
9175         struct nfs4_sequence_args *args;
9176         struct nfs4_sequence_res *res;
9177
9178         args = task->tk_msg.rpc_argp;
9179         res = task->tk_msg.rpc_resp;
9180
9181         nfs4_setup_sequence(clp, args, res, task);
9182 }
9183
9184 static const struct rpc_call_ops nfs41_sequence_ops = {
9185         .rpc_call_done = nfs41_sequence_call_done,
9186         .rpc_call_prepare = nfs41_sequence_prepare,
9187         .rpc_release = nfs41_sequence_release,
9188 };
9189
9190 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9191                 const struct cred *cred,
9192                 struct nfs4_slot *slot,
9193                 bool is_privileged)
9194 {
9195         struct nfs4_sequence_data *calldata;
9196         struct rpc_message msg = {
9197                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9198                 .rpc_cred = cred,
9199         };
9200         struct rpc_task_setup task_setup_data = {
9201                 .rpc_client = clp->cl_rpcclient,
9202                 .rpc_message = &msg,
9203                 .callback_ops = &nfs41_sequence_ops,
9204                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
9205         };
9206         struct rpc_task *ret;
9207
9208         ret = ERR_PTR(-EIO);
9209         if (!refcount_inc_not_zero(&clp->cl_count))
9210                 goto out_err;
9211
9212         ret = ERR_PTR(-ENOMEM);
9213         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9214         if (calldata == NULL)
9215                 goto out_put_clp;
9216         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9217         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9218         msg.rpc_argp = &calldata->args;
9219         msg.rpc_resp = &calldata->res;
9220         calldata->clp = clp;
9221         task_setup_data.callback_data = calldata;
9222
9223         ret = rpc_run_task(&task_setup_data);
9224         if (IS_ERR(ret))
9225                 goto out_err;
9226         return ret;
9227 out_put_clp:
9228         nfs_put_client(clp);
9229 out_err:
9230         nfs41_release_slot(slot);
9231         return ret;
9232 }
9233
9234 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9235 {
9236         struct rpc_task *task;
9237         int ret = 0;
9238
9239         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9240                 return -EAGAIN;
9241         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9242         if (IS_ERR(task))
9243                 ret = PTR_ERR(task);
9244         else
9245                 rpc_put_task_async(task);
9246         dprintk("<-- %s status=%d\n", __func__, ret);
9247         return ret;
9248 }
9249
9250 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9251 {
9252         struct rpc_task *task;
9253         int ret;
9254
9255         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9256         if (IS_ERR(task)) {
9257                 ret = PTR_ERR(task);
9258                 goto out;
9259         }
9260         ret = rpc_wait_for_completion_task(task);
9261         if (!ret)
9262                 ret = task->tk_status;
9263         rpc_put_task(task);
9264 out:
9265         dprintk("<-- %s status=%d\n", __func__, ret);
9266         return ret;
9267 }
9268
9269 struct nfs4_reclaim_complete_data {
9270         struct nfs_client *clp;
9271         struct nfs41_reclaim_complete_args arg;
9272         struct nfs41_reclaim_complete_res res;
9273 };
9274
9275 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9276 {
9277         struct nfs4_reclaim_complete_data *calldata = data;
9278
9279         nfs4_setup_sequence(calldata->clp,
9280                         &calldata->arg.seq_args,
9281                         &calldata->res.seq_res,
9282                         task);
9283 }
9284
9285 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9286 {
9287         switch(task->tk_status) {
9288         case 0:
9289                 wake_up_all(&clp->cl_lock_waitq);
9290                 fallthrough;
9291         case -NFS4ERR_COMPLETE_ALREADY:
9292         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9293                 break;
9294         case -NFS4ERR_DELAY:
9295                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9296                 fallthrough;
9297         case -NFS4ERR_RETRY_UNCACHED_REP:
9298         case -EACCES:
9299                 dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9300                         __func__, task->tk_status, clp->cl_hostname);
9301                 return -EAGAIN;
9302         case -NFS4ERR_BADSESSION:
9303         case -NFS4ERR_DEADSESSION:
9304         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9305                 break;
9306         default:
9307                 nfs4_schedule_lease_recovery(clp);
9308         }
9309         return 0;
9310 }
9311
9312 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9313 {
9314         struct nfs4_reclaim_complete_data *calldata = data;
9315         struct nfs_client *clp = calldata->clp;
9316         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9317
9318         dprintk("--> %s\n", __func__);
9319         if (!nfs41_sequence_done(task, res))
9320                 return;
9321
9322         trace_nfs4_reclaim_complete(clp, task->tk_status);
9323         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9324                 rpc_restart_call_prepare(task);
9325                 return;
9326         }
9327         dprintk("<-- %s\n", __func__);
9328 }
9329
9330 static void nfs4_free_reclaim_complete_data(void *data)
9331 {
9332         struct nfs4_reclaim_complete_data *calldata = data;
9333
9334         kfree(calldata);
9335 }
9336
9337 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9338         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9339         .rpc_call_done = nfs4_reclaim_complete_done,
9340         .rpc_release = nfs4_free_reclaim_complete_data,
9341 };
9342
9343 /*
9344  * Issue a global reclaim complete.
9345  */
9346 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9347                 const struct cred *cred)
9348 {
9349         struct nfs4_reclaim_complete_data *calldata;
9350         struct rpc_message msg = {
9351                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9352                 .rpc_cred = cred,
9353         };
9354         struct rpc_task_setup task_setup_data = {
9355                 .rpc_client = clp->cl_rpcclient,
9356                 .rpc_message = &msg,
9357                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9358                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9359         };
9360         int status = -ENOMEM;
9361
9362         dprintk("--> %s\n", __func__);
9363         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9364         if (calldata == NULL)
9365                 goto out;
9366         calldata->clp = clp;
9367         calldata->arg.one_fs = 0;
9368
9369         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9370         msg.rpc_argp = &calldata->arg;
9371         msg.rpc_resp = &calldata->res;
9372         task_setup_data.callback_data = calldata;
9373         status = nfs4_call_sync_custom(&task_setup_data);
9374 out:
9375         dprintk("<-- %s status=%d\n", __func__, status);
9376         return status;
9377 }
9378
9379 static void
9380 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9381 {
9382         struct nfs4_layoutget *lgp = calldata;
9383         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9384
9385         dprintk("--> %s\n", __func__);
9386         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9387                                 &lgp->res.seq_res, task);
9388         dprintk("<-- %s\n", __func__);
9389 }
9390
9391 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9392 {
9393         struct nfs4_layoutget *lgp = calldata;
9394
9395         dprintk("--> %s\n", __func__);
9396         nfs41_sequence_process(task, &lgp->res.seq_res);
9397         dprintk("<-- %s\n", __func__);
9398 }
9399
9400 static int
9401 nfs4_layoutget_handle_exception(struct rpc_task *task,
9402                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9403 {
9404         struct inode *inode = lgp->args.inode;
9405         struct nfs_server *server = NFS_SERVER(inode);
9406         struct pnfs_layout_hdr *lo;
9407         int nfs4err = task->tk_status;
9408         int err, status = 0;
9409         LIST_HEAD(head);
9410
9411         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9412
9413         nfs4_sequence_free_slot(&lgp->res.seq_res);
9414
9415         switch (nfs4err) {
9416         case 0:
9417                 goto out;
9418
9419         /*
9420          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9421          * on the file. set tk_status to -ENODATA to tell upper layer to
9422          * retry go inband.
9423          */
9424         case -NFS4ERR_LAYOUTUNAVAILABLE:
9425                 status = -ENODATA;
9426                 goto out;
9427         /*
9428          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9429          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9430          */
9431         case -NFS4ERR_BADLAYOUT:
9432                 status = -EOVERFLOW;
9433                 goto out;
9434         /*
9435          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9436          * (or clients) writing to the same RAID stripe except when
9437          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9438          *
9439          * Treat it like we would RECALLCONFLICT -- we retry for a little
9440          * while, and then eventually give up.
9441          */
9442         case -NFS4ERR_LAYOUTTRYLATER:
9443                 if (lgp->args.minlength == 0) {
9444                         status = -EOVERFLOW;
9445                         goto out;
9446                 }
9447                 status = -EBUSY;
9448                 break;
9449         case -NFS4ERR_RECALLCONFLICT:
9450         case -NFS4ERR_RETURNCONFLICT:
9451                 status = -ERECALLCONFLICT;
9452                 break;
9453         case -NFS4ERR_DELEG_REVOKED:
9454         case -NFS4ERR_ADMIN_REVOKED:
9455         case -NFS4ERR_EXPIRED:
9456         case -NFS4ERR_BAD_STATEID:
9457                 exception->timeout = 0;
9458                 spin_lock(&inode->i_lock);
9459                 lo = NFS_I(inode)->layout;
9460                 /* If the open stateid was bad, then recover it. */
9461                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9462                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9463                         spin_unlock(&inode->i_lock);
9464                         exception->state = lgp->args.ctx->state;
9465                         exception->stateid = &lgp->args.stateid;
9466                         break;
9467                 }
9468
9469                 /*
9470                  * Mark the bad layout state as invalid, then retry
9471                  */
9472                 pnfs_mark_layout_stateid_invalid(lo, &head);
9473                 spin_unlock(&inode->i_lock);
9474                 nfs_commit_inode(inode, 0);
9475                 pnfs_free_lseg_list(&head);
9476                 status = -EAGAIN;
9477                 goto out;
9478         }
9479
9480         err = nfs4_handle_exception(server, nfs4err, exception);
9481         if (!status) {
9482                 if (exception->retry)
9483                         status = -EAGAIN;
9484                 else
9485                         status = err;
9486         }
9487 out:
9488         dprintk("<-- %s\n", __func__);
9489         return status;
9490 }
9491
9492 size_t max_response_pages(struct nfs_server *server)
9493 {
9494         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9495         return nfs_page_array_len(0, max_resp_sz);
9496 }
9497
9498 static void nfs4_layoutget_release(void *calldata)
9499 {
9500         struct nfs4_layoutget *lgp = calldata;
9501
9502         dprintk("--> %s\n", __func__);
9503         nfs4_sequence_free_slot(&lgp->res.seq_res);
9504         pnfs_layoutget_free(lgp);
9505         dprintk("<-- %s\n", __func__);
9506 }
9507
9508 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9509         .rpc_call_prepare = nfs4_layoutget_prepare,
9510         .rpc_call_done = nfs4_layoutget_done,
9511         .rpc_release = nfs4_layoutget_release,
9512 };
9513
9514 struct pnfs_layout_segment *
9515 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9516 {
9517         struct inode *inode = lgp->args.inode;
9518         struct nfs_server *server = NFS_SERVER(inode);
9519         struct rpc_task *task;
9520         struct rpc_message msg = {
9521                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9522                 .rpc_argp = &lgp->args,
9523                 .rpc_resp = &lgp->res,
9524                 .rpc_cred = lgp->cred,
9525         };
9526         struct rpc_task_setup task_setup_data = {
9527                 .rpc_client = server->client,
9528                 .rpc_message = &msg,
9529                 .callback_ops = &nfs4_layoutget_call_ops,
9530                 .callback_data = lgp,
9531                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9532         };
9533         struct pnfs_layout_segment *lseg = NULL;
9534         struct nfs4_exception exception = {
9535                 .inode = inode,
9536                 .timeout = *timeout,
9537         };
9538         int status = 0;
9539
9540         dprintk("--> %s\n", __func__);
9541
9542         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9543         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9544
9545         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9546
9547         task = rpc_run_task(&task_setup_data);
9548
9549         status = rpc_wait_for_completion_task(task);
9550         if (status != 0)
9551                 goto out;
9552
9553         if (task->tk_status < 0) {
9554                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9555                 *timeout = exception.timeout;
9556         } else if (lgp->res.layoutp->len == 0) {
9557                 status = -EAGAIN;
9558                 *timeout = nfs4_update_delay(&exception.timeout);
9559         } else
9560                 lseg = pnfs_layout_process(lgp);
9561 out:
9562         trace_nfs4_layoutget(lgp->args.ctx,
9563                         &lgp->args.range,
9564                         &lgp->res.range,
9565                         &lgp->res.stateid,
9566                         status);
9567
9568         rpc_put_task(task);
9569         dprintk("<-- %s status=%d\n", __func__, status);
9570         if (status)
9571                 return ERR_PTR(status);
9572         return lseg;
9573 }
9574
9575 static void
9576 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9577 {
9578         struct nfs4_layoutreturn *lrp = calldata;
9579
9580         dprintk("--> %s\n", __func__);
9581         nfs4_setup_sequence(lrp->clp,
9582                         &lrp->args.seq_args,
9583                         &lrp->res.seq_res,
9584                         task);
9585         if (!pnfs_layout_is_valid(lrp->args.layout))
9586                 rpc_exit(task, 0);
9587 }
9588
9589 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9590 {
9591         struct nfs4_layoutreturn *lrp = calldata;
9592         struct nfs_server *server;
9593
9594         dprintk("--> %s\n", __func__);
9595
9596         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9597                 return;
9598
9599         /*
9600          * Was there an RPC level error? Assume the call succeeded,
9601          * and that we need to release the layout
9602          */
9603         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9604                 lrp->res.lrs_present = 0;
9605                 return;
9606         }
9607
9608         server = NFS_SERVER(lrp->args.inode);
9609         switch (task->tk_status) {
9610         case -NFS4ERR_OLD_STATEID:
9611                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9612                                         &lrp->args.range,
9613                                         lrp->args.inode))
9614                         goto out_restart;
9615                 fallthrough;
9616         default:
9617                 task->tk_status = 0;
9618                 fallthrough;
9619         case 0:
9620                 break;
9621         case -NFS4ERR_DELAY:
9622                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9623                         break;
9624                 goto out_restart;
9625         }
9626         dprintk("<-- %s\n", __func__);
9627         return;
9628 out_restart:
9629         task->tk_status = 0;
9630         nfs4_sequence_free_slot(&lrp->res.seq_res);
9631         rpc_restart_call_prepare(task);
9632 }
9633
9634 static void nfs4_layoutreturn_release(void *calldata)
9635 {
9636         struct nfs4_layoutreturn *lrp = calldata;
9637         struct pnfs_layout_hdr *lo = lrp->args.layout;
9638
9639         dprintk("--> %s\n", __func__);
9640         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9641                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9642         nfs4_sequence_free_slot(&lrp->res.seq_res);
9643         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9644                 lrp->ld_private.ops->free(&lrp->ld_private);
9645         pnfs_put_layout_hdr(lrp->args.layout);
9646         nfs_iput_and_deactive(lrp->inode);
9647         put_cred(lrp->cred);
9648         kfree(calldata);
9649         dprintk("<-- %s\n", __func__);
9650 }
9651
9652 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9653         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9654         .rpc_call_done = nfs4_layoutreturn_done,
9655         .rpc_release = nfs4_layoutreturn_release,
9656 };
9657
9658 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9659 {
9660         struct rpc_task *task;
9661         struct rpc_message msg = {
9662                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9663                 .rpc_argp = &lrp->args,
9664                 .rpc_resp = &lrp->res,
9665                 .rpc_cred = lrp->cred,
9666         };
9667         struct rpc_task_setup task_setup_data = {
9668                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9669                 .rpc_message = &msg,
9670                 .callback_ops = &nfs4_layoutreturn_call_ops,
9671                 .callback_data = lrp,
9672         };
9673         int status = 0;
9674
9675         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9676                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9677                         &task_setup_data.rpc_client, &msg);
9678
9679         dprintk("--> %s\n", __func__);
9680         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9681         if (!sync) {
9682                 if (!lrp->inode) {
9683                         nfs4_layoutreturn_release(lrp);
9684                         return -EAGAIN;
9685                 }
9686                 task_setup_data.flags |= RPC_TASK_ASYNC;
9687         }
9688         if (!lrp->inode)
9689                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9690                                    1);
9691         else
9692                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9693                                    0);
9694         task = rpc_run_task(&task_setup_data);
9695         if (IS_ERR(task))
9696                 return PTR_ERR(task);
9697         if (sync)
9698                 status = task->tk_status;
9699         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9700         dprintk("<-- %s status=%d\n", __func__, status);
9701         rpc_put_task(task);
9702         return status;
9703 }
9704
9705 static int
9706 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9707                 struct pnfs_device *pdev,
9708                 const struct cred *cred)
9709 {
9710         struct nfs4_getdeviceinfo_args args = {
9711                 .pdev = pdev,
9712                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9713                         NOTIFY_DEVICEID4_DELETE,
9714         };
9715         struct nfs4_getdeviceinfo_res res = {
9716                 .pdev = pdev,
9717         };
9718         struct rpc_message msg = {
9719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9720                 .rpc_argp = &args,
9721                 .rpc_resp = &res,
9722                 .rpc_cred = cred,
9723         };
9724         int status;
9725
9726         dprintk("--> %s\n", __func__);
9727         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9728         if (res.notification & ~args.notify_types)
9729                 dprintk("%s: unsupported notification\n", __func__);
9730         if (res.notification != args.notify_types)
9731                 pdev->nocache = 1;
9732
9733         dprintk("<-- %s status=%d\n", __func__, status);
9734
9735         return status;
9736 }
9737
9738 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9739                 struct pnfs_device *pdev,
9740                 const struct cred *cred)
9741 {
9742         struct nfs4_exception exception = { };
9743         int err;
9744
9745         do {
9746                 err = nfs4_handle_exception(server,
9747                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9748                                         &exception);
9749         } while (exception.retry);
9750         return err;
9751 }
9752 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9753
9754 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9755 {
9756         struct nfs4_layoutcommit_data *data = calldata;
9757         struct nfs_server *server = NFS_SERVER(data->args.inode);
9758
9759         nfs4_setup_sequence(server->nfs_client,
9760                         &data->args.seq_args,
9761                         &data->res.seq_res,
9762                         task);
9763 }
9764
9765 static void
9766 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9767 {
9768         struct nfs4_layoutcommit_data *data = calldata;
9769         struct nfs_server *server = NFS_SERVER(data->args.inode);
9770
9771         if (!nfs41_sequence_done(task, &data->res.seq_res))
9772                 return;
9773
9774         switch (task->tk_status) { /* Just ignore these failures */
9775         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9776         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9777         case -NFS4ERR_BADLAYOUT:     /* no layout */
9778         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9779                 task->tk_status = 0;
9780         case 0:
9781                 break;
9782         default:
9783                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9784                         rpc_restart_call_prepare(task);
9785                         return;
9786                 }
9787         }
9788 }
9789
9790 static void nfs4_layoutcommit_release(void *calldata)
9791 {
9792         struct nfs4_layoutcommit_data *data = calldata;
9793
9794         pnfs_cleanup_layoutcommit(data);
9795         nfs_post_op_update_inode_force_wcc(data->args.inode,
9796                                            data->res.fattr);
9797         put_cred(data->cred);
9798         nfs_iput_and_deactive(data->inode);
9799         kfree(data);
9800 }
9801
9802 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9803         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9804         .rpc_call_done = nfs4_layoutcommit_done,
9805         .rpc_release = nfs4_layoutcommit_release,
9806 };
9807
9808 int
9809 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9810 {
9811         struct rpc_message msg = {
9812                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9813                 .rpc_argp = &data->args,
9814                 .rpc_resp = &data->res,
9815                 .rpc_cred = data->cred,
9816         };
9817         struct rpc_task_setup task_setup_data = {
9818                 .task = &data->task,
9819                 .rpc_client = NFS_CLIENT(data->args.inode),
9820                 .rpc_message = &msg,
9821                 .callback_ops = &nfs4_layoutcommit_ops,
9822                 .callback_data = data,
9823         };
9824         struct rpc_task *task;
9825         int status = 0;
9826
9827         dprintk("NFS: initiating layoutcommit call. sync %d "
9828                 "lbw: %llu inode %lu\n", sync,
9829                 data->args.lastbytewritten,
9830                 data->args.inode->i_ino);
9831
9832         if (!sync) {
9833                 data->inode = nfs_igrab_and_active(data->args.inode);
9834                 if (data->inode == NULL) {
9835                         nfs4_layoutcommit_release(data);
9836                         return -EAGAIN;
9837                 }
9838                 task_setup_data.flags = RPC_TASK_ASYNC;
9839         }
9840         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9841         task = rpc_run_task(&task_setup_data);
9842         if (IS_ERR(task))
9843                 return PTR_ERR(task);
9844         if (sync)
9845                 status = task->tk_status;
9846         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9847         dprintk("%s: status %d\n", __func__, status);
9848         rpc_put_task(task);
9849         return status;
9850 }
9851
9852 /*
9853  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9854  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9855  */
9856 static int
9857 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9858                     struct nfs_fsinfo *info,
9859                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9860 {
9861         struct nfs41_secinfo_no_name_args args = {
9862                 .style = SECINFO_STYLE_CURRENT_FH,
9863         };
9864         struct nfs4_secinfo_res res = {
9865                 .flavors = flavors,
9866         };
9867         struct rpc_message msg = {
9868                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9869                 .rpc_argp = &args,
9870                 .rpc_resp = &res,
9871         };
9872         struct nfs4_call_sync_data data = {
9873                 .seq_server = server,
9874                 .seq_args = &args.seq_args,
9875                 .seq_res = &res.seq_res,
9876         };
9877         struct rpc_task_setup task_setup = {
9878                 .rpc_client = server->client,
9879                 .rpc_message = &msg,
9880                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9881                 .callback_data = &data,
9882                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9883         };
9884         const struct cred *cred = NULL;
9885         int status;
9886
9887         if (use_integrity) {
9888                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9889
9890                 cred = nfs4_get_clid_cred(server->nfs_client);
9891                 msg.rpc_cred = cred;
9892         }
9893
9894         dprintk("--> %s\n", __func__);
9895         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9896         status = nfs4_call_sync_custom(&task_setup);
9897         dprintk("<-- %s status=%d\n", __func__, status);
9898
9899         put_cred(cred);
9900
9901         return status;
9902 }
9903
9904 static int
9905 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9906                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9907 {
9908         struct nfs4_exception exception = {
9909                 .interruptible = true,
9910         };
9911         int err;
9912         do {
9913                 /* first try using integrity protection */
9914                 err = -NFS4ERR_WRONGSEC;
9915
9916                 /* try to use integrity protection with machine cred */
9917                 if (_nfs4_is_integrity_protected(server->nfs_client))
9918                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9919                                                           flavors, true);
9920
9921                 /*
9922                  * if unable to use integrity protection, or SECINFO with
9923                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9924                  * disallowed by spec, but exists in deployed servers) use
9925                  * the current filesystem's rpc_client and the user cred.
9926                  */
9927                 if (err == -NFS4ERR_WRONGSEC)
9928                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9929                                                           flavors, false);
9930
9931                 switch (err) {
9932                 case 0:
9933                 case -NFS4ERR_WRONGSEC:
9934                 case -ENOTSUPP:
9935                         goto out;
9936                 default:
9937                         err = nfs4_handle_exception(server, err, &exception);
9938                 }
9939         } while (exception.retry);
9940 out:
9941         return err;
9942 }
9943
9944 static int
9945 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9946                     struct nfs_fsinfo *info)
9947 {
9948         int err;
9949         struct page *page;
9950         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9951         struct nfs4_secinfo_flavors *flavors;
9952         struct nfs4_secinfo4 *secinfo;
9953         int i;
9954
9955         page = alloc_page(GFP_KERNEL);
9956         if (!page) {
9957                 err = -ENOMEM;
9958                 goto out;
9959         }
9960
9961         flavors = page_address(page);
9962         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9963
9964         /*
9965          * Fall back on "guess and check" method if
9966          * the server doesn't support SECINFO_NO_NAME
9967          */
9968         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9969                 err = nfs4_find_root_sec(server, fhandle, info);
9970                 goto out_freepage;
9971         }
9972         if (err)
9973                 goto out_freepage;
9974
9975         for (i = 0; i < flavors->num_flavors; i++) {
9976                 secinfo = &flavors->flavors[i];
9977
9978                 switch (secinfo->flavor) {
9979                 case RPC_AUTH_NULL:
9980                 case RPC_AUTH_UNIX:
9981                 case RPC_AUTH_GSS:
9982                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9983                                         &secinfo->flavor_info);
9984                         break;
9985                 default:
9986                         flavor = RPC_AUTH_MAXFLAVOR;
9987                         break;
9988                 }
9989
9990                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9991                         flavor = RPC_AUTH_MAXFLAVOR;
9992
9993                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9994                         err = nfs4_lookup_root_sec(server, fhandle,
9995                                                    info, flavor);
9996                         if (!err)
9997                                 break;
9998                 }
9999         }
10000
10001         if (flavor == RPC_AUTH_MAXFLAVOR)
10002                 err = -EPERM;
10003
10004 out_freepage:
10005         put_page(page);
10006         if (err == -EACCES)
10007                 return -EPERM;
10008 out:
10009         return err;
10010 }
10011
10012 static int _nfs41_test_stateid(struct nfs_server *server,
10013                 nfs4_stateid *stateid,
10014                 const struct cred *cred)
10015 {
10016         int status;
10017         struct nfs41_test_stateid_args args = {
10018                 .stateid = stateid,
10019         };
10020         struct nfs41_test_stateid_res res;
10021         struct rpc_message msg = {
10022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10023                 .rpc_argp = &args,
10024                 .rpc_resp = &res,
10025                 .rpc_cred = cred,
10026         };
10027         struct rpc_clnt *rpc_client = server->client;
10028
10029         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10030                 &rpc_client, &msg);
10031
10032         dprintk("NFS call  test_stateid %p\n", stateid);
10033         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10034         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10035                         &args.seq_args, &res.seq_res);
10036         if (status != NFS_OK) {
10037                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10038                 return status;
10039         }
10040         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10041         return -res.status;
10042 }
10043
10044 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10045                 int err, struct nfs4_exception *exception)
10046 {
10047         exception->retry = 0;
10048         switch(err) {
10049         case -NFS4ERR_DELAY:
10050         case -NFS4ERR_RETRY_UNCACHED_REP:
10051                 nfs4_handle_exception(server, err, exception);
10052                 break;
10053         case -NFS4ERR_BADSESSION:
10054         case -NFS4ERR_BADSLOT:
10055         case -NFS4ERR_BAD_HIGH_SLOT:
10056         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10057         case -NFS4ERR_DEADSESSION:
10058                 nfs4_do_handle_exception(server, err, exception);
10059         }
10060 }
10061
10062 /**
10063  * nfs41_test_stateid - perform a TEST_STATEID operation
10064  *
10065  * @server: server / transport on which to perform the operation
10066  * @stateid: state ID to test
10067  * @cred: credential
10068  *
10069  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10070  * Otherwise a negative NFS4ERR value is returned if the operation
10071  * failed or the state ID is not currently valid.
10072  */
10073 static int nfs41_test_stateid(struct nfs_server *server,
10074                 nfs4_stateid *stateid,
10075                 const struct cred *cred)
10076 {
10077         struct nfs4_exception exception = {
10078                 .interruptible = true,
10079         };
10080         int err;
10081         do {
10082                 err = _nfs41_test_stateid(server, stateid, cred);
10083                 nfs4_handle_delay_or_session_error(server, err, &exception);
10084         } while (exception.retry);
10085         return err;
10086 }
10087
10088 struct nfs_free_stateid_data {
10089         struct nfs_server *server;
10090         struct nfs41_free_stateid_args args;
10091         struct nfs41_free_stateid_res res;
10092 };
10093
10094 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10095 {
10096         struct nfs_free_stateid_data *data = calldata;
10097         nfs4_setup_sequence(data->server->nfs_client,
10098                         &data->args.seq_args,
10099                         &data->res.seq_res,
10100                         task);
10101 }
10102
10103 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10104 {
10105         struct nfs_free_stateid_data *data = calldata;
10106
10107         nfs41_sequence_done(task, &data->res.seq_res);
10108
10109         switch (task->tk_status) {
10110         case -NFS4ERR_DELAY:
10111                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10112                         rpc_restart_call_prepare(task);
10113         }
10114 }
10115
10116 static void nfs41_free_stateid_release(void *calldata)
10117 {
10118         kfree(calldata);
10119 }
10120
10121 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10122         .rpc_call_prepare = nfs41_free_stateid_prepare,
10123         .rpc_call_done = nfs41_free_stateid_done,
10124         .rpc_release = nfs41_free_stateid_release,
10125 };
10126
10127 /**
10128  * nfs41_free_stateid - perform a FREE_STATEID operation
10129  *
10130  * @server: server / transport on which to perform the operation
10131  * @stateid: state ID to release
10132  * @cred: credential
10133  * @privileged: set to true if this call needs to be privileged
10134  *
10135  * Note: this function is always asynchronous.
10136  */
10137 static int nfs41_free_stateid(struct nfs_server *server,
10138                 const nfs4_stateid *stateid,
10139                 const struct cred *cred,
10140                 bool privileged)
10141 {
10142         struct rpc_message msg = {
10143                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10144                 .rpc_cred = cred,
10145         };
10146         struct rpc_task_setup task_setup = {
10147                 .rpc_client = server->client,
10148                 .rpc_message = &msg,
10149                 .callback_ops = &nfs41_free_stateid_ops,
10150                 .flags = RPC_TASK_ASYNC,
10151         };
10152         struct nfs_free_stateid_data *data;
10153         struct rpc_task *task;
10154
10155         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10156                 &task_setup.rpc_client, &msg);
10157
10158         dprintk("NFS call  free_stateid %p\n", stateid);
10159         data = kmalloc(sizeof(*data), GFP_NOFS);
10160         if (!data)
10161                 return -ENOMEM;
10162         data->server = server;
10163         nfs4_stateid_copy(&data->args.stateid, stateid);
10164
10165         task_setup.callback_data = data;
10166
10167         msg.rpc_argp = &data->args;
10168         msg.rpc_resp = &data->res;
10169         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10170         task = rpc_run_task(&task_setup);
10171         if (IS_ERR(task))
10172                 return PTR_ERR(task);
10173         rpc_put_task(task);
10174         return 0;
10175 }
10176
10177 static void
10178 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10179 {
10180         const struct cred *cred = lsp->ls_state->owner->so_cred;
10181
10182         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10183         nfs4_free_lock_state(server, lsp);
10184 }
10185
10186 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10187                 const nfs4_stateid *s2)
10188 {
10189         if (s1->type != s2->type)
10190                 return false;
10191
10192         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10193                 return false;
10194
10195         if (s1->seqid == s2->seqid)
10196                 return true;
10197
10198         return s1->seqid == 0 || s2->seqid == 0;
10199 }
10200
10201 #endif /* CONFIG_NFS_V4_1 */
10202
10203 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10204                 const nfs4_stateid *s2)
10205 {
10206         return nfs4_stateid_match(s1, s2);
10207 }
10208
10209
10210 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10211         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10212         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10213         .recover_open   = nfs4_open_reclaim,
10214         .recover_lock   = nfs4_lock_reclaim,
10215         .establish_clid = nfs4_init_clientid,
10216         .detect_trunking = nfs40_discover_server_trunking,
10217 };
10218
10219 #if defined(CONFIG_NFS_V4_1)
10220 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10221         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10222         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10223         .recover_open   = nfs4_open_reclaim,
10224         .recover_lock   = nfs4_lock_reclaim,
10225         .establish_clid = nfs41_init_clientid,
10226         .reclaim_complete = nfs41_proc_reclaim_complete,
10227         .detect_trunking = nfs41_discover_server_trunking,
10228 };
10229 #endif /* CONFIG_NFS_V4_1 */
10230
10231 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10232         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10233         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10234         .recover_open   = nfs40_open_expired,
10235         .recover_lock   = nfs4_lock_expired,
10236         .establish_clid = nfs4_init_clientid,
10237 };
10238
10239 #if defined(CONFIG_NFS_V4_1)
10240 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10241         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10242         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10243         .recover_open   = nfs41_open_expired,
10244         .recover_lock   = nfs41_lock_expired,
10245         .establish_clid = nfs41_init_clientid,
10246 };
10247 #endif /* CONFIG_NFS_V4_1 */
10248
10249 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10250         .sched_state_renewal = nfs4_proc_async_renew,
10251         .get_state_renewal_cred = nfs4_get_renew_cred,
10252         .renew_lease = nfs4_proc_renew,
10253 };
10254
10255 #if defined(CONFIG_NFS_V4_1)
10256 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10257         .sched_state_renewal = nfs41_proc_async_sequence,
10258         .get_state_renewal_cred = nfs4_get_machine_cred,
10259         .renew_lease = nfs4_proc_sequence,
10260 };
10261 #endif
10262
10263 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10264         .get_locations = _nfs40_proc_get_locations,
10265         .fsid_present = _nfs40_proc_fsid_present,
10266 };
10267
10268 #if defined(CONFIG_NFS_V4_1)
10269 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10270         .get_locations = _nfs41_proc_get_locations,
10271         .fsid_present = _nfs41_proc_fsid_present,
10272 };
10273 #endif  /* CONFIG_NFS_V4_1 */
10274
10275 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10276         .minor_version = 0,
10277         .init_caps = NFS_CAP_READDIRPLUS
10278                 | NFS_CAP_ATOMIC_OPEN
10279                 | NFS_CAP_POSIX_LOCK,
10280         .init_client = nfs40_init_client,
10281         .shutdown_client = nfs40_shutdown_client,
10282         .match_stateid = nfs4_match_stateid,
10283         .find_root_sec = nfs4_find_root_sec,
10284         .free_lock_state = nfs4_release_lockowner,
10285         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10286         .alloc_seqid = nfs_alloc_seqid,
10287         .call_sync_ops = &nfs40_call_sync_ops,
10288         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10289         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10290         .state_renewal_ops = &nfs40_state_renewal_ops,
10291         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10292 };
10293
10294 #if defined(CONFIG_NFS_V4_1)
10295 static struct nfs_seqid *
10296 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10297 {
10298         return NULL;
10299 }
10300
10301 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10302         .minor_version = 1,
10303         .init_caps = NFS_CAP_READDIRPLUS
10304                 | NFS_CAP_ATOMIC_OPEN
10305                 | NFS_CAP_POSIX_LOCK
10306                 | NFS_CAP_STATEID_NFSV41
10307                 | NFS_CAP_ATOMIC_OPEN_V1
10308                 | NFS_CAP_LGOPEN,
10309         .init_client = nfs41_init_client,
10310         .shutdown_client = nfs41_shutdown_client,
10311         .match_stateid = nfs41_match_stateid,
10312         .find_root_sec = nfs41_find_root_sec,
10313         .free_lock_state = nfs41_free_lock_state,
10314         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10315         .alloc_seqid = nfs_alloc_no_seqid,
10316         .session_trunk = nfs4_test_session_trunk,
10317         .call_sync_ops = &nfs41_call_sync_ops,
10318         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10319         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10320         .state_renewal_ops = &nfs41_state_renewal_ops,
10321         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10322 };
10323 #endif
10324
10325 #if defined(CONFIG_NFS_V4_2)
10326 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10327         .minor_version = 2,
10328         .init_caps = NFS_CAP_READDIRPLUS
10329                 | NFS_CAP_ATOMIC_OPEN
10330                 | NFS_CAP_POSIX_LOCK
10331                 | NFS_CAP_STATEID_NFSV41
10332                 | NFS_CAP_ATOMIC_OPEN_V1
10333                 | NFS_CAP_LGOPEN
10334                 | NFS_CAP_ALLOCATE
10335                 | NFS_CAP_COPY
10336                 | NFS_CAP_OFFLOAD_CANCEL
10337                 | NFS_CAP_COPY_NOTIFY
10338                 | NFS_CAP_DEALLOCATE
10339                 | NFS_CAP_SEEK
10340                 | NFS_CAP_LAYOUTSTATS
10341                 | NFS_CAP_CLONE
10342                 | NFS_CAP_LAYOUTERROR
10343                 | NFS_CAP_READ_PLUS,
10344         .init_client = nfs41_init_client,
10345         .shutdown_client = nfs41_shutdown_client,
10346         .match_stateid = nfs41_match_stateid,
10347         .find_root_sec = nfs41_find_root_sec,
10348         .free_lock_state = nfs41_free_lock_state,
10349         .call_sync_ops = &nfs41_call_sync_ops,
10350         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10351         .alloc_seqid = nfs_alloc_no_seqid,
10352         .session_trunk = nfs4_test_session_trunk,
10353         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10354         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10355         .state_renewal_ops = &nfs41_state_renewal_ops,
10356         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10357 };
10358 #endif
10359
10360 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10361         [0] = &nfs_v4_0_minor_ops,
10362 #if defined(CONFIG_NFS_V4_1)
10363         [1] = &nfs_v4_1_minor_ops,
10364 #endif
10365 #if defined(CONFIG_NFS_V4_2)
10366         [2] = &nfs_v4_2_minor_ops,
10367 #endif
10368 };
10369
10370 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10371 {
10372         ssize_t error, error2, error3;
10373         size_t left = size;
10374
10375         error = generic_listxattr(dentry, list, left);
10376         if (error < 0)
10377                 return error;
10378         if (list) {
10379                 list += error;
10380                 left -= error;
10381         }
10382
10383         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, left);
10384         if (error2 < 0)
10385                 return error2;
10386
10387         if (list) {
10388                 list += error2;
10389                 left -= error2;
10390         }
10391
10392         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, left);
10393         if (error3 < 0)
10394                 return error3;
10395
10396         error += error2 + error3;
10397         if (size && error > size)
10398                 return -ERANGE;
10399         return error;
10400 }
10401
10402 static void nfs4_enable_swap(struct inode *inode)
10403 {
10404         /* The state manager thread must always be running.
10405          * It will notice the client is a swapper, and stay put.
10406          */
10407         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10408
10409         nfs4_schedule_state_manager(clp);
10410 }
10411
10412 static void nfs4_disable_swap(struct inode *inode)
10413 {
10414         /* The state manager thread will now exit once it is
10415          * woken.
10416          */
10417         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10418
10419         set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10420         clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10421         wake_up_var(&clp->cl_state);
10422 }
10423
10424 static const struct inode_operations nfs4_dir_inode_operations = {
10425         .create         = nfs_create,
10426         .lookup         = nfs_lookup,
10427         .atomic_open    = nfs_atomic_open,
10428         .link           = nfs_link,
10429         .unlink         = nfs_unlink,
10430         .symlink        = nfs_symlink,
10431         .mkdir          = nfs_mkdir,
10432         .rmdir          = nfs_rmdir,
10433         .mknod          = nfs_mknod,
10434         .rename         = nfs_rename,
10435         .permission     = nfs_permission,
10436         .getattr        = nfs_getattr,
10437         .setattr        = nfs_setattr,
10438         .listxattr      = nfs4_listxattr,
10439 };
10440
10441 static const struct inode_operations nfs4_file_inode_operations = {
10442         .permission     = nfs_permission,
10443         .getattr        = nfs_getattr,
10444         .setattr        = nfs_setattr,
10445         .listxattr      = nfs4_listxattr,
10446 };
10447
10448 const struct nfs_rpc_ops nfs_v4_clientops = {
10449         .version        = 4,                    /* protocol version */
10450         .dentry_ops     = &nfs4_dentry_operations,
10451         .dir_inode_ops  = &nfs4_dir_inode_operations,
10452         .file_inode_ops = &nfs4_file_inode_operations,
10453         .file_ops       = &nfs4_file_operations,
10454         .getroot        = nfs4_proc_get_root,
10455         .submount       = nfs4_submount,
10456         .try_get_tree   = nfs4_try_get_tree,
10457         .getattr        = nfs4_proc_getattr,
10458         .setattr        = nfs4_proc_setattr,
10459         .lookup         = nfs4_proc_lookup,
10460         .lookupp        = nfs4_proc_lookupp,
10461         .access         = nfs4_proc_access,
10462         .readlink       = nfs4_proc_readlink,
10463         .create         = nfs4_proc_create,
10464         .remove         = nfs4_proc_remove,
10465         .unlink_setup   = nfs4_proc_unlink_setup,
10466         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10467         .unlink_done    = nfs4_proc_unlink_done,
10468         .rename_setup   = nfs4_proc_rename_setup,
10469         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10470         .rename_done    = nfs4_proc_rename_done,
10471         .link           = nfs4_proc_link,
10472         .symlink        = nfs4_proc_symlink,
10473         .mkdir          = nfs4_proc_mkdir,
10474         .rmdir          = nfs4_proc_rmdir,
10475         .readdir        = nfs4_proc_readdir,
10476         .mknod          = nfs4_proc_mknod,
10477         .statfs         = nfs4_proc_statfs,
10478         .fsinfo         = nfs4_proc_fsinfo,
10479         .pathconf       = nfs4_proc_pathconf,
10480         .set_capabilities = nfs4_server_capabilities,
10481         .decode_dirent  = nfs4_decode_dirent,
10482         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10483         .read_setup     = nfs4_proc_read_setup,
10484         .read_done      = nfs4_read_done,
10485         .write_setup    = nfs4_proc_write_setup,
10486         .write_done     = nfs4_write_done,
10487         .commit_setup   = nfs4_proc_commit_setup,
10488         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10489         .commit_done    = nfs4_commit_done,
10490         .lock           = nfs4_proc_lock,
10491         .clear_acl_cache = nfs4_zap_acl_attr,
10492         .close_context  = nfs4_close_context,
10493         .open_context   = nfs4_atomic_open,
10494         .have_delegation = nfs4_have_delegation,
10495         .alloc_client   = nfs4_alloc_client,
10496         .init_client    = nfs4_init_client,
10497         .free_client    = nfs4_free_client,
10498         .create_server  = nfs4_create_server,
10499         .clone_server   = nfs_clone_server,
10500         .enable_swap    = nfs4_enable_swap,
10501         .disable_swap   = nfs4_disable_swap,
10502 };
10503
10504 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10505         .name   = XATTR_NAME_NFSV4_ACL,
10506         .list   = nfs4_xattr_list_nfs4_acl,
10507         .get    = nfs4_xattr_get_nfs4_acl,
10508         .set    = nfs4_xattr_set_nfs4_acl,
10509 };
10510
10511 #ifdef CONFIG_NFS_V4_2
10512 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10513         .prefix = XATTR_USER_PREFIX,
10514         .get    = nfs4_xattr_get_nfs4_user,
10515         .set    = nfs4_xattr_set_nfs4_user,
10516 };
10517 #endif
10518
10519 const struct xattr_handler *nfs4_xattr_handlers[] = {
10520         &nfs4_xattr_nfs4_acl_handler,
10521 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10522         &nfs4_xattr_nfs4_label_handler,
10523 #endif
10524 #ifdef CONFIG_NFS_V4_2
10525         &nfs4_xattr_nfs4_user_handler,
10526 #endif
10527         NULL
10528 };
10529
10530 /*
10531  * Local variables:
10532  *  c-basic-offset: 8
10533  * End:
10534  */